Reproducing Bouchard et al. (2013) from dandiset 000019¶
Paper. Bouchard KE, Mesgarani N, Johnson K, Chang EF (2013). Functional organization of
human sensorimotor cortex for speech articulation. Nature 495, 327-332.
doi:10.1038/nature11911. The claims file
(claims/000019/10.1038-nature11911.yaml) was written from the NIH author manuscript in PubMed
Central (PMC3606666) with its supplement, and checked by a second reader.
Dandiset. DANDI:000019, "Human ECoG speaking consonant-vowel syllables", version 0.220126.2148. The paper cites no version of the dandiset (it predates it). This is the only published version; the draft has the same 31 assets and the same size, so the published version is pinned.
What the dandiset holds. Raw ECoG (256 channels, 3051.76 Hz) of 31 recording blocks from four
subjects (EC2, EC9, GP31, GP33), with a trial table per block (syllable, start, C-V transition
and stop times), a bad flag and an anatomical label per electrode, 3-D electrode coordinates for
two subjects, and a rest period per block. The microphone signal was removed (dandiset
description), and no high gamma is stored, so high gamma is computed here from the raw voltage.
The paper's subjects are EC2, EC33 and EC31; EC2, GP33 and GP31 are taken to be them (see the data
check). EC9 is not in the paper and is not analysed.
Readings. Where the claims file leaves a method detail open it names readings, and every
reading is run; a claim's outcome is the one all its method readings give, and it is blocked as
methods_underspecified when they differ. The first run of this attempt asked five questions
(questions/000019.yaml), answered by a reader on 2026-10-22, and the claims file now names
these readings, all run here:
- the baseline of the z-scores: (a)
pre, -1000 to -500 ms before each C-V transition; (b)whole, the whole block; (c)rest, the rest period the dataset marks as baseline (a reading since the answer; the first run showed it as a variant only); - the CVs averaged: (a) every CV the dataset holds (
all); (b) only CVs with 15 or more renditions (15+); the paper does not say which; - the z-scoring before the PCA:
row, each electrode's whole concatenated (CV x time) series;time, each electrode across CVs at each time; the paper does not say which; - the input projected onto the PCs to give the states: (i) the z-scored matrix given to the PCA; (ii) the baseline z-scored high gamma, as Methods Eq. 5 writes it (added with the answer);
- the acoustic onset of fig3d: the dandiset has no audio, and the hand-marked syllable start may stand in for it as a deviation, with the result against t = 0 shown beside it (the answer).
The fifth answer says the paper gives no way to derive the distances from the central sulcus and Sylvian fissure from grid geometry or unlabelled coordinates, so the spatial claims stay blocked.
| Claim | Tested here |
|---|---|
| data-composition | tested |
| active-electrodes | tested |
| fig1c-e-articulator-electrodes | run; its criterion blocks it (metadata_insufficient) when no subject and numbering puts the six named electrodes on vSMC, as here |
| fig1f-h-phonetic-contrasts | tested |
| fig2a-lips-dorsal | blocked: metadata_insufficient (no positions relative to the central sulcus and Sylvian fissure) |
| fig2b-somatotopic-sequence | blocked: metadata_insufficient |
| suppfig3b-spatial-clustering | blocked: metadata_insufficient |
| suppfig3d-cross-subject | blocked: metadata_insufficient |
| suppfig3f-dv-sequence | blocked: metadata_insufficient |
| suppfig5-intermediate-organization | blocked: metadata_insufficient |
| suppfig6-mixed-tuning | tested |
| fig3d-timing-sequence | tested |
| suppfig7-no-latency-difference | blocked: metadata_insufficient |
| pca-nine-components | tested |
| fig4a-consonant-clusters | tested |
| suppfig11-co-clustering | tested |
| fig4c-d-hierarchy | tested |
| fig4ef-feature-correlations | tested |
| fig5c-separability-timing | tested |
| fig5d-separability-coupling | tested |
| suppfig13-consonant-vowel-distance | tested |
| suppfig9-pc-articulator-weights | tested |
In [1]:
import itertools
import time
import warnings
import h5py
import matplotlib.pyplot as plt
import numpy as np
import pandas as pd
import remfile
from scipy import stats
from scipy.cluster.hierarchy import fcluster, linkage
from scipy.spatial.distance import pdist, squareform
from sklearn.metrics import silhouette_score
import bouchard as B
pd.set_option("display.width", 220)
pd.set_option("display.max_columns", 40)
pd.set_option("display.max_rows", 300)
warnings.filterwarnings("ignore", category=RuntimeWarning)
started = time.time()
In [2]:
# Metadata of every file (EC9 included), and a fingerprint of its first samples, read by streaming.
ASSETS = B.assets()
meta_rows = []
for a in ASSETS:
with h5py.File(remfile.File(a["url"]), "r") as h5:
m = B.read_metadata(h5)
m["fingerprint"] = h5["acquisition/ElectricalSeries/data"][:2000, :4].copy()
m["path"], m["size"] = a["path"], a["size"]
meta_rows.append(m)
print(f"read metadata of {len(meta_rows)} files of dandiset {B.DANDISET_ID} version {B.DANDISET_VERSION} "
f"in {time.time() - started:.0f} s")
read metadata of 31 files of dandiset 000019 version 0.220126.2148 in 29 s
In [3]:
# High gamma of the three subjects of the paper: every block is streamed into memory and processed.
blocks = {}
for s in B.SUBJECTS:
blocks[s] = B.process_blocks([a for a in ASSETS if a["path"].startswith(f"sub-{s}/")],
log=lambda msg: None)
S = {s: B.load_subject(blocks[s]) for s in B.SUBJECTS}
# What each file holds besides the tables: from the processed files in memory, and for EC9 (not
# analysed) from its first file, streamed.
CONTENTS = {a["path"]: b["contents"] for s in B.SUBJECTS for a, b in zip([a for a in ASSETS if a["path"].startswith(f"sub-{s}/")], blocks[s])}
ec9 = next(a for a in ASSETS if a["path"].startswith("sub-EC9/"))
with h5py.File(remfile.File(ec9["url"]), "r") as h5:
CONTENTS[ec9["path"]] = B.describe_file(h5)
print(f"processed {sum(len(v) for v in blocks.values())} blocks in {time.time() - started:.0f} s")
streamed = sum(a["size"] for a in ASSETS if a["path"].split("/")[0][4:] in B.SUBJECTS)
print(f"streamed {streamed / 1e9:.1f} GB of raw ECoG")
processed 24 blocks in 1014 s streamed 39.0 GB of raw ECoG
Data check¶
In [4]:
md = B.dandiset_metadata()
print("Dandiset:", md["name"], "| version", B.DANDISET_VERSION, "| DOI", md.get("doi"))
print("Related resource:", [(r.get("relation"), r.get("identifier"), r.get("url")) for r in md.get("relatedResource", [])])
print("Assets summary: subjects", md["assetsSummary"].get("numberOfSubjects"), "| files", md["assetsSummary"].get("numberOfFiles"))
files = pd.DataFrame([{
"file": m["path"].split("/")[1], "subject": m["subject"], "rate_hz": m["rate"],
"minutes": m["n_samples"] / m["rate"] / 60, "channels": m["n_channels"], "unit": m["unit"],
"conversion": m["conversion"], "trials": len(m["condition"]), "speak": int(m["speak"].sum()),
"bad": int(m["bad"].sum()), "rest_s": sum(b - a for a, b in m["rest"]),
"invalid_intervals": len(m["invalid"]), "has_xyz": bool(np.isfinite(m["x"]).any()),
"mean_x": np.nanmean(m["x"]) if np.isfinite(m["x"]).any() else np.nan,
"electrode_ids": f"{m['electrode_ids'][0]}-{m['electrode_ids'][-1]}"} for m in meta_rows])
print(files.round(2).to_string(index=False))
print("\nFiles and trials per subject:")
print(files.groupby("subject").agg(files=("file", "size"), trials=("trials", "sum"), minutes=("minutes", "sum")).round(1).to_string())
print("\nSampling rates found:", sorted(files.rate_hz.unique()), "(Methods: 3052 Hz)")
print("Trials not marked speak=True, per subject:", (files.trials - files.speak).groupby(files.subject).sum().to_dict())
# What each file holds besides the trial and electrode tables, read from the files.
subj_of = {m["path"]: m["subject"] for m in meta_rows}
objs = pd.DataFrame([{"subject": subj_of[f], "path": o[0], "type": o[1], "shape": o[2]}
for f, c in CONTENTS.items() for o in c["objects"]])
print(f"\nFiles described: {len(CONTENTS)} (every file of EC2, GP31 and GP33, and the first of EC9's {sum(m['subject'] == 'EC9' for m in meta_rows)})")
top = objs[objs.path.str.count("/") <= 2].groupby(["path", "type"]).subject.agg(lambda x: f"{x.nunique()} subjects, {len(x)} files")
print("\nObjects under acquisition, processing, analysis and stimulus (two levels), over all files:")
print(top.to_string())
print("Objects with a microphone or audio name:", sorted({p for p in objs.path if any(k in p.lower() for k in ("audio", "micro", "sound", "speech"))}) or "none")
print("Datasets under processing/, with shapes, one file per subject:")
for s in ["EC2", "EC9", "GP31", "GP33"]:
o = objs[(objs.subject == s) & objs.path.str.startswith("processing/") & objs["shape"].notna()].drop_duplicates("path")
print(f" {s}:", o[["path", "shape"]].values.tolist() or "none",
"| files with processing/:", sum(any(x[0].startswith("processing") for x in c["objects"]) for f, c in CONTENTS.items() if subj_of[f] == s),
"of", sum(subj_of[f] == s for f in CONTENTS), "described")
spec = [f for f, c in CONTENTS.items() if any(x[1] == "Spectrum" for x in c["objects"])][:1]
if spec:
with h5py.File(remfile.File(next(a["url"] for a in ASSETS if a["path"] == spec[0])), "r") as h5:
fr = h5["processing/ecephys/spectrum/frequencies"][:]
pw = h5["processing/ecephys/spectrum/power"]
print(f" {spec[0]}: Spectrum frequencies {fr.min():.2f} to {fr.max():.1f} Hz in {len(fr)} bins; power {pw.shape} "
f"(frequency x channel); a power spectrum of the ECoG channels, not a time series")
first_of = {s: next(c for f, c in CONTENTS.items() if subj_of[f] == s) for s in ["EC2", "EC9", "GP31", "GP33"]}
print("\nElectrode groups (name: location, description) in each described file, grouped:")
for s in ["EC2", "EC9", "GP31", "GP33"]:
kinds = pd.Series({f.split("_ses-")[1]: "; ".join(f"{k}: {v.get('location')}, {v.get('description')}" for k, v in c["electrode_groups"].items())
for f, c in CONTENTS.items() if subj_of[f] == s})
print(f" {s}:", {k: sorted(g.index) for k, g in kinds.groupby(kinds)})
print("Subject fields identical in all described files of a subject:",
all(c["subject"] == first_of[subj_of[f]]["subject"] for f, c in CONTENTS.items()))
print("Subject and description fields, first file per subject:")
for s in ["EC2", "EC9", "GP31", "GP33"]:
c = first_of[s]
print(f" {s}: subject {c['subject']}; electrode groups {c['electrode_groups']}; {c['text']}")
print(f" electrode table columns: {c['electrode_columns']}")
Dandiset: Human ECoG speaking consonant-vowel syllables | version 0.220126.2148 | DOI 10.48324/dandi.000019/0.220126.2148
Related resource: [('dcite:IsDescribedBy', 'DOI:0.1038/nature11911', 'https://doi.org/10.1038/nature11911')]
Assets summary: subjects 4 | files 31
file subject rate_hz minutes channels unit conversion trials speak bad rest_s invalid_intervals has_xyz mean_x electrode_ids
sub-EC2_ses-EC2-B1.nwb EC2 3051.76 10.78 256 volt 0.0 387 387 0 41.00 6 True -57.1 0-255
sub-EC2_ses-EC2-B105.nwb EC2 3051.76 11.62 256 volt 0.0 311 311 0 10.00 5 True -57.1 0-255
sub-EC2_ses-EC2-B15.nwb EC2 3051.76 9.58 256 volt 0.0 382 382 7 57.00 7 True -57.1 0-255
sub-EC2_ses-EC2-B76.nwb EC2 3051.76 11.20 256 volt 0.0 389 389 10 57.00 28 True -57.1 0-255
sub-EC2_ses-EC2-B8.nwb EC2 3051.76 10.02 256 volt 0.0 384 384 7 15.00 4 True -57.1 0-255
sub-EC2_ses-EC2-B89.nwb EC2 3051.76 10.00 256 volt 0.0 397 397 10 61.00 11 True -57.1 0-255
sub-EC2_ses-EC2-B9.nwb EC2 3051.76 9.42 256 volt 0.0 378 378 9 45.00 13 True -57.1 0-255
sub-EC9_ses-EC9-B15.nwb EC9 3051.76 3.25 256 volt 0.0 70 70 0 7.00 25 False NaN 1-256
sub-EC9_ses-EC9-B39.nwb EC9 3051.76 15.62 256 volt 0.0 280 278 1 6.00 12 False NaN 1-256
sub-EC9_ses-EC9-B46.nwb EC9 3051.76 15.92 256 volt 0.0 227 226 1 10.00 1 False NaN 1-256
sub-EC9_ses-EC9-B49.nwb EC9 3051.76 15.62 256 volt 0.0 254 251 4 24.00 0 False NaN 1-256
sub-EC9_ses-EC9-B53.nwb EC9 3051.76 15.78 256 volt 0.0 274 273 5 16.00 17 False NaN 1-256
sub-EC9_ses-EC9-B60.nwb EC9 3051.76 15.90 256 volt 0.0 275 274 2 13.00 9 False NaN 1-256
sub-EC9_ses-EC9-B63.nwb EC9 3051.76 16.37 256 volt 0.0 273 273 3 12.00 10 False NaN 1-256
sub-GP31_ses-GP31-B1.nwb GP31 3051.76 6.70 256 volt 0.0 237 237 9 5.00 0 True -56.8 0-255
sub-GP31_ses-GP31-B2.nwb GP31 3051.76 8.78 256 volt 0.0 315 315 12 20.00 1 True -56.8 0-255
sub-GP31_ses-GP31-B21.nwb GP31 3051.76 7.27 256 volt 0.0 237 237 11 30.00 0 True -56.8 0-255
sub-GP31_ses-GP31-B4.nwb GP31 3051.76 7.75 256 volt 0.0 224 224 9 60.00 0 True -56.8 0-255
sub-GP31_ses-GP31-B6.nwb GP31 3051.76 10.25 256 volt 0.0 377 377 9 3.30 1 True -56.8 0-255
sub-GP31_ses-GP31-B63.nwb GP31 3051.76 7.35 256 volt 0.0 374 374 8 12.00 36 True -56.8 0-255
sub-GP31_ses-GP31-B65.nwb GP31 3051.76 6.47 256 volt 0.0 324 324 8 8.00 0 True -56.8 0-255
sub-GP31_ses-GP31-B67.nwb GP31 3051.76 7.95 256 volt 0.0 429 429 8 4.10 0 True -56.8 0-255
sub-GP31_ses-GP31-B69.nwb GP31 3051.76 8.92 256 volt 0.0 430 430 8 6.00 6 True -56.8 0-255
sub-GP31_ses-GP31-B71.nwb GP31 3051.76 11.08 256 volt 0.0 481 481 8 16.00 0 True -56.8 0-255
sub-GP31_ses-GP31-B78.nwb GP31 3051.76 10.42 256 volt 0.0 424 424 10 18.00 1 True -56.8 0-255
sub-GP31_ses-GP31-B82.nwb GP31 3051.76 11.77 256 volt 0.0 529 529 12 55.00 5 True -56.8 0-255
sub-GP31_ses-GP31-B83.nwb GP31 3051.76 12.00 256 volt 0.0 582 582 9 4.00 0 True -56.8 0-255
sub-GP31_ses-GP31-B9.nwb GP31 3051.76 9.60 256 volt 0.0 322 322 11 50.00 50 True -56.8 0-255
sub-GP33_ses-GP33-B1.nwb GP33 3051.76 9.12 256 volt 0.0 339 339 9 77.97 0 False NaN 1-256
sub-GP33_ses-GP33-B30.nwb GP33 3051.76 10.97 256 volt 0.0 612 612 26 49.57 26 False NaN 1-256
sub-GP33_ses-GP33-B5.nwb GP33 3051.76 11.63 256 volt 0.0 507 507 8 54.62 14 False NaN 1-256
Files and trials per subject:
files trials minutes
subject
EC2 7 2628 72.6
EC9 7 1653 98.5
GP31 14 5285 126.3
GP33 3 1458 31.7
Sampling rates found: [np.float64(3051.7578), np.float64(3051.7578125)] (Methods: 3052 Hz)
Trials not marked speak=True, per subject: {'EC2': 0, 'EC9': 8, 'GP31': 0, 'GP33': 0}
Files described: 25 (every file of EC2, GP31 and GP33, and the first of EC9's 7)
Objects under acquisition, processing, analysis and stimulus (two levels), over all files:
path type
acquisition/ElectricalSeries ElectricalSeries 4 subjects, 25 files
acquisition/ElectricalSeries/data 4 subjects, 25 files
acquisition/ElectricalSeries/electrodes DynamicTableRegion 4 subjects, 25 files
acquisition/ElectricalSeries/starting_time 4 subjects, 25 files
processing/ecephys ProcessingModule 1 subjects, 1 files
processing/ecephys/spectrum Spectrum 1 subjects, 1 files
stimulus/presentation 4 subjects, 25 files
stimulus/templates 4 subjects, 25 files
Objects with a microphone or audio name: none
Datasets under processing/, with shapes, one file per subject:
EC2: [['processing/ecephys/spectrum/frequencies', (15259,)], ['processing/ecephys/spectrum/power', (15259, 256)]] | files with processing/: 1 of 7 described
EC9: none | files with processing/: 0 of 1 described
GP31: none | files with processing/: 0 of 14 described
GP33: none | files with processing/: 0 of 3 described
sub-EC2/sub-EC2_ses-EC2-B1.nwb: Spectrum frequencies 0.00 to 1525.8 Hz in 15259 bins; power (15259, 256) (frequency x channel); a power spectrum of the ECoG channels, not a time series
Electrode groups (name: location, description) in each described file, grouped:
EC2: {'L256Grid electrodes: grid, L256Grid': ['EC2-B1.nwb', 'EC2-B105.nwb', 'EC2-B15.nwb', 'EC2-B76.nwb', 'EC2-B8.nwb', 'EC2-B89.nwb', 'EC2-B9.nwb']}
EC9: {'256Grid electrodes: location, auto_group': ['EC9-B15.nwb']}
GP31: {'L256Grid electrodes: grid, L256Grid': ['GP31-B1.nwb', 'GP31-B2.nwb', 'GP31-B21.nwb', 'GP31-B4.nwb', 'GP31-B6.nwb', 'GP31-B63.nwb', 'GP31-B65.nwb', 'GP31-B67.nwb', 'GP31-B69.nwb', 'GP31-B71.nwb', 'GP31-B78.nwb', 'GP31-B82.nwb', 'GP31-B83.nwb', 'GP31-B9.nwb']}
GP33: {'auto_group: location, auto_group': ['GP33-B1.nwb', 'GP33-B30.nwb', 'GP33-B5.nwb']}
Subject fields identical in all described files of a subject: True
Subject and description fields, first file per subject:
EC2: subject {'species': 'Homo sapiens', 'subject_id': 'EC2'}; electrode groups {'L256Grid electrodes': {'location': 'grid', 'description': 'L256Grid'}}; {'institution': 'University of California, San Francisco', 'lab': 'Chang Lab', 'session_description': 'EC2_B1'}
electrode table columns: {'bad': 'bool', 'filtering': ['none'], 'group_name': ['L256Grid electrodes'], 'id': 'int64', 'imp': 'float64', 'location': ['caudalmiddlefrontal', 'middletemporal', 'parsopercularis', 'postcentral', 'precentral', 'superiortemporal'], 'x': 'float64', 'y': 'float64', 'z': 'float64'}
EC9: subject {'species': 'Homo sapiens', 'subject_id': 'EC9'}; electrode groups {'256Grid electrodes': {'location': 'location', 'description': 'auto_group'}}; {'institution': 'University of California, San Francisco', 'lab': 'Chang Lab', 'session_description': 'EC9_B15'}
electrode table columns: {'bad': 'bool', 'filtering': ['none'], 'group_name': ['256Grid electrodes'], 'id': 'int64', 'imp': 'float64', 'location': ['', 'postcentral', 'precentral'], 'x': 'float64', 'y': 'float64', 'z': 'float64'}
GP31: subject {'species': 'Homo sapiens', 'subject_id': 'GP31'}; electrode groups {'L256Grid electrodes': {'location': 'grid', 'description': 'L256Grid'}}; {'institution': 'University of California, San Francisco', 'lab': 'Chang Lab', 'session_description': 'GP31_B1'}
electrode table columns: {'bad': 'bool', 'filtering': ['none'], 'group_name': ['L256Grid electrodes'], 'id': 'int64', 'imp': 'float64', 'location': ['caudalmiddlefrontal', 'middletemporal', 'parsopercularis', 'postcentral', 'precentral', 'superiortemporal'], 'x': 'float64', 'y': 'float64', 'z': 'float64'}
GP33: subject {'species': 'Homo sapiens', 'subject_id': 'GP33'}; electrode groups {'auto_group': {'location': 'location', 'description': 'auto_group'}}; {'institution': 'University of California, San Francisco', 'lab': 'Chang Lab', 'session_description': 'GP33_B1'}
electrode table columns: {'bad': 'bool', 'filtering': ['none'], 'group_name': ['auto_group'], 'id': 'int64', 'imp': 'float64', 'location': ['caudalmiddlefrontal', 'parsopercularis', 'postcentral', 'precentral', 'superiortemporal'], 'x': 'float64', 'y': 'float64', 'z': 'float64'}
In [5]:
# Duplicates: blocks with the same length, the same trial table, or the same first samples.
pairs = []
for i, j in itertools.combinations(range(len(meta_rows)), 2):
a, b = meta_rows[i], meta_rows[j]
same_n = a["n_samples"] == b["n_samples"]
same_trials = len(a["cv"]) == len(b["cv"]) and np.allclose(a["cv"], b["cv"])
same_data = np.array_equal(a["fingerprint"], b["fingerprint"])
if same_n or same_trials or same_data:
pairs.append((a["path"], b["path"], same_n, same_trials, same_data))
print("Pairs sharing sample count, trial times or first 2000 samples of 4 channels:")
for p in pairs:
print(" ", p)
Pairs sharing sample count, trial times or first 2000 samples of 4 channels:
('sub-EC9/sub-EC9_ses-EC9-B39.nwb', 'sub-EC9/sub-EC9_ses-EC9-B49.nwb', True, False, False)
Only one pair of files shares a property (two EC9 blocks with the same number of samples); their trial times and first samples differ, so no recording was deposited twice.
In [6]:
# Syllables per subject, from the trial labels.
label_rows = []
for m in meta_rows:
for c in m["condition"]:
label_rows.append({"subject": m["subject"], "label": c, "cv": B.parse_label(c)})
labels = pd.DataFrame(label_rows)
for s, g in labels.groupby("subject"):
unparsed = g[g.cv.isna()].label.value_counts().to_dict()
cv = g.dropna().cv
cons = sorted({c for c, v in cv}, key=B.CONSONANTS.index)
counts = cv.value_counts()
print(f"{s}: {len(g)} trials, {counts.size} distinct CVs, {len(cons)} consonants; renditions per CV "
f"{counts.min()} to {counts.max()}; labels that are not CVs of the task: {unparsed}")
missing = [c + v for c in B.CONSONANTS for v in "aiu" if (c, v) not in set(cv)]
print(f" CVs missing: {missing}; CVs with fewer than 15 renditions: "
f"{ {''.join(k): int(n) for k, n in counts.items() if n < 15} }; more than 100: "
f"{ {''.join(k): int(n) for k, n in counts.items() if n > 100} }")
print("Labels merged as misspellings:", labels[labels.label.isin(list(B.SPELLING))].groupby(["subject", "label"]).size().to_dict())
print("Spellings merged: 'she' -> 'shee', 'ha' -> 'haa'. Paper: 19, 19 and 18 consonants, 168 CVs (57 + 57 + 54), 15 to 100 renditions each.")
EC2: 2628 trials, 57 distinct CVs, 19 consonants; renditions per CV 10 to 58; labels that are not CVs of the task: {}
CVs missing: []; CVs with fewer than 15 renditions: {'thi': 10}; more than 100: {}
EC9: 1653 trials, 57 distinct CVs, 19 consonants; renditions per CV 7 to 46; labels that are not CVs of the task: {np.str_('dhee'): 8, np.str_(''): 6, np.str_('dhoo'): 6, np.str_('chee'): 5, np.str_('dhaa'): 5, np.str_('jee'): 2, np.str_('skee'): 1, np.str_('ya'): 1, np.str_('kay'): 1, np.str_('chaa'): 1, np.str_('choo'): 1}
CVs missing: []; CVs with fewer than 15 renditions: {'thi': 12, 'thu': 10, 'tha': 7}; more than 100: {}
GP31: 5285 trials, 55 distinct CVs, 19 consonants; renditions per CV 2 to 107; labels that are not CVs of the task: {}
CVs missing: ['tha', 'thu']; CVs with fewer than 15 renditions: {'thi': 2}; more than 100: {'la': 107, 'di': 106, 'da': 105, 'ga': 105, 'sha': 105, 'ha': 105, 'bi': 105, 'ni': 105, 'hi': 105, 'li': 105, 'yu': 105, 'ka': 104, 'na': 104, 'sa': 104, 'va': 104, 'ti': 104, 'ki': 104, 'mi': 104, 'shi': 104, 'si': 104, 'du': 104, 'nu': 104, 'ba': 103, 'ta': 103, 'shu': 103, 'hu': 103, 'ma': 102, 'ra': 102, 'ya': 102, 'wi': 102, 'tu': 102, 'lu': 102, 'wa': 101}
GP33: 1458 trials, 57 distinct CVs, 19 consonants; renditions per CV 22 to 28; labels that are not CVs of the task: {}
CVs missing: []; CVs with fewer than 15 renditions: {}; more than 100: {}
Labels merged as misspellings: {('EC2', np.str_('she')): 2, ('EC9', np.str_('ha')): 1, ('EC9', np.str_('she')): 1, ('GP33', np.str_('ha')): 2}
Spellings merged: 'she' -> 'shee', 'ha' -> 'haa'. Paper: 19, 19 and 18 consonants, 168 CVs (57 + 57 + 54), 15 to 100 renditions each.
In [7]:
# Electrodes: labels, bad flags, coordinates.
for s in ["EC2", "EC9", "GP31", "GP33"]:
ms = [m for m in meta_rows if m["subject"] == s]
same_loc = all(np.array_equal(m["location"], ms[0]["location"]) for m in ms)
loc = pd.Series(ms[0]["location"]).replace("", "(none)").value_counts().to_dict()
badsets = sorted({tuple(np.where(m["bad"])[0]) for m in ms}, key=len)
print(f"{s}: labels identical in all blocks {same_loc}; labels {loc}")
print(f" bad-channel sets over blocks: {[list(b) for b in badsets]}")
x = ms[0]["x"]
if np.isfinite(x).any():
print(f" x {np.nanmin(x):.1f} to {np.nanmax(x):.1f}, y {np.nanmin(ms[0]['y']):.1f} to {np.nanmax(ms[0]['y']):.1f}, "
f"z {np.nanmin(ms[0]['z']):.1f} to {np.nanmax(ms[0]['z']):.1f} (no unit or frame given); "
f"neighbour spacing median {np.median(np.linalg.norm(np.diff(np.c_[x, ms[0]['y'], ms[0]['z']], axis=0), axis=1)):.2f}")
else:
print(" no coordinates (x, y, z are NaN)")
for s in B.SUBJECTS:
print(f"{s}: vSMC (precentral or postcentral, good in every block) {len(S[s]['vsmc'])} electrodes; "
f"electrodes bad in any block {int((~S[s]['good']).sum())}")
print("vSMC electrodes over the three subjects:", sum(len(S[s]["vsmc"]) for s in B.SUBJECTS), "(Supplementary Figure 9: n = 266)")
EC2: labels identical in all blocks True; labels {'superiortemporal': 51, 'postcentral': 44, 'precentral': 44, 'middletemporal': 37, 'supramarginal': 27, 'parsopercularis': 27, 'caudalmiddlefrontal': 26}
bad-channel sets over blocks: [[], [np.int64(48), np.int64(112), np.int64(113), np.int64(114), np.int64(115), np.int64(123), np.int64(127)], [np.int64(48), np.int64(112), np.int64(113), np.int64(114), np.int64(115), np.int64(123), np.int64(127), np.int64(192), np.int64(240)], [np.int64(48), np.int64(112), np.int64(113), np.int64(114), np.int64(115), np.int64(123), np.int64(127), np.int64(192), np.int64(238), np.int64(240)]]
x -62.6 to -34.6, y -37.5 to 28.4, z -19.0 to 54.0 (no unit or frame given); neighbour spacing median 4.38
EC9: labels identical in all blocks False; labels {'(none)': 177, 'precentral': 41, 'postcentral': 38}
bad-channel sets over blocks: [[], [np.int64(89)], [np.int64(89), np.int64(255)], [np.int64(64), np.int64(65), np.int64(66)], [np.int64(89), np.int64(244), np.int64(245), np.int64(255)], [np.int64(64), np.int64(65), np.int64(66), np.int64(89), np.int64(255)]]
no coordinates (x, y, z are NaN)
GP31: labels identical in all blocks True; labels {'superiortemporal': 59, 'middletemporal': 54, 'postcentral': 44, 'precentral': 42, 'parsopercularis': 24, 'supramarginal': 22, 'caudalmiddlefrontal': 11}
bad-channel sets over blocks: [[np.int64(48), np.int64(112), np.int64(113), np.int64(114), np.int64(115), np.int64(123), np.int64(127), np.int64(255)], [np.int64(48), np.int64(64), np.int64(66), np.int64(112), np.int64(113), np.int64(114), np.int64(115), np.int64(123), np.int64(255)], [np.int64(18), np.int64(48), np.int64(112), np.int64(113), np.int64(114), np.int64(115), np.int64(123), np.int64(127), np.int64(255)], [np.int64(48), np.int64(64), np.int64(66), np.int64(112), np.int64(113), np.int64(114), np.int64(115), np.int64(123), np.int64(127), np.int64(255)], [np.int64(48), np.int64(64), np.int64(65), np.int64(66), np.int64(112), np.int64(113), np.int64(114), np.int64(115), np.int64(123), np.int64(127), np.int64(255)], [np.int64(48), np.int64(64), np.int64(66), np.int64(112), np.int64(113), np.int64(114), np.int64(115), np.int64(123), np.int64(127), np.int64(195), np.int64(201), np.int64(255)], [np.int64(18), np.int64(48), np.int64(64), np.int64(65), np.int64(66), np.int64(112), np.int64(113), np.int64(114), np.int64(115), np.int64(123), np.int64(127), np.int64(255)]]
x -61.6 to -44.3, y -31.1 to 40.2, z -11.9 to 60.5 (no unit or frame given); neighbour spacing median 3.86
GP33: labels identical in all blocks True; labels {'precentral': 77, 'parsopercularis': 58, 'caudalmiddlefrontal': 52, 'superiortemporal': 43, 'postcentral': 26}
bad-channel sets over blocks: [[np.int64(48), np.int64(112), np.int64(113), np.int64(114), np.int64(115), np.int64(123), np.int64(127), np.int64(255)], [np.int64(48), np.int64(112), np.int64(113), np.int64(114), np.int64(115), np.int64(123), np.int64(127), np.int64(192), np.int64(255)], [np.int64(18), np.int64(19), np.int64(48), np.int64(112), np.int64(113), np.int64(114), np.int64(115), np.int64(123), np.int64(127), np.int64(208), np.int64(209), np.int64(210), np.int64(211), np.int64(212), np.int64(213), np.int64(214), np.int64(215), np.int64(216), np.int64(217), np.int64(218), np.int64(219), np.int64(220), np.int64(221), np.int64(222), np.int64(223), np.int64(255)]]
no coordinates (x, y, z are NaN)
EC2: vSMC (precentral or postcentral, good in every block) 86 electrodes; electrodes bad in any block 10
GP31: vSMC (precentral or postcentral, good in every block) 83 electrodes; electrodes bad in any block 14
GP33: vSMC (precentral or postcentral, good in every block) 99 electrodes; electrodes bad in any block 27
vSMC electrodes over the three subjects: 268 (Supplementary Figure 9: n = 266)
The archive's summary counts four subjects. EC2 matches the paper's EC2 by name. GP31 and GP33
are not names the paper uses; the paper's other two subjects are EC31 (left hemisphere) and
EC33 (right). No field in the files links them: the subject fields hold only the species and the
subject id, and the session descriptions name only the block (printed above). In EC2 and GP31 the
electrode group is named L256Grid electrodes, and every electrode has a negative x; both fit a
left hemisphere if the L and the x mean what they usually do. GP33's group is auto_group, with
location location, and GP33 has no coordinates, so its hemisphere cannot be checked. GP31 is the
subject with 18 consonants that have all three vowels (/tha/ and /thu/ are missing, table of
data-composition), which fits the paper's subject with 18 consonants and 54 CVs. That GP31 is EC31
and GP33 is EC33 is an assumption the data are consistent with, not shown by them. EC9 has 23
consonant labels (including /ch/, /dh/, /j/, /sk/), labels on only 79 of 256 electrodes and no
coordinates; it is not one of the paper's subjects and is left out.
Besides the raw series, one file (EC2 block 1) has a processing module holding a power spectrum of the 256 channels (frequency by channel, no time axis); no file holds audio or a microphone series. The kept trials whose epoch overlaps an interval marked as an ECoG artifact are counted once per trial (162); counted once per trial and interval, as in the first version of this check, they are 177.
In [8]:
# Trials kept, trials in invalid intervals, rest periods, and the high-gamma statistics.
rows = []
for s in B.SUBJECTS:
for b in blocks[s]:
cv = b["cv"]
pairs_inv = sum(((cv[b["kept"]] + 0.75 > a0) & (cv[b["kept"]] - 0.5 < a1)).sum() for a0, a1 in b["invalid"])
g = ~b["bad"]
rows.append({"subject": s, "block": b["session"], "trials": len(cv), "kept": len(b["kept"]),
"kept_in_invalid": int(b["in_invalid"].sum()), "trial_interval_overlaps": int(pairs_inv), "rest_s": round(b["rest_seconds"], 1),
"hg_whole_mean_uV": float(np.nanmedian(b["hg_stats"]["whole"][0][g]) * 1e6),
"hg_whole_sd_over_mean": float(np.nanmedian(b["hg_stats"]["whole"][1][g] / b["hg_stats"]["whole"][0][g])),
"pre_mean_over_whole": float(np.nanmedian(b["hg_stats"]["pre"][0][g] / b["hg_stats"]["whole"][0][g])),
"rest_mean_over_whole": float(np.nanmedian(b["hg_stats"]["rest"][0][g] / b["hg_stats"]["whole"][0][g])),
"rest_sd_over_whole": float(np.nanmedian(b["hg_stats"]["rest"][1][g] / b["hg_stats"]["whole"][1][g])),
"env_rate": round(b["env_rate"], 3)})
blk = pd.DataFrame(rows)
print(blk.round(3).to_string(index=False))
print("\nTrials dropped (epoch -1000 to +750 ms outside the recording):", int((blk.trials - blk.kept).sum()))
print("Kept trials whose epoch overlaps an 'ECoG artifact' invalid interval (not excluded; the claims file does not say to):",
int(blk.kept_in_invalid.sum()), "of", int(blk.kept.sum()), "| overlaps counted per trial and interval:", int(blk.trial_interval_overlaps.sum()))
sd_raw = {s: float(np.median([np.std(m["fingerprint"], axis=0).mean() for m in meta_rows if m["subject"] == s])) for s in B.SUBJECTS}
print("Raw series: unit", meta_rows[0]["unit"], "with conversion", meta_rows[0]["conversion"],
"| median SD of the stored values over the first 2000 samples of 4 channels:", {k: f"{v:.1e}" for k, v in sd_raw.items()})
print("hg_whole_mean_uV above is the stored high-gamma amplitude x 1e6, without the conversion factor.")
for s in B.SUBJECTS:
for base in B.BASELINES:
H = S[s]["H"][base][:, S[s]["vsmc"]]
print(f"{s} {base:5s}: per-CV mean z over vSMC electrodes, range {np.nanmin(H):.2f} to {np.nanmax(H):.2f}; "
f"mean over -500 to -400 ms {np.nanmean(H[:, :, :21]):.2f}")
subject block trials kept kept_in_invalid trial_interval_overlaps rest_s hg_whole_mean_uV hg_whole_sd_over_mean pre_mean_over_whole rest_mean_over_whole rest_sd_over_whole env_rate
EC2 EC2_B1 387 387 8 8 41.0 3.874 0.239 0.990 1.017 0.949 400.001
EC2 EC2_B105 311 311 0 0 10.0 3.513 0.259 0.980 0.994 1.073 400.001
EC2 EC2_B15 382 382 4 4 57.0 4.180 0.238 0.992 0.966 0.936 400.002
EC2 EC2_B76 389 389 19 20 57.0 3.746 0.281 0.977 1.063 1.211 400.001
EC2 EC2_B8 384 384 3 3 15.0 3.703 0.243 0.993 1.074 1.069 400.002
EC2 EC2_B89 397 397 6 6 61.0 4.648 0.214 0.996 0.989 0.953 400.002
EC2 EC2_B9 378 378 11 12 45.0 3.732 0.241 0.989 1.015 1.009 400.002
GP31 GP31_B1 237 236 0 0 5.0 2.967 0.218 0.991 1.066 0.941 400.002
GP31 GP31_B2 315 315 1 1 20.0 2.956 0.248 0.990 1.040 0.942 400.002
GP31 GP31_B21 237 237 0 0 30.0 2.043 0.201 0.988 1.034 1.076 400.002
GP31 GP31_B4 224 224 0 0 60.0 2.939 0.227 0.992 1.010 0.980 400.002
GP31 GP31_B6 377 377 1 1 3.3 2.949 0.212 0.991 1.027 1.248 400.001
GP31 GP31_B63 374 374 38 39 12.0 2.688 0.406 1.004 0.851 0.297 400.002
GP31 GP31_B65 324 324 0 0 8.0 2.251 0.177 1.002 0.958 0.978 400.002
GP31 GP31_B67 429 429 0 0 4.1 1.998 0.191 0.999 1.016 1.159 400.002
GP31 GP31_B69 430 430 9 11 6.0 1.912 0.204 0.999 0.979 0.985 400.002
GP31 GP31_B71 481 481 0 0 16.0 1.854 0.204 0.995 0.952 1.000 400.001
GP31 GP31_B78 424 424 0 0 18.0 1.864 0.234 0.986 0.991 0.910 400.002
GP31 GP31_B82 529 529 11 11 55.0 1.975 0.213 0.994 1.006 0.946 400.001
GP31 GP31_B83 582 582 0 0 4.0 2.039 0.250 1.004 1.100 0.898 400.001
GP31 GP31_B9 322 322 15 23 50.0 2.928 1.147 0.938 1.067 1.109 400.002
GP33 GP33_B1 339 339 0 0 78.0 4.442 0.275 0.996 1.000 0.909 400.001
GP33 GP33_B30 612 612 27 27 49.6 2.964 0.297 1.017 0.917 0.852 400.001
GP33 GP33_B5 507 507 9 11 54.6 4.118 0.200 0.992 1.005 1.019 400.001
Trials dropped (epoch -1000 to +750 ms outside the recording): 1
Kept trials whose epoch overlaps an 'ECoG artifact' invalid interval (not excluded; the claims file does not say to): 162 of 9370 | overlaps counted per trial and interval: 177
Raw series: unit volt with conversion 0.001 | median SD of the stored values over the first 2000 samples of 4 channels: {'EC2': '2.5e-05', 'GP31': '6.3e-05', 'GP33': '8.1e-05'}
hg_whole_mean_uV above is the stored high-gamma amplitude x 1e6, without the conversion factor.
EC2 pre : per-CV mean z over vSMC electrodes, range -1.04 to 4.21; mean over -500 to -400 ms -0.10
EC2 whole: per-CV mean z over vSMC electrodes, range -0.92 to 3.38; mean over -500 to -400 ms -0.13
EC2 rest : per-CV mean z over vSMC electrodes, range -1.02 to 5.96; mean over -500 to -400 ms 0.03
GP31 pre : per-CV mean z over vSMC electrodes, range -1.64 to 3.98; mean over -500 to -400 ms -0.02
GP31 whole: per-CV mean z over vSMC electrodes, range -1.63 to 3.57; mean over -500 to -400 ms -0.04
GP31 rest : per-CV mean z over vSMC electrodes, range -1.65 to 7.12; mean over -500 to -400 ms 0.07
GP33 pre : per-CV mean z over vSMC electrodes, range -0.98 to 3.10; mean over -500 to -400 ms -0.02
GP33 whole: per-CV mean z over vSMC electrodes, range -0.99 to 2.82; mean over -500 to -400 ms -0.06
GP33 rest : per-CV mean z over vSMC electrodes, range -0.99 to 3.71; mean over -500 to -400 ms 0.07
Units: the raw series says volts with a conversion factor of 0.001, so by NWB the stored values
times 0.001 are volts. The stored values have an SD of a few 1e-5 (printed above): read as volts
this is tens of microvolts, a usual ECoG amplitude, while with the factor it would be tens of
nanovolts. The data are consistent with values already in volts and a wrong conversion factor;
this does not affect the analysis, which uses z-scores only.
Sampling rate 3051.7578 Hz in all files, the TDT rate the Methods round to 3052 Hz. The pre
baseline (-1000 to -500 ms before each C-V transition) has a median mean within a few percent of
the whole block's. The rest periods are short (3.3 to 78 s per block) and their statistics differ
from the whole block's in both directions by block (median ratios above); z-scores against them
reach higher maxima (ranges above). The rest period is baseline reading (c) of the claims file.
In [9]:
# Clock check: the trial times and the ECoG share a clock if the high gamma tells syllables apart
# at the times the paper says, and not before speech. Leave-one-out nearest-centroid classification
# of the vowel (/a/, /i/, /u/) and of the major oral articulator of the consonant (labial, coronal,
# dorsal; reading (a) of the labels) from the vSMC high-gamma pattern of each CV, at each time.
def loo_accuracy(X, y):
ok = 0
for i in range(len(y)):
mask = np.arange(len(y)) != i
cents = {c: X[mask & (y == c)].mean(0) for c in np.unique(y[mask])}
pred = min(cents, key=lambda c: np.linalg.norm(X[i] - cents[c]))
ok += pred == y[i]
return ok / len(y)
clock = []
for s in B.SUBJECTS:
H = S[s]["H"]["pre"][:, S[s]["vsmc"]]
vow = np.array([v for c, v in S[s]["cvs"]])
maj = np.array([B.MAJOR["a"].get(c, "") for c, v in S[s]["cvs"]])
lab = maj != ""
for t in range(0, len(B.EPOCH_MS), 5):
clock.append({"subject": s, "t_ms": B.EPOCH_MS[t], "vowel": loo_accuracy(H[:, :, t], vow),
"consonant": loo_accuracy(H[lab][:, :, t], maj[lab])})
clock = pd.DataFrame(clock)
summary = clock.groupby("t_ms")[["vowel", "consonant"]].mean()
print("Mean accuracy over subjects (chance 1/3 for vowels; 0.5 to 0.53 for consonants, the largest class share):")
print(summary.iloc[::4].round(2).T.to_string())
for col in ["vowel", "consonant"]:
print(f"{col}: peak {summary[col].max():.2f} at {summary[col].idxmax()} ms; mean before -400 ms {summary[col][summary.index < -400].mean():.2f}")
fig, ax = plt.subplots(figsize=(6, 3))
summary.plot(ax=ax)
ax.axvline(0, color="k", lw=0.5)
ax.set_xlabel("time from C-V transition (ms)"); ax.set_ylabel("LOO accuracy")
plt.show()
Mean accuracy over subjects (chance 1/3 for vowels; 0.5 to 0.53 for consonants, the largest class share): t_ms -500 -400 -300 -200 -100 0 100 200 300 400 500 600 700 vowel 0.38 0.41 0.42 0.35 0.34 0.28 0.63 0.80 0.76 0.74 0.70 0.57 0.48 consonant 0.58 0.62 0.76 0.79 0.89 0.89 0.71 0.44 0.40 0.46 0.45 0.45 0.47 vowel: peak 0.83 at 250 ms; mean before -400 ms 0.39 consonant: peak 0.89 at -75 ms; mean before -400 ms 0.59
The consonant's articulator is decoded best before the C-V transition (0.89 at -75 ms) and the vowel after it (0.83 at 250 ms), and both are near chance at the start of the epoch (vowel 0.39 against 1/3; consonant 0.59 against a largest-class share of about 0.5). This is consistent with the trial times and the ECoG sharing a clock in every subject, with the annotations marking the C-V transition; it does not show the alignment to the millisecond.
