DANDI:000351

Animals were trained on Experiment 1 for 4-11 sessions (7 ± 2.1 sessions, mean ± std) ... (13.3 ± 2.7 sessions in total) ... (4.3 ± 0.8 sessions) ... (8.1 ± 2.5 sessions) ... (2.6 ± 0.8 sessions) ... (5.6 ± 2.4 and 7.1 ± 0.7 sessions for WT and DAT-Cre, respectively)

Methods, Behavioral task, paragraphs 1 to 5

Counted from the dandiset, the means per animal are 7.0 random reward sessions (4 to 11, SD 2.14), 12.71 of Experiment 2, 4.29 of Experiment 3, 7.86 of Experiment 5, 2.43 of Experiment 4, and 5.43 (WT) and 7.0 (DAT-Cre) sequential conditioning sessions, all within the stated SDs. The counts of integer days in the code's ranges are 13.29, 4.29, 8.14 and 2.57, which equal the paper's means for Experiments 2, 5 and 4; this suggests that the paper's counts include sessions on the 7 integer days in the ranges that have no session in the dandiset. The SDs of the Experiment 7 counts, 0.79 (wild type) and 2.45 (DAT-Cre, from datHT-M7's 12 sessions), are close to the paper's 2.4 and 0.7 in the other order, so the paper may have swapped them; the criterion tests only the means.

How the test differed from the paper’s method

• Experiment 5 is counted as the sessions with background rewards in the event log, which equals the count in the range of cumsum_bgdrw_vs_extinction.m (7.86); condition 4 of cumsum_pavlovian.m gives 7.14, also within tolerance.

we excluded rewards with less than 3 s IRI from the previous reward (22.6±2.4%) for the above analyses. Rewards without lick until the next reward (5.7±0.1%) were also excluded from analyses.

Methods, Data analysis, Experiment 1, paragraph 2

Reading (a), mean over the 8 animals, 22.59% of rewards with IRI under 3 s and 5.66% without a lick before the next reward; reading (b), over 56 sessions, 22.59% and 5.68%. Both within tolerance under both readings. The SD over animals is 2.39 and 1.04, and over sessions 7.04 and 4.27; the paper's 0.1 matches neither.

How the test differed from the paper’s method

• The paper does not say how the percentages are averaged; both readings of the criterion are shown and both match.

Every animal showed an increasing sucrose response that reached a high positive asymptote.

Figure 3D and 3E; Results, Tests 1 and 2, paragraph 3

What the paper reports

• one sample t-test against zero, of per-animal correlations, t(7) = 4.40, p = 0.0031 (n = 8 animals)

Per animal r between response and reward number, 0.068 to 0.677 (mean 0.308), positive in 8 of 8; t(7) = 4.41, p = 0.0031 (paper t = 4.40, p = 0.0031). Of 5600 rewards, 4064 are kept by the exclusions, 4013 of them with dF/F.

The experimentally observed correlation between dopamine response to sucrose and the previous IRI was positive, thereby being consistent with ANCCR but not RPE.

Figure 3G; Results, Tests 1 and 2, paragraph 5

What the paper reports

• one sample t-test against zero, of per-animal correlations, t(7) = 5.95, p = 5.7 × 10⁻⁴ (n = 8 animals)

Per animal mean of per-session r between response and previous IRI, 0.029 to 0.405; t(7) = 5.95, p = 5.7e-4 (paper t = 5.95, p = 5.7e-4).

The correlation between dopamine response to a reward and the previous IRI remains positive across sessions.

Figure S8K; Results, Tests 1 and 2, paragraph 5; Table S1

What the paper reports

• change of the per-session correlation over the first 7 sessions: linear regression, t = -0.04, p = 0.9707 (n = 50 sessions from 8 animals)

Linear regression of per-session r on session number over 50 sessions, slope -0.0005, t = -0.04, p = 0.969 (paper t = -0.04, p = 0.9707, n = 50). Mean r over animals is positive in each of sessions 1 to 7 (0.134 to 0.288).

The lick rates generally remained constant across repeated sessions.