The two matrices transcribed from the supplement, printed and drawn as in the figures (1 white, 0 black), so that the transcription can be checked against Supplementary Figures 1d and 12.
In [10]:
# Supplementary Figure 1d (articulators) and Supplementary Figure 12 (phonetic features), as transcribed.
cons_cv = [(c, "a") for c in B.CONSONANTS]
art = pd.DataFrame(B.articulator_matrix(cons_cv)[:, :4].astype(int), index=B.CONSONANTS, columns=B.ARTICULATORS)
feat = pd.DataFrame(np.c_[B.feature_matrix(cons_cv, "location"), B.feature_matrix(cons_cv, "degree")].astype(int),
index=B.CONSONANTS, columns=B.LOCATION + B.DEGREE)
vow = pd.DataFrame({v: np.r_[B.articulator_matrix([("b", v)])[0, 4:], B.feature_matrix([("b", v)], "vowel")[0]].astype(int)
for v in "aiu"}, index=[f"{a} (1d)" for a in B.VOWEL_ARTICULATORS] + [f"{f} (12)" for f in B.VOWEL_FEATURES])
print("Consonant rows of Supplementary Figure 1d (articulators):")
print(art.T.to_string())
print("\nConsonant features of Supplementary Figure 12 (location, then degree and shape):")
print(feat.T.to_string())
print("\nVowel rows of both figures:")
print(vow.to_string())
fig, axes = plt.subplots(1, 3, figsize=(13, 4), gridspec_kw={"width_ratios": [19, 19, 3]})
for ax, (m, title) in zip(axes, [(feat.T, "Suppl. Fig. 12, consonants"), (art.T, "Suppl. Fig. 1d, consonants"), (vow, "vowels")]):
ax.imshow(m.values, cmap="gray", vmin=0, vmax=1, aspect="auto")
ax.set_xticks(range(m.shape[1])); ax.set_xticklabels(m.columns, fontsize=7)
ax.set_yticks(range(m.shape[0])); ax.set_yticklabels(m.index, fontsize=7); ax.set_title(title, fontsize=9)
plt.tight_layout(); plt.show()
Consonant rows of Supplementary Figure 1d (articulators):
b d f g h k l m n p r s sh t th v w y z
Lips 1 0 1 0 0 0 0 1 0 1 0 0 0 0 0 1 1 0 0
Tongue 0 1 0 1 0 1 1 0 1 0 1 1 1 1 1 0 1 1 1
Larynx 1 1 0 1 0 0 1 1 1 0 1 0 0 0 0 1 1 1 1
Jaw 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1
Consonant features of Supplementary Figure 12 (location, then degree and shape):
b d f g h k l m n p r s sh t th v w y z
Bilabial 1 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0 1 0 0
Secondary Labial 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 0 0
Labiodental 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0
Dental 0 0 0 0 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0
Alveolar 0 1 0 0 0 0 1 0 1 0 0 1 0 1 0 0 0 0 1
Post Alveolar 0 0 0 0 0 0 0 0 0 0 1 0 1 0 0 0 0 1 0
Velar 0 0 0 1 0 1 0 0 0 0 0 0 0 0 0 0 1 0 0
Voiced 1 1 0 1 0 0 1 1 1 0 1 0 0 0 0 1 1 1 1
Mandibular 1 1 1 0 0 0 1 1 1 1 1 1 1 1 1 1 1 1 1
Oral Stop 1 1 0 1 0 1 0 1 1 1 0 0 0 1 0 0 0 0 0
Fricative 0 0 1 0 0 0 0 0 0 0 0 1 1 0 1 1 0 0 1
Approximate 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 1 1 0
Nasal Stop 0 0 0 0 0 0 0 1 1 0 0 0 0 0 0 0 0 0 0
Lateral 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0 0 0 0 0
Rhotic 0 0 0 0 0 0 0 0 0 0 1 0 0 0 0 0 0 0 0
Vowel rows of both figures:
a i u
Back Tongue (1d) 1 0 1
High Tongue (1d) 0 1 1
Back Tongue (12) 1 0 1
High Tongue (12) 0 1 1
Lip Rounding (12) 0 0 1
Jaw Open (12) 1 0 0
Shared computations¶
The partial correlations and the GLM (Methods Eqs. 2 to 4) are computed once per baseline and CV set, and the state space (Methods Eq. 5) once per baseline, CV set and z-scoring reading; the states K of each CV are then projected from the PCA input (reading i) or from the baseline z-scored high gamma (reading ii). The articulator windows T_A come from the timing analysis under each of its eight readings (claim fig3d-timing-sequence: threshold set per subject or pooled, over all six articulators or per articulator, and onset at the threshold or at half the peak).
In [11]:
BASES = B.BASELINES # baseline readings (a) pre, (b) whole, (c) rest (claims header)
ZMODES = ("row", "time") # z-scoring before the PCA (claims header)
PROJ = ("i", "ii") # input projected: (i) the PCA input, (ii) Hg as Eq. 5 writes it
ART6 = B.ARTICULATORS + B.VOWEL_ARTICULATORS
TIMING_READINGS = list(itertools.product(("per subject", "pooled"), ("all six", "per articulator"), ("threshold", "half peak")))
CVSETS = ("all", "15+") # (a) every CV, (b) only CVs with at least 15 renditions (claims header)
CVIDX = {(s, cs): np.where(S[s]["counts"] >= (15 if cs == "15+" else 0))[0] for s in B.SUBJECTS for cs in CVSETS}
CVS = {k: [S[k[0]]["cvs"][i] for i in v] for k, v in CVIDX.items()}
AM = {k: B.articulator_matrix(v) for k, v in CVS.items()}
HV = {(s, base, cs): S[s]["H"][base][CVIDX[s, cs]][:, S[s]["vsmc"]]
for s in B.SUBJECTS for base in B.BASELINES for cs in CVSETS}
RHO = {(s, base, cs): B.partial_correlation(HV[s, base, cs], AM[s, cs]) for s in B.SUBJECTS for base in BASES for cs in CVSETS}
GLM = {(s, base, cs): B.glm(HV[s, base, cs], AM[s, cs][:, :4]) for s in B.SUBJECTS for base in BASES for cs in CVSETS}
def timing_rows(base, cs, reading, rho=None):
rho = RHO if rho is None else rho
subj_mode, art_mode, onset_mode = reading
rows = []
for s in B.SUBJECTS:
pool = np.concatenate([rho[x, base, cs] for x in B.SUBJECTS]) if subj_mode == "pooled" else rho[s, base, cs]
for a in range(6):
vals = pool if art_mode == "all six" else pool[:, a]
thr = vals.mean() + 2.5 * vals.std()
for e in range(rho[s, base, cs].shape[0]):
r = rho[s, base, cs][e, a]
if r.max() > thr:
on, pk, off = B.timing(r, thr if onset_mode == "threshold" else r.max() / 2)
rows.append({"subject": s, "articulator": ART6[a], "electrode": e, "onset": on,
"peak": pk, "offset": off, "threshold": thr})
return pd.DataFrame(rows)
TIMING = {(base, cs, r): timing_rows(base, cs, r) for base in BASES for cs in CVSETS for r in TIMING_READINGS}
def articulator_windows(base, cs, reading):
"""T_A per subject: mean onset to mean offset of the subject's electrodes for A (None if none)."""
t = TIMING[base, cs, reading]
out = {}
for s in B.SUBJECTS:
for a in B.ARTICULATORS:
g = t[(t.subject == s) & (t.articulator == a)]
out[s, a] = (g.onset.mean(), g.offset.mean()) if len(g) else None
return out
def weights(base, cs, reading, pmode):
"""Per subject: W (vSMC electrode, 4), NaN for an articulator without a window, and the
electrodes kept by the GLM rule (mean p over T_A below 0.05 for any articulator)."""
T = articulator_windows(base, cs, reading)
out = {}
for s in B.SUBJECTS:
beta, pF, pt, r2 = GLM[s, base, cs]
keep = np.zeros(beta.shape[0], bool)
W = np.full((beta.shape[0], 4), np.nan)
for a, name in enumerate(B.ARTICULATORS):
if T[s, name] is None:
continue
idx = (B.EPOCH_MS >= T[s, name][0]) & (B.EPOCH_MS <= T[s, name][1])
p = (pF[:, idx] if pmode == "F-test" else pt[:, a, idx]).mean(1)
keep |= p < 0.05
W[:, a] = beta[:, a, idx].mean(1)
out[s] = {"W": W, "keep": keep, "windows": {a: T[s, a] for a in B.ARTICULATORS}}
return out
W_READINGS = [(base, cs, r, p) for base in BASES for cs in CVSETS for r in TIMING_READINGS for p in ("F-test", "t-test")]
WEIGHTS = {k: weights(*k) for k in W_READINGS}
STATE = {(s, base, z, cs, w): B.state_space(HV[s, base, cs], z, w)
for s in B.SUBJECTS for base in BASES for z in ZMODES for cs in CVSETS
for w in [(-500, 750), (-375, 620)]}
EP = (-500, 750)
SS_READINGS = list(itertools.product(BASES, ZMODES, PROJ, CVSETS))
KS = {(s, base, z, pj, cs): STATE[s, base, z, cs, EP][2] if pj == "i" else B.project(STATE[s, base, z, cs, EP][0], HV[s, base, cs])
for s in B.SUBJECTS for base, z, pj, cs in SS_READINGS}
# Variant (not a reading of the claims file): the trials whose epoch overlaps an interval the files
# mark as an ECoG artifact are left out of the CV means; the CVs and electrodes are the same.
S_CLEAN = {s: B.load_subject(blocks[s], exclude_invalid=True) for s in B.SUBJECTS}
assert all(S_CLEAN[s]["cvs"] == S[s]["cvs"] and S_CLEAN[s]["counts"].min() > 0 for s in B.SUBJECTS)
HV_CLEAN = {(s, base, cs): S_CLEAN[s]["H"][base][CVIDX[s, cs]][:, S[s]["vsmc"]]
for s in B.SUBJECTS for base in B.BASELINES for cs in CVSETS}
print("Renditions left out by the artifact variant, per subject:", {s: int((S[s]["counts"] - S_CLEAN[s]["counts"]).sum()) for s in B.SUBJECTS},
"| largest share of one CV's renditions:", {s: round(float(((S[s]["counts"] - S_CLEAN[s]["counts"]) / S[s]["counts"]).max()), 2) for s in B.SUBJECTS})
print(f"shared computations done at {time.time() - started:.0f} s")
Renditions left out by the artifact variant, per subject: {'EC2': 51, 'GP31': 75, 'GP33': 36} | largest share of one CV's renditions: {'EC2': 0.11, 'GP31': 0.06, 'GP33': 0.08}
shared computations done at 1030 s
data-composition¶
Three subjects with high-density ECoG (two left, one right hemisphere) read aloud CV syllables made of 18 or 19 consonants followed by /a/, /i/ or /u/, each CV produced 15 to 100 times, with the microphone recorded with the ECoG at 3052 Hz.
In [12]:
comp = []
for s in B.SUBJECTS:
n = S[s]["counts"]
comp.append({"subject": s, "CVs": len(S[s]["cvs"]), "consonants": len({c for c, v in S[s]["cvs"]}),
"vowels": len({v for c, v in S[s]["cvs"]}), "min_renditions": int(n.min()), "max_renditions": int(n.max()),
"CVs_below_15": int((n < 15).sum()), "CVs_above_100": int((n > 100).sum()),
"consonants_with_all_3_vowels_and_15+": len([c for c in B.CONSONANTS if all((c, v) in S[s]["cvs"] and n[S[s]["cvs"].index((c, v))] >= 15 for v in "aiu")]),
"hemisphere_from_x": ("left" if np.nanmean(S[s]["x"]) < 0 else "right") if np.isfinite(S[s]["x"]).any() else "unknown"})
comp = pd.DataFrame(comp)
print(comp.to_string(index=False))
print("Subjects in the dandiset:", sorted(files.subject.unique()), "| of the paper's (EC2, EC31, EC33), by assumed correspondence: 3")
print("Distinct CVs over the three subjects:", comp.CVs.sum(), "(paper 168); with CVs of fewer than 15 renditions left out:",
int((comp.CVs - comp.CVs_below_15).sum()))
print("ECoG sampling rate:", sorted(files.rate_hz.unique()), "Hz (3052 +- 1)")
print("Audio in the files: none (removed for HIPAA, dandiset description)")
subject CVs consonants vowels min_renditions max_renditions CVs_below_15 CVs_above_100 consonants_with_all_3_vowels_and_15+ hemisphere_from_x
EC2 57 19 3 10 58 1 0 18 left
GP31 55 19 3 2 107 1 33 18 left
GP33 57 19 3 22 28 0 0 19 unknown
Subjects in the dandiset: ['EC2', 'EC9', 'GP31', 'GP33'] | of the paper's (EC2, EC31, EC33), by assumed correspondence: 3
Distinct CVs over the three subjects: 169 (paper 168); with CVs of fewer than 15 renditions left out: 167
ECoG sampling rate: [np.float64(3051.7578), np.float64(3051.7578125)] Hz (3052 +- 1)
Audio in the files: none (removed for HIPAA, dandiset description)
Result: partially reproduced. The dandiset holds the three subjects (taking GP31 and GP33 to be EC31 and EC33) and a fourth one, EC9, that is not in the paper. The ECoG rate is 3051.76 Hz. The counts differ from the paper's: 57, 55 and 57 CVs (169 in all, against 168), each subject with 19 consonants, and renditions of 2 to 107 per CV (EC2's /thi/ 10, GP31's /thi/ 2, GP31 up to 107). Leaving out CVs with fewer than 15 renditions gives 56 + 54 + 57 = 167, still not the paper's split. The audio is not in the dandiset, and GP33's hemisphere cannot be checked. In EC2 and GP31, 18 consonants have all three vowels with 15 or more renditions each, and in GP33 19; the paper's split is 19, 19 and 18 consonants.
active-electrodes¶
About 30 active vSMC electrodes were found per subject during syllable articulation, an electrode being active when its high-gamma z-score changed by more than 2 for any syllable.
In [13]:
def overall(outcomes):
"""The outcome all method readings give, or methods_underspecified when they differ."""
u = sorted(set(outcomes))
return u[0] if len(u) == 1 else "differ by reading: " + ", ".join(f"{o} {list(outcomes).count(o)}" for o in u)
def by_factor(df, factors, col="outcome"):
"""Outcome counts for each level of each reading factor, the other factors pooled."""
print(f"Outcome counts by reading factor ({len(df)} readings):")
for f in factors:
print(f" {f}:", {lvl: g[col].value_counts().to_dict() for lvl, g in df.groupby(f, sort=False)})
act = []
for base in B.BASELINES:
for cs in CVSETS:
for s in B.SUBJECTS:
H = HV[s, base, cs]
act.append({"baseline": base, "CVs": cs, "subject": s, "vSMC": H.shape[1],
"(a) z > 2": int((np.nanmax(H, axis=(0, 2)) > 2).sum()),
"(b) |z| > 2": int((np.nanmax(np.abs(H), axis=(0, 2)) > 2).sum())})
act = pd.DataFrame(act)
print(act.to_string(index=False))
am = act.groupby(["baseline", "CVs"])[["(a) z > 2", "(b) |z| > 2"]].mean().round(2)
am["within 27-33 (a or b)"] = am.apply(lambda r: 27 <= r["(a) z > 2"] <= 33 or 27 <= r["(b) |z| > 2"] <= 33, axis=1)
print("\nMean over subjects (criterion 27 to 33):")
print(am.to_string())
for thr in (1.5, 2, 2.5, 3):
print(f"threshold {thr} (all CVs): mean active per subject, pre {np.mean([(np.nanmax(HV[s, 'pre', 'all'], axis=(0, 2)) > thr).sum() for s in B.SUBJECTS]):.1f}, "
f"whole {np.mean([(np.nanmax(HV[s, 'whole', 'all'], axis=(0, 2)) > thr).sum() for s in B.SUBJECTS]):.1f}, "
f"rest {np.mean([(np.nanmax(HV[s, 'rest', 'all'], axis=(0, 2)) > thr).sum() for s in B.SUBJECTS]):.1f}")
res = am.loc[list(BASES)]
per_cs = {cs: "reproduced" if all(res.xs(cs, level=1)["within 27-33 (a or b)"]) else
("partially" if any(res.xs(cs, level=1)["within 27-33 (a or b)"]) else "not reproduced")
for cs in CVSETS}
print("Outcome per CV set:", per_cs)
print("Outcome over readings:", overall(list(per_cs.values())))
print("\nElectrodes active (z > 2) with every CV but not with only the CVs of 15+ renditions, and the CVs above 2 at them:")
for base in BASES:
for s in B.SUBJECTS:
Ha, H15 = HV[s, base, "all"], HV[s, base, "15+"]
only = (np.nanmax(Ha, axis=(0, 2)) > 2) & ~(np.nanmax(H15, axis=(0, 2)) > 2)
drivers = pd.Series([c + v for e in np.where(only)[0] for (c, v), x in zip(CVS[s, "all"], np.nanmax(Ha[:, e], axis=1)) if x > 2])
print(f" {base:5s} {s}: {int(only.sum())} electrodes; CVs above 2 there (electrodes): {drivers.value_counts().to_dict()}",
f"| renditions of those CVs: { {k: int(S[s]['counts'][S[s]['cvs'].index(next(p for p in S[s]['cvs'] if p[0] + p[1] == k))]) for k in drivers.unique()} }")
act_c = pd.DataFrame([{"baseline": base, "CVs": cs, "subject": s,
"(a) z > 2": int((np.nanmax(HV_CLEAN[s, base, cs], axis=(0, 2)) > 2).sum())}
for base in BASES for cs in CVSETS for s in B.SUBJECTS])
print("\nVariant, artifact trials left out (z > 2), active electrodes per subject and mean:")
print(act_c.pivot_table(index=["baseline", "CVs"], columns="subject", values="(a) z > 2").assign(
mean=lambda d: d.mean(axis=1).round(2)).to_string())
baseline CVs subject vSMC (a) z > 2 (b) |z| > 2
pre all EC2 86 15 15
pre all GP31 83 49 49
pre all GP33 99 9 9
pre 15+ EC2 86 14 14
pre 15+ GP31 83 9 9
pre 15+ GP33 99 9 9
whole all EC2 86 15 15
whole all GP31 83 43 43
whole all GP33 99 6 6
whole 15+ EC2 86 15 15
whole 15+ GP31 83 0 0
whole 15+ GP33 99 6 6
rest all EC2 86 30 30
rest all GP31 83 51 51
rest all GP33 99 21 21
rest 15+ EC2 86 29 29
rest 15+ GP31 83 31 31
rest 15+ GP33 99 21 21
Mean over subjects (criterion 27 to 33):
(a) z > 2 (b) |z| > 2 within 27-33 (a or b)
baseline CVs
pre 15+ 10.67 10.67 False
all 24.33 24.33 False
rest 15+ 27.00 27.00 True
all 34.00 34.00 False
whole 15+ 7.00 7.00 False
all 21.33 21.33 False
threshold 1.5 (all CVs): mean active per subject, pre 41.7, whole 38.3, rest 52.0
threshold 2 (all CVs): mean active per subject, pre 24.3, whole 21.3, rest 34.0
threshold 2.5 (all CVs): mean active per subject, pre 11.7, whole 10.0, rest 20.3
threshold 3 (all CVs): mean active per subject, pre 5.3, whole 2.7, rest 14.0
Outcome per CV set: {'all': 'not reproduced', '15+': 'partially'}
Outcome over readings: differ by reading: not reproduced 1, partially 1
Electrodes active (z > 2) with every CV but not with only the CVs of 15+ renditions, and the CVs above 2 at them:
pre EC2: 1 electrodes; CVs above 2 there (electrodes): {'thi': 1} | renditions of those CVs: {'thi': 10}
pre GP31: 40 electrodes; CVs above 2 there (electrodes): {'thi': 40} | renditions of those CVs: {'thi': 2}
pre GP33: 0 electrodes; CVs above 2 there (electrodes): {} | renditions of those CVs: {}
whole EC2: 0 electrodes; CVs above 2 there (electrodes): {} | renditions of those CVs: {}
whole GP31: 43 electrodes; CVs above 2 there (electrodes): {'thi': 43} | renditions of those CVs: {'thi': 2}
whole GP33: 0 electrodes; CVs above 2 there (electrodes): {} | renditions of those CVs: {}
rest EC2: 1 electrodes; CVs above 2 there (electrodes): {'thi': 1} | renditions of those CVs: {'thi': 10}
rest GP31: 20 electrodes; CVs above 2 there (electrodes): {'thi': 20} | renditions of those CVs: {'thi': 2}
rest GP33: 0 electrodes; CVs above 2 there (electrodes): {} | renditions of those CVs: {}
Variant, artifact trials left out (z > 2), active electrodes per subject and mean:
subject EC2 GP31 GP33 mean
baseline CVs
pre 15+ 14.0 3.0 7.0 8.00
all 15.0 46.0 7.0 22.67
rest 15+ 29.0 2.0 14.0 15.00
all 30.0 34.0 14.0 26.00
whole 15+ 15.0 0.0 5.0 6.67
all 15.0 43.0 5.0 21.00
Result: blocked (methods_underspecified). The two readings of the CV set give different
outcomes. With every CV, no baseline gives a mean within 27 to 33: 24.3 per subject with pre
(EC2 15, GP31 49, GP33 9), 21.3 with whole and 34.0 with rest (not reproduced). With only the
CVs of 15 or more renditions, rest gives 27.0 (EC2 29, GP31 31, GP33 21), at the lower edge of
the tolerance, while pre gives 10.7 and whole 7.0 (partially reproduced). Both readings of
the change give the same counts, as no electrode passes on a decrease alone. The count moves
strongly with the threshold (with every CV, 41.7 to 5.3 per subject over thresholds 1.5 to 3 with
pre).
The split between the CV readings comes from one syllable. Every GP31 electrode that is active
with every CV but not with the 15+ CVs is above 2 for /thi/, which GP31 has 2 renditions of (40
electrodes under pre, 43 under whole, 20 under rest); in EC2 one electrode is active through
/thi/ (10 renditions) under pre and rest, and in GP33 none. Leaving out the trials that overlap
intervals marked as ECoG artifacts (a variant, not a reading of the claims file) gives means of
22.7, 21.0 and 26.0 with every CV and 8.0, 6.7 and 15.0 with the 15+ CVs, none within 27 to 33.
GP31's count under rest with the 15+ CVs falls from 31 to 2 in this variant, so the one reading
within tolerance rests on trials the files mark as artifacts.
fig1c-e-articulator-electrodes¶
In one subject, dorsal electrodes (e124, e108) were active for /b/ (lips), mid vSMC electrodes (e129, e133, e105) for /d/ (front tongue), and a more ventral electrode (e104) was most active for /g/ (back tongue).
In [14]:
NAMED = {"b": [124, 108], "d": [129, 133, 105], "g": [104]}
f1 = []
for s in B.SUBJECTS:
for numbering, offset in (("(a) from 1", 1), ("(b) from 0", 0)):
idx = {k: [n - offset for n in v] for k, v in NAMED.items()}
all_idx = sum(idx.values(), [])
on_vsmc = [i in S[s]["vsmc"] for i in all_idx]
for base in BASES:
for measure in ("(i) at -25 ms", "(ii) peak"):
H = S[s]["H"][base]
resp = {}
for c in "bdg":
k = S[s]["cvs"].index((c, "a"))
resp[c] = H[k, :, B.ms_index(-25)] if measure.startswith("(i)") else np.nanmax(H[k], axis=1)
prefers = {i: max("bdg", key=lambda c: resp[c][i]) for i in all_idx}
ok = [prefers[i] == c for c, ii in idx.items() for i in ii]
f1.append({"subject": s, "numbering": numbering, "baseline": base, "measure": measure,
"named on vSMC": f"{sum(on_vsmc)} of 6",
"labels": ",".join(S[s]["location"][i][:4] + ("" if S[s]["good"][i] else "(bad)") for i in all_idx),
"preferences (b b d d d g)": " ".join(prefers[i] for i in all_idx),
"prefer their syllable": f"{sum(ok)} of 6"})
f1 = pd.DataFrame(f1)
print(f1.to_string(index=False))
print("\nSubject and numbering pairs that put all six named electrodes on vSMC:",
f1[f1["named on vSMC"] == "6 of 6"][["subject", "numbering"]].drop_duplicates().values.tolist())
subject numbering baseline measure named on vSMC labels preferences (b b d d d g) prefer their syllable
EC2 (a) from 1 pre (i) at -25 ms 3 of 6 post(bad),post,midd,supe,post,post b b d g d g 5 of 6
EC2 (a) from 1 pre (ii) peak 3 of 6 post(bad),post,midd,supe,post,post b b d g d g 5 of 6
EC2 (a) from 1 whole (i) at -25 ms 3 of 6 post(bad),post,midd,supe,post,post b b d g d g 5 of 6
EC2 (a) from 1 whole (ii) peak 3 of 6 post(bad),post,midd,supe,post,post b b d g d g 5 of 6
EC2 (a) from 1 rest (i) at -25 ms 3 of 6 post(bad),post,midd,supe,post,post b b d g d g 5 of 6
EC2 (a) from 1 rest (ii) peak 3 of 6 post(bad),post,midd,supe,post,post b b d g d g 5 of 6
EC2 (b) from 0 pre (i) at -25 ms 5 of 6 post,post,midd,prec,post,post b b g b b d 2 of 6
EC2 (b) from 0 pre (ii) peak 5 of 6 post,post,midd,prec,post,post b b d b b d 3 of 6
EC2 (b) from 0 whole (i) at -25 ms 5 of 6 post,post,midd,prec,post,post b b g b b d 2 of 6
EC2 (b) from 0 whole (ii) peak 5 of 6 post,post,midd,prec,post,post b b d b b d 3 of 6
EC2 (b) from 0 rest (i) at -25 ms 5 of 6 post,post,midd,prec,post,post b b g b b d 2 of 6
EC2 (b) from 0 rest (ii) peak 5 of 6 post,post,midd,prec,post,post b b d b b d 3 of 6
GP31 (a) from 1 pre (i) at -25 ms 3 of 6 post(bad),post,midd,supe,post,post b b d g g g 4 of 6
GP31 (a) from 1 pre (ii) peak 3 of 6 post(bad),post,midd,supe,post,post b b b g g g 3 of 6
GP31 (a) from 1 whole (i) at -25 ms 3 of 6 post(bad),post,midd,supe,post,post b b d g d g 5 of 6
GP31 (a) from 1 whole (ii) peak 3 of 6 post(bad),post,midd,supe,post,post b b b g g g 3 of 6
GP31 (a) from 1 rest (i) at -25 ms 3 of 6 post(bad),post,midd,supe,post,post b b d g d g 5 of 6
GP31 (a) from 1 rest (ii) peak 3 of 6 post(bad),post,midd,supe,post,post b b g g g g 3 of 6
GP31 (b) from 0 pre (i) at -25 ms 4 of 6 prec,post,midd,supe,post,post b b g b d g 4 of 6
GP31 (b) from 0 pre (ii) peak 4 of 6 prec,post,midd,supe,post,post b b b b g g 3 of 6
GP31 (b) from 0 whole (i) at -25 ms 4 of 6 prec,post,midd,supe,post,post b b g b d d 3 of 6
GP31 (b) from 0 whole (ii) peak 4 of 6 prec,post,midd,supe,post,post b b b b d g 4 of 6
GP31 (b) from 0 rest (i) at -25 ms 4 of 6 prec,post,midd,supe,post,post b b d b d d 4 of 6
GP31 (b) from 0 rest (ii) peak 4 of 6 prec,post,midd,supe,post,post b b g b g g 3 of 6
GP33 (a) from 1 pre (i) at -25 ms 1 of 6 prec(bad),prec,pars,pars,pars,pars b b d d d d 5 of 6
GP33 (a) from 1 pre (ii) peak 1 of 6 prec(bad),prec,pars,pars,pars,pars b b b b d d 3 of 6
GP33 (a) from 1 whole (i) at -25 ms 1 of 6 prec(bad),prec,pars,pars,pars,pars b b d d d d 5 of 6
GP33 (a) from 1 whole (ii) peak 1 of 6 prec(bad),prec,pars,pars,pars,pars b b b b d d 3 of 6
GP33 (a) from 1 rest (i) at -25 ms 1 of 6 prec(bad),prec,pars,pars,pars,pars b b b d d d 4 of 6
GP33 (a) from 1 rest (ii) peak 1 of 6 prec(bad),prec,pars,pars,pars,pars b b b b d d 3 of 6
GP33 (b) from 0 pre (i) at -25 ms 3 of 6 post,prec,pars,pars,prec,pars d d g g d d 1 of 6
GP33 (b) from 0 pre (ii) peak 3 of 6 post,prec,pars,pars,prec,pars d b b b g d 1 of 6
GP33 (b) from 0 whole (i) at -25 ms 3 of 6 post,prec,pars,pars,prec,pars d d g g d d 1 of 6
GP33 (b) from 0 whole (ii) peak 3 of 6 post,prec,pars,pars,prec,pars d b b b g d 1 of 6
GP33 (b) from 0 rest (i) at -25 ms 3 of 6 post,prec,pars,pars,prec,pars d d g g d d 1 of 6
GP33 (b) from 0 rest (ii) peak 3 of 6 post,prec,pars,pars,prec,pars d b b b g d 1 of 6
Subject and numbering pairs that put all six named electrodes on vSMC: []
Result: blocked (metadata_insufficient). No subject, counted from 1 or from 0, has all six named electrodes on vSMC. The most is 5 of 6, in EC2 counted from 0, where e129 is labelled middle temporal; GP31 has at most 4 and GP33 at most 3. The criterion blocks the claim in that case: the dataset does not say which subject Fig. 1 shows or how the paper numbered electrodes. For information, in EC2 counted from 1 (3 of 6 on vSMC, e124 flagged bad in some block), five of the six named electrodes prefer their syllable under both measures and all three baselines; counted from 0, two or three of six do.
fig1f-h-phonetic-contrasts¶
Fricatives with different tongue-tip constriction locations engaged different electrodes; front-tongue consonants of different manner engaged overlapping electrodes with different magnitudes; for /ja/, /ji/, /ju/ the activity was similar before the C-V transition, while a dorsal electrode was active during /u/ only.
The judged part: in at least one subject, a vSMC electrode with a positive lips weight (from the GLM electrodes of claim fig2a-lips-dorsal) has mean high gamma above 2 at +250 ms for /ju/ and below 2 for /ja/ and /ji/. The lips weights and the GLM electrodes are computed without electrode positions, which are only needed for fig2a's spatial judgement (see the summary).
In [15]:
t250 = B.ms_index(250)
fh = []
for key in W_READINGS:
base, cs = key[0], key[1]
hits = {}
for s in B.SUBJECTS:
cvs = CVS[s, cs]
Hj = {v: HV[s, base, cs][cvs.index(("y", v)), :, t250] for v in "aiu"}
w = WEIGHTS[key][s]
cand = w["keep"] & (w["W"][:, 0] > 0)
hit = cand & (Hj["u"] > 2) & (Hj["a"] < 2) & (Hj["i"] < 2)
hits[s] = (int(cand.sum()), int((cand & (Hj["u"] > 2)).sum()), int(hit.sum()),
float(Hj["u"][cand].max()) if cand.any() else np.nan)
fh.append({"baseline": base, "CVs": cs, "timing reading": " / ".join(key[2]), "p-value": key[3],
**{f"{s} lips+ / ju>2 / hits / max /ju/": f"{hits[s][0]} / {hits[s][1]} / {hits[s][2]} / {hits[s][3]:.2f}" for s in B.SUBJECTS},
"ju>2": sum(h[1] for h in hits.values()), "holds": any(h[2] > 0 for h in hits.values())})
fh = pd.DataFrame(fh)
print(fh.to_string(index=False))
print("\nReadings where it holds:", int(fh.holds.sum()), "of", len(fh))
print("Largest /ju/ z at +250 ms among lips-positive GLM electrodes, over all readings and subjects:",
max(float(v.split(" / ")[3]) for c in fh.columns if "max /ju/" in c for v in fh[c]))
print("Readings with a lips-positive GLM electrode above 2 for /ju/:", int((fh["ju>2"] > 0).sum()), "of", len(fh),
"(baselines:", fh[fh["ju>2"] > 0].baseline.value_counts().to_dict(), "); in all of them /ja/ or /ji/ is also above 2 there" if not fh.holds.any() else "")
print("Outcome over readings:", overall(["reproduced" if h else "not reproduced" for h in fh.holds]))
allv = []
for base in BASES:
for cs in CVSETS:
for s in B.SUBJECTS:
cvs = CVS[s, cs]
Hj = {v: HV[s, base, cs][cvs.index(("y", v)), :, t250] for v in "aiu"}
allv.append({"baseline": base, "CVs": cs, "subject": s, "vSMC": len(Hj["u"]), "/ju/ > 2": int((Hj["u"] > 2).sum()),
"/ju/ > 2, /ja/ and /ji/ < 2": int(((Hj["u"] > 2) & (Hj["a"] < 2) & (Hj["i"] < 2)).sum()),
"largest /ju/": round(float(np.nanmax(Hj["u"])), 2)})
print("\nAll vSMC electrodes, without the GLM selection (not the criterion; it separates the selection from the effect):")
print(pd.DataFrame(allv).to_string(index=False))
fig, axes = plt.subplots(1, 3, figsize=(12, 3), sharey=True)
key0 = W_READINGS[0]
for ax, s in zip(axes, B.SUBJECTS):
cvs = CVS[s, "all"]
H = HV[s, "pre", "all"]
w = WEIGHTS[key0][s]
cand = np.where(w["keep"] & (w["W"][:, 0] > 0))[0]
d = H[cvs.index(("y", "u")), cand, t250] - np.maximum(H[cvs.index(("y", "a")), cand, t250], H[cvs.index(("y", "i")), cand, t250])
e = cand[np.argmax(d)]
tr = H[[cvs.index(("y", v)) for v in "aiu"], e]
for v, row in zip("aiu", tr):
ax.plot(B.EPOCH_MS, row, label=f"/j{v}/")
ax.set_title(f"{s} electrode {S[s]['vsmc'][e]}")
print(f"{s}: plotted electrode {S[s]['vsmc'][e]} (largest /ju/ minus /ja/,/ji/ at +250 ms among lips+ GLM electrodes), z at +250 ms: "
+ ", ".join(f"/j{v}/ {row[t250]:.2f}" for v, row in zip("aiu", tr)) + f"; plotted range {tr.min():.2f} to {tr.max():.2f}; "
f"mean |difference| between the three before 0 ms {np.abs(tr[:, B.EPOCH_MS < 0] - tr[:, B.EPOCH_MS < 0].mean(0)).mean():.2f}")
ax.axvline(0, color="k", lw=0.5); ax.axhline(2, color="grey", lw=0.5, ls="--")
axes[0].legend(); axes[0].set_ylabel("z (pre baseline)")
plt.show()
baseline CVs timing reading p-value EC2 lips+ / ju>2 / hits / max /ju/ GP31 lips+ / ju>2 / hits / max /ju/ GP33 lips+ / ju>2 / hits / max /ju/ ju>2 holds
pre all per subject / all six / threshold F-test 11 / 0 / 0 / 0.96 29 / 0 / 0 / 0.40 14 / 0 / 0 / 0.78 0 False
pre all per subject / all six / threshold t-test 8 / 0 / 0 / 0.96 30 / 0 / 0 / 0.40 13 / 0 / 0 / 0.78 0 False
pre all per subject / all six / half peak F-test 8 / 0 / 0 / 0.96 25 / 0 / 0 / 0.40 15 / 0 / 0 / 0.78 0 False
pre all per subject / all six / half peak t-test 6 / 0 / 0 / 0.96 24 / 0 / 0 / 0.40 8 / 0 / 0 / 0.78 0 False
pre all per subject / per articulator / threshold F-test 12 / 0 / 0 / 0.96 30 / 0 / 0 / 0.40 22 / 0 / 0 / 0.70 0 False
pre all per subject / per articulator / threshold t-test 10 / 0 / 0 / 0.99 28 / 0 / 0 / 0.40 13 / 0 / 0 / 0.33 0 False
pre all per subject / per articulator / half peak F-test 11 / 0 / 0 / 1.13 25 / 0 / 0 / 0.40 11 / 0 / 0 / 0.61 0 False
pre all per subject / per articulator / half peak t-test 4 / 0 / 0 / 1.13 23 / 0 / 0 / 0.40 11 / 0 / 0 / 0.27 0 False
pre all pooled / all six / threshold F-test 10 / 0 / 0 / 0.96 29 / 0 / 0 / 0.40 13 / 0 / 0 / 0.61 0 False
pre all pooled / all six / threshold t-test 10 / 0 / 0 / 0.99 33 / 0 / 0 / 0.40 9 / 0 / 0 / 0.37 0 False
pre all pooled / all six / half peak F-test 7 / 0 / 0 / 0.96 24 / 0 / 0 / 0.40 10 / 0 / 0 / 0.61 0 False
pre all pooled / all six / half peak t-test 5 / 0 / 0 / 0.96 24 / 0 / 0 / 0.40 7 / 0 / 0 / 0.25 0 False
pre all pooled / per articulator / threshold F-test 11 / 0 / 0 / 0.96 32 / 0 / 0 / 0.40 22 / 0 / 0 / 0.78 0 False
pre all pooled / per articulator / threshold t-test 10 / 0 / 0 / 1.03 34 / 0 / 0 / 0.40 15 / 0 / 0 / 0.78 0 False
pre all pooled / per articulator / half peak F-test 8 / 0 / 0 / 0.96 25 / 0 / 0 / 0.40 12 / 0 / 0 / 0.61 0 False
pre all pooled / per articulator / half peak t-test 4 / 0 / 0 / 0.96 22 / 0 / 0 / 0.40 9 / 0 / 0 / 0.37 0 False
pre 15+ per subject / all six / threshold F-test 12 / 0 / 0 / 1.03 40 / 0 / 0 / 0.40 14 / 0 / 0 / 0.78 0 False
pre 15+ per subject / all six / threshold t-test 9 / 0 / 0 / 0.99 28 / 0 / 0 / 0.40 13 / 0 / 0 / 0.78 0 False
pre 15+ per subject / all six / half peak F-test 8 / 0 / 0 / 0.96 32 / 0 / 0 / 0.40 15 / 0 / 0 / 0.78 0 False
pre 15+ per subject / all six / half peak t-test 6 / 0 / 0 / 0.96 19 / 0 / 0 / 0.28 8 / 0 / 0 / 0.78 0 False
pre 15+ per subject / per articulator / threshold F-test 13 / 0 / 0 / 0.96 32 / 0 / 0 / 0.40 22 / 0 / 0 / 0.70 0 False
pre 15+ per subject / per articulator / threshold t-test 11 / 0 / 0 / 1.03 30 / 0 / 0 / 0.40 13 / 0 / 0 / 0.33 0 False
pre 15+ per subject / per articulator / half peak F-test 10 / 0 / 0 / 1.13 31 / 0 / 0 / 0.40 11 / 0 / 0 / 0.61 0 False
pre 15+ per subject / per articulator / half peak t-test 4 / 0 / 0 / 1.13 24 / 0 / 0 / 0.40 11 / 0 / 0 / 0.27 0 False
pre 15+ pooled / all six / threshold F-test 11 / 0 / 0 / 1.03 37 / 0 / 0 / 0.40 15 / 0 / 0 / 0.61 0 False
pre 15+ pooled / all six / threshold t-test 9 / 0 / 0 / 0.99 30 / 0 / 0 / 0.40 14 / 0 / 0 / 0.48 0 False
pre 15+ pooled / all six / half peak F-test 8 / 0 / 0 / 0.96 30 / 0 / 0 / 0.40 13 / 0 / 0 / 0.61 0 False
pre 15+ pooled / all six / half peak t-test 6 / 0 / 0 / 0.96 23 / 0 / 0 / 0.40 7 / 0 / 0 / 0.33 0 False
pre 15+ pooled / per articulator / threshold F-test 14 / 0 / 0 / 1.03 39 / 0 / 0 / 0.40 14 / 0 / 0 / 0.61 0 False
pre 15+ pooled / per articulator / threshold t-test 10 / 0 / 0 / 1.03 31 / 0 / 0 / 0.40 12 / 0 / 0 / 0.48 0 False
pre 15+ pooled / per articulator / half peak F-test 10 / 0 / 0 / 1.13 31 / 0 / 0 / 0.40 11 / 0 / 0 / 0.61 0 False
pre 15+ pooled / per articulator / half peak t-test 5 / 0 / 0 / 1.13 23 / 0 / 0 / 0.40 8 / 0 / 0 / 0.27 0 False
whole all per subject / all six / threshold F-test 11 / 0 / 0 / 0.85 30 / 0 / 0 / 0.23 17 / 0 / 0 / 0.67 0 False
whole all per subject / all six / threshold t-test 8 / 0 / 0 / 0.85 23 / 0 / 0 / 0.19 13 / 0 / 0 / 0.67 0 False
whole all per subject / all six / half peak F-test 8 / 0 / 0 / 0.85 28 / 0 / 0 / 0.23 11 / 0 / 0 / 0.47 0 False
whole all per subject / all six / half peak t-test 6 / 0 / 0 / 0.85 25 / 0 / 0 / 0.19 6 / 0 / 0 / 0.18 0 False
whole all per subject / per articulator / threshold F-test 10 / 0 / 0 / 0.85 29 / 0 / 0 / 0.23 23 / 0 / 0 / 0.67 0 False
whole all per subject / per articulator / threshold t-test 13 / 0 / 0 / 1.14 28 / 0 / 0 / 0.19 14 / 0 / 0 / 0.67 0 False
whole all per subject / per articulator / half peak F-test 10 / 0 / 0 / 1.19 25 / 0 / 0 / 0.23 12 / 0 / 0 / 0.47 0 False
whole all per subject / per articulator / half peak t-test 4 / 0 / 0 / 1.19 20 / 0 / 0 / 0.19 9 / 0 / 0 / 0.33 0 False
whole all pooled / all six / threshold F-test 12 / 0 / 0 / 0.85 30 / 0 / 0 / 0.23 13 / 0 / 0 / 0.47 0 False
whole all pooled / all six / threshold t-test 10 / 0 / 0 / 0.92 27 / 0 / 0 / 0.19 10 / 0 / 0 / 0.47 0 False
whole all pooled / all six / half peak F-test 11 / 0 / 0 / 1.19 26 / 0 / 0 / 0.23 12 / 0 / 0 / 0.47 0 False
whole all pooled / all six / half peak t-test 7 / 0 / 0 / 1.19 22 / 0 / 0 / 0.19 6 / 0 / 0 / 0.17 0 False
whole all pooled / per articulator / threshold F-test 11 / 0 / 0 / 0.85 32 / 0 / 0 / 0.23 16 / 0 / 0 / 0.67 0 False
whole all pooled / per articulator / threshold t-test 10 / 0 / 0 / 1.14 31 / 0 / 0 / 0.19 13 / 0 / 0 / 0.33 0 False
whole all pooled / per articulator / half peak F-test 10 / 0 / 0 / 1.19 27 / 0 / 0 / 0.23 11 / 0 / 0 / 0.67 0 False
whole all pooled / per articulator / half peak t-test 7 / 0 / 0 / 1.19 24 / 0 / 0 / 0.19 8 / 0 / 0 / 0.24 0 False
whole 15+ per subject / all six / threshold F-test 11 / 0 / 0 / 1.14 36 / 0 / 0 / 0.23 17 / 0 / 0 / 0.67 0 False
whole 15+ per subject / all six / threshold t-test 9 / 0 / 0 / 0.92 23 / 0 / 0 / 0.19 13 / 0 / 0 / 0.67 0 False
whole 15+ per subject / all six / half peak F-test 8 / 0 / 0 / 0.85 28 / 0 / 0 / 0.34 11 / 0 / 0 / 0.47 0 False
whole 15+ per subject / all six / half peak t-test 6 / 0 / 0 / 0.85 20 / 0 / 0 / 0.19 6 / 0 / 0 / 0.18 0 False
whole 15+ per subject / per articulator / threshold F-test 13 / 0 / 0 / 0.85 32 / 0 / 0 / 0.23 23 / 0 / 0 / 0.67 0 False
whole 15+ per subject / per articulator / threshold t-test 12 / 0 / 0 / 1.14 30 / 0 / 0 / 0.19 14 / 0 / 0 / 0.67 0 False
whole 15+ per subject / per articulator / half peak F-test 10 / 0 / 0 / 1.19 31 / 0 / 0 / 0.34 12 / 0 / 0 / 0.47 0 False
whole 15+ per subject / per articulator / half peak t-test 4 / 0 / 0 / 1.19 22 / 0 / 0 / 0.19 9 / 0 / 0 / 0.33 0 False
whole 15+ pooled / all six / threshold F-test 11 / 0 / 0 / 1.14 41 / 0 / 0 / 0.34 15 / 0 / 0 / 0.47 0 False
whole 15+ pooled / all six / threshold t-test 10 / 0 / 0 / 1.14 34 / 0 / 0 / 0.34 14 / 0 / 0 / 0.37 0 False
whole 15+ pooled / all six / half peak F-test 8 / 0 / 0 / 0.85 29 / 0 / 0 / 0.34 13 / 0 / 0 / 0.47 0 False
whole 15+ pooled / all six / half peak t-test 6 / 0 / 0 / 0.85 21 / 0 / 0 / 0.34 7 / 0 / 0 / 0.20 0 False
whole 15+ pooled / per articulator / threshold F-test 15 / 0 / 0 / 1.14 40 / 0 / 0 / 0.34 14 / 0 / 0 / 0.47 0 False
whole 15+ pooled / per articulator / threshold t-test 11 / 0 / 0 / 1.14 32 / 0 / 0 / 0.34 11 / 0 / 0 / 0.37 0 False
whole 15+ pooled / per articulator / half peak F-test 10 / 0 / 0 / 1.19 32 / 0 / 0 / 0.34 12 / 0 / 0 / 0.47 0 False
whole 15+ pooled / per articulator / half peak t-test 5 / 0 / 0 / 1.19 21 / 0 / 0 / 0.19 7 / 0 / 0 / 0.24 0 False
rest all per subject / all six / threshold F-test 10 / 1 / 0 / 2.41 27 / 0 / 0 / 0.94 17 / 0 / 0 / 1.56 1 False
rest all per subject / all six / threshold t-test 9 / 0 / 0 / 1.25 25 / 0 / 0 / 0.83 18 / 0 / 0 / 1.56 0 False
rest all per subject / all six / half peak F-test 8 / 0 / 0 / 1.18 24 / 0 / 0 / 0.94 12 / 0 / 0 / 1.56 0 False
rest all per subject / all six / half peak t-test 6 / 0 / 0 / 1.18 19 / 0 / 0 / 0.83 7 / 0 / 0 / 0.52 0 False
rest all per subject / per articulator / threshold F-test 11 / 0 / 0 / 1.18 24 / 0 / 0 / 0.83 17 / 0 / 0 / 1.00 0 False
rest all per subject / per articulator / threshold t-test 10 / 1 / 0 / 3.54 26 / 0 / 0 / 0.83 14 / 0 / 0 / 0.48 1 False
rest all per subject / per articulator / half peak F-test 9 / 0 / 0 / 1.18 24 / 0 / 0 / 0.94 10 / 0 / 0 / 1.56 0 False
rest all per subject / per articulator / half peak t-test 6 / 1 / 0 / 3.54 25 / 0 / 0 / 0.83 6 / 0 / 0 / 0.48 1 False
rest all pooled / all six / threshold F-test 11 / 0 / 0 / 1.18 22 / 0 / 0 / 0.94 12 / 0 / 0 / 1.00 0 False
rest all pooled / all six / threshold t-test 12 / 0 / 0 / 1.71 25 / 0 / 0 / 0.83 10 / 0 / 0 / 1.00 0 False
rest all pooled / all six / half peak F-test 11 / 1 / 0 / 2.30 20 / 0 / 0 / 0.94 11 / 0 / 0 / 1.00 1 False
rest all pooled / all six / half peak t-test 5 / 1 / 0 / 2.30 19 / 0 / 0 / 0.83 8 / 0 / 0 / 1.00 1 False
rest all pooled / per articulator / threshold F-test 12 / 0 / 0 / 1.18 28 / 0 / 0 / 0.94 18 / 0 / 0 / 1.00 0 False
rest all pooled / per articulator / threshold t-test 13 / 1 / 0 / 3.54 25 / 0 / 0 / 0.83 15 / 0 / 0 / 0.48 1 False
rest all pooled / per articulator / half peak F-test 11 / 1 / 0 / 2.30 27 / 0 / 0 / 0.94 10 / 0 / 0 / 1.56 1 False
rest all pooled / per articulator / half peak t-test 5 / 2 / 0 / 3.54 19 / 0 / 0 / 0.83 6 / 0 / 0 / 0.48 2 False
rest 15+ per subject / all six / threshold F-test 12 / 2 / 0 / 3.54 33 / 0 / 0 / 0.94 17 / 0 / 0 / 1.56 2 False
rest 15+ per subject / all six / threshold t-test 8 / 0 / 0 / 1.18 26 / 0 / 0 / 0.83 18 / 0 / 0 / 1.56 0 False
rest 15+ per subject / all six / half peak F-test 8 / 0 / 0 / 1.18 25 / 0 / 0 / 0.83 12 / 0 / 0 / 1.56 0 False
rest 15+ per subject / all six / half peak t-test 6 / 0 / 0 / 1.18 18 / 0 / 0 / 0.83 7 / 0 / 0 / 0.52 0 False
rest 15+ per subject / per articulator / threshold F-test 12 / 1 / 0 / 3.54 24 / 0 / 0 / 0.83 17 / 0 / 0 / 1.00 1 False
rest 15+ per subject / per articulator / threshold t-test 9 / 0 / 0 / 1.18 26 / 0 / 0 / 0.83 14 / 0 / 0 / 0.48 0 False
rest 15+ per subject / per articulator / half peak F-test 9 / 0 / 0 / 1.18 25 / 0 / 0 / 0.83 10 / 0 / 0 / 1.56 0 False
rest 15+ per subject / per articulator / half peak t-test 5 / 0 / 0 / 1.18 20 / 0 / 0 / 0.83 6 / 0 / 0 / 0.48 0 False
rest 15+ pooled / all six / threshold F-test 11 / 1 / 0 / 3.54 33 / 0 / 0 / 0.94 12 / 0 / 0 / 1.00 1 False
rest 15+ pooled / all six / threshold t-test 8 / 0 / 0 / 1.18 26 / 0 / 0 / 0.83 13 / 0 / 0 / 1.00 0 False
rest 15+ pooled / all six / half peak F-test 9 / 0 / 0 / 1.18 25 / 0 / 0 / 0.83 11 / 0 / 0 / 1.00 0 False
rest 15+ pooled / all six / half peak t-test 6 / 0 / 0 / 1.18 18 / 0 / 0 / 0.83 8 / 0 / 0 / 0.84 0 False
rest 15+ pooled / per articulator / threshold F-test 12 / 0 / 0 / 1.18 28 / 0 / 0 / 0.94 12 / 0 / 0 / 1.00 0 False
rest 15+ pooled / per articulator / threshold t-test 11 / 0 / 0 / 1.71 22 / 0 / 0 / 0.83 10 / 0 / 0 / 1.00 0 False
rest 15+ pooled / per articulator / half peak F-test 10 / 1 / 0 / 2.30 25 / 0 / 0 / 0.83 9 / 0 / 0 / 1.00 1 False
rest 15+ pooled / per articulator / half peak t-test 6 / 1 / 0 / 2.30 17 / 0 / 0 / 0.83 7 / 0 / 0 / 1.00 1 False
Readings where it holds: 0 of 96
Largest /ju/ z at +250 ms among lips-positive GLM electrodes, over all readings and subjects: 3.54
Readings with a lips-positive GLM electrode above 2 for /ju/: 13 of 96 (baselines: {'rest': 13} ); in all of them /ja/ or /ji/ is also above 2 there
Outcome over readings: not reproduced
All vSMC electrodes, without the GLM selection (not the criterion; it separates the selection from the effect):
baseline CVs subject vSMC /ju/ > 2 /ju/ > 2, /ja/ and /ji/ < 2 largest /ju/
pre all EC2 86 0 0 1.13
pre all GP31 83 0 0 0.40
pre all GP33 99 0 0 0.91
pre 15+ EC2 86 0 0 1.13
pre 15+ GP31 83 0 0 0.40
pre 15+ GP33 99 0 0 0.91
whole all EC2 86 0 0 1.19
whole all GP31 83 0 0 0.34
whole all GP33 99 0 0 0.84
whole 15+ EC2 86 0 0 1.19
whole 15+ GP31 83 0 0 0.34
whole 15+ GP33 99 0 0 0.84
rest all EC2 86 8 1 3.54
rest all GP31 83 0 0 0.94
rest all GP33 99 0 0 1.74
rest 15+ EC2 86 8 1 3.54
rest 15+ GP31 83 0 0 0.94
rest 15+ GP33 99 0 0 1.74
EC2: plotted electrode 107 (largest /ju/ minus /ja/,/ji/ at +250 ms among lips+ GLM electrodes), z at +250 ms: /ja/ -0.18, /ji/ -0.37, /ju/ 0.96; plotted range -0.56 to 1.03; mean |difference| between the three before 0 ms 0.18
GP31: plotted electrode 108 (largest /ju/ minus /ja/,/ji/ at +250 ms among lips+ GLM electrodes), z at +250 ms: /ja/ -0.22, /ji/ -0.36, /ju/ 0.15; plotted range -0.39 to 0.29; mean |difference| between the three before 0 ms 0.12 GP33: plotted electrode 110 (largest /ju/ minus /ja/,/ji/ at +250 ms among lips+ GLM electrodes), z at +250 ms: /ja/ -0.02, /ji/ -0.38, /ju/ 0.61; plotted range -0.81 to 1.14; mean |difference| between the three before 0 ms 0.29
Result: not reproduced. Under all 96 readings (three baselines, two CV sets, eight timing
readings for T_A, two p-value readings) no GLM electrode with a positive lips weight has z above 2
for /ju/ at +250 ms and below 2 for /ja/ and /ji/. With the rest baseline some do exceed 2 for
/ju/ (13 of 96 readings, largest 3.54), but at each of them /ja/ or /ji/ is above 2 as well; under
pre and whole none reaches 2. The /ju/ responses of the plotted electrodes are larger than
/ja/ and /ji/ at +250 ms, but under the threshold. Without the GLM selection, among all vSMC
electrodes, one EC2 electrode under rest (both CV sets) has /ju/ above 2 and /ja/ and /ji/ below
2; under pre and whole no vSMC electrode in any subject reaches 2 for /ju/ (largest 1.19), so
there the result does not come from the selection. The other parts of the claim are not judged
(no measure in the paper).
Spatial claims (blocked)¶
fig2a-lips-dorsal, fig2b-somatotopic-sequence, suppfig3b-spatial-clustering, suppfig3d-cross-subject, suppfig3f-dv-sequence, suppfig5-intermediate-organization and suppfig7-no-latency-difference all need each electrode's anterior-posterior distance from the central sulcus and dorsal-ventral distance from the Sylvian fissure (Methods). The dandiset gives neither. EC2 and GP31 have x, y, z coordinates with no unit, frame or description; GP33, one of the three subjects, has none. The labels say precentral or postcentral, not how far from either sulcus, and no electrode is labelled as on the guenon. Asked whether the paper gives a way to obtain these distances from the grid geometry, the reader answered that it does not: the paper measures them on each subject's MRI reconstruction. These claims stay blocked as metadata_insufficient.
The GLM electrode selection and weights of fig2a-lips-dorsal do not need positions and are computed above. For information only (not an outcome, and not a reading of the claim), the median z coordinate of the top 10% of each articulator's weights in the two subjects that have coordinates, under the first reading:
In [16]:
key0 = W_READINGS[0]
print("reading:", key0)
for s in ("EC2", "GP31"):
w = WEIGHTS[key0][s]
z = S[s]["z"][S[s]["vsmc"]]
kept = np.where(w["keep"])[0]
out = []
for a, name in enumerate(B.ARTICULATORS):
if np.isnan(w["W"][kept, a]).all():
out.append(f"{name} no window"); continue
top = kept[np.argsort(-w["W"][kept, a])[:max(1, len(kept) // 10)]]
out.append(f"{name} {np.median(z[top]):.1f}")
print(f"{s}: {len(kept)} GLM electrodes; median z of top 10% weights: " + ", ".join(out))
reading: ('pre', 'all', ('per subject', 'all six', 'threshold'), 'F-test')
EC2: 36 GLM electrodes; median z of top 10% weights: Lips 36.6, Tongue 15.6, Larynx 11.3, Jaw 19.9
GP31: 36 GLM electrodes; median z of top 10% weights: Lips 44.3, Tongue no window, Larynx 22.8, Jaw 28.8
suppfig6-mixed-tuning¶
Single electrodes show a clear preferred articulator and also carry weights for other articulators.
In [17]:
def mixed_tuning(Wk):
"""The criterion on weights Wk (electrode, 4): holds, group means, electrodes without a positive weight."""
Wk = np.nan_to_num(Wk, nan=-np.inf)
pos = np.where(Wk > 0, Wk, 0)
has = pos.sum(1) > 0
pct = 100 * pos[has] / pos[has].sum(1, keepdims=True)
pref = np.argmax(Wk[has], axis=1)
groups, ok = {}, True
for a in range(4):
g = pct[pref == a]
if len(g) == 0:
groups[B.ARTICULATORS[a]] = "none"
continue
m = g.mean(0)
ok &= bool(np.argmax(m) == a and (np.delete(m, a) > 0).any())
groups[B.ARTICULATORS[a]] = f"n={len(g)} " + "/".join(f"{x:.0f}" for x in m)
return ok, groups, int((~has).sum())
rng6 = np.random.default_rng(6)
s6 = []
for key in W_READINGS:
Wk = np.concatenate([WEIGHTS[key][s]["W"][WEIGHTS[key][s]["keep"]] for s in B.SUBJECTS])
no_window = int(np.isnan(Wk).any(1).sum())
# Null: each articulator's weights permuted across the subject's GLM electrodes, which breaks any
# tuning of single electrodes while keeping each articulator's distribution of weights.
Wnull = np.concatenate([np.column_stack([rng6.permutation(col) for col in WEIGHTS[key][s]["W"][WEIGHTS[key][s]["keep"]].T])
for s in B.SUBJECTS])
null_ok = mixed_tuning(Wnull)[0]
ok, groups, no_pos = mixed_tuning(Wk)
s6.append({"baseline": key[0], "CVs": key[1], "timing reading": " / ".join(key[2]), "p-value": key[3],
"electrodes": int(len(Wk)), "with an articulator lacking T_A": no_window,
"no positive weight": no_pos, **groups, "holds": ok, "holds on permuted weights": null_ok})
s6 = pd.DataFrame(s6)
print("Group means of positive weights as % (Lips/Tongue/Larynx/Jaw), electrodes grouped by largest W:")
print(s6.to_string(index=False))
print("\nReadings where it holds:", int(s6.holds.sum()), "of", len(s6))
for a, name in enumerate(B.ARTICULATORS):
v = [float(x.split(" ")[1].split("/")[a]) for x in s6[name] if x != "none"]
print(f"{name} group: preferred share {min(v):.0f} to {max(v):.0f}% over the {len(v)} readings with such electrodes")
print("GLM electrodes over the three subjects:", s6.electrodes.min(), "to", s6.electrodes.max())
print("Outcome over readings:", overall(["reproduced" if h else "not reproduced" for h in s6.holds]))
print("Null, the same test on weights permuted across electrodes (one permutation per reading): holds in",
int(s6["holds on permuted weights"].sum()), "of", len(s6))
Group means of positive weights as % (Lips/Tongue/Larynx/Jaw), electrodes grouped by largest W:
baseline CVs timing reading p-value electrodes with an articulator lacking T_A no positive weight Lips Tongue Larynx Jaw holds holds on permuted weights
pre all per subject / all six / threshold F-test 103 36 0 n=30 74/2/4/21 n=6 0/79/8/13 n=17 9/9/65/17 n=50 8/7/7/78 True True
pre all per subject / all six / threshold t-test 103 35 0 n=29 76/1/5/18 n=8 0/83/6/11 n=14 9/10/68/13 n=52 7/7/5/82 True True
pre all per subject / all six / half peak F-test 82 27 0 n=26 73/1/4/22 n=6 1/85/7/8 n=9 13/8/71/8 n=41 10/9/2/79 True True
pre all per subject / all six / half peak t-test 62 26 0 n=22 75/1/3/21 n=3 3/85/2/10 n=5 17/2/70/11 n=32 10/7/2/81 True True
pre all per subject / per articulator / threshold F-test 110 0 0 n=30 65/10/6/19 n=14 16/65/8/12 n=13 5/16/68/12 n=53 8/10/4/77 True True
pre all per subject / per articulator / threshold t-test 92 0 0 n=23 63/14/4/19 n=10 20/68/5/6 n=12 4/12/71/13 n=47 6/11/3/80 True True
pre all per subject / per articulator / half peak F-test 75 0 0 n=24 69/9/3/18 n=7 4/74/5/16 n=7 15/14/63/8 n=37 9/14/5/72 True True
pre all per subject / per articulator / half peak t-test 58 0 0 n=18 67/13/4/16 n=5 25/53/5/16 n=6 12/12/67/9 n=29 7/13/3/76 True True
pre all pooled / all six / threshold F-test 101 34 0 n=31 71/1/4/24 n=4 0/91/4/6 n=14 5/2/75/18 n=52 9/5/5/82 True True
pre all pooled / all six / threshold t-test 99 39 0 n=24 74/0/4/22 n=5 2/82/4/12 n=11 6/1/81/12 n=59 10/4/4/83 True True
pre all pooled / all six / half peak F-test 76 25 0 n=26 73/1/3/23 n=4 0/92/3/4 n=8 11/4/75/9 n=38 8/6/2/84 True True
pre all pooled / all six / half peak t-test 62 25 0 n=21 75/0/3/22 n=3 5/79/2/14 n=5 9/1/81/9 n=33 9/4/3/84 True True
pre all pooled / per articulator / threshold F-test 108 0 0 n=28 68/8/5/19 n=12 8/77/8/7 n=14 13/6/65/16 n=54 11/8/6/75 True True
pre all pooled / per articulator / threshold t-test 101 0 0 n=24 69/7/5/18 n=7 13/67/6/14 n=12 11/5/68/15 n=58 9/8/6/77 True True
pre all pooled / per articulator / half peak F-test 73 0 0 n=25 70/8/3/20 n=4 2/96/2/0 n=7 11/6/73/10 n=37 11/11/3/74 True True
pre all pooled / per articulator / half peak t-test 60 0 0 n=19 70/9/4/17 n=4 18/60/1/22 n=6 10/7/71/11 n=31 7/11/3/79 True True
pre 15+ per subject / all six / threshold F-test 114 47 0 n=34 71/2/6/21 n=5 0/82/2/16 n=23 14/7/63/16 n=52 10/6/4/80 True True
pre 15+ per subject / all six / threshold t-test 98 30 0 n=30 76/1/5/18 n=6 0/91/2/7 n=19 9/9/70/11 n=43 7/8/4/81 True True
pre 15+ per subject / all six / half peak F-test 90 35 0 n=30 73/1/5/20 n=6 1/80/11/8 n=10 11/7/73/9 n=44 12/8/2/78 True True
pre 15+ per subject / all six / half peak t-test 56 20 0 n=22 79/1/4/16 n=3 3/84/3/10 n=3 14/4/80/2 n=28 9/8/2/81 True False
pre 15+ per subject / per articulator / threshold F-test 116 0 0 n=30 69/7/6/18 n=15 15/64/8/13 n=16 5/14/67/14 n=55 9/8/6/78 True True
pre 15+ per subject / per articulator / threshold t-test 97 0 0 n=24 70/8/6/16 n=9 23/59/11/7 n=15 4/10/71/15 n=49 8/7/6/80 True True
pre 15+ per subject / per articulator / half peak F-test 81 0 0 n=25 72/6/5/17 n=5 6/67/7/19 n=9 12/8/66/15 n=42 11/10/5/74 True True
pre 15+ per subject / per articulator / half peak t-test 58 0 0 n=19 74/7/6/13 n=4 29/55/6/10 n=5 13/11/68/8 n=30 8/9/3/79 True True
pre 15+ pooled / all six / threshold F-test 115 44 1 n=34 72/2/6/20 n=7 6/66/8/20 n=21 6/6/72/17 n=52 9/6/6/79 True True
pre 15+ pooled / all six / threshold t-test 107 36 1 n=27 75/0/6/19 n=9 5/70/8/16 n=20 5/7/73/15 n=50 8/5/6/80 True True
pre 15+ pooled / all six / half peak F-test 90 32 0 n=30 74/1/4/21 n=7 7/66/16/11 n=8 9/0/82/8 n=45 9/8/4/78 True True
pre 15+ pooled / all six / half peak t-test 63 24 0 n=21 78/0/4/18 n=4 12/72/11/5 n=5 11/7/74/9 n=33 9/6/3/83 True True
pre 15+ pooled / per articulator / threshold F-test 117 0 0 n=34 70/5/7/18 n=13 9/74/5/11 n=14 8/9/69/14 n=56 10/5/6/79 True True
pre 15+ pooled / per articulator / threshold t-test 100 0 0 n=26 73/4/6/17 n=9 9/80/4/6 n=15 8/9/72/12 n=50 8/6/5/80 True True
pre 15+ pooled / per articulator / half peak F-test 86 0 0 n=29 72/4/4/20 n=6 6/73/4/17 n=10 12/8/74/6 n=41 10/9/3/77 True True
pre 15+ pooled / per articulator / half peak t-test 61 0 0 n=19 75/5/4/16 n=5 12/68/2/18 n=6 9/6/76/10 n=31 9/8/2/81 True True
whole all per subject / all six / threshold F-test 107 36 0 n=33 71/3/5/21 n=6 0/78/9/13 n=18 9/10/64/17 n=50 8/7/7/78 True True
whole all per subject / all six / threshold t-test 95 30 0 n=26 74/1/3/21 n=7 0/77/6/17 n=13 11/11/63/15 n=49 5/7/6/82 True True
whole all per subject / all six / half peak F-test 77 29 0 n=28 72/1/4/23 n=5 2/89/6/3 n=7 14/5/68/13 n=37 10/9/3/77 True True
whole all per subject / all six / half peak t-test 58 26 0 n=22 74/1/2/23 n=2 5/93/3/0 n=3 17/0/61/21 n=31 10/7/2/81 True True
whole all per subject / per articulator / threshold F-test 105 0 0 n=29 65/11/2/22 n=12 9/74/9/7 n=14 12/7/68/13 n=50 9/9/4/78 True True
whole all per subject / per articulator / threshold t-test 93 0 0 n=21 65/10/2/22 n=4 15/78/7/0 n=11 12/6/65/17 n=57 9/10/4/78 True True
whole all per subject / per articulator / half peak F-test 72 0 0 n=26 72/7/2/19 n=4 8/83/1/9 n=6 12/8/73/7 n=36 10/12/4/74 True True
whole all per subject / per articulator / half peak t-test 53 0 0 n=17 70/9/2/18 n=3 22/63/1/14 n=5 14/9/68/9 n=28 8/11/3/78 True True
whole all pooled / all six / threshold F-test 106 35 1 n=30 71/2/4/24 n=3 0/87/5/8 n=15 7/5/75/13 n=57 10/6/5/80 True True
whole all pooled / all six / threshold t-test 96 35 0 n=22 73/1/3/24 n=6 2/77/3/18 n=11 6/2/79/14 n=57 10/4/3/83 True True
whole all pooled / all six / half peak F-test 85 27 0 n=30 72/2/4/22 n=4 0/92/3/4 n=9 12/3/75/10 n=42 9/6/2/82 True True
whole all pooled / all six / half peak t-test 61 23 0 n=20 76/0/2/22 n=3 4/81/5/10 n=4 2/1/84/13 n=34 10/6/3/82 True True
whole all pooled / per articulator / threshold F-test 106 0 0 n=31 66/11/5/17 n=9 2/80/8/10 n=13 4/11/62/23 n=53 10/9/7/74 True True
whole all pooled / per articulator / threshold t-test 100 0 0 n=25 65/12/6/17 n=6 10/70/15/5 n=13 7/7/64/22 n=56 8/11/6/75 True True
whole all pooled / per articulator / half peak F-test 73 0 0 n=28 66/10/5/18 n=4 8/71/14/8 n=7 14/8/67/11 n=34 9/11/6/74 True True
whole all pooled / per articulator / half peak t-test 60 0 0 n=20 67/12/3/18 n=4 17/60/1/22 n=4 15/0/66/19 n=32 8/10/7/75 True True
whole 15+ per subject / all six / threshold F-test 111 41 0 n=33 70/2/7/20 n=6 0/85/2/12 n=18 13/10/62/15 n=54 10/6/5/78 True True
whole 15+ per subject / all six / threshold t-test 88 25 0 n=26 76/1/5/18 n=6 0/92/2/6 n=13 12/8/65/15 n=43 7/7/3/83 True True
whole 15+ per subject / all six / half peak F-test 78 30 0 n=29 75/1/4/20 n=5 2/85/10/3 n=6 12/5/73/10 n=38 11/9/4/76 True True
whole 15+ per subject / all six / half peak t-test 53 21 0 n=23 78/1/3/19 n=2 5/92/3/0 n=1 12/0/59/30 n=27 8/8/3/81 True True
whole 15+ per subject / per articulator / threshold F-test 114 0 0 n=29 70/7/3/19 n=13 11/73/9/7 n=16 12/8/65/16 n=56 10/7/5/78 True True
whole 15+ per subject / per articulator / threshold t-test 95 0 0 n=24 72/6/3/19 n=4 15/77/8/0 n=14 10/6/62/21 n=53 9/6/5/79 True True
whole 15+ per subject / per articulator / half peak F-test 78 0 0 n=27 72/6/4/18 n=4 8/80/3/9 n=8 16/6/70/9 n=39 13/9/5/73 True True
whole 15+ per subject / per articulator / half peak t-test 54 0 0 n=19 72/7/5/16 n=2 19/74/1/5 n=5 15/9/67/9 n=28 9/8/3/80 True True
whole 15+ pooled / all six / threshold F-test 116 47 1 n=35 71/2/6/21 n=9 7/67/13/13 n=16 7/5/75/13 n=55 10/5/6/78 True True
whole 15+ pooled / all six / threshold t-test 111 39 1 n=27 73/0/4/23 n=11 6/70/13/11 n=17 6/9/73/12 n=55 8/5/6/80 True True
whole 15+ pooled / all six / half peak F-test 89 31 0 n=29 75/1/4/20 n=7 8/68/13/11 n=8 11/1/79/9 n=45 11/8/4/77 True True
whole 15+ pooled / all six / half peak t-test 56 22 0 n=22 79/0/3/18 n=3 16/65/11/7 n=3 1/10/82/6 n=28 10/6/2/82 True True
whole 15+ pooled / per articulator / threshold F-test 116 0 0 n=33 69/7/3/21 n=10 9/79/5/8 n=13 10/6/73/11 n=60 11/6/6/77 True True
whole 15+ pooled / per articulator / threshold t-test 99 0 0 n=23 70/6/4/20 n=5 12/81/4/3 n=16 7/5/73/15 n=55 10/6/5/79 True True
whole 15+ pooled / per articulator / half peak F-test 86 0 0 n=31 69/7/5/19 n=4 2/90/1/7 n=10 13/7/71/9 n=41 12/9/3/76 True True
whole 15+ pooled / per articulator / half peak t-test 55 0 0 n=19 72/7/4/17 n=3 10/76/0/14 n=6 10/3/72/15 n=27 10/7/2/80 True True
rest all per subject / all six / threshold F-test 102 34 0 n=28 70/1/4/25 n=9 5/71/17/6 n=13 6/12/62/20 n=52 8/8/7/77 True True
rest all per subject / all six / threshold t-test 103 33 0 n=20 71/1/4/25 n=14 10/73/9/8 n=14 5/14/62/18 n=55 7/7/5/80 True True
rest all per subject / all six / half peak F-test 75 26 0 n=27 71/2/3/25 n=5 5/83/8/4 n=8 13/12/60/16 n=35 9/9/4/78 True True
rest all per subject / all six / half peak t-test 53 20 0 n=19 75/1/2/22 n=4 11/68/16/4 n=2 16/0/54/30 n=28 12/6/2/81 True True
rest all per subject / per articulator / threshold F-test 99 0 0 n=23 67/9/4/20 n=9 11/67/13/9 n=13 2/16/61/21 n=54 11/8/10/70 True True
rest all per subject / per articulator / threshold t-test 96 0 0 n=20 69/10/4/17 n=7 14/72/11/4 n=13 2/13/67/18 n=56 9/8/9/74 True True
rest all per subject / per articulator / half peak F-test 71 0 0 n=23 72/6/3/19 n=3 6/80/11/3 n=6 12/8/60/20 n=39 11/11/9/69 True True
rest all per subject / per articulator / half peak t-test 57 0 0 n=17 71/8/3/17 n=2 23/51/11/15 n=6 12/4/64/20 n=32 12/9/6/73 True True
rest all pooled / all six / threshold F-test 93 26 0 n=26 72/2/3/23 n=8 2/80/2/16 n=14 6/12/65/18 n=45 8/6/4/81 True True
rest all pooled / all six / threshold t-test 94 28 0 n=19 76/0/3/21 n=8 2/83/3/12 n=17 5/15/65/15 n=50 10/6/2/81 True True
rest all pooled / all six / half peak F-test 74 22 0 n=25 73/2/2/22 n=6 4/83/9/4 n=6 11/14/63/13 n=37 11/8/3/79 True True
rest all pooled / all six / half peak t-test 57 20 0 n=17 78/1/2/20 n=3 9/86/3/1 n=4 10/14/68/8 n=33 10/7/2/81 True True
rest all pooled / per articulator / threshold F-test 105 0 0 n=24 69/6/4/21 n=10 15/63/14/8 n=12 3/15/58/24 n=59 10/7/9/74 True True
rest all pooled / per articulator / threshold t-test 95 0 1 n=19 70/7/5/18 n=6 20/66/14/0 n=12 4/9/63/23 n=57 9/7/7/76 True True
rest all pooled / per articulator / half peak F-test 75 0 0 n=22 72/4/3/20 n=3 7/79/11/3 n=7 13/6/59/22 n=43 12/9/9/71 True True
rest all pooled / per articulator / half peak t-test 52 0 0 n=14 76/6/3/15 n=1 31/48/21/0 n=4 11/8/63/18 n=33 11/7/8/74 True True
rest 15+ per subject / all six / threshold F-test 106 0 1 n=27 71/5/3/21 n=10 8/69/16/7 n=13 15/12/52/20 n=55 10/9/5/76 True True
rest 15+ per subject / all six / threshold t-test 93 0 1 n=21 70/5/4/21 n=17 13/72/7/8 n=10 9/10/60/21 n=44 8/9/2/80 True True
rest 15+ per subject / all six / half peak F-test 74 0 0 n=27 68/6/4/23 n=6 10/74/10/7 n=6 12/17/62/9 n=35 10/9/5/76 True True
rest 15+ per subject / all six / half peak t-test 52 0 0 n=18 72/4/2/22 n=6 19/61/11/9 n=1 34/0/49/17 n=27 11/6/2/82 True True
rest 15+ per subject / per articulator / threshold F-test 99 0 0 n=25 72/4/3/22 n=8 16/65/17/2 n=9 5/17/58/21 n=57 9/7/11/74 True True
rest 15+ per subject / per articulator / threshold t-test 94 0 0 n=23 73/4/3/20 n=5 10/77/13/0 n=11 1/10/66/23 n=55 9/5/9/77 True True
rest 15+ per subject / per articulator / half peak F-test 71 0 0 n=24 72/2/4/22 n=3 6/77/14/3 n=4 10/10/68/12 n=40 11/8/12/69 True True
rest 15+ per subject / per articulator / half peak t-test 52 0 0 n=17 74/3/4/19 none n=3 10/6/73/11 n=32 10/4/9/77 True True
rest 15+ pooled / all six / threshold F-test 102 0 0 n=28 69/6/4/21 n=13 7/67/21/5 n=10 9/10/61/21 n=51 10/8/6/76 True True
rest 15+ pooled / all six / threshold t-test 96 0 0 n=22 70/5/4/21 n=18 9/71/13/7 n=12 4/13/67/16 n=44 9/9/4/78 True True
rest 15+ pooled / all six / half peak F-test 78 0 0 n=28 67/6/4/23 n=8 9/70/16/5 n=5 8/13/68/12 n=37 9/10/5/76 True True
rest 15+ pooled / all six / half peak t-test 57 0 0 n=18 72/4/2/22 n=7 16/58/9/17 n=4 9/15/64/12 n=28 10/7/2/80 True True
rest 15+ pooled / per articulator / threshold F-test 99 0 0 n=27 70/5/3/22 n=7 6/82/8/4 n=10 9/16/58/17 n=55 9/7/4/80 True True
rest 15+ pooled / per articulator / threshold t-test 90 0 0 n=20 72/4/2/21 n=7 6/71/17/6 n=11 7/15/61/16 n=52 8/7/3/81 True True
rest 15+ pooled / per articulator / half peak F-test 71 0 0 n=28 68/4/3/24 n=3 0/94/6/0 n=4 17/10/59/14 n=36 9/6/6/79 True False
rest 15+ pooled / per articulator / half peak t-test 49 0 0 n=18 72/3/2/23 n=1 24/71/0/5 n=2 22/0/59/20 n=28 10/4/4/83 True True
Readings where it holds: 96 of 96
Lips group: preferred share 63 to 79% over the 96 readings with such electrodes
Tongue group: preferred share 48 to 96% over the 95 readings with such electrodes
Larynx group: preferred share 49 to 84% over the 96 readings with such electrodes
Jaw group: preferred share 69 to 84% over the 96 readings with such electrodes
GLM electrodes over the three subjects: 49 to 117
Outcome over readings: reproduced
Null, the same test on weights permuted across electrodes (one permutation per reading): holds in 94 of 96
Result: reproduced. Under all 96 readings each group's preferred articulator has the largest mean share of positive weight (Lips 63 to 79%, Tongue 48 to 96%, Larynx 49 to 84%, Jaw 69 to 84%), and the other articulators keep a non-zero share (table). In one reading no electrode prefers the tongue, so that group is absent there. The GLM electrodes number 49 to 117 over the three subjects. In readings where an articulator has no window T_A in a subject, its weights there are missing and those electrodes are grouped on the other articulators. The test barely discriminates: on weights permuted across each subject's electrodes, which removes any tuning of single electrodes, it holds in 94 of 96 readings, because grouping electrodes by their largest weight puts the preferred articulator first almost by construction. The outcome follows the criterion; that single electrodes have a clear preferred articulator beyond what the grouping gives is not shown by it.
fig3d-timing-sequence¶
Articulator correlations follow a staggered sequence consistent across subjects; lip and tongue correlations begin well before sound onset, jaw and larynx align to the C-V transition, and the high and back tongue vowel features peak near the vowel midpoint (~250 ms).
Acoustic onset: the dandiset has no audio, so the onset of each CV is taken from the hand-marked syllable start of the trial table (mean over renditions, then over CVs). The criterion allows this substitute since the reader's answer, as a deviation from the prototype and thresholding of the Methods, with the result against t = 0 shown beside it.