Figure S8B; Supplementary Note 3, point 1; Table S1

What the paper reports

• total lick rate across sessions: linear regression, t = -0.79, p = 0.4389 (n = 50 sessions from 8 animals)
• consummatory lick rate across sessions: linear regression, t = -0.20, p = 0.8397 (n = 50 sessions from 8 animals)
• non-consummatory lick rate across sessions: linear regression, t = -0.58, p = 0.5618 (n = 50 sessions from 8 animals)

Over 50 sessions, total rate slope -0.038 licks/s per session (t = -0.78, p = 0.4389), consummatory -0.008 (t = -0.20, p = 0.8397), non-consummatory -0.027 (t = -0.58, p = 0.5618); the paper's p values.

The consummatory lick rate is not correlated with dopamine reward response.

Figure S8C and S8D; Table S1

What the paper reports

• one sample t-test of per-animal correlations, t = 0.23, p = 0.8245 (n = 8 animals)

Per animal r from -0.20 to 0.15 (mean 0.010) over 281 to 686 rewards; t(7) = 0.23, p = 0.824 (paper t = 0.23, p = 0.8245). 0 to 14 rewards per animal with an infinite or missing lick rate are dropped, as the code drops infinite values.

There was no change in the baseline dopamine activity across sessions.

Figure S8F and S8G; Methods, Data analysis, Experiment 1, paragraph 3; Table S1

What the paper reports

• one sample t-test of per-animal correlations, t = 0.11, p = 0.919 (n = 8 animals)

With peaks kept by the Methods' rule (more than 0.5 s from any lick and more than 2 s from any reward), mean r of peak size with session -0.034, t(7) = -0.13, p = 0.90; with the code's rule, mean r 0.018, t(7) = 0.07, p = 0.95 (paper t = 0.11, p = 0.919). Per animal r from -0.98 to 0.97. Sessions keep a median of 194 peaks (5 to 516) under the Methods' rule.

How the test differed from the paper’s method

• Peaks are found with scipy.signal.find_peaks (distance 50 samples, height and prominence as the code), in place of Matlab's findpeaks; NaN samples are set to the session minimum so that they form no peak.

Dopamine baseline activity change across sessions is uncorrelated with dopamine reward response change across sessions.

Figure S8H; Supplementary Note 3, point 4

Across the 8 animals, r between the baseline peak size correlation and the reward response correlation is -0.10 (p = 0.81) with the Methods' rule and -0.09 (p = 0.83) with the code's. The reward response correlation with session is positive in every animal (0.23 to 0.92).

The non-consummatory exploratory licks made by the animals depended on the average IRI for a given session. Thus, animals learn the average reward rate of an environment within at most two sessions.

Figure S8L; Table S1

What the paper reports

• non-consummatory lick rate against IRI condition: linear regression, t = -2.73, p = 0.0147 (n = 18 sessions from 6 animals)
• consummatory lick rate against IRI condition: linear regression, t = 1.00, p = 0.3338 (n = 18 sessions from 6 animals)

With the IRI in seconds, non-consummatory slope -0.0195 licks/s per s, t = -2.73, p = 0.0147, and consummatory t = 1.00, p = 0.334 (the paper's values, n = 18); with the rank of the condition, t = -2.61 (p = 0.019) and t = 1.34 (p = 0.199). Every animal has two sessions at each of 6, 12 and 30 s.

Dopamine cue response considerably leads behavior across animals. ... Behavioral learning is much more abrupt than dopaminergic learning (paired t test for abruptness of change; t(6) = 9.06; two-tailed p = 1.0×10-4; paired t test for change trial; t(6) = -2.93; two-tailed p = 0.0263; n=7 animals)

Figure 4C to 4F; Results, Tests 3-7, paragraph 1

What the paper reports

• abruptness of change, lick against dopamine: paired t-test, t(6) = 9.06, p = 10⁻⁴ (n = 7 animals)
• change trial, lick against dopamine: paired t-test, t(6) = -2.93, p = 0.0263 (n = 7 animals)

Methods' lick measure (0 to 2 s), lick abruptness 0.85 against dopamine 0.26, paired t = 5.06, p = 0.0023; change trial lick 337 against dopamine 177, t = 3.31, p = 0.016. Code's lick measure (cue and trace), t = 9.09, p = 1.0e-4 and t = 2.93, p = 0.026, the paper's values with the change-trial t as lick minus dopamine (the paper prints -2.93). Trials used, round(1.5 x trials before learnedday), 191 to 900. Of them 0 to 3 per animal have a NaN smoothed dF/F sample from -1 to 2 s and keep the nansum of the others, as aucsignal.m does; leaving them out gives t = 5.15, 3.31, 9.15 and 2.93.