In [18]:
onsets = []
for s in B.SUBJECTS:
for b in blocks[s]:
k = b["kept"]
for lab, st, cv in zip(b["condition"][k], b["start"][k], b["cv"][k]):
p = B.parse_label(lab)
if p:
onsets.append({"subject": s, "cv": p, "onset_ms": 1000 * (st - cv)})
onsets = pd.DataFrame(onsets)
per_cv = onsets.groupby(["subject", "cv"]).onset_ms.mean()
ACOUSTIC_ONSET = per_cv.mean()
print(f"Acoustic onset, substitute from the hand-marked syllable starts: mean over {len(per_cv)} CVs {ACOUSTIC_ONSET:.1f} ms "
f"(per subject {per_cv.groupby(level=0).mean().round(1).to_dict()}; CV range {per_cv.min():.0f} to {per_cv.max():.0f} ms)")
def fig3d_table(TIM):
"""The criterion under every reading, from timing tables {(baseline, CVs, reading): rows}."""
f3 = []
for base in BASES:
for cs in CVSETS:
for reading in TIMING_READINGS:
t = TIM[base, cs, reading]
for anova in ("onset", "peak"):
F, p = stats.f_oneway(*[t[t.articulator == a][anova] for a in ART6])
mo = t.groupby("articulator").onset.mean()
mp = t.groupby("articulator").peak.mean()
c1 = mo["Lips"] < ACOUSTIC_ONSET and mo["Tongue"] < ACOUSTIC_ONSET
c2 = min(mo["Jaw"], mo["Larynx"]) > max(mo["Lips"], mo["Tongue"])
c3 = all(abs(mp[a] - 250) <= 25 for a in B.VOWEL_ARTICULATORS)
c1_0 = mo["Lips"] < 0 and mo["Tongue"] < 0
n_ok, n_ok0 = c1 + c2 + c3, c1_0 + c2 + c3
outcome = ("reproduced" if p < 1e-10 and n_ok == 3 else
"partially" if p < 1e-10 and n_ok == 2 else "not reproduced")
outcome0 = ("reproduced" if p < 1e-10 and n_ok0 == 3 else
"partially" if p < 1e-10 and n_ok0 == 2 else "not reproduced")
f3.append({"baseline": base, "CVs": cs, "timing reading": " / ".join(reading), "ANOVA on": anova,
"n": len(t), **{f"n {a[:6]}": int((t.articulator == a).sum()) for a in ART6},
"F": round(F, 1), "p": p, **{f"onset {a[:6]}": round(mo.get(a, np.nan)) for a in ART6},
"peak Back": round(mp["Back Tongue"]), "peak High": round(mp["High Tongue"]),
"lips,tongue < acoustic": c1, "jaw,larynx later": c2, "vowel peaks 250+-25": c3,
"outcome": outcome, "lips,tongue < 0": c1_0, "outcome vs t = 0": outcome0})
return pd.DataFrame(f3)
f3 = fig3d_table(TIMING)
print(f3.to_string(index=False))
print("\nOutcome over readings (substitute acoustic onset):", overall(f3.outcome))
print("Beside it, against t = 0:", overall(f3["outcome vs t = 0"]), "| readings with lips and tongue onsets before 0 ms:",
int(f3["lips,tongue < 0"].sum()), "of", len(f3), "| before the substitute onset:", int(f3["lips,tongue < acoustic"].sum()))
print("Ranges over readings: mean onset Lips", f3["onset Lips"].min(), "to", f3["onset Lips"].max(),
"| Tongue", f3["onset Tongue"].min(), "to", f3["onset Tongue"].max(), "| Jaw", f3["onset Jaw"].min(), "to", f3["onset Jaw"].max(),
"| Larynx", f3["onset Larynx"].min(), "to", f3["onset Larynx"].max(), "| peak Back", f3["peak Back"].min(), "to", f3["peak Back"].max(),
"| peak High", f3["peak High"].min(), "to", f3["peak High"].max())
print("ANOVA F", f3.F.min(), "to", f3.F.max(), "| readings with p < 1e-10:", int((f3.p < 1e-10).sum()), "of", len(f3),
"| conditions holding per reading:", (f3["lips,tongue < acoustic"].astype(int) + f3["jaw,larynx later"] + f3["vowel peaks 250+-25"]).value_counts().to_dict())
print("Total electrodes over articulators and subjects: from", f3.n.min(), "to", f3.n.max(), "(paper 211 +- 21)")
print("Electrodes per articulator over readings (min to max):", {a: (int(f3[f"n {a[:6]}"].min()), int(f3[f"n {a[:6]}"].max())) for a in ART6},
"| Tongue under the 'all six' readings:", (int(f3[f3["timing reading"].str.contains("all six")]["n Tongue"].min()),
int(f3[f3["timing reading"].str.contains("all six")]["n Tongue"].max())))
print("Per subject, first reading (pre, all CVs):", TIMING["pre", "all", TIMING_READINGS[0]].groupby(["subject", "articulator"]).size().unstack().to_dict("index"))
for s in B.SUBJECTS:
t = TIMING["pre", "all", TIMING_READINGS[0]]
print(s, "mean onsets (ms):", t[t.subject == s].groupby("articulator").onset.mean().round().to_dict())
fig, ax = plt.subplots(figsize=(6, 3))
t = TIMING["pre", "all", TIMING_READINGS[0]]
for i, a in enumerate(ART6):
g = t[t.articulator == a]
ax.errorbar(g.onset.mean(), i, xerr=g.onset.sem(), fmt="<", color="C0")
ax.errorbar(g.peak.mean(), i, xerr=g.peak.sem(), fmt="o", color="C1")
ax.errorbar(g.offset.mean(), i, xerr=g.offset.sem(), fmt=">", color="C2")
ax.axvline(ACOUSTIC_ONSET, color="grey", ls="--", label="substitute acoustic onset"); ax.axvline(0, color="k", lw=0.5, label="C-V transition")
ax.legend(fontsize=7)
ax.set_yticks(range(6)); ax.set_yticklabels(ART6); ax.set_xlabel("ms (first reading, pre baseline, all CVs)")
plt.show()
# Variant (not a reading): the artifact trials left out of the CV means.
RHO_CLEAN = {(s, base, cs): B.partial_correlation(HV_CLEAN[s, base, cs], AM[s, cs]) for s in B.SUBJECTS for base in BASES for cs in CVSETS}
f3c = fig3d_table({(base, cs, r): timing_rows(base, cs, r, RHO_CLEAN) for base in BASES for cs in CVSETS for r in TIMING_READINGS})
print("Variant, artifact trials left out: outcome over readings", overall(f3c.outcome), "| against t = 0", overall(f3c["outcome vs t = 0"]),
"| conditions holding per reading", (f3c["lips,tongue < acoustic"].astype(int) + f3c["jaw,larynx later"] + f3c["vowel peaks 250+-25"]).value_counts().to_dict(),
"| mean onset Lips", f3c["onset Lips"].min(), "to", f3c["onset Lips"].max(), "| Tongue", f3c["onset Tongue"].min(), "to", f3c["onset Tongue"].max(),
"| Jaw", f3c["onset Jaw"].min(), "to", f3c["onset Jaw"].max(), "| peak High", f3c["peak High"].min(), "to", f3c["peak High"].max(),
"| electrodes", f3c.n.min(), "to", f3c.n.max())
Acoustic onset, substitute from the hand-marked syllable starts: mean over 169 CVs -109.6 ms (per subject {'EC2': -86.6, 'GP31': -113.2, 'GP33': -129.2}; CV range -257 to -11 ms)
baseline CVs timing reading ANOVA on n n Lips n Tongue n Larynx n Jaw n Back T n High T F p onset Lips onset Tongue onset Larynx onset Jaw onset Back T onset High T peak Back peak High lips,tongue < acoustic jaw,larynx later vowel peaks 250+-25 outcome lips,tongue < 0 outcome vs t = 0
pre all per subject / all six / threshold onset 175 31 5 18 34 38 49 22.5 3.008100e-17 -173 160 -13 -136 165 224 235 305 False False False not reproduced False not reproduced
pre all per subject / all six / threshold peak 175 31 5 18 34 38 49 27.2 4.093779e-20 -173 160 -13 -136 165 224 235 305 False False False not reproduced False not reproduced
pre all per subject / all six / half peak onset 175 31 5 18 34 38 49 27.3 3.599985e-20 -233 93 -91 -260 112 166 235 305 False False False not reproduced False not reproduced
pre all per subject / all six / half peak peak 175 31 5 18 34 38 49 27.2 4.093779e-20 -233 93 -91 -260 112 166 235 305 False False False not reproduced False not reproduced
pre all per subject / per articulator / threshold onset 224 33 43 39 31 38 40 10.6 3.852976e-09 -83 89 55 -134 234 211 297 278 False False False not reproduced False not reproduced
pre all per subject / per articulator / threshold peak 224 33 43 39 31 38 40 11.6 6.157008e-10 -83 89 55 -134 234 211 297 278 False False False not reproduced False not reproduced
pre all per subject / per articulator / half peak onset 224 33 43 39 31 38 40 13.9 9.129367e-12 -132 52 12 -270 183 147 297 278 False False False not reproduced False not reproduced
pre all per subject / per articulator / half peak peak 224 33 43 39 31 38 40 11.6 6.157008e-10 -132 52 12 -270 183 147 297 278 False False False not reproduced False not reproduced
pre all pooled / all six / threshold onset 176 25 3 18 38 35 57 24.7 1.222061e-18 -187 213 98 -125 169 206 245 282 False False False not reproduced False not reproduced
pre all pooled / all six / threshold peak 176 25 3 18 38 35 57 29.5 1.648552e-21 -187 213 98 -125 169 206 245 282 False False False not reproduced False not reproduced
pre all pooled / all six / half peak onset 176 25 3 18 38 35 57 27.8 1.627926e-20 -244 162 6 -248 109 132 245 282 False False False not reproduced False not reproduced
pre all pooled / all six / half peak peak 176 25 3 18 38 35 57 29.5 1.648552e-21 -244 162 6 -248 109 132 245 282 False False False not reproduced False not reproduced
pre all pooled / per articulator / threshold onset 197 33 41 31 31 28 33 11.9 5.274175e-10 -154 119 58 -117 179 195 262 284 False False False not reproduced False not reproduced
pre all pooled / per articulator / threshold peak 197 33 41 31 31 28 33 13.5 2.575740e-11 -154 119 58 -117 179 195 262 284 False False False not reproduced False not reproduced
pre all pooled / per articulator / half peak onset 197 33 41 31 31 28 33 15.1 1.830459e-12 -211 80 -5 -253 120 126 262 284 False False False not reproduced False not reproduced
pre all pooled / per articulator / half peak peak 197 33 41 31 31 28 33 13.5 2.575740e-11 -211 80 -5 -253 120 126 262 284 False False False not reproduced False not reproduced
pre 15+ per subject / all six / threshold onset 165 26 5 14 29 39 52 19.9 2.265976e-15 -154 148 -30 -138 165 219 246 306 False False False not reproduced False not reproduced
pre 15+ per subject / all six / threshold peak 165 26 5 14 29 39 52 24.5 3.178201e-18 -154 148 -30 -138 165 219 246 306 False False False not reproduced False not reproduced
pre 15+ per subject / all six / half peak onset 165 26 5 14 29 39 52 25.3 1.058468e-18 -244 93 -111 -259 76 143 246 306 False False False not reproduced False not reproduced
pre 15+ per subject / all six / half peak peak 165 26 5 14 29 39 52 24.5 3.178201e-18 -244 93 -111 -259 76 143 246 306 False False False not reproduced False not reproduced
pre 15+ per subject / per articulator / threshold onset 185 30 36 26 30 28 35 10.4 9.551120e-09 -54 163 47 -138 197 232 285 309 False False False not reproduced False not reproduced
pre 15+ per subject / per articulator / threshold peak 185 30 36 26 30 28 35 11.7 8.208530e-10 -54 163 47 -138 197 232 285 309 False False False not reproduced False not reproduced
pre 15+ per subject / per articulator / half peak onset 185 30 36 26 30 28 35 12.6 1.767613e-10 -137 125 -18 -260 89 144 285 309 False False False not reproduced False not reproduced
pre 15+ per subject / per articulator / half peak peak 185 30 36 26 30 28 35 11.7 8.208530e-10 -137 125 -18 -260 89 144 285 309 False False False not reproduced False not reproduced
pre 15+ pooled / all six / threshold onset 165 24 2 12 31 39 57 30.2 1.588162e-21 -191 220 38 -128 187 228 271 317 False False False not reproduced False not reproduced
pre 15+ pooled / all six / threshold peak 165 24 2 12 31 39 57 39.2 4.073747e-26 -191 220 38 -128 187 228 271 317 False False False not reproduced False not reproduced
pre 15+ pooled / all six / half peak onset 165 24 2 12 31 39 57 34.0 1.643266e-23 -267 192 -69 -253 102 139 271 317 False False False not reproduced False not reproduced
pre 15+ pooled / all six / half peak peak 165 24 2 12 31 39 57 39.2 4.073747e-26 -267 192 -69 -253 102 139 271 317 False False False not reproduced False not reproduced
pre 15+ pooled / per articulator / threshold onset 180 25 40 27 31 28 29 13.9 2.099851e-11 -171 139 58 -131 201 214 287 309 False False False not reproduced False not reproduced
pre 15+ pooled / per articulator / threshold peak 180 25 40 27 31 28 29 14.7 4.932962e-12 -171 139 58 -131 201 214 287 309 False False False not reproduced False not reproduced
pre 15+ pooled / per articulator / half peak onset 180 25 40 27 31 28 29 16.0 5.764995e-13 -239 100 -11 -253 92 123 287 309 False False False not reproduced False not reproduced
pre 15+ pooled / per articulator / half peak peak 180 25 40 27 31 28 29 14.7 4.932962e-12 -239 100 -11 -253 92 123 287 309 False False False not reproduced False not reproduced
whole all per subject / all six / threshold onset 179 31 5 16 35 43 49 22.4 2.642003e-17 -161 148 7 -136 177 223 242 304 False False False not reproduced False not reproduced
whole all per subject / all six / threshold peak 179 31 5 16 35 43 49 26.3 9.768148e-20 -161 148 7 -136 177 223 242 304 False False False not reproduced False not reproduced
whole all per subject / all six / half peak onset 179 31 5 16 35 43 49 28.0 9.734750e-21 -223 92 -71 -265 128 161 242 304 False False False not reproduced False not reproduced
whole all per subject / all six / half peak peak 179 31 5 16 35 43 49 26.3 9.768148e-20 -223 92 -71 -265 128 161 242 304 False False False not reproduced False not reproduced
whole all per subject / per articulator / threshold onset 226 32 41 40 29 42 42 9.9 1.500497e-08 -97 107 78 -131 196 217 255 281 False False False not reproduced False not reproduced
whole all per subject / per articulator / threshold peak 226 32 41 40 29 42 42 9.9 1.433881e-08 -97 107 78 -131 196 217 255 281 False False False not reproduced False not reproduced
whole all per subject / per articulator / half peak onset 226 32 41 40 29 42 42 12.8 6.842306e-11 -148 73 39 -261 148 154 255 281 False False False not reproduced False not reproduced
whole all per subject / per articulator / half peak peak 226 32 41 40 29 42 42 9.9 1.433881e-08 -148 73 39 -261 148 154 255 281 False False False not reproduced False not reproduced
whole all pooled / all six / threshold onset 173 26 5 15 38 34 55 22.5 3.260289e-17 -158 36 112 -126 190 197 270 276 False False False not reproduced False not reproduced
whole all pooled / all six / threshold peak 173 26 5 15 38 34 55 27.4 3.522387e-20 -158 36 112 -126 190 197 270 276 False False False not reproduced False not reproduced
whole all pooled / all six / half peak onset 173 26 5 15 38 34 55 26.1 1.874297e-19 -222 -26 25 -246 132 123 270 276 False False False not reproduced True not reproduced
whole all pooled / all six / half peak peak 173 26 5 15 38 34 55 27.4 3.522387e-20 -222 -26 25 -246 132 123 270 276 False False False not reproduced True not reproduced
whole all pooled / per articulator / threshold onset 199 34 42 33 30 28 32 11.5 9.489350e-10 -131 133 75 -116 196 193 279 286 False False False not reproduced False not reproduced
whole all pooled / per articulator / threshold peak 199 34 42 33 30 28 32 13.4 3.392311e-11 -131 133 75 -116 196 193 279 286 False False False not reproduced False not reproduced
whole all pooled / per articulator / half peak onset 199 34 42 33 30 28 32 14.9 2.319526e-12 -195 96 16 -243 129 130 279 286 False False False not reproduced False not reproduced
whole all pooled / per articulator / half peak peak 199 34 42 33 30 28 32 13.4 3.392311e-11 -195 96 16 -243 129 130 279 286 False False False not reproduced False not reproduced
whole 15+ per subject / all six / threshold onset 162 26 5 11 30 40 50 18.4 2.575473e-14 -153 147 -2 -136 164 210 243 302 False False False not reproduced False not reproduced
whole 15+ per subject / all six / threshold peak 162 26 5 11 30 40 50 22.8 3.886384e-17 -153 147 -2 -136 164 210 243 302 False False False not reproduced False not reproduced
whole 15+ per subject / all six / half peak onset 162 26 5 11 30 40 50 24.8 2.607336e-18 -247 92 -98 -265 77 137 243 302 False False False not reproduced False not reproduced
whole 15+ per subject / all six / half peak peak 162 26 5 11 30 40 50 22.8 3.886384e-17 -247 92 -98 -265 77 137 243 302 False False False not reproduced False not reproduced
whole 15+ per subject / per articulator / threshold onset 188 29 35 28 30 31 35 11.9 5.317144e-10 -76 176 65 -137 221 227 301 307 False False False not reproduced False not reproduced
whole 15+ per subject / per articulator / threshold peak 188 29 35 28 30 31 35 13.6 2.820393e-11 -76 176 65 -137 221 227 301 307 False False False not reproduced False not reproduced
whole 15+ per subject / per articulator / half peak onset 188 29 35 28 30 31 35 14.6 4.650912e-12 -159 141 -6 -259 112 142 301 307 False False False not reproduced False not reproduced
whole 15+ per subject / per articulator / half peak peak 188 29 35 28 30 31 35 13.6 2.820393e-11 -159 141 -6 -259 112 142 301 307 False False False not reproduced False not reproduced
whole 15+ pooled / all six / threshold onset 163 24 3 11 30 40 55 26.5 2.404551e-19 -190 168 23 -122 175 218 256 307 False False False not reproduced False not reproduced
whole 15+ pooled / all six / threshold peak 163 24 3 11 30 40 55 32.4 1.405508e-22 -190 168 23 -122 175 218 256 307 False False False not reproduced False not reproduced
whole 15+ pooled / all six / half peak onset 163 24 3 11 30 40 55 30.1 2.158521e-21 -270 88 -76 -252 89 127 256 307 False False False not reproduced False not reproduced
whole 15+ pooled / all six / half peak peak 163 24 3 11 30 40 55 32.4 1.405508e-22 -270 88 -76 -252 89 127 256 307 False False False not reproduced False not reproduced
whole 15+ pooled / per articulator / threshold onset 187 26 43 29 30 29 30 13.7 2.357204e-11 -159 148 72 -122 210 225 294 314 False False False not reproduced False not reproduced
whole 15+ pooled / per articulator / threshold peak 187 26 43 29 30 29 30 14.8 3.841425e-12 -159 148 72 -122 210 225 294 314 False False False not reproduced False not reproduced
whole 15+ pooled / per articulator / half peak onset 187 26 43 29 30 29 30 15.8 6.300431e-13 -228 105 6 -252 103 130 294 314 False False False not reproduced False not reproduced
whole 15+ pooled / per articulator / half peak peak 187 26 43 29 30 29 30 14.8 3.841425e-12 -228 105 6 -252 103 130 294 314 False False False not reproduced False not reproduced
rest all per subject / all six / threshold onset 168 26 6 14 37 35 50 21.7 1.268418e-16 -154 229 10 -129 136 214 203 296 False False False not reproduced False not reproduced
rest all per subject / all six / threshold peak 168 26 6 14 37 35 50 26.7 1.253267e-19 -154 229 10 -129 136 214 203 296 False False False not reproduced False not reproduced
rest all per subject / all six / half peak onset 168 26 6 14 37 35 50 27.3 5.526509e-20 -229 179 -68 -254 67 159 203 296 False False False not reproduced False not reproduced
rest all per subject / all six / half peak peak 168 26 6 14 37 35 50 26.7 1.253267e-19 -229 179 -68 -254 67 159 203 296 False False False not reproduced False not reproduced
rest all per subject / per articulator / threshold onset 202 30 38 33 31 34 36 10.8 3.166727e-09 -101 143 72 -138 168 231 227 302 False False False not reproduced False not reproduced
rest all per subject / per articulator / threshold peak 202 30 38 33 31 34 36 11.0 2.379807e-09 -101 143 72 -138 168 231 227 302 False False False not reproduced False not reproduced
rest all per subject / per articulator / half peak onset 202 30 38 33 31 34 36 14.0 1.106925e-11 -167 106 21 -262 102 167 227 302 False False False not reproduced False not reproduced
rest all per subject / per articulator / half peak peak 202 30 38 33 31 34 36 11.0 2.379807e-09 -167 106 21 -262 102 167 227 302 False False False not reproduced False not reproduced
rest all pooled / all six / threshold onset 170 22 6 15 42 30 55 22.6 3.404576e-17 -171 210 86 -129 142 196 224 278 False False False not reproduced False not reproduced
rest all pooled / all six / threshold peak 170 22 6 15 42 30 55 26.8 9.072746e-20 -171 210 86 -129 142 196 224 278 False False False not reproduced False not reproduced
rest all pooled / all six / half peak onset 170 22 6 15 42 30 55 26.7 1.146495e-19 -252 150 9 -238 79 127 224 278 False False False not reproduced False not reproduced
rest all pooled / all six / half peak peak 170 22 6 15 42 30 55 26.8 9.072746e-20 -252 150 9 -238 79 127 224 278 False False False not reproduced False not reproduced
rest all pooled / per articulator / threshold onset 192 29 43 32 35 23 30 9.4 5.225132e-08 -86 168 131 -129 121 168 215 276 False False False not reproduced False not reproduced
rest all pooled / per articulator / threshold peak 192 29 43 32 35 23 30 10.1 1.432819e-08 -86 168 131 -129 121 168 215 276 False False False not reproduced False not reproduced
rest all pooled / per articulator / half peak onset 192 29 43 32 35 23 30 13.6 2.598851e-11 -153 133 70 -249 56 116 215 276 False False False not reproduced False not reproduced
rest all pooled / per articulator / half peak peak 192 29 43 32 35 23 30 10.1 1.432819e-08 -153 133 70 -249 56 116 215 276 False False False not reproduced False not reproduced
rest 15+ per subject / all six / threshold onset 168 24 7 10 35 40 52 18.4 1.861795e-14 -126 290 -1 -138 145 198 235 300 False False False not reproduced False not reproduced
rest 15+ per subject / all six / threshold peak 168 24 7 10 35 40 52 22.9 2.456985e-17 -126 290 -1 -138 145 198 235 300 False False False not reproduced False not reproduced
rest 15+ per subject / all six / half peak onset 168 24 7 10 35 40 52 24.3 3.269460e-18 -218 235 -90 -256 53 124 235 300 False False False not reproduced False not reproduced
rest 15+ per subject / all six / half peak peak 168 24 7 10 35 40 52 22.9 2.456985e-17 -218 235 -90 -256 53 124 235 300 False False False not reproduced False not reproduced
rest 15+ per subject / per articulator / threshold onset 193 27 43 26 34 28 35 12.4 2.117814e-10 -106 185 94 -138 176 208 272 311 False False False not reproduced False not reproduced
rest 15+ per subject / per articulator / threshold peak 193 27 43 26 34 28 35 13.9 1.519627e-11 -106 185 94 -138 176 208 272 311 False False False not reproduced False not reproduced
rest 15+ per subject / per articulator / half peak onset 193 27 43 26 34 28 35 15.2 1.633913e-12 -180 145 39 -252 69 129 272 311 False False False not reproduced False not reproduced
rest 15+ per subject / per articulator / half peak peak 193 27 43 26 34 28 35 13.9 1.519627e-11 -180 145 39 -252 69 129 272 311 False False False not reproduced False not reproduced
rest 15+ pooled / all six / threshold onset 163 22 5 11 34 39 52 23.0 2.631448e-17 -155 213 49 -142 174 195 268 301 False False False not reproduced False not reproduced
rest 15+ pooled / all six / threshold peak 163 22 5 11 34 39 52 29.8 3.393746e-21 -155 213 49 -142 174 195 268 301 False False False not reproduced False not reproduced
rest 15+ pooled / all six / half peak onset 163 22 5 11 34 39 52 26.1 3.821200e-19 -247 148 -39 -247 86 103 268 301 False False False not reproduced False not reproduced
rest 15+ pooled / all six / half peak peak 163 22 5 11 34 39 52 29.8 3.393746e-21 -247 148 -39 -247 86 103 268 301 False False False not reproduced False not reproduced
rest 15+ pooled / per articulator / threshold onset 187 26 45 28 33 26 29 16.0 4.700755e-13 -133 242 114 -140 166 212 263 325 False False False not reproduced False not reproduced
rest 15+ pooled / per articulator / threshold peak 187 26 45 28 33 26 29 16.6 1.751393e-13 -133 242 114 -140 166 212 263 325 False False False not reproduced False not reproduced
rest 15+ pooled / per articulator / half peak onset 187 26 45 28 33 26 29 19.5 1.594970e-15 -213 204 32 -249 52 99 263 325 False False False not reproduced False not reproduced
rest 15+ pooled / per articulator / half peak peak 187 26 45 28 33 26 29 16.6 1.751393e-13 -213 204 32 -249 52 99 263 325 False False False not reproduced False not reproduced
Outcome over readings (substitute acoustic onset): not reproduced
Beside it, against t = 0: not reproduced | readings with lips and tongue onsets before 0 ms: 2 of 96 | before the substitute onset: 0
Ranges over readings: mean onset Lips -270 to -54 | Tongue -26 to 290 | Jaw -270 to -116 | Larynx -111 to 131 | peak Back 203 to 301 | peak High 276 to 325
ANOVA F 9.4 to 39.2 | readings with p < 1e-10: 76 of 96 | conditions holding per reading: {0: 96}
Total electrodes over articulators and subjects: from 162 to 226 (paper 211 +- 21)
Electrodes per articulator over readings (min to max): {'Lips': (22, 34), 'Tongue': (2, 45), 'Larynx': (10, 40), 'Jaw': (29, 42), 'Back Tongue': (23, 43), 'High Tongue': (29, 57)} | Tongue under the 'all six' readings: (2, 7)
Per subject, first reading (pre, all CVs): {'EC2': {'Back Tongue': 11.0, 'High Tongue': 22.0, 'Jaw': 13.0, 'Larynx': 3.0, 'Lips': 3.0, 'Tongue': 1.0}, 'GP31': {'Back Tongue': 14.0, 'High Tongue': 12.0, 'Jaw': 6.0, 'Larynx': 7.0, 'Lips': 23.0, 'Tongue': nan}, 'GP33': {'Back Tongue': 13.0, 'High Tongue': 15.0, 'Jaw': 15.0, 'Larynx': 8.0, 'Lips': 5.0, 'Tongue': 4.0}}
EC2 mean onsets (ms): {'Back Tongue': 59.0, 'High Tongue': 166.0, 'Jaw': -130.0, 'Larynx': 172.0, 'Lips': -122.0, 'Tongue': -225.0}
GP31 mean onsets (ms): {'Back Tongue': 312.0, 'High Tongue': 448.0, 'Jaw': -178.0, 'Larynx': 0.0, 'Lips': -229.0}
GP33 mean onsets (ms): {'Back Tongue': 96.0, 'High Tongue': 129.0, 'Jaw': -125.0, 'Larynx': -94.0, 'Lips': 50.0, 'Tongue': 256.0}
Variant, artifact trials left out: outcome over readings not reproduced | against t = 0 not reproduced | conditions holding per reading {0: 92, 1: 4} | mean onset Lips -338 to -125 | Tongue -6 to 289 | Jaw -291 to -96 | peak High 268 to 303 | electrodes 156 to 227
Result: not reproduced. Under every one of the 96 readings none of the three order conditions holds, against the substitute acoustic onset (-110 ms) and against t = 0 alike. The lips correlations begin before the substitute onset (mean lips onset -270 to -54 ms), but the tongue correlations begin at -26 to +290 ms, so lips and tongue both precede the substitute onset in no reading and precede t = 0 in 2 of 96; the jaw begins as early as the lips (-270 to -116 ms), not later; and the high-tongue peak is at 276 to 325 ms, outside 250 +- 25 ms. The ANOVA across the six articulators has F = 9.4 to 39.2, with p below 1e-10 in 76 of the 96 readings. Electrode counts are 162 to 226 (paper 211). Tongue electrodes number 2 to 7 under the readings with one threshold for all six articulators (2 to 45 over all readings), so the tongue onsets of those readings rest on few electrodes. Leaving out the trials that overlap intervals marked as ECoG artifacts (a variant, not a reading) gives the same outcome: one condition holds in 4 of 96 readings and none in the others, with mean lips onsets of -338 to -125 ms and tongue onsets of -6 to 289 ms.
pca-nine-components¶
The cortical state-space has nine spatial PCs in every subject, the eigenvalues falling fast up to the ninth and slowly after, together explaining about 60% of the variance.
In [19]:
# The variance explained does not depend on the input projected (readings i and ii).
pc = []
for base in BASES:
for z in ZMODES:
for cs in CVSETS:
for w in [(-500, 750), (-375, 620)]:
fr = {s: STATE[s, base, z, cs, w][1] for s in B.SUBJECTS}
pc.append({"baseline": base, "z-scoring": z, "CVs": cs, "window": f"{w[0]} to {w[1]}",
**{s: round(fr[s][:9].sum(), 3) for s in B.SUBJECTS},
"mean": round(np.mean([fr[s][:9].sum() for s in B.SUBJECTS]), 3)})
pc = pd.DataFrame(pc)
pc["within 0.54-0.66"] = pc["mean"].between(0.54, 0.66)
print(pc.to_string(index=False))
pc["outcome"] = np.where(pc["within 0.54-0.66"], "reproduced", "not reproduced")
print("\nOutcome over readings:", overall(pc.outcome))
by_factor(pc, ["baseline", "z-scoring", "CVs", "window"])
fig, axes = plt.subplots(1, 2, figsize=(10, 3))
for ax, z in zip(axes, ZMODES):
for s in B.SUBJECTS:
ax.plot(np.arange(1, 21), 100 * STATE[s, "pre", z, "all", EP][1][:20], "o-", ms=3, label=s)
ax.axvline(9, color="grey", ls="--"); ax.set_title(f"pre baseline, {z} z-scoring, all CVs"); ax.set_xlabel("PC"); ax.set_ylabel("% variance")
axes[0].legend(); plt.show()
print("% variance of PCs 1 to 12, pre baseline, all CVs, epoch (all values plotted are within 0 to 40%):")
for z in ZMODES:
for s in B.SUBJECTS:
print(z, s, np.round(100 * STATE[s, "pre", z, "all", EP][1][:12], 1))
baseline z-scoring CVs window EC2 GP31 GP33 mean within 0.54-0.66
pre row all -500 to 750 0.607 0.734 0.536 0.625 True
pre row all -375 to 620 0.624 0.743 0.549 0.639 True
pre row 15+ -500 to 750 0.605 0.665 0.536 0.602 True
pre row 15+ -375 to 620 0.621 0.675 0.549 0.615 True
pre time all -500 to 750 0.440 0.677 0.468 0.528 False
pre time all -375 to 620 0.458 0.681 0.475 0.538 False
pre time 15+ -500 to 750 0.438 0.562 0.468 0.490 False
pre time 15+ -375 to 620 0.454 0.565 0.475 0.498 False
whole row all -500 to 750 0.608 0.743 0.539 0.630 True
whole row all -375 to 620 0.625 0.752 0.552 0.643 True
whole row 15+ -500 to 750 0.606 0.651 0.539 0.598 True
whole row 15+ -375 to 620 0.621 0.659 0.552 0.611 True
whole time all -500 to 750 0.440 0.691 0.473 0.535 False
whole time all -375 to 620 0.459 0.695 0.481 0.545 True
whole time 15+ -500 to 750 0.438 0.560 0.473 0.490 False
whole time 15+ -375 to 620 0.454 0.562 0.481 0.499 False
rest row all -500 to 750 0.608 0.790 0.552 0.650 True
rest row all -375 to 620 0.624 0.792 0.566 0.661 False
rest row 15+ -500 to 750 0.603 0.786 0.552 0.647 True
rest row 15+ -375 to 620 0.619 0.787 0.566 0.657 True
rest time all -500 to 750 0.440 0.738 0.489 0.556 True
rest time all -375 to 620 0.459 0.735 0.497 0.564 True
rest time 15+ -500 to 750 0.436 0.705 0.489 0.543 True
rest time 15+ -375 to 620 0.453 0.700 0.497 0.550 True
Outcome over readings: differ by reading: not reproduced 8, reproduced 16
Outcome counts by reading factor (24 readings):
baseline: {'pre': {'reproduced': 4, 'not reproduced': 4}, 'whole': {'reproduced': 5, 'not reproduced': 3}, 'rest': {'reproduced': 7, 'not reproduced': 1}}
z-scoring: {'row': {'reproduced': 11, 'not reproduced': 1}, 'time': {'not reproduced': 7, 'reproduced': 5}}
CVs: {'all': {'reproduced': 8, 'not reproduced': 4}, '15+': {'reproduced': 8, 'not reproduced': 4}}
window: {'-500 to 750': {'reproduced': 8, 'not reproduced': 4}, '-375 to 620': {'reproduced': 8, 'not reproduced': 4}}
% variance of PCs 1 to 12, pre baseline, all CVs, epoch (all values plotted are within 0 to 40%): row EC2 [27.3 9.2 5.6 5.2 4.1 3.1 2.6 1.8 1.7 1.6 1.5 1.4] row GP31 [38.8 8.9 6.7 4.8 3.8 3.6 2.5 2.3 2. 1.8 1.5 1.4] row GP33 [22.4 8.1 5.2 4.8 4. 2.8 2.2 2.2 1.9 1.8 1.7 1.5] time EC2 [13.6 7. 5.1 4.3 3.6 3.4 2.5 2.4 2.1 1.9 1.8 1.6] time GP31 [36.9 8.2 4.9 4.1 3.4 2.8 2.7 2.4 2.2 1.7 1.6 1.5] time GP33 [16.7 7.5 5.7 4.3 3.4 2.6 2.4 2.1 2. 1.9 1.7 1.5]
Result: blocked (methods_underspecified). With row z-scoring, the first nine PCs explain a
mean of 0.60 to 0.66 over subjects, within 0.54 to 0.66 in 11 of 12 readings (rest, every CV and
the -375 to 620 ms window gives 0.661). With time z-scoring, the mean is 0.49 to 0.56, within the
tolerance under the rest baseline (4 readings) and once under whole, and below it in the other
7 readings. The readings give different outcomes (16 within, 8 not). The
z-scoring decides most of the split (row 11 of 12 within, time 5 of 12); the CV set and the
window do not change the count (8 of 12 within each). The variance explained does
not depend on the input projected. The spectra show no sharp change at the ninth eigenvalue in any
subject (printed values); that part is not judged.
fig4a-consonant-clusters¶
At the consonant time (-25 ms) k-means clustering of the cortical state-space groups syllables by major oral articulator, and three clusters are more parsimonious than two or four (mean silhouette 0.47 against 0.45 and 0.43).
In [20]:
tc, tv = B.ms_index(-25), B.ms_index(250)
sil = []
for base, z, pj, cs in SS_READINGS:
for s in B.SUBJECTS:
K = KS[s, base, z, pj, cs][:, :, tc]
row = {"baseline": base, "z-scoring": z, "projection": pj, "CVs": cs, "subject": s}
for k in (2, 3, 4):
row[f"k={k}"] = silhouette_score(K, B.kmeans(K, k, seed=40 + k))
sil.append(row)
sil = pd.DataFrame(sil)
print(sil.round(3).to_string(index=False))
means = sil.groupby(["baseline", "z-scoring", "projection", "CVs"])[["k=2", "k=3", "k=4"]].mean()
means["k=3 highest"] = means.idxmax(axis=1) == "k=3"
means["all within 10%"] = (abs(means["k=2"] - 0.45) <= 0.045) & (abs(means["k=3"] - 0.47) <= 0.047) & (abs(means["k=4"] - 0.43) <= 0.043)
print("\nMean over subjects (paper 0.45, 0.47, 0.43):")
print(means.round(3).to_string())
means["outcome"] = ["reproduced" if h and w else "partially" if h else "not reproduced"
for h, w in zip(means["k=3 highest"], means["all within 10%"])]
print("Outcome over readings:", overall(means.outcome))
by_factor(means.reset_index(), ["baseline", "z-scoring", "projection", "CVs"])
print("Reading with the highest mean silhouette, by k, over the 24 readings:", means[["k=2", "k=3", "k=4"]].idxmax(axis=1).value_counts().to_dict())
baseline z-scoring projection CVs subject k=2 k=3 k=4
pre row i all EC2 0.259 0.280 0.270
pre row i all GP31 0.782 0.410 0.252
pre row i all GP33 0.158 0.153 0.171
pre row i 15+ EC2 0.263 0.274 0.273
pre row i 15+ GP31 0.370 0.239 0.227
pre row i 15+ GP33 0.158 0.153 0.171
pre row ii all EC2 0.363 0.345 0.352
pre row ii all GP31 0.516 0.497 0.312
pre row ii all GP33 0.232 0.231 0.249
pre row ii 15+ EC2 0.362 0.362 0.369
pre row ii 15+ GP31 0.507 0.311 0.330
pre row ii 15+ GP33 0.232 0.231 0.249
pre time i all EC2 0.255 0.233 0.234
pre time i all GP31 0.759 0.326 0.223
pre time i all GP33 0.137 0.138 0.147
pre time i 15+ EC2 0.195 0.232 0.247
pre time i 15+ GP31 0.303 0.209 0.210
pre time i 15+ GP33 0.137 0.138 0.147
pre time ii all EC2 0.350 0.328 0.360
pre time ii all GP31 0.512 0.499 0.310
pre time ii all GP33 0.231 0.231 0.248
pre time ii 15+ EC2 0.349 0.360 0.362
pre time ii 15+ GP31 0.488 0.296 0.324
pre time ii 15+ GP33 0.231 0.231 0.248
whole row i all EC2 0.256 0.278 0.267
whole row i all GP31 0.801 0.377 0.249
whole row i all GP33 0.158 0.160 0.167
whole row i 15+ EC2 0.261 0.272 0.272
whole row i 15+ GP31 0.329 0.228 0.219
whole row i 15+ GP33 0.158 0.160 0.167
whole row ii all EC2 0.358 0.337 0.347
whole row ii all GP31 0.793 0.439 0.306
whole row ii all GP33 0.219 0.224 0.242
whole row ii 15+ EC2 0.358 0.360 0.365
whole row ii 15+ GP31 0.435 0.301 0.311
whole row ii 15+ GP33 0.219 0.224 0.242
whole time i all EC2 0.253 0.233 0.230
whole time i all GP31 0.776 0.307 0.230
whole time i all GP33 0.135 0.136 0.163
whole time i 15+ EC2 0.190 0.228 0.243
whole time i 15+ GP31 0.244 0.201 0.201
whole time i 15+ GP33 0.135 0.136 0.163
whole time ii all EC2 0.345 0.328 0.353
whole time ii all GP31 0.783 0.433 0.302
whole time ii all GP33 0.218 0.231 0.240
whole time ii 15+ EC2 0.345 0.356 0.357
whole time ii 15+ GP31 0.430 0.295 0.301
whole time ii 15+ GP33 0.218 0.231 0.240
rest row i all EC2 0.261 0.277 0.261
rest row i all GP31 0.350 0.353 0.359
rest row i all GP33 0.161 0.159 0.175
rest row i 15+ EC2 0.260 0.269 0.264
rest row i 15+ GP31 0.333 0.343 0.381
rest row i 15+ GP33 0.161 0.159 0.175
rest row ii all EC2 0.386 0.344 0.347
rest row ii all GP31 0.659 0.411 0.408
rest row ii all GP33 0.221 0.233 0.249
rest row ii 15+ EC2 0.384 0.345 0.347
rest row ii 15+ GP31 0.675 0.412 0.431
rest row ii 15+ GP33 0.221 0.233 0.249
rest time i all EC2 0.251 0.225 0.228
rest time i all GP31 0.605 0.618 0.325
rest time i all GP33 0.130 0.157 0.168
rest time i 15+ EC2 0.187 0.231 0.241
rest time i 15+ GP31 0.596 0.292 0.294
rest time i 15+ GP33 0.130 0.157 0.168
rest time ii all EC2 0.361 0.316 0.335
rest time ii all GP31 0.654 0.402 0.406
rest time ii all GP33 0.228 0.235 0.248
rest time ii 15+ EC2 0.358 0.318 0.335
rest time ii 15+ GP31 0.671 0.409 0.426
rest time ii 15+ GP33 0.228 0.235 0.248
Mean over subjects (paper 0.45, 0.47, 0.43):
k=2 k=3 k=4 k=3 highest all within 10%
baseline z-scoring projection CVs
pre row i 15+ 0.264 0.222 0.223 False False
all 0.400 0.281 0.231 False False
ii 15+ 0.367 0.301 0.316 False False
all 0.370 0.358 0.304 False False
time i 15+ 0.211 0.193 0.202 False False
all 0.384 0.232 0.201 False False
ii 15+ 0.356 0.296 0.311 False False
all 0.364 0.353 0.306 False False
rest row i 15+ 0.251 0.257 0.273 False False
all 0.257 0.263 0.265 False False
ii 15+ 0.427 0.330 0.343 False False
all 0.422 0.329 0.335 False False
time i 15+ 0.304 0.227 0.234 False False
all 0.329 0.333 0.241 True False
ii 15+ 0.419 0.321 0.336 False False
all 0.414 0.318 0.330 False False
whole row i 15+ 0.249 0.220 0.219 False False
all 0.405 0.272 0.228 False False
ii 15+ 0.337 0.295 0.306 False False
all 0.457 0.333 0.298 False False
time i 15+ 0.190 0.188 0.202 False False
all 0.388 0.225 0.208 False False
ii 15+ 0.331 0.294 0.299 False False
all 0.449 0.331 0.298 False False
Outcome over readings: differ by reading: not reproduced 23, partially 1
Outcome counts by reading factor (24 readings):
baseline: {'pre': {'not reproduced': 8}, 'rest': {'not reproduced': 7, 'partially': 1}, 'whole': {'not reproduced': 8}}
z-scoring: {'row': {'not reproduced': 12}, 'time': {'not reproduced': 11, 'partially': 1}}
projection: {'i': {'not reproduced': 11, 'partially': 1}, 'ii': {'not reproduced': 12}}
CVs: {'15+': {'not reproduced': 12}, 'all': {'not reproduced': 11, 'partially': 1}}
Reading with the highest mean silhouette, by k, over the 24 readings: {'k=2': 20, 'k=4': 3, 'k=3': 1}
Result: blocked (methods_underspecified). In 23 of the 24 readings three clusters do not
have the highest mean silhouette (not reproduced). In one, rest baseline, time z-scoring,
projection (i) and every CV, k = 3 is highest by a small margin (0.333 against 0.329 for k = 2),
with values outside tolerance (partially reproduced), so the readings differ. Over the 24
readings, k = 2 has the highest mean silhouette in 20, k = 4 in 3 and k = 3 in 1. The silhouettes are
lower than the paper's under every reading (k = 3: 0.19 to 0.36 against 0.47). GP31's k = 2
silhouette is 0.35 to 0.80 with every CV, where its /thi/ has two renditions, and 0.24 to 0.68 with
only CVs of 15 or more renditions.
suppfig11-co-clustering¶
Clusters of the cortical state-space match the major articulators at the consonant time (median co-clustering 0.50 against 0.32 by chance) and the vowels at the vowel time (0.65 against 0.37), across subjects.
In [21]:
rng = np.random.default_rng(11)
cc = []
for base, z, pj, cs in SS_READINGS:
obs = {"cons a": [], "cons b": [], "vowel": []}
null = {"cons a": [], "cons b": [], "vowel": []}
for s in B.SUBJECTS:
K = KS[s, base, z, pj, cs]
nc_c = B.kmeans(K[:, :, tc], 3, seed=21)
nc_v = B.kmeans(K[:, :, tv], 3, seed=22)
for name, nc, labs in (("cons a", nc_c, [B.MAJOR["a"].get(c) for c, v in CVS[s, cs]]),
("cons b", nc_c, [B.MAJOR["b"].get(c) for c, v in CVS[s, cs]]),
("vowel", nc_v, [v for c, v in CVS[s, cs]])):
keep = np.array([l is not None for l in labs])
ac = np.array([l for l in labs if l is not None])
ncs = nc[keep]
obs[name].append(B.co_clustering(ncs, ac))
null[name].append(np.mean([B.co_clustering(ncs, rng.permutation(ac)) for _ in range(1000)], axis=0))
row = {"baseline": base, "z-scoring": z, "projection": pj, "CVs": cs}
test_ok = {}
for name in obs:
o, n = np.concatenate(obs[name]), np.concatenate(null[name])
p = stats.wilcoxon(o, n).pvalue
row[f"{name}: n, obs, null, p"] = f"{len(o)}, {np.median(o):.3f}, {np.median(n):.3f}, {p:.1e}"
test_ok[name] = bool(p < 1e-10 and np.median(o - n) > 0)
for cons in ("cons a", "cons b"):
ok_tests = test_ok[cons] and test_ok["vowel"] # each label reading with its own consonant test
mo = {k: np.median(np.concatenate(obs[k])) for k in (cons, "vowel")}
mn = {k: np.median(np.concatenate(null[k])) for k in (cons, "vowel")}
within = abs(mo[cons] - 0.50) <= 0.05 and abs(mn[cons] - 0.32) <= 0.032 and abs(mo["vowel"] - 0.65) <= 0.065 and abs(mn["vowel"] - 0.37) <= 0.037
row[f"outcome ({cons[-1]})"] = "reproduced" if ok_tests and within else ("partially" if ok_tests else "not reproduced")
cc.append(row)
cc = pd.DataFrame(cc)
print(cc.to_string(index=False))
print("Expected: consonants 0.50 +- 0.05 against 0.32 +- 0.032; vowels 0.65 +- 0.065 against 0.37 +- 0.037; p < 1e-10")
print("Outcome over readings:", overall(list(cc["outcome (a)"]) + list(cc["outcome (b)"])))
cc_long = pd.concat([cc.assign(labels=lr, outcome=cc[f"outcome ({lr})"]) for lr in ("a", "b")])
by_factor(cc_long, ["baseline", "z-scoring", "projection", "CVs", "labels"])
for name in ("cons a", "cons b", "vowel"):
col = cc[f"{name}: n, obs, null, p"].str.split(", ", expand=True).astype(float)
print(f"{name}: observed median {col[1].min():.3f} to {col[1].max():.3f}, null {col[2].min():.3f} to {col[2].max():.3f}, p {col[3].min():.0e} to {col[3].max():.0e}")
baseline z-scoring projection CVs cons a: n, obs, null, p cons b: n, obs, null, p vowel: n, obs, null, p outcome (a) outcome (b)
pre row i all 133, 0.750, 0.395, 2.2e-16 160, 0.667, 0.385, 1.0e-17 169, 0.486, 0.386, 3.8e-11 partially partially
pre row i 15+ 131, 0.750, 0.338, 2.5e-19 158, 0.711, 0.336, 2.6e-20 167, 0.444, 0.383, 4.7e-08 not reproduced not reproduced
pre row ii all 133, 0.878, 0.388, 3.8e-19 160, 0.767, 0.346, 2.2e-20 169, 0.578, 0.349, 5.5e-18 partially partially
pre row ii 15+ 131, 0.880, 0.344, 6.1e-21 158, 0.784, 0.350, 4.5e-22 167, 0.615, 0.351, 5.4e-19 partially partially
pre time i all 133, 0.720, 0.355, 5.3e-16 160, 0.645, 0.349, 7.7e-16 169, 0.512, 0.355, 2.8e-16 partially partially
pre time i 15+ 131, 0.706, 0.347, 1.8e-18 158, 0.595, 0.348, 1.1e-18 167, 0.583, 0.348, 1.7e-18 partially partially
pre time ii all 133, 0.878, 0.387, 3.7e-19 160, 0.754, 0.342, 2.5e-20 169, 0.698, 0.348, 2.1e-16 partially partially
pre time ii 15+ 131, 0.880, 0.347, 1.2e-20 158, 0.784, 0.356, 8.5e-22 167, 0.698, 0.343, 2.4e-20 partially partially
whole row i all 133, 0.783, 0.392, 1.5e-17 160, 0.667, 0.385, 1.7e-19 169, 0.493, 0.389, 1.6e-11 partially partially
whole row i 15+ 131, 0.750, 0.347, 2.8e-19 158, 0.651, 0.336, 2.8e-20 167, 0.409, 0.387, 2.9e-08 not reproduced not reproduced
whole row ii all 133, 0.889, 0.394, 7.3e-20 160, 0.784, 0.350, 8.3e-22 169, 0.583, 0.352, 9.8e-18 partially partially
whole row ii 15+ 131, 0.880, 0.342, 2.0e-20 158, 0.784, 0.347, 8.6e-22 167, 0.667, 0.359, 2.0e-19 partially partially
whole time i all 133, 0.720, 0.348, 4.9e-17 160, 0.645, 0.367, 4.5e-18 169, 0.512, 0.357, 2.8e-16 partially partially
whole time i 15+ 131, 0.720, 0.339, 3.5e-19 158, 0.636, 0.346, 2.5e-19 167, 0.650, 0.356, 1.4e-19 partially partially
whole time ii all 133, 0.878, 0.381, 7.4e-20 160, 0.786, 0.342, 5.0e-22 169, 0.565, 0.348, 2.1e-17 partially partially
whole time ii 15+ 131, 0.880, 0.342, 2.0e-20 158, 0.800, 0.357, 1.8e-21 167, 0.650, 0.346, 6.8e-19 partially partially
rest row i all 133, 0.744, 0.393, 5.3e-18 160, 0.667, 0.389, 7.5e-19 169, 0.493, 0.387, 8.0e-09 not reproduced not reproduced
rest row i 15+ 131, 0.750, 0.360, 3.3e-18 158, 0.706, 0.349, 3.2e-20 167, 0.492, 0.386, 4.7e-07 not reproduced not reproduced
rest row ii all 133, 0.897, 0.412, 8.5e-21 160, 0.800, 0.363, 3.7e-22 169, 0.558, 0.363, 1.9e-17 partially partially
rest row ii 15+ 131, 0.909, 0.411, 1.2e-20 158, 0.800, 0.347, 9.3e-23 167, 0.558, 0.361, 1.0e-17 partially partially
rest time i all 133, 0.633, 0.391, 3.3e-14 160, 0.611, 0.378, 2.7e-13 169, 0.484, 0.343, 3.9e-11 partially partially
rest time i 15+ 131, 0.732, 0.384, 1.4e-18 158, 0.706, 0.370, 1.7e-19 167, 0.492, 0.430, 3.1e-10 not reproduced not reproduced
rest time ii all 133, 0.897, 0.406, 2.6e-20 160, 0.778, 0.355, 1.4e-22 169, 0.585, 0.362, 1.2e-17 partially partially
rest time ii 15+ 131, 0.909, 0.395, 3.5e-20 158, 0.792, 0.345, 1.6e-22 167, 0.585, 0.360, 2.1e-17 partially partially
Expected: consonants 0.50 +- 0.05 against 0.32 +- 0.032; vowels 0.65 +- 0.065 against 0.37 +- 0.037; p < 1e-10
Outcome over readings: differ by reading: not reproduced 10, partially 38
Outcome counts by reading factor (48 readings):
baseline: {'pre': {'partially': 14, 'not reproduced': 2}, 'whole': {'partially': 14, 'not reproduced': 2}, 'rest': {'partially': 10, 'not reproduced': 6}}
z-scoring: {'row': {'partially': 16, 'not reproduced': 8}, 'time': {'partially': 22, 'not reproduced': 2}}
projection: {'i': {'partially': 14, 'not reproduced': 10}, 'ii': {'partially': 24}}
CVs: {'all': {'partially': 22, 'not reproduced': 2}, '15+': {'partially': 16, 'not reproduced': 8}}
labels: {'a': {'partially': 19, 'not reproduced': 5}, 'b': {'partially': 19, 'not reproduced': 5}}
cons a: observed median 0.633 to 0.909, null 0.338 to 0.412, p 6e-21 to 3e-14
cons b: observed median 0.595 to 0.800, null 0.336 to 0.389, p 9e-23 to 3e-13
vowel: observed median 0.409 to 0.698, null 0.343 to 0.430, p 2e-20 to 5e-07
Result: blocked (methods_underspecified). The state-space clusters match the major
articulators more than chance under every reading (consonant p from 9e-23 to 3e-13). The observed
consonant medians (0.633 to 0.909 with labels (a), 0.595 to 0.800 with labels (b)) are above the
paper's 0.50 +- 0.05, and the vowel medians (0.41 to 0.70) are mostly below its 0.65 +- 0.065, so
38 of the 48 readings (24 readings by two label readings) give partially reproduced. In five
readings, all with projection (i) (row with 15+ CVs under each baseline, rest with row and
every CV, and rest with time and 15+ CVs), the vowel test misses the p < 1e-10 bound (p from
3e-10 to 5e-07) and gives not reproduced under both label readings. Projection (ii) gives partially
reproduced in all 24 of its readings. Each label reading is now judged with its own consonant test
and the vowel test; the outcomes are the same as before. The nulls (0.336 to 0.412 for consonants,
0.343 to 0.430 for vowels) are near the paper's 0.32 and 0.37.
fig4c-d-hierarchy¶
Hierarchical clustering of the state-space puts the major oral articulator at the top tier for consonants; for vowels the first split is lip rounding (/u/ against /a/ and /i/).
In [22]:
hc = []
for base, z, pj, cs in SS_READINGS:
for s in B.SUBJECTS:
K = KS[s, base, z, pj, cs]
cons_cut = fcluster(linkage(K[:, :, tc], "ward"), 3, "maxclust")
vow_cut = fcluster(linkage(K[:, :, tv], "ward"), 2, "maxclust")
row = {"baseline": base, "z-scoring": z, "projection": pj, "CVs": cs, "subject": s}
for lr in ("a", "b"):
labs = np.array([B.MAJOR[lr].get(c, "") for c, v in CVS[s, cs]])
lab = labs != ""
maj = {k: pd.Series(labs[lab & (cons_cut == k)]).mode().tolist() for k in (1, 2, 3)}
distinct = all(len(v) == 1 for v in maj.values()) and len({v[0] for v in maj.values()}) == 3
home = {a: pd.Series(cons_cut[lab & (labs == a)]).mode()[0] for a in ("labial", "coronal", "dorsal")}
purity = np.mean([cons_cut[i] == home[labs[i]] for i in np.where(lab)[0]])
row[f"cons ({lr}) majorities"] = "/".join(v[0] if len(v) == 1 else "tie" for v in maj.values())
row[f"cons ({lr}) in home"] = round(purity, 2)
row[f"cons ({lr}) holds"] = bool(distinct and purity >= 2 / 3)
row[f"cons ({lr}) co-clust"] = round(np.median(B.co_clustering(cons_cut[lab], labs[lab])), 2)
vow = np.array([v for c, v in CVS[s, cs]])
ok = any((vow_cut[vow == "u"] == ku).mean() > 0.5 and (vow_cut[vow != "u"] == 3 - ku).mean() > 0.5 for ku in (1, 2))
row["a/i/u in cluster 1"] = "/".join(f"{(vow_cut[vow == v] == 1).mean():.2f}" for v in "aiu")
row["vowel holds"] = bool(ok)
hc.append(row)
hc = pd.DataFrame(hc)
print(hc.to_string(index=False))
oc = []
for (base, z, pj, cs), g in hc.groupby(["baseline", "z-scoring", "projection", "CVs"]):
for lr in ("a", "b"):
both = g[f"cons ({lr}) holds"] & g["vowel holds"]
oc.append({"baseline": base, "z-scoring": z, "projection": pj, "CVs": cs, "labels": lr, "subjects with both": int(both.sum()),
"outcome": "reproduced" if both.all() else ("partially" if both.any() else "not reproduced")})
oc = pd.DataFrame(oc)
print(oc.to_string(index=False))
print("Outcome over readings:", overall(oc.outcome))
by_factor(oc, ["baseline", "z-scoring", "projection", "CVs", "labels"])
baseline z-scoring projection CVs subject cons (a) majorities cons (a) in home cons (a) holds cons (a) co-clust cons (b) majorities cons (b) in home cons (b) holds cons (b) co-clust a/i/u in cluster 1 vowel holds
pre row i all EC2 coronal/labial/dorsal 0.98 True 0.98 coronal/labial/dorsal 0.93 True 0.92 0.74/0.58/0.11 True
pre row i all GP31 labial/coronal/coronal 0.91 False 0.80 labial/coronal/coronal 0.90 False 0.77 0.94/0.95/1.00 False
pre row i all GP33 coronal/coronal/labial 0.64 False 0.70 coronal/coronal/labial 0.63 False 0.67 0.53/0.84/0.26 True
pre row i 15+ EC2 labial/coronal/coronal 0.80 False 0.63 labial/coronal/coronal 0.81 False 0.61 0.58/0.56/0.11 True
pre row i 15+ GP31 labial/coronal/coronal 0.71 False 0.70 labial/coronal/coronal 0.73 False 0.71 0.94/1.00/1.00 False
pre row i 15+ GP33 coronal/coronal/labial 0.64 False 0.70 coronal/coronal/labial 0.63 False 0.67 0.53/0.84/0.26 True
pre row ii all EC2 coronal/dorsal/labial 1.00 True 1.00 coronal/dorsal/labial 0.94 True 0.94 0.00/0.00/0.84 True
pre row ii all GP31 coronal/labial/coronal 0.91 False 0.80 coronal/labial/coronal 0.92 False 0.78 1.00/0.95/1.00 False
pre row ii all GP33 coronal/labial/labial 0.76 False 0.77 coronal/labial/labial 0.72 False 0.73 0.11/0.95/0.05 True
pre row ii 15+ EC2 coronal/dorsal/labial 1.00 True 1.00 coronal/dorsal/labial 0.94 True 0.94 0.00/0.00/0.84 True
pre row ii 15+ GP31 labial/coronal/coronal 0.76 False 0.74 labial/dorsal/coronal 0.75 True 0.72 0.94/1.00/1.00 False
pre row ii 15+ GP33 coronal/labial/labial 0.76 False 0.77 coronal/labial/labial 0.72 False 0.73 0.11/0.95/0.05 True
pre time i all EC2 coronal/labial/coronal 0.73 False 0.67 coronal/labial/coronal 0.74 False 0.67 0.00/0.00/0.89 True
pre time i all GP31 labial/coronal/coronal 0.88 False 0.78 labial/coronal/coronal 0.88 False 0.75 0.94/0.95/1.00 False
pre time i all GP33 coronal/coronal/coronal 0.56 False 0.32 coronal/coronal/coronal 0.57 False 0.31 0.21/0.16/0.32 False
pre time i 15+ EC2 coronal/labial/coronal 0.73 False 0.65 coronal/labial/coronal 0.74 False 0.65 0.00/0.00/0.84 True
pre time i 15+ GP31 labial/coronal/coronal 0.69 False 0.67 labial/dorsal/coronal 0.69 True 0.65 0.94/1.00/1.00 False
pre time i 15+ GP33 coronal/coronal/coronal 0.56 False 0.32 coronal/coronal/coronal 0.57 False 0.31 0.21/0.16/0.32 False
pre time ii all EC2 coronal/dorsal/labial 1.00 True 1.00 coronal/dorsal/labial 0.94 True 0.94 1.00/0.00/0.00 False
pre time ii all GP31 coronal/labial/coronal 0.91 False 0.80 coronal/labial/coronal 0.88 False 0.75 1.00/0.95/1.00 False
pre time ii all GP33 labial/coronal/coronal 0.80 False 0.77 labial/coronal/coronal 0.80 False 0.71 0.16/0.00/0.37 False
pre time ii 15+ EC2 coronal/dorsal/labial 1.00 True 1.00 coronal/dorsal/labial 0.94 True 0.94 1.00/0.00/0.00 True
pre time ii 15+ GP31 labial/tie/coronal 0.79 False 0.77 labial/dorsal/coronal 0.76 True 0.75 0.94/1.00/1.00 False
pre time ii 15+ GP33 labial/coronal/coronal 0.80 False 0.77 labial/coronal/coronal 0.80 False 0.71 0.16/0.00/0.37 False
whole row i all EC2 coronal/labial/dorsal 0.98 True 0.98 coronal/labial/dorsal 0.93 True 0.92 0.84/0.79/0.21 True
whole row i all GP31 labial/coronal/coronal 0.91 False 0.80 labial/coronal/coronal 0.90 False 0.77 1.00/0.95/1.00 False
whole row i all GP33 labial/coronal/coronal 0.60 False 0.56 labial/coronal/coronal 0.61 False 0.55 0.05/0.00/0.05 False
whole row i 15+ EC2 labial/coronal/coronal 0.80 False 0.63 labial/coronal/coronal 0.81 False 0.61 0.58/0.44/0.11 True
whole row i 15+ GP31 labial/coronal/coronal 0.67 False 0.59 labial/coronal/coronal 0.63 False 0.52 0.94/1.00/1.00 False
whole row i 15+ GP33 labial/coronal/coronal 0.60 False 0.56 labial/coronal/coronal 0.61 False 0.55 0.05/0.00/0.05 False
whole row ii all EC2 coronal/labial/dorsal 0.98 True 0.98 coronal/labial/dorsal 0.93 True 0.92 0.00/0.00/1.00 True
whole row ii all GP31 labial/coronal/coronal 0.91 False 0.80 labial/coronal/coronal 0.87 False 0.74 1.00/0.95/1.00 False
whole row ii all GP33 coronal/labial/labial 0.73 False 0.73 coronal/labial/labial 0.67 False 0.62 0.11/0.00/0.11 False
whole row ii 15+ EC2 coronal/dorsal/labial 1.00 True 1.00 coronal/dorsal/labial 0.94 True 0.94 0.00/0.00/1.00 True
whole row ii 15+ GP31 labial/coronal/coronal 0.76 False 0.67 labial/coronal/coronal 0.71 False 0.58 0.94/1.00/1.00 False
whole row ii 15+ GP33 coronal/labial/labial 0.73 False 0.73 coronal/labial/labial 0.67 False 0.62 0.11/0.00/0.11 False
whole time i all EC2 coronal/labial/coronal 0.73 False 0.67 coronal/labial/coronal 0.74 False 0.67 0.00/0.00/0.42 False
whole time i all GP31 labial/coronal/coronal 0.93 False 0.82 labial/coronal/coronal 0.90 False 0.77 1.00/0.95/1.00 False
whole time i all GP33 coronal/coronal/coronal 0.53 False 0.32 coronal/coronal/coronal 0.56 False 0.30 0.11/0.00/0.05 False
whole time i 15+ EC2 coronal/labial/coronal 0.73 False 0.65 coronal/labial/coronal 0.74 False 0.65 0.00/0.00/0.84 True
whole time i 15+ GP31 coronal/labial/coronal 0.60 False 0.51 coronal/labial/coronal 0.59 False 0.51 0.94/1.00/1.00 False
whole time i 15+ GP33 coronal/coronal/coronal 0.53 False 0.32 coronal/coronal/coronal 0.56 False 0.30 0.11/0.00/0.05 False
whole time ii all EC2 coronal/labial/dorsal 1.00 True 1.00 coronal/labial/dorsal 0.96 True 0.96 1.00/0.00/0.00 False
whole time ii all GP31 labial/coronal/coronal 0.91 False 0.80 labial/coronal/coronal 0.88 False 0.75 1.00/0.95/1.00 False
whole time ii all GP33 coronal/labial/labial 0.73 False 0.72 coronal/labial/labial 0.70 False 0.69 0.16/0.00/0.63 True
whole time ii 15+ EC2 coronal/dorsal/labial 1.00 True 1.00 coronal/dorsal/labial 0.94 True 0.94 1.00/0.00/0.00 True
whole time ii 15+ GP31 labial/coronal/coronal 0.76 False 0.74 labial/dorsal/coronal 0.75 True 0.72 0.94/1.00/1.00 False
whole time ii 15+ GP33 coronal/labial/labial 0.73 False 0.72 coronal/labial/labial 0.70 False 0.69 0.16/0.00/0.63 True
rest row i all EC2 coronal/labial/dorsal 0.98 True 0.98 coronal/labial/dorsal 0.93 True 0.92 0.74/0.53/0.05 True
rest row i all GP31 coronal/labial/coronal 0.86 False 0.71 coronal/labial/coronal 0.88 False 0.71 1.00/0.95/1.00 False
rest row i all GP33 labial/coronal/coronal 0.71 False 0.73 labial/coronal/coronal 0.70 False 0.65 0.89/1.00/0.89 False
rest row i 15+ EC2 coronal/labial/dorsal 0.95 True 0.95 coronal/labial/dorsal 0.91 True 0.89 0.74/0.56/0.05 True
rest row i 15+ GP31 coronal/tie/labial 0.86 False 0.73 coronal/tie/labial 0.88 False 0.72 1.00/0.94/1.00 False
rest row i 15+ GP33 labial/coronal/coronal 0.71 False 0.73 labial/coronal/coronal 0.70 False 0.65 0.89/1.00/0.89 False
rest row ii all EC2 coronal/labial/dorsal 1.00 True 1.00 coronal/labial/dorsal 0.93 True 0.91 1.00/0.05/0.00 True
rest row ii all GP31 tie/coronal/labial 0.84 False 0.71 tie/coronal/labial 0.87 False 0.71 1.00/0.95/1.00 False
rest row ii all GP33 labial/coronal/coronal 0.80 False 0.82 labial/coronal/coronal 0.78 False 0.72 0.11/0.00/0.11 False
rest row ii 15+ EC2 coronal/labial/dorsal 0.98 True 0.98 coronal/labial/dorsal 0.91 True 0.91 1.00/0.00/0.00 True
rest row ii 15+ GP31 tie/coronal/labial 0.86 False 0.73 tie/coronal/labial 0.88 False 0.72 1.00/0.94/1.00 False
rest row ii 15+ GP33 labial/coronal/coronal 0.80 False 0.82 labial/coronal/coronal 0.78 False 0.72 0.11/0.00/0.11 False
rest time i all EC2 coronal/labial/coronal 0.73 False 0.67 coronal/labial/coronal 0.74 False 0.67 1.00/0.53/0.16 True
rest time i all GP31 tie/coronal/coronal 0.93 False 0.49 tie/coronal/coronal 0.94 False 0.49 1.00/0.95/1.00 False
rest time i all GP33 labial/tie/coronal 0.78 False 0.38 labial/tie/coronal 0.78 False 0.38 0.89/1.00/0.89 False
rest time i 15+ EC2 coronal/labial/coronal 0.73 False 0.67 coronal/labial/coronal 0.75 False 0.61 0.00/0.44/0.79 True
rest time i 15+ GP31 tie/coronal/labial 0.86 False 0.73 tie/coronal/labial 0.88 False 0.72 0.17/0.11/0.11 False
rest time i 15+ GP33 labial/tie/coronal 0.78 False 0.38 labial/tie/coronal 0.78 False 0.38 0.89/1.00/0.89 False
rest time ii all EC2 coronal/labial/dorsal 0.98 True 0.98 coronal/labial/dorsal 0.93 True 0.92 1.00/0.05/0.00 True
rest time ii all GP31 tie/coronal/labial 0.84 False 0.71 tie/coronal/labial 0.87 False 0.71 1.00/0.95/1.00 False
rest time ii all GP33 labial/coronal/coronal 0.78 False 0.81 labial/coronal/coronal 0.74 False 0.71 0.11/0.00/0.11 False
rest time ii 15+ EC2 coronal/labial/dorsal 1.00 True 1.00 coronal/labial/dorsal 0.94 True 0.94 1.00/0.00/0.00 True
rest time ii 15+ GP31 tie/coronal/labial 0.86 False 0.73 tie/coronal/labial 0.88 False 0.72 1.00/0.94/1.00 False
rest time ii 15+ GP33 labial/coronal/coronal 0.78 False 0.81 labial/coronal/coronal 0.74 False 0.71 0.11/0.00/0.11 False
baseline z-scoring projection CVs labels subjects with both outcome
pre row i 15+ a 0 not reproduced
pre row i 15+ b 0 not reproduced
pre row i all a 1 partially
pre row i all b 1 partially
pre row ii 15+ a 1 partially
pre row ii 15+ b 1 partially
pre row ii all a 1 partially
pre row ii all b 1 partially
pre time i 15+ a 0 not reproduced
pre time i 15+ b 0 not reproduced
pre time i all a 0 not reproduced
pre time i all b 0 not reproduced
pre time ii 15+ a 1 partially
pre time ii 15+ b 1 partially
pre time ii all a 0 not reproduced
pre time ii all b 0 not reproduced
rest row i 15+ a 1 partially
rest row i 15+ b 1 partially
rest row i all a 1 partially
rest row i all b 1 partially
rest row ii 15+ a 1 partially
rest row ii 15+ b 1 partially
rest row ii all a 1 partially
rest row ii all b 1 partially
rest time i 15+ a 0 not reproduced
rest time i 15+ b 0 not reproduced
rest time i all a 0 not reproduced
rest time i all b 0 not reproduced
rest time ii 15+ a 1 partially
rest time ii 15+ b 1 partially
rest time ii all a 1 partially
rest time ii all b 1 partially
whole row i 15+ a 0 not reproduced
whole row i 15+ b 0 not reproduced
whole row i all a 1 partially
whole row i all b 1 partially
whole row ii 15+ a 1 partially
whole row ii 15+ b 1 partially
whole row ii all a 1 partially
whole row ii all b 1 partially
whole time i 15+ a 0 not reproduced
whole time i 15+ b 0 not reproduced
whole time i all a 0 not reproduced
whole time i all b 0 not reproduced
whole time ii 15+ a 1 partially
whole time ii 15+ b 1 partially
whole time ii all a 0 not reproduced
whole time ii all b 0 not reproduced
Outcome over readings: differ by reading: not reproduced 20, partially 28
Outcome counts by reading factor (48 readings):
baseline: {'pre': {'not reproduced': 8, 'partially': 8}, 'rest': {'partially': 12, 'not reproduced': 4}, 'whole': {'not reproduced': 8, 'partially': 8}}
z-scoring: {'row': {'partially': 20, 'not reproduced': 4}, 'time': {'not reproduced': 16, 'partially': 8}}
projection: {'i': {'not reproduced': 16, 'partially': 8}, 'ii': {'partially': 20, 'not reproduced': 4}}
CVs: {'15+': {'partially': 14, 'not reproduced': 10}, 'all': {'partially': 14, 'not reproduced': 10}}
labels: {'a': {'partially': 14, 'not reproduced': 10}, 'b': {'partially': 14, 'not reproduced': 10}}
Result: blocked (methods_underspecified). Both splits hold together in at most one subject,
EC2 (14 of the 24 readings), so no reading gives reproduced; 28 of the 48 readings (with the two
label readings) give partially reproduced and 20 not reproduced. Under time z-scoring with
projection (i) no subject meets both. The vowel split (/u/ against /a/ and /i/) never holds in GP31. By reading factor, projection
(ii) gives partially reproduced in 20 of 24 readings and projection (i) in 8 of 24; the CV set and
the label reading do not change the counts (14 of 24 each).
fig4ef-feature-correlations¶
Over early times (-375 to 120 ms) constriction location correlates with the state-space structure more than constriction degree, which correlates more than the upcoming vowel; over later times (125 to 620 ms) vowel configuration correlates most.
In [23]:
FEATS = ("location", "degree", "vowel")
def feature_r(base, z, pj, cs, metric="euclidean"):
"""R(t) per subject and feature set (Methods Eq. 7): {(subject, set): R over epoch times}."""
out = {}
for s in B.SUBJECTS:
K = KS[s, base, z, pj, cs]
dK = np.array([pdist(K[:, :, t]) for t in range(K.shape[2])])
for f in FEATS:
dP = pdist(B.feature_matrix(CVS[s, cs], f), metric)
out[s, f] = np.array([np.corrcoef(d, dP)[0, 1] for d in dK])
return out
early = (B.EPOCH_MS >= -375) & (B.EPOCH_MS < 120)
late = (B.EPOCH_MS >= 125) & (B.EPOCH_MS < 620)
RT = {k: feature_r(*k) for k in SS_READINGS}
fe = []
for (base, z, pj, cs), R in RT.items():
for metric in ("euclidean", "hamming"):
Rm = R if metric == "euclidean" else feature_r(base, z, pj, cs, "hamming")
e = {f: np.concatenate([Rm[s, f][early] for s in B.SUBJECTS]) for f in FEATS}
l = {f: np.concatenate([Rm[s, f][late] for s in B.SUBJECTS]) for f in FEATS}
tests = {"early loc>deg": (e["location"], e["degree"]), "early deg>vow": (e["degree"], e["vowel"]),
"late vow>loc": (l["vowel"], l["location"]), "late vow>deg": (l["vowel"], l["degree"])}
row = {"baseline": base, "z-scoring": z, "projection": pj, "CVs": cs, "distance": metric, "n": len(e["location"])}
n_ok = 0
for name, (a, b) in tests.items():
p = stats.wilcoxon(a, b).pvalue
ok = bool(np.median(a - b) > 0 and p < 1e-10)
row[name] = f"{np.median(a - b):+.3f} p={p:.0e} {'Y' if ok else 'n'}"
n_ok += ok
row["med early L/D/V"] = "/".join(f"{np.median(e[f]):.2f}" for f in FEATS)
row["med late L/D/V"] = "/".join(f"{np.median(l[f]):.2f}" for f in FEATS)
row["outcome"] = "reproduced" if n_ok == 4 else ("partially" if n_ok else "not reproduced")
fe.append(row)
fe = pd.DataFrame(fe)
print(fe.to_string(index=False))
print("Outcome over readings (Euclidean; Hamming is a variant):", overall(fe[fe.distance == "euclidean"].outcome))
by_factor(fe[fe.distance == "euclidean"], ["baseline", "z-scoring", "projection", "CVs"])
baseline z-scoring projection CVs distance n early loc>deg early deg>vow late vow>loc late vow>deg med early L/D/V med late L/D/V outcome
pre row i all euclidean 297 +0.100 p=1e-32 Y +0.047 p=2e-41 Y -0.039 p=1e-02 n +0.011 p=8e-03 n 0.13/0.06/0.00 0.05/0.07/0.03 partially
pre row i all hamming 297 +0.083 p=3e-24 Y +0.043 p=4e-38 Y -0.031 p=1e-01 n -0.007 p=9e-03 n 0.10/0.04/-0.00 0.04/0.08/0.02 partially
pre row i 15+ euclidean 297 +0.083 p=5e-16 Y +0.061 p=3e-46 Y +0.080 p=4e-20 Y +0.020 p=1e-06 n 0.14/0.08/0.01 0.02/0.09/0.08 partially
pre row i 15+ hamming 297 +0.059 p=2e-07 n +0.057 p=3e-41 Y +0.066 p=8e-15 Y +0.012 p=2e-01 n 0.11/0.07/0.00 0.02/0.09/0.08 partially
pre row ii all euclidean 297 +0.141 p=3e-43 Y +0.045 p=4e-36 Y -0.027 p=2e-07 n +0.026 p=6e-20 Y 0.18/0.06/0.00 0.05/0.07/0.06 partially
pre row ii all hamming 297 +0.113 p=2e-40 Y +0.044 p=4e-31 Y -0.025 p=2e-06 n +0.006 p=2e-05 n 0.14/0.04/-0.00 0.04/0.07/0.04 partially
pre row ii 15+ euclidean 297 +0.145 p=6e-40 Y +0.063 p=7e-40 Y +0.147 p=3e-33 Y +0.082 p=7e-31 Y 0.21/0.07/0.00 0.02/0.08/0.14 reproduced
pre row ii 15+ hamming 297 +0.110 p=1e-35 Y +0.055 p=1e-32 Y +0.136 p=7e-29 Y +0.070 p=1e-17 Y 0.16/0.06/0.00 0.02/0.08/0.13 reproduced
pre time i all euclidean 297 +0.097 p=6e-36 Y +0.043 p=3e-41 Y -0.029 p=9e-07 n +0.020 p=1e-12 Y 0.12/0.06/0.01 0.05/0.07/0.05 partially
pre time i all hamming 297 +0.073 p=2e-28 Y +0.043 p=1e-37 Y -0.027 p=1e-05 n +0.003 p=1e-03 n 0.10/0.04/-0.00 0.04/0.08/0.04 partially
pre time i 15+ euclidean 297 +0.052 p=8e-16 Y +0.069 p=5e-45 Y +0.114 p=2e-29 Y +0.052 p=5e-21 Y 0.13/0.09/0.01 0.02/0.09/0.11 reproduced
pre time i 15+ hamming 297 +0.029 p=2e-08 n +0.057 p=7e-40 Y +0.101 p=1e-23 Y +0.045 p=2e-10 n 0.11/0.08/0.01 0.02/0.09/0.10 partially
pre time ii all euclidean 297 +0.147 p=8e-44 Y +0.043 p=1e-34 Y -0.023 p=1e-07 n +0.027 p=3e-24 Y 0.18/0.06/0.00 0.05/0.06/0.06 partially
pre time ii all hamming 297 +0.117 p=2e-41 Y +0.044 p=5e-29 Y -0.020 p=8e-07 n +0.006 p=7e-07 n 0.14/0.04/-0.00 0.04/0.06/0.05 partially
pre time ii 15+ euclidean 297 +0.151 p=2e-40 Y +0.063 p=1e-37 Y +0.144 p=1e-32 Y +0.084 p=2e-32 Y 0.22/0.08/0.00 0.02/0.07/0.14 reproduced
pre time ii 15+ hamming 297 +0.117 p=6e-37 Y +0.052 p=4e-30 Y +0.136 p=2e-28 Y +0.070 p=2e-18 Y 0.17/0.07/0.00 0.02/0.08/0.13 reproduced
whole row i all euclidean 297 +0.101 p=1e-32 Y +0.044 p=2e-40 Y -0.043 p=1e-02 n +0.013 p=9e-03 n 0.13/0.05/0.00 0.05/0.07/0.03 partially
whole row i all hamming 297 +0.082 p=4e-24 Y +0.043 p=5e-37 Y -0.036 p=1e-01 n -0.004 p=9e-03 n 0.10/0.03/-0.00 0.04/0.08/0.02 partially
whole row i 15+ euclidean 297 +0.075 p=8e-15 Y +0.063 p=1e-45 Y +0.083 p=9e-21 Y +0.027 p=6e-07 n 0.15/0.08/0.01 0.02/0.09/0.08 partially
whole row i 15+ hamming 297 +0.051 p=2e-07 n +0.057 p=3e-40 Y +0.066 p=1e-15 Y +0.019 p=1e-01 n 0.11/0.07/0.00 0.02/0.09/0.08 partially
whole row ii all euclidean 297 +0.135 p=2e-44 Y +0.043 p=6e-35 Y -0.030 p=3e-07 n +0.027 p=9e-21 Y 0.17/0.06/0.00 0.05/0.06/0.05 partially
whole row ii all hamming 297 +0.105 p=6e-42 Y +0.040 p=2e-29 Y -0.028 p=4e-06 n +0.007 p=5e-06 n 0.13/0.04/-0.00 0.04/0.07/0.04 partially
whole row ii 15+ euclidean 297 +0.146 p=7e-41 Y +0.065 p=6e-40 Y +0.146 p=4e-33 Y +0.089 p=5e-33 Y 0.21/0.08/0.00 0.02/0.07/0.14 reproduced
whole row ii 15+ hamming 297 +0.114 p=9e-38 Y +0.047 p=2e-31 Y +0.136 p=1e-28 Y +0.082 p=7e-21 Y 0.16/0.07/0.00 0.02/0.07/0.14 reproduced
whole time i all euclidean 297 +0.096 p=4e-36 Y +0.042 p=3e-40 Y -0.029 p=2e-06 n +0.023 p=2e-13 Y 0.12/0.06/0.00 0.05/0.07/0.05 partially
whole time i all hamming 297 +0.076 p=2e-29 Y +0.039 p=2e-36 Y -0.028 p=2e-05 n +0.005 p=2e-04 n 0.10/0.04/-0.00 0.04/0.08/0.04 partially
whole time i 15+ euclidean 297 +0.060 p=3e-14 Y +0.067 p=1e-44 Y +0.120 p=2e-29 Y +0.063 p=2e-23 Y 0.13/0.10/0.01 0.02/0.09/0.11 reproduced
whole time i 15+ hamming 297 +0.034 p=6e-08 n +0.053 p=4e-39 Y +0.110 p=7e-24 Y +0.054 p=4e-12 Y 0.11/0.08/0.01 0.02/0.09/0.11 partially
whole time ii all euclidean 297 +0.137 p=8e-45 Y +0.040 p=3e-32 Y -0.030 p=2e-07 n +0.027 p=6e-25 Y 0.17/0.06/0.00 0.05/0.06/0.06 partially
whole time ii all hamming 297 +0.109 p=7e-43 Y +0.039 p=1e-26 Y -0.026 p=2e-06 n +0.007 p=1e-07 n 0.13/0.04/-0.00 0.04/0.06/0.05 partially
whole time ii 15+ euclidean 297 +0.145 p=3e-41 Y +0.067 p=3e-37 Y +0.147 p=9e-33 Y +0.085 p=4e-35 Y 0.22/0.08/0.00 0.02/0.07/0.14 reproduced
whole time ii 15+ hamming 297 +0.116 p=5e-39 Y +0.045 p=3e-28 Y +0.138 p=4e-28 Y +0.074 p=9e-22 Y 0.17/0.07/0.00 0.02/0.07/0.13 reproduced
rest row i all euclidean 297 +0.066 p=2e-19 Y +0.047 p=7e-35 Y +0.022 p=8e-05 n +0.021 p=3e-02 n 0.10/0.06/0.01 0.02/0.07/0.02 partially
rest row i all hamming 297 +0.039 p=2e-10 n +0.040 p=1e-30 Y +0.017 p=5e-04 n -0.000 p=2e-02 n 0.07/0.05/-0.00 0.02/0.08/0.02 partially
rest row i 15+ euclidean 297 +0.032 p=3e-05 n +0.053 p=1e-38 Y +0.075 p=5e-18 Y +0.028 p=1e-03 n 0.09/0.07/0.01 0.01/0.08/0.04 partially
rest row i 15+ hamming 297 +0.011 p=9e-02 n +0.049 p=4e-31 Y +0.078 p=3e-15 Y +0.018 p=9e-01 n 0.07/0.06/0.00 -0.00/0.09/0.04 partially
rest row ii all euclidean 297 +0.111 p=2e-33 Y +0.036 p=3e-23 Y +0.043 p=1e-13 Y +0.035 p=7e-22 Y 0.13/0.05/0.00 0.02/0.06/0.05 reproduced
rest row ii all hamming 297 +0.091 p=5e-26 Y +0.029 p=3e-17 Y +0.037 p=2e-12 Y +0.018 p=3e-06 n 0.10/0.04/-0.00 0.02/0.06/0.04 partially
rest row ii 15+ euclidean 297 +0.109 p=4e-23 Y +0.038 p=6e-24 Y +0.111 p=9e-32 Y +0.045 p=3e-25 Y 0.13/0.06/0.00 0.01/0.07/0.05 reproduced
rest row ii 15+ hamming 297 +0.073 p=1e-16 Y +0.027 p=4e-15 Y +0.109 p=3e-27 Y +0.036 p=5e-11 Y 0.10/0.04/0.00 0.01/0.06/0.05 reproduced
rest time i all euclidean 297 +0.054 p=6e-23 Y +0.045 p=2e-36 Y +0.027 p=2e-10 n +0.034 p=3e-16 Y 0.10/0.06/0.01 0.03/0.07/0.04 partially
rest time i all hamming 297 +0.037 p=1e-13 Y +0.040 p=1e-31 Y +0.020 p=1e-09 n +0.018 p=1e-05 n 0.08/0.05/0.00 0.02/0.07/0.04 partially
rest time i 15+ euclidean 297 +0.003 p=3e-04 n +0.056 p=4e-39 Y +0.107 p=9e-29 Y +0.059 p=4e-22 Y 0.10/0.07/0.01 0.01/0.09/0.06 partially
rest time i 15+ hamming 297 -0.009 p=5e-02 n +0.045 p=5e-32 Y +0.103 p=8e-24 Y +0.049 p=8e-10 n 0.08/0.07/0.01 0.00/0.09/0.06 partially
rest time ii all euclidean 297 +0.118 p=1e-34 Y +0.043 p=2e-21 Y +0.042 p=8e-15 Y +0.037 p=2e-26 Y 0.14/0.06/0.00 0.02/0.05/0.05 reproduced
rest time ii all hamming 297 +0.097 p=5e-28 Y +0.034 p=3e-15 Y +0.035 p=1e-13 Y +0.018 p=8e-09 n 0.10/0.05/-0.00 0.01/0.06/0.04 partially
rest time ii 15+ euclidean 297 +0.114 p=3e-24 Y +0.044 p=2e-22 Y +0.112 p=1e-31 Y +0.050 p=2e-29 Y 0.14/0.06/0.00 0.00/0.06/0.05 reproduced
rest time ii 15+ hamming 297 +0.082 p=8e-19 Y +0.033 p=1e-13 Y +0.109 p=2e-27 Y +0.038 p=6e-13 Y 0.10/0.05/0.00 0.00/0.06/0.04 reproduced
Outcome over readings (Euclidean; Hamming is a variant): differ by reading: partially 14, reproduced 10
Outcome counts by reading factor (24 readings):
baseline: {'pre': {'partially': 5, 'reproduced': 3}, 'whole': {'partially': 5, 'reproduced': 3}, 'rest': {'partially': 4, 'reproduced': 4}}
z-scoring: {'row': {'partially': 8, 'reproduced': 4}, 'time': {'partially': 6, 'reproduced': 6}}
projection: {'i': {'partially': 10, 'reproduced': 2}, 'ii': {'reproduced': 8, 'partially': 4}}
CVs: {'all': {'partially': 10, 'reproduced': 2}, '15+': {'reproduced': 8, 'partially': 4}}
Result: blocked (methods_underspecified). In the early window, constriction location
correlates with the state-space distances more than constriction degree, and degree more than the
vowel, with p < 1e-10 under every reading with Euclidean distances except two (rest, projection
(i), 15+ CVs, where location over degree has p = 3e-05 and 3e-04). In the late window the vowel
features correlate more than both other sets with p < 1e-10 in 10 of the 24 readings (reproduced):
every 15+ reading with projection (ii), pre and whole with time, projection (i) and 15+ CVs,
and rest with projection (ii) and every CV. In the other 14 the late comparisons fail in part
(partially reproduced), mostly with every CV, where location correlates as much as or more than
the vowel features. By reading factor, projection (ii) gives 8 of the 10 reproduced readings and the 15+ CVs 8 of
10. Hamming distances, a variant, are in the table and are not judged.
fig5c-separability-timing¶
Cluster separability, averaged over subjects, rises from about -300 ms and peaks about 200 ms before the C-V transition for consonants, and rises from about 50 ms and peaks at +250 ms for vowels.