Dopamine cue response reaches 93% of its response on the final session (~100% if normalized relative to average cue response on session n and n-1).

Figure S10C; Supplementary Note 5, paragraph 2

Reading (a), mean of per-animal ratios, cue response on the learned session over the last session 0.934 and over the mean of the last two 0.934; reading (b), ratio of means, 0.913 and 0.947. All within tolerance. Per animal the first ratio ranges from 0.59 to 1.33. The Fig 4 measure gives 0.985 and 0.955. One trial of the sessions compared has a NaN dF/F sample; without it reading (a) gives 0.938 and 0.938.

Behavior is learned abruptly by all animals, but the dopaminergic cue response shows little to no change. ... Paired t test for abruptness of change; t(6) = 22.92; two-tailed p = 4.52×10-7; one-sample t test for ΔAIC against a null of zero; t(6) = 7.49 for lick

Figure 4G to 4I; Results, Tests 3-7, paragraph 2

What the paper reports

• abruptness of change, lick against dopamine: paired t-test, t(6) = 22.92, p = 4.5 × 10⁻⁷ (n = 7 animals)
• delta AIC for licking against zero: one sample t-test, t(6) = 7.49, p = 2.9 × 10⁻⁴ (n = 7 animals)

Lick timing abruptness 0.233 against dopamine 0.064, paired t = 15.34, p = 4.9e-6 (paper t = 22.92). Changing model better than constant for lick timing in all 7 animals, mean delta AIC 382.8, t = 7.49, p = 2.9e-4 with k = 4 (the paper's t), and t = 7.53 with k = 3. 12 to 77 trials per animal without licks in cue and trace are dropped, as the code does. 0 to 2 trials per animal have a NaN smoothed dF/F sample from -1 to 2 s; leaving them out gives abruptness t = 15.40 and delta AIC t = 7.46. The abruptness t is smaller than the paper's while the delta AIC t equals it, for a reason not found here; HJ-FP-M4 lacks Day30, one of its three sessions before the change, and has 88 trials before it where the others have 143 to 150.

How the test differed from the paper’s method

• The changing model is fitted by bounded least squares (scipy) from 20 random start points with a fixed seed, keeping the best fit; Matlab's fit starts custom models at random points, so the code's fits are not repeated exactly.
• dF/F is divided by 100 before the dopamine AUC, as the files store percent and the Methods define dF/F as the fraction (470 - fitted 405)/fitted 405 (Methods, Fiber photometry). The scale for which the code's bounds were set is not stated; see fig4hi-delay-change-dopamine-unchanged.

Experimentally, we observed that while the animals learned the new delay rapidly, dopaminergic cue response showed no significant change (Fig 4 G-I).

Figure 4G to 4I; Results, Tests 3-7, paragraph 2

What the paper reports

• delta AIC for dopamine against zero: one sample t-test, t(6) = -0.86, p = 0.4244 (n = 7 animals)

Dopamine delta AIC (constant minus changing) per animal -168.5 to 32.7, mean -18.7; t(6) = -0.73, p = 0.49 with k = 4 (paper t = -0.86, p = 0.42), and t = -0.65, p = 0.54 with k = 3. No change is found. Without HJ-FP-M3 (delta AIC -168.5), t(5) = 0.99, p = 0.37. Leaving out trials with a NaN dF/F sample gives t = -0.73. With the code's bounds, which the criterion fixes, this test cannot detect a change in these data. On dF/F as a fraction, the scale used, b is at most 10 where 5 of 7 animals have mean responses of 13.2 to 44.6; three animals have delta AIC of exactly -6.0, the changing model collapsed to the constant one, and in HJ-FP-M3 the changing model fits worse than the constant one. On dF/F in percent, the stored scale, the bounded model cannot fit the data (delta AIC -272 to -817 in every animal, t = -7.74), so that variant does not test the scale. Fitted without bounds the changing model is better in every animal (delta AIC 14 to 38, t = 7.84, p = 2.3e-4). The outcome rests on the assumption that the paper's fit was bounded as here; the paper's t of -0.86, close to -0.73, suggests it was. Which dF/F scale the bounds were set for is asked of the authors (findings).