In [24]:
rng5 = np.random.default_rng(5)
N_BOOT = 1000
def cluster_labels(s, cs, K, which, kind):
"""Labels per CV (per time and CV for k-means); -1 for CVs without a label."""
cvs = CVS[s, cs]
if kind == "kmeans":
return np.array([B.kmeans(K[:, :, t], 3, seed=1000 + t, n_init=10) for t in range(K.shape[2])])
if which == "vowel":
return np.array(["aiu".index(v) for c, v in cvs])
lr = kind[-1]
names = {"labial": 0, "coronal": 1, "dorsal": 2}
out = []
for c, v in cvs:
m = B.MAJOR[lr].get(c)
out.append(3 if kind.startswith("four") and c in B.SIBILANTS else (names[m] if m else -1))
return np.array(out)
SEP_KINDS = {"consonant": ["groups a", "groups b", "four a", "four b", "kmeans"], "vowel": ["groups", "kmeans"]}
def separability_curves(base, z, pj, cs):
out = {}
kmeans_cache = {}
for which, kinds in SEP_KINDS.items():
for kind in kinds:
curves, boots = [], []
for s in B.SUBJECTS:
K = KS[s, base, z, pj, cs]
if kind == "kmeans":
if s not in kmeans_cache:
kmeans_cache[s] = cluster_labels(s, cs, K, which, kind)
lab = kmeans_cache[s]
else:
lab = cluster_labels(s, cs, K, which, kind)
keep = (lab >= 0) if lab.ndim == 1 else np.ones(K.shape[0], bool)
Ks = K[keep]
labk = lab[keep] if lab.ndim == 1 else lab[:, keep]
D = np.array([squareform(pdist(Ks[:, :, t])) for t in range(Ks.shape[2])])
curves.append(B.separability(D, labk)[0])
wts = rng5.multinomial(len(Ks), np.ones(len(Ks)) / len(Ks), size=N_BOOT)
boots.append(B.separability(D, labk, wts))
out[which, kind] = (np.mean(curves, 0), np.mean(boots, 0).std(0), curves)
return out
SEP = {k: separability_curves(*k) for k in SS_READINGS}
def onset_peak(curve, se):
"""Peak time, onset (a) first time above the level at -500 ms by 2 bootstrap s.e., onset (b)
first time of the run up to the peak above half the peak value."""
p = int(np.argmax(curve))
above = np.where(curve[:p + 1] > curve[0] + 2 * se[:p + 1])[0]
on_a = B.EPOCH_MS[above[0]] if len(above) else np.nan
below = np.where(curve[:p + 1] < curve[p] / 2)[0]
on_b = B.EPOCH_MS[below[-1] + 1] if len(below) else np.nan
return B.EPOCH_MS[p], on_a, on_b
sp = []
for (base, z, pj, cs), d in SEP.items():
for ck in SEP_KINDS["consonant"]:
vk = "kmeans" if ck == "kmeans" else "groups"
cpk, con_a, con_b = onset_peak(d["consonant", ck][0], d["consonant", ck][1])
vpk, von_a, von_b = onset_peak(d["vowel", vk][0], d["vowel", vk][1])
peaks_ok = abs(cpk + 200) <= 20 and abs(vpk - 250) <= 25
ons_ok = {"a": abs(con_a + 300) <= 30 and abs(von_a - 50) <= 10, "b": abs(con_b + 300) <= 30 and abs(von_b - 50) <= 10}
sp.append({"baseline": base, "z-scoring": z, "projection": pj, "CVs": cs, "clusters": f"{ck} | {vk}",
"cons peak": cpk, "cons onset a": con_a, "cons onset b": con_b,
"vowel peak": vpk, "vowel onset a": von_a, "vowel onset b": von_b,
"peaks ok": peaks_ok, "onset ok a": ons_ok["a"], "onset ok b": ons_ok["b"],
"outcome": "reproduced" if peaks_ok and any(ons_ok.values()) else ("partially" if peaks_ok else "not reproduced")})
sp = pd.DataFrame(sp)
print(sp.to_string(index=False))
print("Expected: consonant peak -200 +- 20, onset -300 +- 30; vowel peak 250 +- 25, onset 50 +- 10 ms")
print("Outcome over readings:", overall(sp.outcome))
by_factor(sp, ["baseline", "z-scoring", "projection", "CVs", "clusters"])
g = sp[~sp.clusters.str.startswith("kmeans")]
print("Phonetic-group readings: consonant peak", g["cons peak"].min(), "to", g["cons peak"].max(), "| vowel peak", g["vowel peak"].min(), "to", g["vowel peak"].max(),
"| consonant onset (a)", g["cons onset a"].min(), "to", g["cons onset a"].max(), "| (b)", g["cons onset b"].min(), "to", g["cons onset b"].max(),
"| vowel onset (a)", g["vowel onset a"].min(), "to", g["vowel onset a"].max(), "| (b)", g["vowel onset b"].min(), "to", g["vowel onset b"].max())
g = sp[sp.clusters.str.startswith("kmeans")]
print("k-means readings: peak of the single curve", g["cons peak"].min(), "to", g["cons peak"].max())
fig, axes = plt.subplots(1, 2, figsize=(11, 3))
for ax, key in zip(axes, [("pre", "row", "i", "all"), ("pre", "time", "i", "all")]):
for (which, kind), (c, se, _) in SEP[key].items():
if kind != "kmeans":
ax.plot(B.EPOCH_MS, c, label=f"{which} {kind}")
ax.axvline(0, color="k", lw=0.5); ax.set_title(", ".join(key)); ax.set_xlabel("ms")
axes[0].legend(fontsize=7); axes[0].set_ylabel("separability")
plt.show()
for key, d in SEP.items():
print(key, "curve ranges:", {f"{w} {k}": (round(float(c.min()), 2), round(float(c.max()), 2)) for (w, k), (c, se, _) in d.items()})
baseline z-scoring projection CVs clusters cons peak cons onset a cons onset b vowel peak vowel onset a vowel onset b peaks ok onset ok a onset ok b outcome
pre row i all groups a | groups 5 -240.0 -185.0 200 105.0 95.0 False False False not reproduced
pre row i all groups b | groups 5 -260.0 -195.0 200 105.0 95.0 False False False not reproduced
pre row i all four a | groups -65 -260.0 -210.0 200 105.0 95.0 False False False not reproduced
pre row i all four b | groups -65 -270.0 -230.0 200 105.0 95.0 False False False not reproduced
pre row i all kmeans | kmeans 290 -165.0 240.0 290 -165.0 240.0 False False False not reproduced
pre row i 15+ groups a | groups 5 -260.0 -190.0 205 80.0 90.0 False False False not reproduced
pre row i 15+ groups b | groups 5 -275.0 -200.0 205 80.0 90.0 False False False not reproduced
pre row i 15+ four a | groups -60 -270.0 -220.0 205 80.0 90.0 False False False not reproduced
pre row i 15+ four b | groups -60 -275.0 -235.0 205 80.0 90.0 False False False not reproduced
pre row i 15+ kmeans | kmeans -135 -375.0 -395.0 -135 -375.0 -395.0 False False False not reproduced
pre row ii all groups a | groups -60 -275.0 -195.0 200 75.0 95.0 False False False not reproduced
pre row ii all groups b | groups -60 -295.0 -205.0 200 75.0 95.0 False False False not reproduced
pre row ii all four a | groups -70 -300.0 -220.0 200 75.0 95.0 False False False not reproduced
pre row ii all four b | groups -70 -325.0 -235.0 200 75.0 95.0 False False False not reproduced
pre row ii all kmeans | kmeans 285 -280.0 240.0 285 -280.0 240.0 False False False not reproduced
pre row ii 15+ groups a | groups -60 -290.0 -200.0 205 65.0 95.0 False False False not reproduced
pre row ii 15+ groups b | groups -60 -320.0 -210.0 205 65.0 95.0 False False False not reproduced
pre row ii 15+ four a | groups -70 -330.0 -225.0 205 65.0 95.0 False False False not reproduced
pre row ii 15+ four b | groups -70 -365.0 -235.0 205 65.0 95.0 False False False not reproduced
pre row ii 15+ kmeans | kmeans -120 -375.0 -375.0 -120 -375.0 -375.0 False False False not reproduced
pre time i all groups a | groups 5 -235.0 -230.0 225 75.0 100.0 False False False not reproduced
pre time i all groups b | groups 10 -260.0 -255.0 225 75.0 100.0 False False False not reproduced
pre time i all four a | groups -95 -270.0 -270.0 225 75.0 100.0 False False False not reproduced
pre time i all four b | groups -145 -280.0 -275.0 225 75.0 100.0 False False False not reproduced
pre time i all kmeans | kmeans 310 NaN 240.0 310 NaN 240.0 False False False not reproduced
pre time i 15+ groups a | groups 5 -255.0 -240.0 235 60.0 100.0 False False False not reproduced
pre time i 15+ groups b | groups -65 -280.0 -265.0 235 60.0 100.0 False True False not reproduced
pre time i 15+ four a | groups -90 -290.0 -280.0 235 60.0 100.0 False True False not reproduced
pre time i 15+ four b | groups -95 -295.0 -285.0 235 60.0 100.0 False True False not reproduced
pre time i 15+ kmeans | kmeans 330 325.0 NaN 330 325.0 NaN False False False not reproduced
pre time ii all groups a | groups -60 -280.0 -200.0 220 70.0 95.0 False False False not reproduced
pre time ii all groups b | groups -60 -300.0 -210.0 220 70.0 95.0 False False False not reproduced
pre time ii all four a | groups -65 -305.0 -225.0 220 70.0 95.0 False False False not reproduced
pre time ii all four b | groups -65 -325.0 -235.0 220 70.0 95.0 False False False not reproduced
pre time ii all kmeans | kmeans 290 -275.0 240.0 290 -275.0 240.0 False False False not reproduced
pre time ii 15+ groups a | groups -60 -300.0 -205.0 215 60.0 95.0 False True False not reproduced
pre time ii 15+ groups b | groups -60 -325.0 -215.0 215 60.0 95.0 False True False not reproduced
pre time ii 15+ four a | groups -70 -345.0 -225.0 215 60.0 95.0 False False False not reproduced
pre time ii 15+ four b | groups -70 -375.0 -240.0 215 60.0 95.0 False False False not reproduced
pre time ii 15+ kmeans | kmeans -120 -375.0 -375.0 -120 -375.0 -375.0 False False False not reproduced
whole row i all groups a | groups 5 -230.0 -180.0 205 105.0 95.0 False False False not reproduced
whole row i all groups b | groups 5 -255.0 -195.0 205 105.0 95.0 False False False not reproduced
whole row i all four a | groups -70 -255.0 -210.0 205 105.0 95.0 False False False not reproduced
whole row i all four b | groups -75 -265.0 -230.0 205 105.0 95.0 False False False not reproduced
whole row i all kmeans | kmeans 290 -50.0 245.0 290 -50.0 245.0 False False False not reproduced
whole row i 15+ groups a | groups 5 -260.0 -190.0 205 80.0 90.0 False False False not reproduced
whole row i 15+ groups b | groups 5 -275.0 -200.0 205 80.0 90.0 False False False not reproduced
whole row i 15+ four a | groups -70 -265.0 -220.0 205 80.0 90.0 False False False not reproduced
whole row i 15+ four b | groups -75 -275.0 -235.0 205 80.0 90.0 False False False not reproduced
whole row i 15+ kmeans | kmeans 10 -365.0 -410.0 10 -365.0 -410.0 False False False not reproduced
whole row ii all groups a | groups -60 -270.0 -195.0 205 75.0 95.0 False False False not reproduced
whole row ii all groups b | groups -60 -285.0 -205.0 205 75.0 95.0 False False False not reproduced
whole row ii all four a | groups -75 -290.0 -220.0 205 75.0 95.0 False False False not reproduced
whole row ii all four b | groups -75 -310.0 -230.0 205 75.0 95.0 False False False not reproduced
whole row ii all kmeans | kmeans 300 -260.0 245.0 300 -260.0 245.0 False False False not reproduced
whole row ii 15+ groups a | groups -60 -285.0 -200.0 210 65.0 95.0 False False False not reproduced
whole row ii 15+ groups b | groups -65 -315.0 -210.0 210 65.0 95.0 False False False not reproduced
whole row ii 15+ four a | groups -75 -315.0 -220.0 210 65.0 95.0 False False False not reproduced
whole row ii 15+ four b | groups -75 -365.0 -230.0 210 65.0 95.0 False False False not reproduced
whole row ii 15+ kmeans | kmeans 10 -375.0 -390.0 10 -365.0 -390.0 False False False not reproduced
whole time i all groups a | groups 5 -225.0 -225.0 230 80.0 105.0 False False False not reproduced
whole time i all groups b | groups 10 -260.0 -250.0 230 80.0 105.0 False False False not reproduced
whole time i all four a | groups -90 -270.0 -270.0 230 80.0 105.0 False False False not reproduced
whole time i all four b | groups -145 -280.0 -270.0 230 80.0 105.0 False False False not reproduced
whole time i all kmeans | kmeans 285 NaN 245.0 285 NaN 245.0 False False False not reproduced
whole time i 15+ groups a | groups 5 -255.0 -235.0 235 60.0 100.0 False False False not reproduced
whole time i 15+ groups b | groups 5 -280.0 -260.0 235 60.0 100.0 False True False not reproduced
whole time i 15+ four a | groups -90 -290.0 -275.0 235 60.0 100.0 False True False not reproduced
whole time i 15+ four b | groups -90 -305.0 -280.0 235 60.0 100.0 False True False not reproduced
whole time i 15+ kmeans | kmeans 360 -365.0 NaN 360 -365.0 NaN False False False not reproduced
whole time ii all groups a | groups -60 -275.0 -195.0 220 65.0 95.0 False False False not reproduced
whole time ii all groups b | groups -60 -290.0 -210.0 220 65.0 95.0 False False False not reproduced
whole time ii all four a | groups -75 -295.0 -220.0 220 65.0 95.0 False False False not reproduced
whole time ii all four b | groups -75 -310.0 -235.0 220 65.0 95.0 False False False not reproduced
whole time ii all kmeans | kmeans 300 -265.0 245.0 300 -265.0 245.0 False False False not reproduced
whole time ii 15+ groups a | groups -60 -290.0 -200.0 215 60.0 95.0 False True False not reproduced
whole time ii 15+ groups b | groups -65 -315.0 -215.0 215 60.0 95.0 False True False not reproduced
whole time ii 15+ four a | groups -80 -330.0 -225.0 215 60.0 95.0 False True False not reproduced
whole time ii 15+ four b | groups -75 -370.0 -235.0 215 60.0 95.0 False False False not reproduced
whole time ii 15+ kmeans | kmeans 10 -375.0 -375.0 10 -375.0 -375.0 False False False not reproduced
rest row i all groups a | groups 5 -185.0 -175.0 160 100.0 100.0 False False False not reproduced
rest row i all groups b | groups 10 -205.0 -195.0 160 100.0 100.0 False False False not reproduced
rest row i all four a | groups -70 -215.0 -210.0 160 100.0 100.0 False False False not reproduced
rest row i all four b | groups -70 -235.0 -225.0 160 100.0 100.0 False False False not reproduced
rest row i all kmeans | kmeans 370 NaN 335.0 370 NaN 335.0 False False False not reproduced
rest row i 15+ groups a | groups 10 -190.0 -170.0 155 90.0 95.0 False False False not reproduced
rest row i 15+ groups b | groups 10 -205.0 -195.0 155 90.0 95.0 False False False not reproduced
rest row i 15+ four a | groups -70 -220.0 -210.0 155 90.0 95.0 False False False not reproduced
rest row i 15+ four b | groups -75 -235.0 -225.0 155 90.0 95.0 False False False not reproduced
rest row i 15+ kmeans | kmeans -170 NaN -170.0 -170 NaN -170.0 False False False not reproduced
rest row ii all groups a | groups -5 -220.0 -185.0 165 70.0 95.0 False False False not reproduced
rest row ii all groups b | groups 5 -235.0 -200.0 165 70.0 95.0 False False False not reproduced
rest row ii all four a | groups -70 -260.0 -215.0 165 70.0 95.0 False False False not reproduced
rest row ii all four b | groups -70 -275.0 -230.0 165 70.0 95.0 False False False not reproduced
rest row ii all kmeans | kmeans 375 NaN 215.0 375 NaN 215.0 False False False not reproduced
rest row ii 15+ groups a | groups 0 -215.0 -185.0 210 65.0 95.0 False False False not reproduced
rest row ii 15+ groups b | groups 5 -240.0 -200.0 210 65.0 95.0 False False False not reproduced
rest row ii 15+ four a | groups -70 -270.0 -215.0 210 65.0 95.0 False False False not reproduced
rest row ii 15+ four b | groups -75 -285.0 -230.0 210 65.0 95.0 False False False not reproduced
rest row ii 15+ kmeans | kmeans -155 NaN -165.0 -155 NaN -165.0 False False False not reproduced
rest time i all groups a | groups 5 -180.0 -215.0 230 80.0 100.0 False False False not reproduced
rest time i all groups b | groups 15 -205.0 -235.0 230 80.0 100.0 False False False not reproduced
rest time i all four a | groups -150 -245.0 -260.0 230 80.0 100.0 False False False not reproduced
rest time i all four b | groups -155 -255.0 -260.0 230 80.0 100.0 False False False not reproduced
rest time i all kmeans | kmeans 375 NaN 320.0 375 NaN 320.0 False False False not reproduced
rest time i 15+ groups a | groups 5 -180.0 -225.0 235 70.0 100.0 False False False not reproduced
rest time i 15+ groups b | groups -65 -215.0 -250.0 235 70.0 100.0 False False False not reproduced
rest time i 15+ four a | groups -85 -250.0 -265.0 235 70.0 100.0 False False False not reproduced
rest time i 15+ four b | groups -150 -265.0 -270.0 235 70.0 100.0 False False False not reproduced
rest time i 15+ kmeans | kmeans 345 NaN 235.0 345 NaN 235.0 False False False not reproduced
rest time ii all groups a | groups -10 -225.0 -195.0 220 60.0 95.0 False False False not reproduced
rest time ii all groups b | groups -65 -245.0 -205.0 220 60.0 95.0 False False False not reproduced
rest time ii all four a | groups -70 -270.0 -220.0 220 60.0 95.0 False True False not reproduced
rest time ii all four b | groups -70 -280.0 -235.0 220 60.0 95.0 False True False not reproduced
rest time ii all kmeans | kmeans 370 NaN 215.0 370 NaN 215.0 False False False not reproduced
rest time ii 15+ groups a | groups -5 -225.0 -195.0 220 55.0 90.0 False False False not reproduced
rest time ii 15+ groups b | groups -65 -250.0 -205.0 220 55.0 90.0 False False False not reproduced
rest time ii 15+ four a | groups -70 -280.0 -220.0 220 55.0 90.0 False True False not reproduced
rest time ii 15+ four b | groups -75 -290.0 -235.0 220 55.0 90.0 False True False not reproduced
rest time ii 15+ kmeans | kmeans -150 NaN -160.0 -150 NaN -160.0 False False False not reproduced
Expected: consonant peak -200 +- 20, onset -300 +- 30; vowel peak 250 +- 25, onset 50 +- 10 ms
Outcome over readings: not reproduced
Outcome counts by reading factor (120 readings):
baseline: {'pre': {'not reproduced': 40}, 'whole': {'not reproduced': 40}, 'rest': {'not reproduced': 40}}
z-scoring: {'row': {'not reproduced': 60}, 'time': {'not reproduced': 60}}
projection: {'i': {'not reproduced': 60}, 'ii': {'not reproduced': 60}}
CVs: {'all': {'not reproduced': 60}, '15+': {'not reproduced': 60}}
clusters: {'groups a | groups': {'not reproduced': 24}, 'groups b | groups': {'not reproduced': 24}, 'four a | groups': {'not reproduced': 24}, 'four b | groups': {'not reproduced': 24}, 'kmeans | kmeans': {'not reproduced': 24}}
Phonetic-group readings: consonant peak -155 to 15 | vowel peak 155 to 235 | consonant onset (a) -375.0 to -180.0 | (b) -285.0 to -170.0 | vowel onset (a) 55.0 to 105.0 | (b) 90.0 to 105.0
k-means readings: peak of the single curve -170 to 375
('pre', 'row', 'i', 'all') curve ranges: {'consonant groups a': (-0.54, 2.61), 'consonant groups b': (-0.43, 2.25), 'consonant four a': (-0.35, 2.66), 'consonant four b': (-0.34, 2.3), 'consonant kmeans': (2.38, 13.86), 'vowel groups': (-0.04, 0.82), 'vowel kmeans': (2.38, 13.86)}
('pre', 'row', 'i', '15+') curve ranges: {'consonant groups a': (-0.13, 3.12), 'consonant groups b': (-0.15, 2.67), 'consonant four a': (-0.27, 2.96), 'consonant four b': (-0.25, 2.57), 'consonant kmeans': (1.85, 4.19), 'vowel groups': (-0.0, 0.88), 'vowel kmeans': (1.85, 4.19)}
('pre', 'row', 'ii', 'all') curve ranges: {'consonant groups a': (-0.12, 1.14), 'consonant groups b': (-0.09, 0.96), 'consonant four a': (-0.07, 1.21), 'consonant four b': (-0.06, 1.04), 'consonant kmeans': (0.57, 2.93), 'vowel groups': (-0.02, 0.39), 'vowel kmeans': (0.57, 2.93)}
('pre', 'row', 'ii', '15+') curve ranges: {'consonant groups a': (-0.03, 1.22), 'consonant groups b': (-0.03, 1.02), 'consonant four a': (-0.04, 1.25), 'consonant four b': (-0.03, 1.07), 'consonant kmeans': (0.44, 1.38), 'vowel groups': (-0.01, 0.4), 'vowel kmeans': (0.44, 1.38)}
('pre', 'time', 'i', 'all') curve ranges: {'consonant groups a': (-0.51, 1.77), 'consonant groups b': (-0.41, 1.57), 'consonant four a': (-0.34, 2.0), 'consonant four b': (-0.3, 1.83), 'consonant kmeans': (2.66, 12.63), 'vowel groups': (-0.04, 1.3), 'vowel kmeans': (2.66, 12.63)}
('pre', 'time', 'i', '15+') curve ranges: {'consonant groups a': (-0.13, 2.14), 'consonant groups b': (-0.13, 1.87), 'consonant four a': (-0.22, 2.28), 'consonant four b': (-0.16, 2.06), 'consonant kmeans': (2.17, 4.51), 'vowel groups': (0.0, 1.41), 'vowel kmeans': (2.17, 4.51)}
('pre', 'time', 'ii', 'all') curve ranges: {'consonant groups a': (-0.11, 1.16), 'consonant groups b': (-0.09, 0.98), 'consonant four a': (-0.07, 1.19), 'consonant four b': (-0.06, 1.03), 'consonant kmeans': (0.57, 2.97), 'vowel groups': (-0.02, 0.45), 'vowel kmeans': (0.57, 2.97)}
('pre', 'time', 'ii', '15+') curve ranges: {'consonant groups a': (-0.03, 1.22), 'consonant groups b': (-0.03, 1.03), 'consonant four a': (-0.04, 1.24), 'consonant four b': (-0.03, 1.07), 'consonant kmeans': (0.43, 1.36), 'vowel groups': (-0.01, 0.46), 'vowel kmeans': (0.43, 1.36)}
('whole', 'row', 'i', 'all') curve ranges: {'consonant groups a': (-0.58, 2.56), 'consonant groups b': (-0.46, 2.22), 'consonant four a': (-0.33, 2.66), 'consonant four b': (-0.33, 2.3), 'consonant kmeans': (2.34, 13.98), 'vowel groups': (-0.04, 0.81), 'vowel kmeans': (2.34, 13.98)}
('whole', 'row', 'i', '15+') curve ranges: {'consonant groups a': (-0.13, 3.16), 'consonant groups b': (-0.16, 2.71), 'consonant four a': (-0.22, 3.0), 'consonant four b': (-0.22, 2.61), 'consonant kmeans': (1.81, 4.07), 'vowel groups': (-0.01, 0.87), 'vowel kmeans': (1.81, 4.07)}
('whole', 'row', 'ii', 'all') curve ranges: {'consonant groups a': (-0.12, 1.04), 'consonant groups b': (-0.09, 0.87), 'consonant four a': (-0.06, 1.13), 'consonant four b': (-0.05, 0.96), 'consonant kmeans': (0.56, 2.86), 'vowel groups': (-0.02, 0.36), 'vowel kmeans': (0.56, 2.86)}
('whole', 'row', 'ii', '15+') curve ranges: {'consonant groups a': (-0.03, 1.13), 'consonant groups b': (-0.03, 0.94), 'consonant four a': (-0.03, 1.17), 'consonant four b': (-0.02, 1.0), 'consonant kmeans': (0.41, 1.26), 'vowel groups': (-0.01, 0.37), 'vowel kmeans': (0.41, 1.26)}
('whole', 'time', 'i', 'all') curve ranges: {'consonant groups a': (-0.53, 1.76), 'consonant groups b': (-0.44, 1.57), 'consonant four a': (-0.32, 2.04), 'consonant four b': (-0.29, 1.85), 'consonant kmeans': (2.74, 12.29), 'vowel groups': (-0.04, 1.3), 'vowel kmeans': (2.74, 12.29)}
('whole', 'time', 'i', '15+') curve ranges: {'consonant groups a': (-0.12, 2.19), 'consonant groups b': (-0.14, 1.91), 'consonant four a': (-0.19, 2.36), 'consonant four b': (-0.15, 2.13), 'consonant kmeans': (2.19, 4.31), 'vowel groups': (-0.01, 1.42), 'vowel kmeans': (2.19, 4.31)}
('whole', 'time', 'ii', 'all') curve ranges: {'consonant groups a': (-0.11, 1.05), 'consonant groups b': (-0.09, 0.89), 'consonant four a': (-0.06, 1.11), 'consonant four b': (-0.05, 0.95), 'consonant kmeans': (0.56, 2.88), 'vowel groups': (-0.02, 0.42), 'vowel kmeans': (0.56, 2.88)}
('whole', 'time', 'ii', '15+') curve ranges: {'consonant groups a': (-0.03, 1.13), 'consonant groups b': (-0.03, 0.95), 'consonant four a': (-0.03, 1.16), 'consonant four b': (-0.02, 1.0), 'consonant kmeans': (0.4, 1.22), 'vowel groups': (-0.01, 0.43), 'vowel kmeans': (0.4, 1.22)}
('rest', 'row', 'i', 'all') curve ranges: {'consonant groups a': (-0.48, 2.53), 'consonant groups b': (-0.44, 2.16), 'consonant four a': (-0.28, 2.63), 'consonant four b': (-0.28, 2.36), 'consonant kmeans': (3.24, 14.11), 'vowel groups': (-0.02, 0.77), 'vowel kmeans': (3.24, 14.11)}
('rest', 'row', 'i', '15+') curve ranges: {'consonant groups a': (-0.31, 3.03), 'consonant groups b': (-0.24, 2.56), 'consonant four a': (-0.23, 2.9), 'consonant four b': (-0.22, 2.66), 'consonant kmeans': (2.85, 11.31), 'vowel groups': (-0.01, 0.79), 'vowel kmeans': (2.85, 11.31)}
('rest', 'row', 'ii', 'all') curve ranges: {'consonant groups a': (-0.13, 1.19), 'consonant groups b': (-0.1, 1.0), 'consonant four a': (-0.07, 1.3), 'consonant four b': (-0.07, 1.13), 'consonant kmeans': (0.94, 4.14), 'vowel groups': (-0.02, 0.41), 'vowel kmeans': (0.94, 4.14)}
('rest', 'row', 'ii', '15+') curve ranges: {'consonant groups a': (-0.06, 1.27), 'consonant groups b': (-0.05, 1.05), 'consonant four a': (-0.05, 1.35), 'consonant four b': (-0.06, 1.18), 'consonant kmeans': (0.82, 3.63), 'vowel groups': (-0.01, 0.42), 'vowel kmeans': (0.82, 3.63)}
('rest', 'time', 'i', 'all') curve ranges: {'consonant groups a': (-0.45, 1.74), 'consonant groups b': (-0.39, 1.53), 'consonant four a': (-0.25, 2.06), 'consonant four b': (-0.24, 2.04), 'consonant kmeans': (3.09, 13.65), 'vowel groups': (-0.02, 1.21), 'vowel kmeans': (3.09, 13.65)}
('rest', 'time', 'i', '15+') curve ranges: {'consonant groups a': (-0.27, 2.08), 'consonant groups b': (-0.17, 1.79), 'consonant four a': (-0.2, 2.23), 'consonant four b': (-0.17, 2.14), 'consonant kmeans': (2.48, 7.98), 'vowel groups': (-0.01, 1.3), 'vowel kmeans': (2.48, 7.98)}
('rest', 'time', 'ii', 'all') curve ranges: {'consonant groups a': (-0.12, 1.17), 'consonant groups b': (-0.1, 0.99), 'consonant four a': (-0.07, 1.32), 'consonant four b': (-0.07, 1.16), 'consonant kmeans': (0.95, 4.37), 'vowel groups': (-0.02, 0.51), 'vowel kmeans': (0.95, 4.37)}
('rest', 'time', 'ii', '15+') curve ranges: {'consonant groups a': (-0.06, 1.24), 'consonant groups b': (-0.05, 1.04), 'consonant four a': (-0.06, 1.36), 'consonant four b': (-0.06, 1.2), 'consonant kmeans': (0.88, 3.56), 'vowel groups': (-0.02, 0.52), 'vowel kmeans': (0.88, 3.56)}
Result: not reproduced. Under none of the 120 readings (24 baseline, z-scoring, projection and CV readings by five cluster readings) are both peaks within tolerance. With the phonetic groups the consonant separability peaks at -155 to +15 ms (paper -200 +- 20) and the vowel separability at 155 to 235 ms (paper 250 +- 25, within it in some readings); with k-means at each time one curve serves both, peaking anywhere from -170 to 375 ms. The consonant onsets (reading a, -375 to -180 ms) reach the paper's -300 +- 30 ms under some readings, and the vowel onsets (55 to 105 ms) are mostly later than its 50 +- 10 ms. The bootstrap no longer pairs a token with its own copies, which changes only the s.e. used by onset reading (a); no reading gives both peaks within tolerance under any reading factor (table of counts).
fig5d-separability-coupling¶
Cluster separability and the correlation of the state-space with phonetic features are tightly linked over time for consonants and vowels in each subject (R2 0.42 to 0.98, all P < 1e-10).
In [25]:
c5 = []
for (base, z, pj, cs), d in SEP.items():
R = RT[base, z, pj, cs]
for ck in SEP_KINDS["consonant"]:
vk = "kmeans" if ck == "kmeans" else "groups"
for rmode in ("(a) matching set", "(b) mean of three"):
vals = []
for i, s in enumerate(B.SUBJECTS):
for which, kind, fset in (("consonant", ck, "location"), ("vowel", vk, "vowel")):
r = R[s, fset] if rmode.startswith("(a)") else np.mean([R[s, f] for f in FEATS], 0)
res = stats.pearsonr(d[which, kind][2][i], r)
vals.append((res.statistic ** 2, res.pvalue, res.statistic))
r2 = np.array([v[0] for v in vals]); ps = np.array([v[1] for v in vals]); rs = np.array([v[2] for v in vals])
ok = bool((r2 >= 0.378).all() and (ps < 1e-10).all() and (rs > 0).all())
part = bool((rs > 0).all() and (ps < 1e-10).all())
c5.append({"baseline": base, "z-scoring": z, "projection": pj, "CVs": cs, "clusters": f"{ck} | {vk}", "R": rmode,
"R2 (EC2 c,v GP31 c,v GP33 c,v)": " ".join(f"{x:.2f}" for x in r2),
"max p": ps.max(), "min r": round(rs.min(), 2),
"outcome": "reproduced" if ok else ("partially" if part else "not reproduced")})
c5 = pd.DataFrame(c5)
print(c5.to_string(index=False))
print("\nOutcome over readings:", overall(c5.outcome))
by_factor(c5, ["baseline", "z-scoring", "projection", "CVs", "clusters", "R"])
print("Outcomes by reading of the clusters and of R:")
print(c5.groupby(["clusters", "R"]).outcome.value_counts().unstack(fill_value=0).to_string())
baseline z-scoring projection CVs clusters R R2 (EC2 c,v GP31 c,v GP33 c,v) max p min r outcome
pre row i all groups a | groups (a) matching set 0.90 0.95 0.69 0.97 0.57 0.88 8.557880e-48 0.76 reproduced
pre row i all groups a | groups (b) mean of three 0.18 0.29 0.83 0.31 0.12 0.13 2.801544e-08 -0.56 not reproduced
pre row i all groups b | groups (a) matching set 0.89 0.95 0.67 0.97 0.48 0.88 1.350045e-37 0.70 reproduced
pre row i all groups b | groups (b) mean of three 0.16 0.29 0.84 0.31 0.09 0.13 1.873684e-06 -0.56 not reproduced
pre row i all four a | groups (a) matching set 0.85 0.95 0.79 0.97 0.67 0.88 2.200816e-61 0.82 reproduced
pre row i all four a | groups (b) mean of three 0.12 0.29 0.81 0.31 0.09 0.13 1.769221e-06 -0.56 not reproduced
pre row i all four b | groups (a) matching set 0.84 0.95 0.73 0.97 0.59 0.88 1.428680e-49 0.77 reproduced
pre row i all four b | groups (b) mean of three 0.12 0.29 0.83 0.31 0.08 0.13 7.360150e-06 -0.56 not reproduced
pre row i all kmeans | kmeans (a) matching set 0.66 0.02 0.12 0.01 0.07 0.01 1.423802e-01 -0.34 not reproduced
pre row i all kmeans | kmeans (b) mean of three 0.31 0.31 0.00 0.00 0.06 0.06 3.588361e-01 -0.25 not reproduced
pre row i 15+ groups a | groups (a) matching set 0.93 0.95 0.51 0.85 0.57 0.88 1.757800e-40 0.71 reproduced
pre row i 15+ groups a | groups (b) mean of three 0.20 0.27 0.37 0.11 0.12 0.13 1.159030e-07 -0.33 not reproduced
pre row i 15+ groups b | groups (a) matching set 0.92 0.95 0.50 0.85 0.48 0.88 1.350045e-37 0.70 reproduced
pre row i 15+ groups b | groups (b) mean of three 0.19 0.27 0.38 0.11 0.09 0.13 1.873684e-06 -0.33 not reproduced
pre row i 15+ four a | groups (a) matching set 0.88 0.95 0.58 0.85 0.67 0.88 2.037010e-49 0.76 reproduced
pre row i 15+ four a | groups (b) mean of three 0.13 0.27 0.41 0.11 0.09 0.13 1.769221e-06 -0.33 not reproduced
pre row i 15+ four b | groups (a) matching set 0.87 0.95 0.52 0.85 0.59 0.88 5.230691e-42 0.72 reproduced
pre row i 15+ four b | groups (b) mean of three 0.14 0.27 0.39 0.11 0.08 0.13 7.360150e-06 -0.33 not reproduced
pre row i 15+ kmeans | kmeans (a) matching set 0.73 0.04 0.04 0.32 0.07 0.01 1.231100e-01 -0.56 not reproduced
pre row i 15+ kmeans | kmeans (b) mean of three 0.34 0.34 0.02 0.02 0.06 0.06 2.047243e-02 -0.25 not reproduced
pre row ii all groups a | groups (a) matching set 0.91 0.91 0.86 0.98 0.65 0.94 4.364097e-59 0.81 reproduced
pre row ii all groups a | groups (b) mean of three 0.02 0.59 0.89 0.48 0.02 0.37 2.163991e-02 -0.69 not reproduced
pre row ii all groups b | groups (a) matching set 0.90 0.91 0.86 0.98 0.66 0.94 6.388438e-60 0.81 reproduced
pre row ii all groups b | groups (b) mean of three 0.03 0.59 0.91 0.48 0.02 0.37 1.276696e-02 -0.69 not reproduced
pre row ii all four a | groups (a) matching set 0.88 0.91 0.88 0.98 0.74 0.94 1.230339e-74 0.86 reproduced
pre row ii all four a | groups (b) mean of three 0.01 0.59 0.88 0.48 0.04 0.37 1.680837e-01 -0.69 not reproduced
pre row ii all four b | groups (a) matching set 0.87 0.91 0.87 0.98 0.73 0.94 8.405368e-72 0.85 reproduced
pre row ii all four b | groups (b) mean of three 0.01 0.59 0.90 0.48 0.04 0.37 1.459390e-01 -0.69 not reproduced
pre row ii all kmeans | kmeans (a) matching set 0.81 0.06 0.02 0.02 0.08 0.07 1.754536e-02 -0.26 not reproduced
pre row ii all kmeans | kmeans (b) mean of three 0.12 0.12 0.00 0.00 0.01 0.01 5.605257e-01 -0.11 not reproduced
pre row ii 15+ groups a | groups (a) matching set 0.91 0.91 0.58 0.90 0.65 0.94 2.117134e-49 0.76 reproduced
pre row ii 15+ groups a | groups (b) mean of three 0.02 0.60 0.27 0.14 0.02 0.37 3.835496e-02 -0.37 not reproduced
pre row ii 15+ groups b | groups (a) matching set 0.90 0.91 0.61 0.90 0.66 0.94 2.740828e-53 0.78 reproduced
pre row ii 15+ groups b | groups (b) mean of three 0.02 0.60 0.30 0.14 0.02 0.37 2.315223e-02 -0.37 not reproduced
pre row ii 15+ four a | groups (a) matching set 0.88 0.91 0.62 0.90 0.74 0.94 2.572838e-54 0.79 reproduced
pre row ii 15+ four a | groups (b) mean of three 0.00 0.60 0.30 0.14 0.04 0.37 3.599467e-01 -0.37 not reproduced
pre row ii 15+ four b | groups (a) matching set 0.88 0.91 0.62 0.90 0.73 0.94 7.186415e-55 0.79 reproduced
pre row ii 15+ four b | groups (b) mean of three 0.00 0.60 0.31 0.14 0.04 0.37 3.074522e-01 -0.37 not reproduced
pre row ii 15+ kmeans | kmeans (a) matching set 0.80 0.07 0.29 0.47 0.08 0.07 3.217501e-05 -0.69 not reproduced
pre row ii 15+ kmeans | kmeans (b) mean of three 0.11 0.11 0.08 0.08 0.01 0.01 8.226222e-02 -0.11 not reproduced
pre time i all groups a | groups (a) matching set 0.92 0.97 0.67 0.97 0.45 0.92 8.814722e-34 0.67 reproduced
pre time i all groups a | groups (b) mean of three 0.00 0.64 0.72 0.19 0.02 0.29 3.573545e-01 -0.44 not reproduced
pre time i all groups b | groups (a) matching set 0.92 0.97 0.65 0.97 0.40 0.92 4.407971e-29 0.63 reproduced
pre time i all groups b | groups (b) mean of three 0.00 0.64 0.72 0.19 0.02 0.29 5.124998e-01 -0.44 not reproduced
pre time i all four a | groups (a) matching set 0.78 0.97 0.72 0.97 0.59 0.92 9.214609e-50 0.77 reproduced
pre time i all four a | groups (b) mean of three 0.04 0.64 0.73 0.19 0.02 0.29 4.232537e-02 -0.44 not reproduced
pre time i all four b | groups (a) matching set 0.78 0.97 0.68 0.97 0.52 0.92 2.242277e-41 0.72 reproduced
pre time i all four b | groups (b) mean of three 0.03 0.64 0.72 0.19 0.02 0.29 3.959789e-02 -0.44 not reproduced
pre time i all kmeans | kmeans (a) matching set 0.25 0.01 0.20 0.03 0.15 0.01 2.145276e-01 -0.45 not reproduced
pre time i all kmeans | kmeans (b) mean of three 0.14 0.14 0.07 0.07 0.07 0.07 2.471032e-05 -0.27 not reproduced
pre time i 15+ groups a | groups (a) matching set 0.95 0.98 0.49 0.91 0.45 0.92 8.814722e-34 0.67 reproduced
pre time i 15+ groups a | groups (b) mean of three 0.00 0.64 0.36 0.08 0.02 0.29 8.787047e-01 -0.29 not reproduced
pre time i 15+ groups b | groups (a) matching set 0.94 0.98 0.48 0.91 0.40 0.92 4.407971e-29 0.63 reproduced
pre time i 15+ groups b | groups (b) mean of three 0.00 0.64 0.34 0.08 0.02 0.29 9.157462e-01 -0.29 not reproduced
pre time i 15+ four a | groups (a) matching set 0.80 0.98 0.53 0.91 0.59 0.92 1.002650e-42 0.73 reproduced
pre time i 15+ four a | groups (b) mean of three 0.03 0.64 0.36 0.08 0.02 0.29 4.232537e-02 -0.29 not reproduced
pre time i 15+ four b | groups (a) matching set 0.79 0.98 0.46 0.91 0.52 0.92 3.803138e-35 0.68 reproduced
pre time i 15+ four b | groups (b) mean of three 0.02 0.64 0.30 0.08 0.02 0.29 4.329357e-02 -0.29 not reproduced
pre time i 15+ kmeans | kmeans (a) matching set 0.26 0.01 0.00 0.32 0.15 0.01 6.795617e-01 -0.57 not reproduced
pre time i 15+ kmeans | kmeans (b) mean of three 0.16 0.16 0.00 0.00 0.07 0.07 4.054647e-01 -0.26 not reproduced
pre time ii all groups a | groups (a) matching set 0.93 0.91 0.86 0.99 0.67 0.94 3.756906e-62 0.82 reproduced
pre time ii all groups a | groups (b) mean of three 0.01 0.65 0.89 0.48 0.04 0.34 1.247366e-01 -0.69 not reproduced
pre time ii all groups b | groups (a) matching set 0.92 0.91 0.86 0.99 0.68 0.94 2.465903e-63 0.82 reproduced
pre time ii all groups b | groups (b) mean of three 0.01 0.65 0.91 0.48 0.04 0.34 8.902795e-02 -0.69 not reproduced
pre time ii all four a | groups (a) matching set 0.91 0.91 0.88 0.99 0.76 0.94 4.797057e-78 0.87 reproduced
pre time ii all four a | groups (b) mean of three 0.00 0.65 0.88 0.48 0.05 0.34 7.438878e-01 -0.69 not reproduced
pre time ii all four b | groups (a) matching set 0.91 0.91 0.87 0.99 0.74 0.94 1.894696e-75 0.86 reproduced
pre time ii all four b | groups (b) mean of three 0.00 0.65 0.91 0.48 0.05 0.34 6.413088e-01 -0.69 not reproduced
pre time ii all kmeans | kmeans (a) matching set 0.72 0.02 0.03 0.02 0.05 0.07 3.285534e-02 -0.27 not reproduced
pre time ii all kmeans | kmeans (b) mean of three 0.16 0.16 0.00 0.00 0.01 0.01 3.960150e-01 -0.11 not reproduced
pre time ii 15+ groups a | groups (a) matching set 0.93 0.91 0.60 0.91 0.67 0.94 3.818786e-51 0.77 reproduced
pre time ii 15+ groups a | groups (b) mean of three 0.01 0.67 0.28 0.13 0.04 0.34 2.359537e-01 -0.36 not reproduced
pre time ii 15+ groups b | groups (a) matching set 0.92 0.91 0.63 0.91 0.68 0.94 3.704163e-55 0.79 reproduced
pre time ii 15+ groups b | groups (b) mean of three 0.01 0.67 0.30 0.13 0.04 0.34 1.782315e-01 -0.36 not reproduced
pre time ii 15+ four a | groups (a) matching set 0.90 0.91 0.63 0.91 0.76 0.94 5.111766e-56 0.80 reproduced
pre time ii 15+ four a | groups (b) mean of three 0.00 0.67 0.30 0.13 0.05 0.34 9.717886e-01 -0.36 not reproduced
pre time ii 15+ four b | groups (a) matching set 0.90 0.91 0.64 0.91 0.74 0.94 8.707229e-57 0.80 reproduced
pre time ii 15+ four b | groups (b) mean of three 0.00 0.67 0.31 0.13 0.05 0.34 8.936653e-01 -0.36 not reproduced
pre time ii 15+ kmeans | kmeans (a) matching set 0.70 0.01 0.31 0.47 0.05 0.07 8.133747e-02 -0.69 not reproduced
pre time ii 15+ kmeans | kmeans (b) mean of three 0.16 0.16 0.09 0.09 0.01 0.01 7.499047e-02 -0.11 not reproduced
whole row i all groups a | groups (a) matching set 0.89 0.94 0.87 0.97 0.56 0.87 9.930735e-47 0.75 reproduced
whole row i all groups a | groups (b) mean of three 0.17 0.27 0.82 0.30 0.14 0.12 2.172713e-08 -0.55 not reproduced
whole row i all groups b | groups (a) matching set 0.87 0.94 0.86 0.97 0.48 0.87 1.024428e-36 0.69 reproduced
whole row i all groups b | groups (b) mean of three 0.16 0.27 0.85 0.30 0.10 0.12 2.374137e-07 -0.55 not reproduced
whole row i all four a | groups (a) matching set 0.84 0.94 0.87 0.97 0.67 0.87 4.037112e-61 0.82 reproduced
whole row i all four a | groups (b) mean of three 0.12 0.27 0.82 0.30 0.10 0.12 2.758712e-07 -0.55 not reproduced
whole row i all four b | groups (a) matching set 0.83 0.94 0.84 0.97 0.58 0.87 4.945629e-49 0.76 reproduced
whole row i all four b | groups (b) mean of three 0.12 0.27 0.85 0.30 0.09 0.12 2.074469e-06 -0.55 not reproduced
whole row i all kmeans | kmeans (a) matching set 0.65 0.02 0.09 0.06 0.07 0.01 6.554510e-02 -0.29 not reproduced
whole row i all kmeans | kmeans (b) mean of three 0.30 0.30 0.05 0.05 0.08 0.08 5.225319e-04 -0.29 not reproduced
whole row i 15+ groups a | groups (a) matching set 0.92 0.95 0.62 0.88 0.56 0.87 9.930735e-47 0.75 reproduced
whole row i 15+ groups a | groups (b) mean of three 0.20 0.26 0.45 0.13 0.14 0.12 2.172713e-08 -0.36 not reproduced
whole row i 15+ groups b | groups (a) matching set 0.91 0.95 0.61 0.88 0.48 0.87 1.024428e-36 0.69 reproduced
whole row i 15+ groups b | groups (b) mean of three 0.19 0.26 0.47 0.13 0.10 0.12 2.374137e-07 -0.36 not reproduced
whole row i 15+ four a | groups (a) matching set 0.87 0.95 0.60 0.88 0.67 0.87 9.137639e-52 0.78 reproduced
whole row i 15+ four a | groups (b) mean of three 0.14 0.26 0.44 0.13 0.10 0.12 2.758712e-07 -0.36 not reproduced
whole row i 15+ four b | groups (a) matching set 0.86 0.95 0.56 0.88 0.58 0.87 7.965225e-46 0.75 reproduced
whole row i 15+ four b | groups (b) mean of three 0.14 0.26 0.44 0.13 0.09 0.12 2.074469e-06 -0.36 not reproduced
whole row i 15+ kmeans | kmeans (a) matching set 0.74 0.04 0.16 0.36 0.07 0.01 6.554510e-02 -0.60 not reproduced
whole row i 15+ kmeans | kmeans (b) mean of three 0.34 0.34 0.08 0.08 0.08 0.08 6.258169e-06 -0.29 not reproduced
whole row ii all groups a | groups (a) matching set 0.90 0.91 0.92 0.98 0.65 0.94 1.501843e-58 0.81 reproduced
whole row ii all groups a | groups (b) mean of three 0.03 0.57 0.86 0.45 0.04 0.32 9.793249e-03 -0.67 not reproduced
whole row ii all groups b | groups (a) matching set 0.90 0.91 0.93 0.98 0.65 0.94 1.646605e-58 0.81 reproduced
whole row ii all groups b | groups (b) mean of three 0.03 0.57 0.88 0.45 0.04 0.32 6.432049e-03 -0.67 not reproduced
whole row ii all four a | groups (a) matching set 0.87 0.91 0.92 0.98 0.74 0.94 2.599212e-75 0.86 reproduced
whole row ii all four a | groups (b) mean of three 0.01 0.57 0.86 0.45 0.06 0.32 1.312153e-01 -0.67 not reproduced
whole row ii all four b | groups (a) matching set 0.87 0.91 0.92 0.98 0.72 0.94 1.168611e-71 0.85 reproduced
whole row ii all four b | groups (b) mean of three 0.01 0.57 0.89 0.45 0.06 0.32 1.146401e-01 -0.67 not reproduced
whole row ii all kmeans | kmeans (a) matching set 0.81 0.07 0.02 0.04 0.04 0.06 1.458139e-02 -0.27 not reproduced
whole row ii all kmeans | kmeans (b) mean of three 0.12 0.12 0.01 0.01 0.01 0.01 2.212496e-01 -0.11 not reproduced
whole row ii 15+ groups a | groups (a) matching set 0.91 0.91 0.68 0.91 0.65 0.94 1.501843e-58 0.81 reproduced
whole row ii 15+ groups a | groups (b) mean of three 0.02 0.59 0.35 0.17 0.04 0.32 2.686008e-02 -0.41 not reproduced
whole row ii 15+ groups b | groups (a) matching set 0.90 0.91 0.71 0.91 0.65 0.94 1.646605e-58 0.81 reproduced
whole row ii 15+ groups b | groups (b) mean of three 0.02 0.59 0.40 0.17 0.04 0.32 1.726786e-02 -0.41 not reproduced
whole row ii 15+ four a | groups (a) matching set 0.88 0.91 0.68 0.91 0.74 0.94 1.264664e-63 0.82 reproduced
whole row ii 15+ four a | groups (b) mean of three 0.00 0.59 0.37 0.17 0.06 0.32 3.017194e-01 -0.41 not reproduced
whole row ii 15+ four b | groups (a) matching set 0.88 0.91 0.70 0.91 0.72 0.94 2.141893e-66 0.83 reproduced
whole row ii 15+ four b | groups (b) mean of three 0.01 0.59 0.40 0.17 0.06 0.32 2.571645e-01 -0.41 not reproduced
whole row ii 15+ kmeans | kmeans (a) matching set 0.81 0.07 0.47 0.48 0.04 0.06 1.207573e-03 -0.69 not reproduced
whole row ii 15+ kmeans | kmeans (b) mean of three 0.11 0.11 0.19 0.19 0.01 0.01 7.273542e-02 -0.11 not reproduced
whole time i all groups a | groups (a) matching set 0.91 0.97 0.75 0.98 0.45 0.91 2.275136e-34 0.67 reproduced
whole time i all groups a | groups (b) mean of three 0.00 0.64 0.73 0.17 0.04 0.26 2.853585e-01 -0.41 not reproduced
whole time i all groups b | groups (a) matching set 0.91 0.97 0.73 0.98 0.40 0.91 4.645786e-29 0.63 reproduced
whole time i all groups b | groups (b) mean of three 0.00 0.64 0.75 0.17 0.03 0.26 3.914154e-01 -0.41 not reproduced
whole time i all four a | groups (a) matching set 0.78 0.97 0.74 0.98 0.59 0.91 2.295098e-50 0.77 reproduced
whole time i all four a | groups (b) mean of three 0.04 0.64 0.75 0.17 0.03 0.26 9.102167e-03 -0.41 not reproduced
whole time i all four b | groups (a) matching set 0.78 0.97 0.70 0.98 0.52 0.91 1.865308e-41 0.72 reproduced
whole time i all four b | groups (b) mean of three 0.03 0.64 0.74 0.17 0.03 0.26 1.197155e-02 -0.41 not reproduced
whole time i all kmeans | kmeans (a) matching set 0.26 0.01 0.15 0.05 0.11 0.01 2.030027e-01 -0.39 not reproduced
whole time i all kmeans | kmeans (b) mean of three 0.14 0.14 0.08 0.08 0.05 0.05 2.083728e-04 -0.28 not reproduced
whole time i 15+ groups a | groups (a) matching set 0.94 0.98 0.55 0.92 0.45 0.91 2.275136e-34 0.67 reproduced
whole time i 15+ groups a | groups (b) mean of three 0.00 0.64 0.40 0.07 0.04 0.26 8.694233e-01 -0.26 not reproduced
whole time i 15+ groups b | groups (a) matching set 0.94 0.98 0.53 0.92 0.40 0.91 4.645786e-29 0.63 reproduced
whole time i 15+ groups b | groups (b) mean of three 0.00 0.64 0.38 0.07 0.03 0.26 9.740930e-01 -0.26 not reproduced
whole time i 15+ four a | groups (a) matching set 0.80 0.98 0.51 0.92 0.59 0.91 5.534198e-41 0.72 reproduced
whole time i 15+ four a | groups (b) mean of three 0.02 0.64 0.36 0.07 0.03 0.26 1.269556e-02 -0.26 not reproduced
whole time i 15+ four b | groups (a) matching set 0.79 0.98 0.45 0.92 0.52 0.91 7.923060e-34 0.67 reproduced
whole time i 15+ four b | groups (b) mean of three 0.02 0.64 0.31 0.07 0.03 0.26 5.135526e-02 -0.26 not reproduced
whole time i 15+ kmeans | kmeans (a) matching set 0.28 0.02 0.00 0.33 0.11 0.01 2.946195e-01 -0.57 not reproduced
whole time i 15+ kmeans | kmeans (b) mean of three 0.23 0.23 0.00 0.00 0.05 0.05 9.145159e-01 -0.23 not reproduced
whole time ii all groups a | groups (a) matching set 0.93 0.91 0.93 0.99 0.67 0.94 1.621938e-61 0.82 reproduced
whole time ii all groups a | groups (b) mean of three 0.01 0.63 0.86 0.45 0.06 0.29 1.166388e-01 -0.67 not reproduced
whole time ii all groups b | groups (a) matching set 0.92 0.91 0.93 0.99 0.67 0.94 1.164756e-61 0.82 reproduced
whole time ii all groups b | groups (b) mean of three 0.01 0.63 0.89 0.45 0.06 0.29 8.983255e-02 -0.67 not reproduced
whole time ii all four a | groups (a) matching set 0.91 0.91 0.92 0.99 0.76 0.94 1.177567e-78 0.87 reproduced
whole time ii all four a | groups (b) mean of three 0.00 0.63 0.86 0.45 0.08 0.29 6.795010e-01 -0.67 not reproduced
whole time ii all four b | groups (a) matching set 0.91 0.91 0.92 0.99 0.74 0.94 4.557709e-75 0.86 reproduced
whole time ii all four b | groups (b) mean of three 0.00 0.63 0.90 0.45 0.07 0.29 5.904175e-01 -0.67 not reproduced
whole time ii all kmeans | kmeans (a) matching set 0.73 0.02 0.02 0.04 0.03 0.06 2.452778e-02 -0.24 not reproduced
whole time ii all kmeans | kmeans (b) mean of three 0.16 0.16 0.01 0.01 0.01 0.01 2.075319e-01 -0.12 not reproduced
whole time ii 15+ groups a | groups (a) matching set 0.93 0.91 0.69 0.92 0.67 0.94 1.621938e-61 0.82 reproduced
whole time ii 15+ groups a | groups (b) mean of three 0.01 0.66 0.36 0.16 0.06 0.29 2.144277e-01 -0.40 not reproduced
whole time ii 15+ groups b | groups (a) matching set 0.92 0.91 0.72 0.92 0.67 0.94 1.164756e-61 0.82 reproduced
whole time ii 15+ groups b | groups (b) mean of three 0.01 0.66 0.40 0.16 0.06 0.29 1.717178e-01 -0.40 not reproduced
whole time ii 15+ four a | groups (a) matching set 0.90 0.91 0.70 0.92 0.76 0.94 2.342622e-66 0.83 reproduced
whole time ii 15+ four a | groups (b) mean of three 0.00 0.66 0.38 0.16 0.08 0.29 9.741504e-01 -0.40 not reproduced
whole time ii 15+ four b | groups (a) matching set 0.90 0.91 0.71 0.92 0.74 0.94 3.103412e-69 0.84 reproduced
whole time ii 15+ four b | groups (b) mean of three 0.00 0.66 0.41 0.16 0.07 0.29 8.482887e-01 -0.40 not reproduced
whole time ii 15+ kmeans | kmeans (a) matching set 0.70 0.01 0.49 0.48 0.03 0.06 8.021031e-02 -0.70 not reproduced
whole time ii 15+ kmeans | kmeans (b) mean of three 0.16 0.16 0.20 0.20 0.01 0.01 5.312232e-02 -0.12 not reproduced
rest row i all groups a | groups (a) matching set 0.88 0.94 0.42 0.70 0.55 0.82 2.133976e-31 0.65 reproduced
rest row i all groups a | groups (b) mean of three 0.20 0.26 0.38 0.19 0.18 0.10 2.424806e-07 -0.44 not reproduced
rest row i all groups b | groups (a) matching set 0.86 0.94 0.30 0.70 0.47 0.82 5.445767e-21 0.55 partially
rest row i all groups b | groups (b) mean of three 0.18 0.26 0.35 0.19 0.13 0.10 2.424806e-07 -0.44 not reproduced
rest row i all four a | groups (a) matching set 0.82 0.94 0.27 0.70 0.65 0.82 8.925815e-19 0.52 partially
rest row i all four a | groups (b) mean of three 0.14 0.26 0.28 0.19 0.12 0.10 2.424806e-07 -0.44 not reproduced
rest row i all four b | groups (a) matching set 0.81 0.94 0.14 0.70 0.57 0.82 9.019966e-10 0.37 not reproduced
rest row i all four b | groups (b) mean of three 0.14 0.26 0.21 0.19 0.10 0.10 3.396701e-07 -0.44 not reproduced
rest row i all kmeans | kmeans (a) matching set 0.59 0.01 0.29 0.06 0.13 0.00 2.895399e-01 -0.53 not reproduced
rest row i all kmeans | kmeans (b) mean of three 0.31 0.31 0.22 0.22 0.08 0.08 6.103611e-06 -0.46 not reproduced
rest row i 15+ groups a | groups (a) matching set 0.91 0.95 0.63 0.73 0.55 0.82 4.888179e-45 0.74 reproduced
rest row i 15+ groups a | groups (b) mean of three 0.20 0.26 0.51 0.23 0.18 0.10 2.424806e-07 -0.48 not reproduced
rest row i 15+ groups b | groups (a) matching set 0.89 0.95 0.49 0.73 0.47 0.82 6.667915e-36 0.68 reproduced
rest row i 15+ groups b | groups (b) mean of three 0.20 0.26 0.44 0.23 0.13 0.10 2.424806e-07 -0.48 not reproduced
rest row i 15+ four a | groups (a) matching set 0.83 0.95 0.49 0.73 0.65 0.82 3.743932e-38 0.70 reproduced
rest row i 15+ four a | groups (b) mean of three 0.14 0.26 0.42 0.23 0.12 0.10 2.424806e-07 -0.48 not reproduced
rest row i 15+ four b | groups (a) matching set 0.82 0.95 0.33 0.73 0.57 0.82 3.621067e-23 0.57 partially
rest row i 15+ four b | groups (b) mean of three 0.14 0.26 0.31 0.23 0.10 0.10 3.396701e-07 -0.48 not reproduced
rest row i 15+ kmeans | kmeans (a) matching set 0.69 0.03 0.18 0.00 0.13 0.00 6.365091e-01 -0.42 not reproduced
rest row i 15+ kmeans | kmeans (b) mean of three 0.35 0.35 0.15 0.15 0.08 0.08 6.103611e-06 -0.39 not reproduced
rest row ii all groups a | groups (a) matching set 0.91 0.91 0.48 0.72 0.65 0.92 1.641750e-37 0.69 reproduced
rest row ii all groups a | groups (b) mean of three 0.02 0.63 0.36 0.18 0.08 0.27 4.289733e-02 -0.42 not reproduced
rest row ii all groups b | groups (a) matching set 0.91 0.91 0.37 0.72 0.65 0.92 1.370849e-26 0.61 partially
rest row ii all groups b | groups (b) mean of three 0.02 0.63 0.32 0.18 0.08 0.27 2.176771e-02 -0.42 not reproduced
rest row ii all four a | groups (a) matching set 0.88 0.91 0.28 0.72 0.74 0.92 1.801708e-19 0.53 partially
rest row ii all four a | groups (b) mean of three 0.00 0.63 0.24 0.18 0.09 0.27 2.970818e-01 -0.42 not reproduced
rest row ii all four b | groups (a) matching set 0.88 0.91 0.16 0.72 0.72 0.92 5.747181e-11 0.40 partially
rest row ii all four b | groups (b) mean of three 0.01 0.63 0.17 0.18 0.09 0.27 2.617314e-01 -0.42 not reproduced
rest row ii all kmeans | kmeans (a) matching set 0.80 0.05 0.26 0.13 0.01 0.06 2.359165e-01 -0.51 not reproduced
rest row ii all kmeans | kmeans (b) mean of three 0.12 0.12 0.18 0.18 0.02 0.02 1.517899e-02 -0.43 not reproduced
rest row ii 15+ groups a | groups (a) matching set 0.90 0.90 0.65 0.71 0.65 0.92 1.708128e-58 0.81 reproduced
rest row ii 15+ groups a | groups (b) mean of three 0.01 0.65 0.44 0.20 0.08 0.27 2.355117e-01 -0.44 not reproduced
rest row ii 15+ groups b | groups (a) matching set 0.90 0.90 0.52 0.71 0.65 0.92 1.349821e-41 0.72 reproduced
rest row ii 15+ groups b | groups (b) mean of three 0.01 0.65 0.38 0.20 0.08 0.27 1.351648e-01 -0.44 not reproduced
rest row ii 15+ four a | groups (a) matching set 0.86 0.90 0.45 0.71 0.74 0.92 5.407298e-34 0.67 reproduced
rest row ii 15+ four a | groups (b) mean of three 0.00 0.65 0.32 0.20 0.09 0.27 9.881077e-01 -0.44 not reproduced
rest row ii 15+ four b | groups (a) matching set 0.86 0.90 0.31 0.71 0.72 0.92 1.040219e-21 0.56 partially
rest row ii 15+ four b | groups (b) mean of three 0.00 0.65 0.23 0.20 0.09 0.27 9.072754e-01 -0.44 not reproduced
rest row ii 15+ kmeans | kmeans (a) matching set 0.80 0.06 0.13 0.01 0.01 0.06 2.359165e-01 -0.36 not reproduced
rest row ii 15+ kmeans | kmeans (b) mean of three 0.08 0.08 0.11 0.11 0.02 0.02 1.517899e-02 -0.34 not reproduced
rest time i all groups a | groups (a) matching set 0.89 0.97 0.42 0.89 0.46 0.89 1.450655e-31 0.65 reproduced
rest time i all groups a | groups (b) mean of three 0.00 0.63 0.36 0.23 0.06 0.23 4.883826e-01 -0.48 not reproduced
rest time i all groups b | groups (a) matching set 0.89 0.97 0.31 0.89 0.40 0.89 5.519352e-22 0.56 partially
rest time i all groups b | groups (b) mean of three 0.00 0.63 0.30 0.23 0.05 0.23 6.321657e-01 -0.48 not reproduced
rest time i all four a | groups (a) matching set 0.74 0.97 0.28 0.89 0.57 0.89 1.225408e-19 0.53 partially
rest time i all four a | groups (b) mean of three 0.03 0.63 0.27 0.23 0.04 0.23 8.108904e-03 -0.48 not reproduced
rest time i all four b | groups (a) matching set 0.73 0.97 0.18 0.89 0.49 0.89 1.903761e-12 0.43 partially
rest time i all four b | groups (b) mean of three 0.02 0.63 0.21 0.23 0.03 0.23 2.544771e-02 -0.48 not reproduced
rest time i all kmeans | kmeans (a) matching set 0.26 0.01 0.15 0.10 0.17 0.01 1.828130e-01 -0.41 not reproduced
rest time i all kmeans | kmeans (b) mean of three 0.14 0.14 0.16 0.16 0.09 0.09 1.097735e-06 -0.40 not reproduced
rest time i 15+ groups a | groups (a) matching set 0.93 0.97 0.52 0.90 0.46 0.89 3.594869e-35 0.68 reproduced
rest time i 15+ groups a | groups (b) mean of three 0.00 0.63 0.35 0.19 0.06 0.23 8.856142e-01 -0.44 not reproduced
rest time i 15+ groups b | groups (a) matching set 0.92 0.97 0.39 0.90 0.40 0.89 7.883490e-29 0.63 reproduced
rest time i 15+ groups b | groups (b) mean of three 0.00 0.63 0.27 0.19 0.05 0.23 7.093804e-01 -0.44 not reproduced
rest time i 15+ four a | groups (a) matching set 0.77 0.97 0.40 0.90 0.57 0.89 1.670712e-29 0.63 reproduced
rest time i 15+ four a | groups (b) mean of three 0.02 0.63 0.26 0.19 0.04 0.23 3.629098e-02 -0.44 not reproduced
rest time i 15+ four b | groups (a) matching set 0.76 0.97 0.28 0.90 0.49 0.89 1.776402e-19 0.53 partially
rest time i 15+ four b | groups (b) mean of three 0.01 0.63 0.19 0.19 0.03 0.23 1.175378e-01 -0.44 not reproduced
rest time i 15+ kmeans | kmeans (a) matching set 0.26 0.02 0.15 0.01 0.17 0.01 2.170003e-01 -0.41 not reproduced
rest time i 15+ kmeans | kmeans (b) mean of three 0.20 0.20 0.17 0.17 0.09 0.09 1.097735e-06 -0.41 not reproduced
rest time ii all groups a | groups (a) matching set 0.94 0.91 0.49 0.71 0.67 0.92 4.199108e-38 0.70 reproduced
rest time ii all groups a | groups (b) mean of three 0.00 0.71 0.37 0.18 0.10 0.24 5.469896e-01 -0.43 not reproduced
rest time ii all groups b | groups (a) matching set 0.93 0.91 0.37 0.71 0.67 0.92 4.244516e-27 0.61 partially
rest time ii all groups b | groups (b) mean of three 0.00 0.71 0.33 0.18 0.10 0.24 3.958366e-01 -0.43 not reproduced
rest time ii all four a | groups (a) matching set 0.92 0.91 0.29 0.71 0.76 0.92 4.376317e-20 0.54 partially
rest time ii all four a | groups (b) mean of three 0.00 0.71 0.25 0.18 0.12 0.24 6.499623e-01 -0.43 not reproduced
rest time ii all four b | groups (a) matching set 0.92 0.91 0.17 0.71 0.74 0.92 2.070404e-11 0.41 partially
rest time ii all four b | groups (b) mean of three 0.00 0.71 0.17 0.18 0.11 0.24 7.481806e-01 -0.43 not reproduced
rest time ii all kmeans | kmeans (a) matching set 0.70 0.00 0.26 0.15 0.00 0.05 3.184273e-01 -0.51 not reproduced
rest time ii all kmeans | kmeans (b) mean of three 0.17 0.17 0.19 0.19 0.02 0.02 1.686258e-02 -0.43 not reproduced
rest time ii 15+ groups a | groups (a) matching set 0.93 0.89 0.64 0.71 0.67 0.92 2.402508e-57 0.80 reproduced
rest time ii 15+ groups a | groups (b) mean of three 0.00 0.71 0.43 0.19 0.10 0.24 8.321848e-01 -0.43 not reproduced
rest time ii 15+ groups b | groups (a) matching set 0.92 0.89 0.51 0.71 0.67 0.92 5.433325e-41 0.72 reproduced
rest time ii 15+ groups b | groups (b) mean of three 0.00 0.71 0.38 0.19 0.10 0.24 6.531606e-01 -0.43 not reproduced
rest time ii 15+ four a | groups (a) matching set 0.91 0.89 0.44 0.71 0.76 0.92 1.461749e-33 0.67 reproduced
rest time ii 15+ four a | groups (b) mean of three 0.00 0.71 0.32 0.19 0.12 0.24 3.754385e-01 -0.43 not reproduced
rest time ii 15+ four b | groups (a) matching set 0.91 0.89 0.31 0.71 0.74 0.92 1.914441e-21 0.55 partially
rest time ii 15+ four b | groups (b) mean of three 0.00 0.71 0.23 0.19 0.11 0.24 4.641456e-01 -0.43 not reproduced
rest time ii 15+ kmeans | kmeans (a) matching set 0.68 0.00 0.14 0.00 0.00 0.05 3.409214e-01 -0.37 not reproduced
rest time ii 15+ kmeans | kmeans (b) mean of three 0.15 0.15 0.13 0.13 0.02 0.02 1.686258e-02 -0.36 not reproduced
Outcome over readings: differ by reading: not reproduced 145, partially 15, reproduced 80
Outcome counts by reading factor (240 readings):
baseline: {'pre': {'not reproduced': 48, 'reproduced': 32}, 'whole': {'not reproduced': 48, 'reproduced': 32}, 'rest': {'not reproduced': 49, 'reproduced': 16, 'partially': 15}}
z-scoring: {'row': {'not reproduced': 73, 'reproduced': 40, 'partially': 7}, 'time': {'not reproduced': 72, 'reproduced': 40, 'partially': 8}}
projection: {'i': {'not reproduced': 73, 'reproduced': 40, 'partially': 7}, 'ii': {'not reproduced': 72, 'reproduced': 40, 'partially': 8}}
CVs: {'all': {'not reproduced': 73, 'reproduced': 36, 'partially': 11}, '15+': {'not reproduced': 72, 'reproduced': 44, 'partially': 4}}
clusters: {'groups a | groups': {'reproduced': 24, 'not reproduced': 24}, 'groups b | groups': {'not reproduced': 24, 'reproduced': 20, 'partially': 4}, 'four a | groups': {'not reproduced': 24, 'reproduced': 20, 'partially': 4}, 'four b | groups': {'not reproduced': 25, 'reproduced': 16, 'partially': 7}, 'kmeans | kmeans': {'not reproduced': 48}}
R: {'(a) matching set': {'reproduced': 80, 'not reproduced': 25, 'partially': 15}, '(b) mean of three': {'not reproduced': 120}}
Outcomes by reading of the clusters and of R:
outcome not reproduced partially reproduced
clusters R
four a | groups (a) matching set 0 4 20
(b) mean of three 24 0 0
four b | groups (a) matching set 1 7 16
(b) mean of three 24 0 0
groups a | groups (a) matching set 0 0 24
(b) mean of three 24 0 0
groups b | groups (a) matching set 0 4 20
(b) mean of three 24 0 0
kmeans | kmeans (a) matching set 24 0 0
(b) mean of three 24 0 0
Result: blocked (methods_underspecified). With the phonetic groups as clusters and the matching feature set for R (constriction location for consonants, vowel features for vowels), the six R2 values lie within 0.38 to 1 with p < 1e-10 in 80 of 96 readings, are positive and significant with some below 0.38 in 15, and fail in 1 (R2 0.14 to 0.99 over these readings). With the mean of the three feature sets, or with k-means clusters at each time, the coupling fails in every reading. The readings of the claim give different outcomes, and two of them decide the split: the reading of R (with the mean of three feature sets all 120 readings fail) and the reading of the clusters (with k-means clusters all 48 readings fail, the 24 with the matching feature set among them). Every level of the baseline, z-scoring, projection and CV set has both reproduced and not reproduced readings.
suppfig13-consonant-vowel-distance¶
Consonant and vowel representations occupy distinct regions of the state-space; consonant-vowel distances are larger than consonant-consonant and vowel-vowel distances, and consonant-consonant distances are larger than vowel-vowel ones.
In [26]:
cvd = []
for base, z, pj, cs in SS_READINGS:
for reading in ("(a) mean of cross pairings", "(b) i at -25, j at +250"):
CC, VV, CVd = [], [], []
for s in B.SUBJECTS:
K = KS[s, base, z, pj, cs]
Kc, Kv = K[:, :, tc], K[:, :, tv]
i, j = np.triu_indices(len(Kc), 1)
CC.append(np.linalg.norm(Kc[i] - Kc[j], axis=1))
VV.append(np.linalg.norm(Kv[i] - Kv[j], axis=1))
d1 = np.linalg.norm(Kc[i] - Kv[j], axis=1)
d2 = np.linalg.norm(Kc[j] - Kv[i], axis=1)
CVd.append((d1 + d2) / 2 if reading.startswith("(a)") else d1)
CC, VV, CVd = map(np.concatenate, (CC, VV, CVd))
row = {"baseline": base, "z-scoring": z, "projection": pj, "CVs": cs, "CV distance": reading, "pairs": len(CC),
"median CC": round(np.median(CC), 2), "median VV": round(np.median(VV), 2), "median CV": round(np.median(CVd), 2)}
ok = True
for name, (a, b) in {"CV>CC": (CVd, CC), "CV>VV": (CVd, VV), "CC>VV": (CC, VV)}.items():
p = min(1.0, 3 * stats.wilcoxon(a, b).pvalue)
row[name] = f"{np.median(a - b):+.2f} p={p:.0e}"
ok &= bool(np.median(a - b) > 0 and p < 1e-10)
row["orderings hold"] = ok
# The criterion defines only reproduced. A reading where the orderings hold and only the pair
# count differs is counted as partially reproduced (the effect holds, the data differ), by the
# protocol's general definition; one where an ordering fails, as not reproduced.
row["outcome"] = ("reproduced" if ok and len(CC) == 4623 else "partially" if ok else "not reproduced")
cvd.append(row)
cvd = pd.DataFrame(cvd)
print(cvd.to_string(index=False))
print("Pairs within subjects (all CVs):", {s: len(CVS[s, 'all']) * (len(CVS[s, 'all']) - 1) // 2 for s in B.SUBJECTS},
"| 15+ CVs:", {s: len(CVS[s, '15+']) * (len(CVS[s, '15+']) - 1) // 2 for s in B.SUBJECTS}, "(paper n = 4623)")
print("Outcome over readings:", overall(cvd.outcome))
print("Readings meeting the criterion as written (orderings and n = 4623):", int(((cvd.outcome == "reproduced")).sum()), "of", len(cvd),
"| orderings hold:", int(cvd["orderings hold"].sum()))
by_factor(cvd, ["baseline", "z-scoring", "projection", "CVs", "CV distance"])
baseline z-scoring projection CVs CV distance pairs median CC median VV median CV CV>CC CV>VV CC>VV orderings hold outcome
pre row i all (a) mean of cross pairings 4677 8.78 5.88 10.39 +1.17 p=0e+00 +3.98 p=0e+00 +2.44 p=0e+00 True partially
pre row i all (b) i at -25, j at +250 4677 8.78 5.88 10.28 +1.25 p=0e+00 +3.59 p=0e+00 +2.44 p=0e+00 True partially
pre row i 15+ (a) mean of cross pairings 4567 9.33 6.37 11.11 +1.33 p=0e+00 +4.33 p=0e+00 +2.65 p=0e+00 True partially
pre row i 15+ (b) i at -25, j at +250 4567 9.33 6.37 10.80 +1.41 p=0e+00 +3.87 p=0e+00 +2.65 p=0e+00 True partially
pre row ii all (a) mean of cross pairings 4677 2.53 1.45 3.27 +0.38 p=0e+00 +1.59 p=0e+00 +0.98 p=0e+00 True partially
pre row ii all (b) i at -25, j at +250 4677 2.53 1.45 3.18 +0.46 p=0e+00 +1.45 p=0e+00 +0.98 p=0e+00 True partially
pre row ii 15+ (a) mean of cross pairings 4567 2.48 1.44 3.23 +0.37 p=0e+00 +1.58 p=0e+00 +0.97 p=0e+00 True partially
pre row ii 15+ (b) i at -25, j at +250 4567 2.48 1.44 3.13 +0.46 p=0e+00 +1.45 p=0e+00 +0.97 p=0e+00 True partially
pre time i all (a) mean of cross pairings 4677 8.52 7.38 7.97 -0.46 p=4e-38 +0.58 p=0e+00 +0.96 p=0e+00 False not reproduced
pre time i all (b) i at -25, j at +250 4677 8.52 7.38 7.96 -0.43 p=2e-18 +0.65 p=0e+00 +0.96 p=0e+00 False not reproduced
pre time i 15+ (a) mean of cross pairings 4567 9.02 7.93 8.50 -0.46 p=4e-34 +0.49 p=2e-37 +0.84 p=0e+00 False not reproduced
pre time i 15+ (b) i at -25, j at +250 4567 9.02 7.93 8.47 -0.41 p=9e-18 +0.58 p=9e-40 +0.84 p=0e+00 False not reproduced
pre time ii all (a) mean of cross pairings 4677 2.55 1.53 3.18 +0.30 p=0e+00 +1.44 p=0e+00 +0.95 p=0e+00 True partially
pre time ii all (b) i at -25, j at +250 4677 2.55 1.53 3.10 +0.39 p=0e+00 +1.32 p=0e+00 +0.95 p=0e+00 True partially
pre time ii 15+ (a) mean of cross pairings 4567 2.51 1.51 3.14 +0.31 p=0e+00 +1.44 p=0e+00 +0.94 p=0e+00 True partially
pre time ii 15+ (b) i at -25, j at +250 4567 2.51 1.51 3.05 +0.40 p=0e+00 +1.33 p=0e+00 +0.94 p=0e+00 True partially
whole row i all (a) mean of cross pairings 4677 8.68 5.84 10.31 +1.14 p=0e+00 +3.94 p=0e+00 +2.41 p=0e+00 True partially
whole row i all (b) i at -25, j at +250 4677 8.68 5.84 10.18 +1.22 p=0e+00 +3.58 p=0e+00 +2.41 p=0e+00 True partially
whole row i 15+ (a) mean of cross pairings 4567 9.38 6.41 11.16 +1.33 p=0e+00 +4.34 p=0e+00 +2.67 p=0e+00 True partially
whole row i 15+ (b) i at -25, j at +250 4567 9.38 6.41 10.86 +1.41 p=0e+00 +3.90 p=0e+00 +2.67 p=0e+00 True partially
whole row ii all (a) mean of cross pairings 4677 2.33 1.36 3.06 +0.35 p=0e+00 +1.46 p=0e+00 +0.90 p=0e+00 True partially
whole row ii all (b) i at -25, j at +250 4677 2.33 1.36 2.94 +0.43 p=0e+00 +1.34 p=0e+00 +0.90 p=0e+00 True partially
whole row ii 15+ (a) mean of cross pairings 4567 2.29 1.35 3.03 +0.34 p=0e+00 +1.45 p=0e+00 +0.89 p=0e+00 True partially
whole row ii 15+ (b) i at -25, j at +250 4567 2.29 1.35 2.89 +0.42 p=0e+00 +1.34 p=0e+00 +0.89 p=0e+00 True partially
whole time i all (a) mean of cross pairings 4677 8.47 7.39 7.93 -0.47 p=1e-36 +0.55 p=0e+00 +0.94 p=0e+00 False not reproduced
whole time i all (b) i at -25, j at +250 4677 8.47 7.39 7.95 -0.43 p=2e-17 +0.62 p=0e+00 +0.94 p=0e+00 False not reproduced
whole time i 15+ (a) mean of cross pairings 4567 9.03 8.01 8.53 -0.43 p=4e-31 +0.46 p=2e-32 +0.78 p=8e-45 False not reproduced
whole time i 15+ (b) i at -25, j at +250 4567 9.03 8.01 8.50 -0.38 p=2e-15 +0.56 p=2e-36 +0.78 p=8e-45 False not reproduced
whole time ii all (a) mean of cross pairings 4677 2.34 1.44 2.97 +0.29 p=0e+00 +1.31 p=0e+00 +0.86 p=0e+00 True partially
whole time ii all (b) i at -25, j at +250 4677 2.34 1.44 2.86 +0.36 p=0e+00 +1.21 p=0e+00 +0.86 p=0e+00 True partially
whole time ii 15+ (a) mean of cross pairings 4567 2.31 1.43 2.94 +0.29 p=0e+00 +1.31 p=0e+00 +0.85 p=0e+00 True partially
whole time ii 15+ (b) i at -25, j at +250 4567 2.31 1.43 2.82 +0.37 p=0e+00 +1.21 p=0e+00 +0.85 p=0e+00 True partially
rest row i all (a) mean of cross pairings 4677 9.02 6.05 10.62 +0.95 p=0e+00 +3.86 p=0e+00 +2.48 p=0e+00 True partially
rest row i all (b) i at -25, j at +250 4677 9.02 6.05 10.38 +1.12 p=0e+00 +3.40 p=0e+00 +2.48 p=0e+00 True partially
rest row i 15+ (a) mean of cross pairings 4567 9.47 6.22 11.02 +1.09 p=0e+00 +4.13 p=0e+00 +2.70 p=0e+00 True partially
rest row i 15+ (b) i at -25, j at +250 4567 9.47 6.22 10.73 +1.19 p=0e+00 +3.66 p=0e+00 +2.70 p=0e+00 True partially
rest row ii all (a) mean of cross pairings 4677 2.78 1.65 3.62 +0.43 p=0e+00 +1.75 p=0e+00 +1.02 p=0e+00 True partially
rest row ii all (b) i at -25, j at +250 4677 2.78 1.65 3.47 +0.54 p=0e+00 +1.54 p=0e+00 +1.02 p=0e+00 True partially
rest row ii 15+ (a) mean of cross pairings 4567 2.75 1.65 3.58 +0.41 p=0e+00 +1.75 p=0e+00 +1.03 p=0e+00 True partially
rest row ii 15+ (b) i at -25, j at +250 4567 2.75 1.65 3.43 +0.52 p=0e+00 +1.55 p=0e+00 +1.03 p=0e+00 True partially
rest time i all (a) mean of cross pairings 4677 8.66 7.58 8.25 -0.44 p=3e-33 +0.42 p=1e-35 +0.77 p=0e+00 False not reproduced
rest time i all (b) i at -25, j at +250 4677 8.66 7.58 8.16 -0.37 p=7e-15 +0.57 p=3e-42 +0.77 p=0e+00 False not reproduced
rest time i 15+ (a) mean of cross pairings 4567 8.99 7.92 8.55 -0.40 p=1e-27 +0.35 p=1e-25 +0.67 p=2e-35 False not reproduced
rest time i 15+ (b) i at -25, j at +250 4567 8.99 7.92 8.49 -0.32 p=1e-11 +0.46 p=2e-30 +0.67 p=2e-35 False not reproduced
rest time ii all (a) mean of cross pairings 4677 2.81 1.77 3.52 +0.34 p=0e+00 +1.56 p=0e+00 +0.98 p=0e+00 True partially
rest time ii all (b) i at -25, j at +250 4677 2.81 1.77 3.40 +0.45 p=0e+00 +1.34 p=0e+00 +0.98 p=0e+00 True partially
rest time ii 15+ (a) mean of cross pairings 4567 2.78 1.76 3.48 +0.34 p=0e+00 +1.55 p=0e+00 +0.97 p=0e+00 True partially
rest time ii 15+ (b) i at -25, j at +250 4567 2.78 1.76 3.35 +0.44 p=0e+00 +1.34 p=0e+00 +0.97 p=0e+00 True partially
Pairs within subjects (all CVs): {'EC2': 1596, 'GP31': 1485, 'GP33': 1596} | 15+ CVs: {'EC2': 1540, 'GP31': 1431, 'GP33': 1596} (paper n = 4623)
Outcome over readings: differ by reading: not reproduced 12, partially 36
Readings meeting the criterion as written (orderings and n = 4623): 0 of 48 | orderings hold: 36
Outcome counts by reading factor (48 readings):
baseline: {'pre': {'partially': 12, 'not reproduced': 4}, 'whole': {'partially': 12, 'not reproduced': 4}, 'rest': {'partially': 12, 'not reproduced': 4}}
z-scoring: {'row': {'partially': 24}, 'time': {'not reproduced': 12, 'partially': 12}}
projection: {'i': {'partially': 12, 'not reproduced': 12}, 'ii': {'partially': 24}}
CVs: {'all': {'partially': 18, 'not reproduced': 6}, '15+': {'partially': 18, 'not reproduced': 6}}
CV distance: {'(a) mean of cross pairings': {'partially': 18, 'not reproduced': 6}, '(b) i at -25, j at +250': {'partially': 18, 'not reproduced': 6}}
Result: blocked (methods_underspecified). Under row z-scoring, and under time z-scoring
with projection (ii), consonant-vowel distances exceed consonant-consonant ones, which exceed
vowel-vowel ones, with corrected p < 1e-10. Under time z-scoring with projection (i) the
consonant-vowel distances are shorter than the consonant-consonant ones, in every baseline and CV
set (12 of 12 readings). The number of pairs is 4677 with every CV and 4567 with CVs of 15 or more
renditions, never the paper's 4623, so no reading meets the criterion as written (0 of 48). The
criterion defines only reproduced; this attempt counts a reading where all orderings hold and only
the pair count differs as partially reproduced (36 of 48), by the protocol's general definition of
an effect that holds in part, and one where an ordering fails as not reproduced (12). The pair
count is fixed by the dataset's syllables (data-composition), while the orderings depend on the
z-scoring and projection readings, which the paper does not fix; the outcome is blocked for that.
suppfig9-pc-articulator-weights¶
The first three PCs are weighted toward tongue over lips, and toward lips over jaw and larynx, while jaw and larynx do not differ.
Note on the method: by Methods Eq. 6, log10|PCw_ij| = log10|M_ij| + log10|W_jA|, so for a given electrode the PC term is the same for every articulator and cancels in the paired comparisons; the tests give the same result in every PC.
In [27]:
pairs6 = list(itertools.combinations(range(4), 2))
s9 = []
for (base, cs, reading, pmode) in W_READINGS:
for z in ZMODES:
for eset in ("all vSMC", "GLM electrodes"):
if eset == "all vSMC" and pmode == "t-test":
continue # the p-value reading only changes which electrodes are kept
vals = []
for s in B.SUBJECTS:
w = WEIGHTS[base, cs, reading, pmode][s]
M = STATE[s, base, z, cs, EP][0]
sel = w["keep"] if eset == "GLM electrodes" else np.ones(len(w["keep"]), bool)
vals.append(np.log10(np.abs(M[sel, :3, None] * w["W"][sel, None, :])))
V = np.concatenate(vals) # (electrode, PC, articulator)
n_in = len(V)
V = V[np.isfinite(V).all(axis=(1, 2))] # electrodes without a window T_A for an articulator are dropped
per_pc, meds = [], []
for pci in range(3):
p = {(a, b): (np.median(V[:, pci, a] - V[:, pci, b]), min(1.0, 6 * stats.wilcoxon(V[:, pci, a], V[:, pci, b]).pvalue))
for a, b in pairs6}
def greater(a, b, alpha):
m, pv = p[(a, b)] if (a, b) in p else (-p[(b, a)][0], p[(b, a)][1])
return bool(m > 0 and pv < alpha)
L, T, X, J = 0, 1, 2, 3
ra = greater(L, J, 0.001) and greater(L, X, 0.001)
rb = ra and greater(T, J, 0.001) and greater(T, X, 0.001)
per_pc.append(greater(T, L, 0.05) and (ra or rb) and p[(X, J)][1] > 0.05)
meds.append("/".join(f"{np.median(V[:, pci, a]):.2f}" for a in range(4)))
n_ok = sum(per_pc)
s9.append({"baseline": base, "CVs": cs, "timing reading": " / ".join(reading), "p-value": pmode, "z-scoring": z,
"electrodes": eset, "n": len(V), "dropped": n_in - len(V), "PC1 medians L/T/X/J": meds[0], "PCs holding": "".join("Y" if x else "n" for x in per_pc),
"outcome": "reproduced" if n_ok == 3 else ("partially" if n_ok else "not reproduced")})
s9 = pd.DataFrame(s9)
print(s9.to_string(index=False))
print("\nOutcome over readings:", overall(s9.outcome))
print("Electrodes dropped for lack of a window T_A: 0 in", int((s9.dropped == 0).sum()), "readings, up to", int(s9.dropped.max()),
"| n from", int(s9.n.min()), "to", int(s9.n.max()))
base, cs, reading, pmode, z = "pre", "all", TIMING_READINGS[2], "F-test", "row"
V = np.concatenate([np.log10(np.abs(STATE[s, base, z, cs, EP][0][:, :3, None] * WEIGHTS[base, cs, reading, pmode][s]["W"][:, None, :])) for s in B.SUBJECTS])
V = V[np.isfinite(V).all(axis=(1, 2))]
for pci in range(3):
ps = {f"{B.ARTICULATORS[a][:2]}-{B.ARTICULATORS[b][:2]}": f"{np.median(V[:, pci, a] - V[:, pci, b]):+.2f} p={min(1, 6 * stats.wilcoxon(V[:, pci, a], V[:, pci, b]).pvalue):.0e}" for a, b in pairs6}
print(f"PC{pci + 1} ({base}, {cs}, {reading}, all vSMC, n={len(V)}): median paired differences and corrected p {ps}")
baseline CVs timing reading p-value z-scoring electrodes n dropped PC1 medians L/T/X/J PCs holding outcome
pre all per subject / all six / threshold F-test row all vSMC 185 83 -2.27/-2.47/-2.56/-2.22 nnn not reproduced
pre all per subject / all six / threshold F-test row GLM electrodes 67 36 -1.59/-2.13/-1.93/-1.42 nnn not reproduced
pre all per subject / all six / threshold F-test time all vSMC 185 83 -2.22/-2.40/-2.47/-2.19 nnn not reproduced
pre all per subject / all six / threshold F-test time GLM electrodes 67 36 -1.64/-2.16/-1.94/-1.45 nnn not reproduced
pre all per subject / all six / threshold t-test row GLM electrodes 68 35 -1.61/-2.13/-1.99/-1.40 nnn not reproduced
pre all per subject / all six / threshold t-test time GLM electrodes 68 35 -1.66/-2.16/-1.95/-1.45 nnn not reproduced
pre all per subject / all six / half peak F-test row all vSMC 185 83 -2.37/-2.48/-2.55/-2.33 nnn not reproduced
pre all per subject / all six / half peak F-test row GLM electrodes 55 27 -1.56/-2.06/-2.04/-1.42 nnn not reproduced
pre all per subject / all six / half peak F-test time all vSMC 185 83 -2.29/-2.40/-2.44/-2.27 nnn not reproduced
pre all per subject / all six / half peak F-test time GLM electrodes 55 27 -1.61/-2.06/-2.09/-1.47 nnn not reproduced
pre all per subject / all six / half peak t-test row GLM electrodes 36 26 -1.38/-2.07/-2.11/-1.25 nnn not reproduced
pre all per subject / all six / half peak t-test time GLM electrodes 36 26 -1.43/-2.03/-2.14/-1.30 nnn not reproduced
pre all per subject / per articulator / threshold F-test row all vSMC 268 0 -2.34/-2.36/-2.50/-2.19 nnn not reproduced
pre all per subject / per articulator / threshold F-test row GLM electrodes 110 0 -1.91/-2.10/-2.29/-1.74 nnn not reproduced
pre all per subject / per articulator / threshold F-test time all vSMC 268 0 -2.29/-2.33/-2.48/-2.19 nnn not reproduced
pre all per subject / per articulator / threshold F-test time GLM electrodes 110 0 -1.89/-2.07/-2.26/-1.76 nnn not reproduced
pre all per subject / per articulator / threshold t-test row GLM electrodes 92 0 -1.92/-2.08/-2.12/-1.60 nnn not reproduced
pre all per subject / per articulator / threshold t-test time GLM electrodes 92 0 -1.90/-2.05/-2.10/-1.60 nnn not reproduced
pre all per subject / per articulator / half peak F-test row all vSMC 268 0 -2.37/-2.40/-2.53/-2.23 nnn not reproduced
pre all per subject / per articulator / half peak F-test row GLM electrodes 75 0 -1.73/-2.00/-2.26/-1.65 nnn not reproduced
pre all per subject / per articulator / half peak F-test time all vSMC 268 0 -2.33/-2.37/-2.49/-2.21 nnn not reproduced
pre all per subject / per articulator / half peak F-test time GLM electrodes 75 0 -1.78/-1.99/-2.23/-1.68 nnn not reproduced
pre all per subject / per articulator / half peak t-test row GLM electrodes 58 0 -1.80/-1.99/-2.12/-1.63 nnn not reproduced
pre all per subject / per articulator / half peak t-test time GLM electrodes 58 0 -1.85/-2.03/-2.16/-1.68 nnn not reproduced
pre all pooled / all six / threshold F-test row all vSMC 185 83 -2.34/-2.43/-2.55/-2.23 nnn not reproduced
pre all pooled / all six / threshold F-test row GLM electrodes 67 34 -1.44/-2.17/-2.05/-1.41 nnn not reproduced
pre all pooled / all six / threshold F-test time all vSMC 185 83 -2.19/-2.36/-2.50/-2.21 nnn not reproduced
pre all pooled / all six / threshold F-test time GLM electrodes 67 34 -1.48/-2.20/-2.04/-1.44 nnn not reproduced
pre all pooled / all six / threshold t-test row GLM electrodes 60 39 -1.44/-2.17/-2.02/-1.37 nnn not reproduced
pre all pooled / all six / threshold t-test time GLM electrodes 60 39 -1.46/-2.20/-2.03/-1.35 nnn not reproduced
pre all pooled / all six / half peak F-test row all vSMC 185 83 -2.39/-2.51/-2.59/-2.30 nnn not reproduced
pre all pooled / all six / half peak F-test row GLM electrodes 51 25 -1.40/-2.20/-2.08/-1.42 nnn not reproduced
pre all pooled / all six / half peak F-test time all vSMC 185 83 -2.25/-2.42/-2.54/-2.28 nnn not reproduced
pre all pooled / all six / half peak F-test time GLM electrodes 51 25 -1.46/-2.24/-2.12/-1.45 nnn not reproduced
pre all pooled / all six / half peak t-test row GLM electrodes 37 25 -1.39/-2.21/-2.13/-1.26 nnn not reproduced
pre all pooled / all six / half peak t-test time GLM electrodes 37 25 -1.46/-2.25/-2.16/-1.31 nnn not reproduced
pre all pooled / per articulator / threshold F-test row all vSMC 268 0 -2.30/-2.33/-2.44/-2.20 nnn not reproduced
pre all pooled / per articulator / threshold F-test row GLM electrodes 108 0 -1.73/-2.12/-2.26/-1.63 nnn not reproduced
pre all pooled / per articulator / threshold F-test time all vSMC 268 0 -2.23/-2.28/-2.44/-2.20 nnn not reproduced
pre all pooled / per articulator / threshold F-test time GLM electrodes 108 0 -1.77/-2.13/-2.24/-1.68 nnn not reproduced
pre all pooled / per articulator / threshold t-test row GLM electrodes 101 0 -1.77/-2.15/-2.10/-1.58 nnn not reproduced
pre all pooled / per articulator / threshold t-test time GLM electrodes 101 0 -1.81/-2.18/-2.05/-1.57 nnn not reproduced
pre all pooled / per articulator / half peak F-test row all vSMC 268 0 -2.36/-2.38/-2.48/-2.23 nnn not reproduced
pre all pooled / per articulator / half peak F-test row GLM electrodes 73 0 -1.53/-2.07/-2.24/-1.61 nnn not reproduced
pre all pooled / per articulator / half peak F-test time all vSMC 268 0 -2.34/-2.32/-2.47/-2.22 nnn not reproduced
pre all pooled / per articulator / half peak F-test time GLM electrodes 73 0 -1.57/-2.07/-2.26/-1.62 nnn not reproduced
pre all pooled / per articulator / half peak t-test row GLM electrodes 60 0 -1.53/-2.10/-2.10/-1.55 nnn not reproduced
pre all pooled / per articulator / half peak t-test time GLM electrodes 60 0 -1.57/-2.12/-2.13/-1.60 nnn not reproduced
pre 15+ per subject / all six / threshold F-test row all vSMC 185 83 -2.31/-2.50/-2.55/-2.25 nnn not reproduced
pre 15+ per subject / all six / threshold F-test row GLM electrodes 67 47 -1.59/-2.13/-1.93/-1.41 nnn not reproduced
pre 15+ per subject / all six / threshold F-test time all vSMC 185 83 -2.21/-2.43/-2.48/-2.26 nnn not reproduced
pre 15+ per subject / all six / threshold F-test time GLM electrodes 67 47 -1.64/-2.17/-1.94/-1.46 nnn not reproduced
pre 15+ per subject / all six / threshold t-test row GLM electrodes 68 30 -1.62/-2.15/-1.97/-1.41 nnn not reproduced
pre 15+ per subject / all six / threshold t-test time GLM electrodes 68 30 -1.65/-2.17/-1.94/-1.45 nnn not reproduced
pre 15+ per subject / all six / half peak F-test row all vSMC 185 83 -2.40/-2.48/-2.63/-2.38 nnn not reproduced
pre 15+ per subject / all six / half peak F-test row GLM electrodes 55 35 -1.58/-2.06/-2.06/-1.42 nnn not reproduced
pre 15+ per subject / all six / half peak F-test time all vSMC 185 83 -2.35/-2.42/-2.47/-2.39 nnn not reproduced
pre 15+ per subject / all six / half peak F-test time GLM electrodes 55 35 -1.62/-2.10/-2.10/-1.47 nnn not reproduced
pre 15+ per subject / all six / half peak t-test row GLM electrodes 36 20 -1.38/-2.06/-2.08/-1.28 nnn not reproduced
pre 15+ per subject / all six / half peak t-test time GLM electrodes 36 20 -1.43/-2.06/-2.16/-1.29 nnn not reproduced
pre 15+ per subject / per articulator / threshold F-test row all vSMC 268 0 -2.34/-2.39/-2.58/-2.31 nnn not reproduced
pre 15+ per subject / per articulator / threshold F-test row GLM electrodes 116 0 -1.92/-2.10/-2.30/-1.72 nnn not reproduced
pre 15+ per subject / per articulator / threshold F-test time all vSMC 268 0 -2.28/-2.36/-2.55/-2.30 nnn not reproduced
pre 15+ per subject / per articulator / threshold F-test time GLM electrodes 116 0 -1.89/-2.15/-2.29/-1.75 nnn not reproduced
pre 15+ per subject / per articulator / threshold t-test row GLM electrodes 97 0 -1.89/-2.12/-2.23/-1.61 nnn not reproduced
pre 15+ per subject / per articulator / threshold t-test time GLM electrodes 97 0 -1.86/-2.15/-2.23/-1.66 nnn not reproduced
pre 15+ per subject / per articulator / half peak F-test row all vSMC 268 0 -2.38/-2.48/-2.62/-2.38 nnn not reproduced
pre 15+ per subject / per articulator / half peak F-test row GLM electrodes 81 0 -1.73/-2.09/-2.28/-1.66 nnn not reproduced
pre 15+ per subject / per articulator / half peak F-test time all vSMC 268 0 -2.36/-2.41/-2.59/-2.38 nnn not reproduced
pre 15+ per subject / per articulator / half peak F-test time GLM electrodes 81 0 -1.74/-2.12/-2.34/-1.70 nnn not reproduced
pre 15+ per subject / per articulator / half peak t-test row GLM electrodes 58 0 -1.67/-2.08/-2.28/-1.56 nnn not reproduced
pre 15+ per subject / per articulator / half peak t-test time GLM electrodes 58 0 -1.68/-2.09/-2.30/-1.60 nnn not reproduced
pre 15+ pooled / all six / threshold F-test row all vSMC 185 83 -2.27/-2.42/-2.53/-2.27 nnn not reproduced
pre 15+ pooled / all six / threshold F-test row GLM electrodes 71 44 -1.49/-2.04/-1.98/-1.41 nnn not reproduced
pre 15+ pooled / all six / threshold F-test time all vSMC 185 83 -2.18/-2.39/-2.51/-2.23 nnn not reproduced
pre 15+ pooled / all six / threshold F-test time GLM electrodes 71 44 -1.48/-2.08/-2.02/-1.45 nnn not reproduced
pre 15+ pooled / all six / threshold t-test row GLM electrodes 71 36 -1.55/-2.08/-1.98/-1.41 nnn not reproduced
pre 15+ pooled / all six / threshold t-test time GLM electrodes 71 36 -1.53/-2.13/-2.01/-1.45 nnn not reproduced
pre 15+ pooled / all six / half peak F-test row all vSMC 185 83 -2.40/-2.46/-2.61/-2.36 nnn not reproduced
pre 15+ pooled / all six / half peak F-test row GLM electrodes 58 32 -1.44/-2.16/-2.05/-1.43 nnn not reproduced
pre 15+ pooled / all six / half peak F-test time all vSMC 185 83 -2.30/-2.42/-2.50/-2.38 nnn not reproduced
pre 15+ pooled / all six / half peak F-test time GLM electrodes 58 32 -1.49/-2.16/-2.08/-1.49 nnn not reproduced
pre 15+ pooled / all six / half peak t-test row GLM electrodes 39 24 -1.35/-2.03/-2.06/-1.24 nnn not reproduced
pre 15+ pooled / all six / half peak t-test time GLM electrodes 39 24 -1.44/-2.06/-2.18/-1.28 nnn not reproduced
pre 15+ pooled / per articulator / threshold F-test row all vSMC 268 0 -2.32/-2.31/-2.56/-2.33 nnn not reproduced
pre 15+ pooled / per articulator / threshold F-test row GLM electrodes 117 0 -1.59/-2.07/-2.19/-1.66 nnn not reproduced
pre 15+ pooled / per articulator / threshold F-test time all vSMC 268 0 -2.26/-2.27/-2.51/-2.30 nnn not reproduced
pre 15+ pooled / per articulator / threshold F-test time GLM electrodes 117 0 -1.67/-2.06/-2.20/-1.71 nnn not reproduced
pre 15+ pooled / per articulator / threshold t-test row GLM electrodes 100 0 -1.51/-2.07/-2.00/-1.59 nnn not reproduced
pre 15+ pooled / per articulator / threshold t-test time GLM electrodes 100 0 -1.56/-2.06/-2.01/-1.63 nnn not reproduced
pre 15+ pooled / per articulator / half peak F-test row all vSMC 268 0 -2.39/-2.39/-2.58/-2.37 nnn not reproduced
pre 15+ pooled / per articulator / half peak F-test row GLM electrodes 86 0 -1.53/-2.04/-2.12/-1.59 nnn not reproduced
pre 15+ pooled / per articulator / half peak F-test time all vSMC 268 0 -2.29/-2.33/-2.55/-2.38 nnn not reproduced
pre 15+ pooled / per articulator / half peak F-test time GLM electrodes 86 0 -1.53/-1.99/-2.16/-1.62 nnn not reproduced
pre 15+ pooled / per articulator / half peak t-test row GLM electrodes 61 0 -1.45/-2.09/-2.16/-1.56 nnn not reproduced
pre 15+ pooled / per articulator / half peak t-test time GLM electrodes 61 0 -1.46/-2.11/-2.13/-1.60 nnn not reproduced
whole all per subject / all six / threshold F-test row all vSMC 185 83 -2.31/-2.52/-2.57/-2.25 nnn not reproduced
whole all per subject / all six / threshold F-test row GLM electrodes 71 36 -1.63/-2.16/-2.06/-1.44 nnn not reproduced
whole all per subject / all six / threshold F-test time all vSMC 185 83 -2.26/-2.40/-2.53/-2.26 nnn not reproduced
whole all per subject / all six / threshold F-test time GLM electrodes 71 36 -1.75/-2.17/-2.07/-1.50 nnn not reproduced
whole all per subject / all six / threshold t-test row GLM electrodes 65 30 -1.62/-2.14/-2.06/-1.42 nnn not reproduced
whole all per subject / all six / threshold t-test time GLM electrodes 65 30 -1.72/-2.16/-2.05/-1.42 nnn not reproduced
whole all per subject / all six / half peak F-test row all vSMC 185 83 -2.40/-2.47/-2.61/-2.34 nnn not reproduced
whole all per subject / all six / half peak F-test row GLM electrodes 48 29 -1.54/-2.06/-2.22/-1.44 nnn not reproduced
whole all per subject / all six / half peak F-test time all vSMC 185 83 -2.36/-2.41/-2.52/-2.30 nnn not reproduced
whole all per subject / all six / half peak F-test time GLM electrodes 48 29 -1.56/-2.07/-2.26/-1.47 nnn not reproduced
whole all per subject / all six / half peak t-test row GLM electrodes 32 26 -1.39/-2.08/-2.29/-1.26 nnn not reproduced
whole all per subject / all six / half peak t-test time GLM electrodes 32 26 -1.40/-2.05/-2.32/-1.30 nnn not reproduced
whole all per subject / per articulator / threshold F-test row all vSMC 268 0 -2.32/-2.37/-2.51/-2.18 nnn not reproduced
whole all per subject / per articulator / threshold F-test row GLM electrodes 105 0 -1.80/-2.10/-2.28/-1.64 nnn not reproduced
whole all per subject / per articulator / threshold F-test time all vSMC 268 0 -2.30/-2.28/-2.50/-2.15 nnn not reproduced
whole all per subject / per articulator / threshold F-test time GLM electrodes 105 0 -1.85/-2.11/-2.24/-1.68 nnn not reproduced
whole all per subject / per articulator / threshold t-test row GLM electrodes 93 0 -1.80/-2.16/-2.18/-1.51 nnn not reproduced
whole all per subject / per articulator / threshold t-test time GLM electrodes 93 0 -1.85/-2.12/-2.15/-1.57 nnn not reproduced
whole all per subject / per articulator / half peak F-test row all vSMC 268 0 -2.39/-2.36/-2.59/-2.29 nnn not reproduced
whole all per subject / per articulator / half peak F-test row GLM electrodes 72 0 -1.59/-2.12/-2.32/-1.60 nnn not reproduced
whole all per subject / per articulator / half peak F-test time all vSMC 268 0 -2.36/-2.34/-2.55/-2.24 nnn not reproduced
whole all per subject / per articulator / half peak F-test time GLM electrodes 72 0 -1.64/-2.11/-2.36/-1.66 nnn not reproduced
whole all per subject / per articulator / half peak t-test row GLM electrodes 53 0 -1.59/-2.17/-2.31/-1.59 nnn not reproduced
whole all per subject / per articulator / half peak t-test time GLM electrodes 53 0 -1.64/-2.20/-2.28/-1.57 nnn not reproduced
whole all pooled / all six / threshold F-test row all vSMC 185 83 -2.38/-2.43/-2.57/-2.26 nnn not reproduced
whole all pooled / all six / threshold F-test row GLM electrodes 71 35 -1.45/-2.16/-2.06/-1.46 nnn not reproduced
whole all pooled / all six / threshold F-test time all vSMC 185 83 -2.27/-2.34/-2.52/-2.26 nnn not reproduced
whole all pooled / all six / threshold F-test time GLM electrodes 71 35 -1.53/-2.16/-2.09/-1.51 nnn not reproduced
whole all pooled / all six / threshold t-test row GLM electrodes 61 35 -1.44/-2.18/-2.05/-1.39 nnn not reproduced
whole all pooled / all six / threshold t-test time GLM electrodes 61 35 -1.45/-2.20/-2.03/-1.38 nnn not reproduced
whole all pooled / all six / half peak F-test row all vSMC 185 83 -2.39/-2.49/-2.61/-2.32 nnn not reproduced
whole all pooled / all six / half peak F-test row GLM electrodes 58 27 -1.47/-2.21/-2.05/-1.45 nnn not reproduced
whole all pooled / all six / half peak F-test time all vSMC 185 83 -2.31/-2.43/-2.56/-2.30 nnn not reproduced
whole all pooled / all six / half peak F-test time GLM electrodes 58 27 -1.54/-2.24/-2.13/-1.48 nnn not reproduced
whole all pooled / all six / half peak t-test row GLM electrodes 38 23 -1.41/-2.21/-2.03/-1.28 nnn not reproduced
whole all pooled / all six / half peak t-test time GLM electrodes 38 23 -1.47/-2.23/-2.15/-1.32 nnn not reproduced
whole all pooled / per articulator / threshold F-test row all vSMC 268 0 -2.34/-2.38/-2.49/-2.24 nnn not reproduced
whole all pooled / per articulator / threshold F-test row GLM electrodes 106 0 -1.78/-2.12/-2.26/-1.65 nnn not reproduced
whole all pooled / per articulator / threshold F-test time all vSMC 268 0 -2.27/-2.35/-2.50/-2.20 nnn not reproduced
whole all pooled / per articulator / threshold F-test time GLM electrodes 106 0 -1.79/-2.13/-2.29/-1.70 nnn not reproduced
whole all pooled / per articulator / threshold t-test row GLM electrodes 100 0 -1.78/-2.14/-2.18/-1.61 nnn not reproduced
whole all pooled / per articulator / threshold t-test time GLM electrodes 100 0 -1.84/-2.16/-2.12/-1.60 nnn not reproduced
whole all pooled / per articulator / half peak F-test row all vSMC 268 0 -2.39/-2.37/-2.53/-2.29 nnn not reproduced
whole all pooled / per articulator / half peak F-test row GLM electrodes 73 0 -1.57/-2.11/-2.20/-1.61 nnn not reproduced
whole all pooled / per articulator / half peak F-test time all vSMC 268 0 -2.36/-2.36/-2.52/-2.25 nnn not reproduced
whole all pooled / per articulator / half peak F-test time GLM electrodes 73 0 -1.63/-2.09/-2.23/-1.68 nnn not reproduced
whole all pooled / per articulator / half peak t-test row GLM electrodes 60 0 -1.55/-2.07/-2.09/-1.56 nnn not reproduced
whole all pooled / per articulator / half peak t-test time GLM electrodes 60 0 -1.59/-2.08/-2.16/-1.58 nnn not reproduced
whole 15+ per subject / all six / threshold F-test row all vSMC 185 83 -2.36/-2.52/-2.61/-2.29 nnn not reproduced
whole 15+ per subject / all six / threshold F-test row GLM electrodes 70 41 -1.64/-2.19/-2.08/-1.44 nnn not reproduced
whole 15+ per subject / all six / threshold F-test time all vSMC 185 83 -2.26/-2.44/-2.54/-2.29 nnn not reproduced
whole 15+ per subject / all six / threshold F-test time GLM electrodes 70 41 -1.71/-2.17/-2.10/-1.49 nnn not reproduced
whole 15+ per subject / all six / threshold t-test row GLM electrodes 63 25 -1.60/-2.19/-2.08/-1.41 nnn not reproduced
whole 15+ per subject / all six / threshold t-test time GLM electrodes 63 25 -1.63/-2.17/-2.09/-1.42 nnn not reproduced
whole 15+ per subject / all six / half peak F-test row all vSMC 185 83 -2.45/-2.48/-2.66/-2.38 nnn not reproduced
whole 15+ per subject / all six / half peak F-test row GLM electrodes 48 30 -1.54/-2.06/-2.24/-1.44 nnn not reproduced
whole 15+ per subject / all six / half peak F-test time all vSMC 185 83 -2.36/-2.43/-2.56/-2.40 nnn not reproduced
whole 15+ per subject / all six / half peak F-test time GLM electrodes 48 30 -1.56/-2.10/-2.29/-1.46 nnn not reproduced
whole 15+ per subject / all six / half peak t-test row GLM electrodes 32 21 -1.39/-2.08/-2.31/-1.27 nnn not reproduced
whole 15+ per subject / all six / half peak t-test time GLM electrodes 32 21 -1.40/-2.08/-2.32/-1.30 nnn not reproduced
whole 15+ per subject / per articulator / threshold F-test row all vSMC 268 0 -2.30/-2.39/-2.64/-2.29 nnn not reproduced
whole 15+ per subject / per articulator / threshold F-test row GLM electrodes 114 0 -1.89/-2.11/-2.34/-1.69 nnn not reproduced
whole 15+ per subject / per articulator / threshold F-test time all vSMC 268 0 -2.29/-2.35/-2.58/-2.30 nnn not reproduced
whole 15+ per subject / per articulator / threshold F-test time GLM electrodes 114 0 -1.87/-2.15/-2.28/-1.77 nnn not reproduced
whole 15+ per subject / per articulator / threshold t-test row GLM electrodes 95 0 -1.75/-2.18/-2.25/-1.57 nnn not reproduced
whole 15+ per subject / per articulator / threshold t-test time GLM electrodes 95 0 -1.75/-2.26/-2.22/-1.63 nnn not reproduced
whole 15+ per subject / per articulator / half peak F-test row all vSMC 268 0 -2.38/-2.43/-2.65/-2.39 nnn not reproduced
whole 15+ per subject / per articulator / half peak F-test row GLM electrodes 78 0 -1.59/-2.14/-2.32/-1.67 nnn not reproduced
whole 15+ per subject / per articulator / half peak F-test time all vSMC 268 0 -2.36/-2.42/-2.61/-2.40 nnn not reproduced
whole 15+ per subject / per articulator / half peak F-test time GLM electrodes 78 0 -1.65/-2.15/-2.39/-1.71 nnn not reproduced
whole 15+ per subject / per articulator / half peak t-test row GLM electrodes 54 0 -1.53/-2.15/-2.24/-1.59 nnn not reproduced
whole 15+ per subject / per articulator / half peak t-test time GLM electrodes 54 0 -1.55/-2.17/-2.23/-1.62 nnn not reproduced
whole 15+ pooled / all six / threshold F-test row all vSMC 185 83 -2.28/-2.38/-2.56/-2.33 nnn not reproduced
whole 15+ pooled / all six / threshold F-test row GLM electrodes 69 47 -1.49/-2.03/-2.02/-1.45 nnn not reproduced
whole 15+ pooled / all six / threshold F-test time all vSMC 185 83 -2.23/-2.31/-2.52/-2.31 nnn not reproduced
whole 15+ pooled / all six / threshold F-test time GLM electrodes 69 47 -1.53/-2.08/-2.05/-1.49 nnn not reproduced
whole 15+ pooled / all six / threshold t-test row GLM electrodes 72 39 -1.57/-2.03/-2.01/-1.45 nnn not reproduced
whole 15+ pooled / all six / threshold t-test time GLM electrodes 72 39 -1.59/-2.10/-2.04/-1.49 nnn not reproduced
whole 15+ pooled / all six / half peak F-test row all vSMC 185 83 -2.41/-2.41/-2.64/-2.37 nnn not reproduced
whole 15+ pooled / all six / half peak F-test row GLM electrodes 58 31 -1.46/-2.19/-2.07/-1.46 nnn not reproduced
whole 15+ pooled / all six / half peak F-test time all vSMC 185 83 -2.32/-2.45/-2.53/-2.40 nnn not reproduced
whole 15+ pooled / all six / half peak F-test time GLM electrodes 58 31 -1.55/-2.14/-2.10/-1.51 nnn not reproduced
whole 15+ pooled / all six / half peak t-test row GLM electrodes 34 22 -1.38/-2.10/-2.14/-1.36 nnn not reproduced
whole 15+ pooled / all six / half peak t-test time GLM electrodes 34 22 -1.41/-2.09/-2.25/-1.36 nnn not reproduced
whole 15+ pooled / per articulator / threshold F-test row all vSMC 268 0 -2.34/-2.40/-2.59/-2.31 nnn not reproduced
whole 15+ pooled / per articulator / threshold F-test row GLM electrodes 116 0 -1.60/-2.17/-2.25/-1.68 nnn not reproduced
whole 15+ pooled / per articulator / threshold F-test time all vSMC 268 0 -2.29/-2.33/-2.54/-2.27 nnn not reproduced
whole 15+ pooled / per articulator / threshold F-test time GLM electrodes 116 0 -1.69/-2.19/-2.20/-1.76 nnn not reproduced
whole 15+ pooled / per articulator / threshold t-test row GLM electrodes 99 0 -1.57/-2.18/-2.13/-1.60 nnn not reproduced
whole 15+ pooled / per articulator / threshold t-test time GLM electrodes 99 0 -1.59/-2.21/-2.12/-1.64 nnn not reproduced
whole 15+ pooled / per articulator / half peak F-test row all vSMC 268 0 -2.38/-2.41/-2.58/-2.40 nnn not reproduced
whole 15+ pooled / per articulator / half peak F-test row GLM electrodes 86 0 -1.54/-2.14/-2.17/-1.63 nnn not reproduced
whole 15+ pooled / per articulator / half peak F-test time all vSMC 268 0 -2.33/-2.38/-2.59/-2.40 nnn not reproduced
whole 15+ pooled / per articulator / half peak F-test time GLM electrodes 86 0 -1.59/-2.11/-2.18/-1.68 nnn not reproduced
whole 15+ pooled / per articulator / half peak t-test row GLM electrodes 55 0 -1.45/-2.24/-2.19/-1.59 nnn not reproduced
whole 15+ pooled / per articulator / half peak t-test time GLM electrodes 55 0 -1.50/-2.27/-2.19/-1.63 nnn not reproduced
rest all per subject / all six / threshold F-test row all vSMC 185 83 -2.31/-2.30/-2.54/-2.24 nnn not reproduced
rest all per subject / all six / threshold F-test row GLM electrodes 68 34 -1.54/-2.01/-2.00/-1.40 nnn not reproduced
rest all per subject / all six / threshold F-test time all vSMC 185 83 -2.25/-2.28/-2.45/-2.15 nnn not reproduced
rest all per subject / all six / threshold F-test time GLM electrodes 68 34 -1.61/-2.06/-2.06/-1.45 nnn not reproduced
rest all per subject / all six / threshold t-test row GLM electrodes 70 33 -1.59/-2.00/-2.05/-1.39 nnn not reproduced
rest all per subject / all six / threshold t-test time GLM electrodes 70 33 -1.67/-2.00/-2.03/-1.44 nnn not reproduced
rest all per subject / all six / half peak F-test row all vSMC 185 83 -2.36/-2.37/-2.57/-2.34 nnn not reproduced
rest all per subject / all six / half peak F-test row GLM electrodes 49 26 -1.48/-2.02/-2.12/-1.42 nnn not reproduced
rest all per subject / all six / half peak F-test time all vSMC 185 83 -2.27/-2.32/-2.50/-2.28 nnn not reproduced
rest all per subject / all six / half peak F-test time GLM electrodes 49 26 -1.51/-2.09/-2.15/-1.42 nnn not reproduced
rest all per subject / all six / half peak t-test row GLM electrodes 33 20 -1.35/-1.99/-2.17/-1.21 nnn not reproduced
rest all per subject / all six / half peak t-test time GLM electrodes 33 20 -1.34/-2.08/-2.21/-1.29 nnn not reproduced
rest all per subject / per articulator / threshold F-test row all vSMC 268 0 -2.28/-2.37/-2.44/-2.16 nnn not reproduced
rest all per subject / per articulator / threshold F-test row GLM electrodes 99 0 -1.81/-2.12/-2.13/-1.65 nnn not reproduced
rest all per subject / per articulator / threshold F-test time all vSMC 268 0 -2.22/-2.33/-2.42/-2.13 nnn not reproduced
rest all per subject / per articulator / threshold F-test time GLM electrodes 99 0 -1.89/-2.13/-2.11/-1.69 nnn not reproduced
rest all per subject / per articulator / threshold t-test row GLM electrodes 96 0 -1.94/-2.14/-2.10/-1.49 nnn not reproduced
rest all per subject / per articulator / threshold t-test time GLM electrodes 96 0 -1.95/-2.17/-2.03/-1.59 nnn not reproduced
rest all per subject / per articulator / half peak F-test row all vSMC 268 0 -2.32/-2.38/-2.46/-2.20 nnn not reproduced
rest all per subject / per articulator / half peak F-test row GLM electrodes 71 0 -1.65/-2.06/-2.10/-1.62 nnn not reproduced
rest all per subject / per articulator / half peak F-test time all vSMC 268 0 -2.30/-2.33/-2.45/-2.15 nnn not reproduced
rest all per subject / per articulator / half peak F-test time GLM electrodes 71 0 -1.64/-2.09/-2.15/-1.62 nnn not reproduced
rest all per subject / per articulator / half peak t-test row GLM electrodes 57 0 -1.57/-2.18/-2.09/-1.55 nnn not reproduced
rest all per subject / per articulator / half peak t-test time GLM electrodes 57 0 -1.58/-2.15/-2.07/-1.60 nnn not reproduced
rest all pooled / all six / threshold F-test row all vSMC 185 83 -2.28/-2.29/-2.48/-2.28 nnn not reproduced
rest all pooled / all six / threshold F-test row GLM electrodes 67 26 -1.45/-2.03/-1.98/-1.40 nnn not reproduced
rest all pooled / all six / threshold F-test time all vSMC 185 83 -2.21/-2.24/-2.46/-2.21 nnn not reproduced
rest all pooled / all six / threshold F-test time GLM electrodes 67 26 -1.46/-2.11/-2.02/-1.43 nnn not reproduced
rest all pooled / all six / threshold t-test row GLM electrodes 66 28 -1.46/-2.05/-1.98/-1.38 nnn not reproduced
rest all pooled / all six / threshold t-test time GLM electrodes 66 28 -1.46/-2.14/-1.98/-1.40 nnn not reproduced
rest all pooled / all six / half peak F-test row all vSMC 185 83 -2.37/-2.32/-2.53/-2.30 nnn not reproduced
rest all pooled / all six / half peak F-test row GLM electrodes 52 22 -1.40/-2.06/-2.00/-1.42 nnn not reproduced
rest all pooled / all six / half peak F-test time all vSMC 185 83 -2.29/-2.28/-2.48/-2.24 nnn not reproduced
rest all pooled / all six / half peak F-test time GLM electrodes 52 22 -1.48/-2.10/-2.07/-1.43 nnn not reproduced
rest all pooled / all six / half peak t-test row GLM electrodes 37 20 -1.38/-2.03/-1.98/-1.23 nnn not reproduced
rest all pooled / all six / half peak t-test time GLM electrodes 37 20 -1.46/-2.05/-2.10/-1.34 nnn not reproduced
rest all pooled / per articulator / threshold F-test row all vSMC 268 0 -2.31/-2.33/-2.48/-2.19 nnn not reproduced
rest all pooled / per articulator / threshold F-test row GLM electrodes 105 0 -1.92/-2.11/-2.17/-1.64 nnn not reproduced
rest all pooled / per articulator / threshold F-test time all vSMC 268 0 -2.26/-2.27/-2.45/-2.14 nnn not reproduced
rest all pooled / per articulator / threshold F-test time GLM electrodes 105 0 -1.90/-2.12/-2.10/-1.67 nnn not reproduced
rest all pooled / per articulator / threshold t-test row GLM electrodes 95 0 -1.95/-2.13/-2.11/-1.48 nnn not reproduced
rest all pooled / per articulator / threshold t-test time GLM electrodes 95 0 -1.99/-2.15/-2.08/-1.58 nnn not reproduced
rest all pooled / per articulator / half peak F-test row all vSMC 268 0 -2.31/-2.34/-2.49/-2.19 nnn not reproduced
rest all pooled / per articulator / half peak F-test row GLM electrodes 75 0 -1.70/-2.13/-2.15/-1.58 nnn not reproduced
rest all pooled / per articulator / half peak F-test time all vSMC 268 0 -2.29/-2.29/-2.48/-2.14 nnn not reproduced
rest all pooled / per articulator / half peak F-test time GLM electrodes 75 0 -1.71/-2.13/-2.21/-1.63 nnn not reproduced
rest all pooled / per articulator / half peak t-test row GLM electrodes 52 0 -1.55/-2.18/-2.11/-1.44 nnn not reproduced
rest all pooled / per articulator / half peak t-test time GLM electrodes 52 0 -1.57/-2.16/-2.12/-1.55 nnn not reproduced
rest 15+ per subject / all six / threshold F-test row all vSMC 268 0 -2.30/-2.35/-2.54/-2.22 nnn not reproduced
rest 15+ per subject / all six / threshold F-test row GLM electrodes 106 0 -1.60/-2.14/-2.12/-1.59 nnn not reproduced
rest 15+ per subject / all six / threshold F-test time all vSMC 268 0 -2.25/-2.30/-2.47/-2.19 nnn not reproduced
rest 15+ per subject / all six / threshold F-test time GLM electrodes 106 0 -1.67/-2.19/-2.16/-1.67 nnn not reproduced
rest 15+ per subject / all six / threshold t-test row GLM electrodes 93 0 -1.57/-2.01/-2.18/-1.52 nnn not reproduced
rest 15+ per subject / all six / threshold t-test time GLM electrodes 93 0 -1.63/-2.06/-2.18/-1.61 nnn not reproduced
rest 15+ per subject / all six / half peak F-test row all vSMC 268 0 -2.36/-2.44/-2.57/-2.23 nnn not reproduced
rest 15+ per subject / all six / half peak F-test row GLM electrodes 74 0 -1.52/-2.17/-2.20/-1.58 nnn not reproduced
rest 15+ per subject / all six / half peak F-test time all vSMC 268 0 -2.29/-2.41/-2.54/-2.24 nnn not reproduced
rest 15+ per subject / all six / half peak F-test time GLM electrodes 74 0 -1.53/-2.20/-2.23/-1.65 nnn not reproduced
rest 15+ per subject / all six / half peak t-test row GLM electrodes 52 0 -1.44/-2.12/-2.21/-1.47 nnn not reproduced
rest 15+ per subject / all six / half peak t-test time GLM electrodes 52 0 -1.41/-2.15/-2.26/-1.53 nnn not reproduced
rest 15+ per subject / per articulator / threshold F-test row all vSMC 268 0 -2.28/-2.38/-2.50/-2.19 nnn not reproduced
rest 15+ per subject / per articulator / threshold F-test row GLM electrodes 99 0 -1.78/-2.15/-2.26/-1.57 nnn not reproduced
rest 15+ per subject / per articulator / threshold F-test time all vSMC 268 0 -2.23/-2.31/-2.48/-2.16 nnn not reproduced
rest 15+ per subject / per articulator / threshold F-test time GLM electrodes 99 0 -1.87/-2.15/-2.31/-1.61 nnn not reproduced
rest 15+ per subject / per articulator / threshold t-test row GLM electrodes 94 0 -1.84/-2.16/-2.28/-1.49 nnn not reproduced
rest 15+ per subject / per articulator / threshold t-test time GLM electrodes 94 0 -1.89/-2.16/-2.31/-1.58 nnn not reproduced
rest 15+ per subject / per articulator / half peak F-test row all vSMC 268 0 -2.35/-2.42/-2.51/-2.23 nnn not reproduced
rest 15+ per subject / per articulator / half peak F-test row GLM electrodes 71 0 -1.61/-2.21/-2.20/-1.59 nnn not reproduced
rest 15+ per subject / per articulator / half peak F-test time all vSMC 268 0 -2.30/-2.36/-2.52/-2.24 nnn not reproduced
rest 15+ per subject / per articulator / half peak F-test time GLM electrodes 71 0 -1.60/-2.16/-2.26/-1.65 nnn not reproduced
rest 15+ per subject / per articulator / half peak t-test row GLM electrodes 52 0 -1.54/-2.27/-2.27/-1.47 nnn not reproduced
rest 15+ per subject / per articulator / half peak t-test time GLM electrodes 52 0 -1.56/-2.28/-2.27/-1.54 nnn not reproduced
rest 15+ pooled / all six / threshold F-test row all vSMC 268 0 -2.26/-2.31/-2.50/-2.21 nnn not reproduced
rest 15+ pooled / all six / threshold F-test row GLM electrodes 102 0 -1.52/-2.06/-2.11/-1.59 nnn not reproduced
rest 15+ pooled / all six / threshold F-test time all vSMC 268 0 -2.21/-2.27/-2.48/-2.20 nnn not reproduced
rest 15+ pooled / all six / threshold F-test time GLM electrodes 102 0 -1.56/-2.11/-2.15/-1.64 nnn not reproduced
rest 15+ pooled / all six / threshold t-test row GLM electrodes 96 0 -1.53/-1.96/-2.09/-1.58 nnn not reproduced
rest 15+ pooled / all six / threshold t-test time GLM electrodes 96 0 -1.56/-2.08/-2.13/-1.63 nnn not reproduced
rest 15+ pooled / all six / half peak F-test row all vSMC 268 0 -2.35/-2.42/-2.54/-2.24 nnn not reproduced
rest 15+ pooled / all six / half peak F-test row GLM electrodes 78 0 -1.49/-2.13/-2.14/-1.51 nnn not reproduced
rest 15+ pooled / all six / half peak F-test time all vSMC 268 0 -2.32/-2.35/-2.49/-2.21 nnn not reproduced
rest 15+ pooled / all six / half peak F-test time GLM electrodes 78 0 -1.53/-2.14/-2.17/-1.60 nnn not reproduced
rest 15+ pooled / all six / half peak t-test row GLM electrodes 57 0 -1.43/-2.12/-2.14/-1.48 nnn not reproduced
rest 15+ pooled / all six / half peak t-test time GLM electrodes 57 0 -1.46/-2.12/-2.21/-1.60 nnn not reproduced
rest 15+ pooled / per articulator / threshold F-test row all vSMC 268 0 -2.29/-2.38/-2.51/-2.23 nnn not reproduced
rest 15+ pooled / per articulator / threshold F-test row GLM electrodes 99 0 -1.50/-2.12/-2.17/-1.57 nnn not reproduced
rest 15+ pooled / per articulator / threshold F-test time all vSMC 268 0 -2.22/-2.30/-2.46/-2.20 nnn not reproduced
rest 15+ pooled / per articulator / threshold F-test time GLM electrodes 99 0 -1.53/-2.14/-2.16/-1.61 nnn not reproduced
rest 15+ pooled / per articulator / threshold t-test row GLM electrodes 90 0 -1.47/-2.15/-2.12/-1.48 nnn not reproduced
rest 15+ pooled / per articulator / threshold t-test time GLM electrodes 90 0 -1.49/-2.13/-2.10/-1.59 nnn not reproduced
rest 15+ pooled / per articulator / half peak F-test row all vSMC 268 0 -2.35/-2.43/-2.54/-2.24 nnn not reproduced
rest 15+ pooled / per articulator / half peak F-test row GLM electrodes 71 0 -1.44/-2.19/-2.25/-1.50 nnn not reproduced
rest 15+ pooled / per articulator / half peak F-test time all vSMC 268 0 -2.29/-2.37/-2.52/-2.21 nnn not reproduced
rest 15+ pooled / per articulator / half peak F-test time GLM electrodes 71 0 -1.46/-2.21/-2.29/-1.60 nnn not reproduced
rest 15+ pooled / per articulator / half peak t-test row GLM electrodes 49 0 -1.38/-2.25/-2.25/-1.44 nnn not reproduced
rest 15+ pooled / per articulator / half peak t-test time GLM electrodes 49 0 -1.41/-2.28/-2.25/-1.45 nnn not reproduced
Outcome over readings: not reproduced
Electrodes dropped for lack of a window T_A: 0 in 168 readings, up to 83 | n from 32 to 268
PC1 (pre, all, ('per subject', 'per articulator', 'threshold'), all vSMC, n=268): median paired differences and corrected p {'Li-To': '+0.03 p=1e+00', 'Li-La': '+0.13 p=1e-03', 'Li-Ja': '-0.08 p=1e-01', 'To-La': '+0.09 p=3e-02', 'To-Ja': '-0.14 p=2e-04', 'La-Ja': '-0.28 p=5e-11'}
PC2 (pre, all, ('per subject', 'per articulator', 'threshold'), all vSMC, n=268): median paired differences and corrected p {'Li-To': '+0.03 p=1e+00', 'Li-La': '+0.13 p=1e-03', 'Li-Ja': '-0.08 p=1e-01', 'To-La': '+0.09 p=3e-02', 'To-Ja': '-0.14 p=2e-04', 'La-Ja': '-0.28 p=5e-11'}
PC3 (pre, all, ('per subject', 'per articulator', 'threshold'), all vSMC, n=268): median paired differences and corrected p {'Li-To': '+0.03 p=1e+00', 'Li-La': '+0.13 p=1e-03', 'Li-Ja': '-0.08 p=1e-01', 'To-La': '+0.09 p=3e-02', 'To-Ja': '-0.14 p=2e-04', 'La-Ja': '-0.28 p=5e-11'}
Result: not reproduced. Under all 288 readings the condition fails in all three PCs. In the
reading printed in full (all 268 vSMC electrodes, pre, every CV), the tongue weights are not
larger than the lips weights (median paired difference of log10 weights, lips minus tongue, +0.03,
corrected p = 1), and jaw and larynx differ (corrected p = 5e-11). The input projected does not
enter this test, which uses the PCs only. Electrodes without a window T_A for an articulator are
dropped: none in 168 readings, up to 83 in the others (GP31 lacks a tongue window in some timing
readings), leaving 32 to 268 electrodes.
Summary¶
| Claim | Outcome |
|---|---|
| data-composition | partially reproduced |
| active-electrodes | blocked: methods_underspecified |
| fig1c-e-articulator-electrodes | blocked: metadata_insufficient |
| fig1f-h-phonetic-contrasts | not reproduced |
| fig2a-lips-dorsal | blocked: metadata_insufficient |
| fig2b-somatotopic-sequence | blocked: metadata_insufficient |
| suppfig3b-spatial-clustering | blocked: metadata_insufficient |
| suppfig3d-cross-subject | blocked: metadata_insufficient |
| suppfig3f-dv-sequence | blocked: metadata_insufficient |
| suppfig5-intermediate-organization | blocked: metadata_insufficient |
| suppfig6-mixed-tuning | reproduced |
| fig3d-timing-sequence | not reproduced |
| suppfig7-no-latency-difference | blocked: metadata_insufficient |
| pca-nine-components | blocked: methods_underspecified |
| fig4a-consonant-clusters | blocked: methods_underspecified |
| suppfig11-co-clustering | blocked: methods_underspecified |
| fig4c-d-hierarchy | blocked: methods_underspecified |
| fig4ef-feature-correlations | blocked: methods_underspecified |
| fig5c-separability-timing | not reproduced |
| fig5d-separability-coupling | blocked: methods_underspecified |
| suppfig13-consonant-vowel-distance | blocked: methods_underspecified |
| suppfig9-pc-articulator-weights | not reproduced |
A failure to reproduce here may be a failure of this attempt, not of the paper: see the reasoning and the differences from the paper's method.
How the analysis differs from the paper's. High gamma is computed here from the raw ECoG, as the Methods describe, with a filter bank whose widths the paper does not give (a standard choice) and with the dataset's bad-channel flags in place of the paper's inspection. The baseline for the z-scores is not given by the paper; the three readings of the claims file are run, the rest period the dataset marks among them. vSMC electrodes are those labelled precentral or postcentral (268, against the paper's 266). The acoustic onset of fig3d is the hand-marked syllable start, a substitute the criterion allows as a deviation since the audio is not in the dandiset; the result against t = 0 has the same outcome. The questions asked by this attempt were answered from the paper, which fixes none of the details asked about: whether CVs with fewer than 15 renditions are kept, how the state-space input is z-scored, and which input is projected onto the PCs are all left open, and every reading is run. The blocked outcomes of active-electrodes and of the state-space claims other than fig5d come from these readings giving different results; the paper cannot settle them. That of fig5d comes from the claims file's readings of R and of the clusters, which split it at every level of the readings above. The GLM weights used by fig1f-h, suppfig6 and suppfig9 come from claim fig2a-lips-dorsal, which is blocked for its spatial judgement only; the weights need no positions, so these claims are tested. Two variants that are not readings of the claims file are shown beside the results: leaving out trials that overlap intervals marked as ECoG artifacts (active-electrodes, fig3d), and the suppfig6 test on permuted weights.
What was found about the dataset. It holds four subjects; EC2, GP31 and GP33 fit the paper's
EC2, EC31 and EC33 (assumed, not shown), and EC9 is not in the paper. The audio was removed. No
electrode has a position relative to the central sulcus or the Sylvian fissure, and one of the
three subjects (GP33) has no coordinates at all, so seven spatial findings cannot be tested. The
coordinates of EC2 and GP31 have no unit or frame. GP31 has two renditions of /thi/ and none of
/tha/ or /thu/, and up to 107 renditions of other CVs; EC2 has 10 of /thi/; the paper states 15 to
100 per CV. Two labels are misspelt (she, ha). Electrode ids start at 0 in EC2 and GP31 and at 1
in GP33 and EC9. 162 of the 9370 kept trials overlap intervals marked as ECoG artifacts, and
leaving them out moves the active-electrode count of the one reading within tolerance out of it.
The subject fields hold only species and id, and nothing links GP31 and GP33 to the paper's EC31
and EC33; GP33's electrode group is auto_group. One file (EC2 block 1) holds a power spectrum
in a processing module.
In [28]:
print(f"notebook ran in {(time.time() - started) / 60:.1f} min")
notebook ran in 21.7 min