How the test differed from the paper’s method

• The model fit and the dF/F scale are as in fig4hi-delay-change-behavior-changes.
• Assumed that the paper's fit applied the code's bounds (A -2 to 50, b -10 to 10) to dopamine values on a scale where they bind as they do here. Neither the Methods nor the code states the scale of the dF/F the Matlab scripts read from the authors' local files. The criterion fixes the bounds; the two scales the dandiset supports (fraction and percent) both give no change, and the paper's t (-0.86) is close to the one found here (-0.73), while the unbounded fit, outside the criterion, finds a change in every animal.

Omission response is not negative with respect to pre-cue baseline for any animal after cue duration change but was negative during extinction.

Figure 4J and S10B; Results, Tests 3-7, paragraph 2; Methods, Experiments 2-5, paragraph 4

What the paper reports

• cumulative omission-window response, delay change against extinction: paired t-test, t = 3.37, p = 0.015 (n = 7 animals)

Cumulative response at the old reward time on the first 8 s session is at or above 0 in 7 of 7 animals; the extinction omission response is negative on average (Methods' measure, mean -87.25, 6 of 7 negative; code's, -45.75, 5 of 7). Paired t = 3.03, p = 0.023 (Methods) and t = 3.42, p = 0.014 (code; paper t = 3.37, p = 0.0150). The criterion's reading (b), negative on average, is met; reading (a), in every animal, is not (F1 is positive under both measures). Leaving out the 0 to 4 trials per session with a NaN dF/F sample in the windows gives t = 3.01, p = 0.024.

Across all animals, the dopaminergic cue response remains significantly positive despite abrupt behavioral learning of extinction (paired t test for abruptness of change; t(6) = 5.67; two-tailed p = 0.0013; paired t test for change trial; t(6) = -2.40; two-tailed p = 0.0531; n=7 animals).

Figure 4J to 4L

What the paper reports

• abruptness of change, lick against dopamine: paired t-test, t(6) = 5.67, p = 0.0013 (n = 7 animals)
• change trial, lick against dopamine: paired t-test, t(6) = -2.40, p = 0.0531 (n = 7 animals)

Lick abruptness exceeds dopamine abruptness with the Methods' lick measure (1.09 against 0.13, t = 3.28, p = 0.017) and the code's (0.47 against 0.13, t = 9.73, p = 6.8e-5; paper t = 5.67, p = 0.0013). The change-trial comparison gives t = -2.95, p = 0.026 (Methods) and t = -2.40, p = 0.0531 (code, the paper's values). 125 trials per animal, 0 to 4 of them with a NaN dF/F sample from -1 to 2 s; without them the abruptness t values are 3.43 and 9.39. The code variant's abruptness t (9.7) is larger than the paper's 5.67 while its change-trial t equals the paper's, for a reason not found here.

As predicted by ANCCR but not RPE, dopamine cue response remained significantly positive well after animals cease to behaviorally respond to the cues (Fig 4J-L)

Figure 4J to 4L; Results, Tests 3-7, paragraph 3

(a) Last extinction session, mean CS+ response positive across animals, t(6) = 4.31, p = 0.0050 (4.33 without the 2 trials with a NaN dF/F sample); (b) trials after the lick change trial, t(6) = 4.16, p = 0.0060 with the code's lick measure and t(6) = 3.78, p = 0.0092 with the Methods'; (c) positive at p < 0.05 within the last session in 7 of 7 animals.

How the test differed from the paper’s method

• For reading (b) the trials after the lick change trial run to the end of the extinction sessions; the change trial is taken from both the code's and the Methods' lick measures, and both give every animal a positive mean after it (2.2 to 30.6).

We did not observe any negative dopamine response to the intermediate cue regardless of how baseline was measured, and instead observed a positive but weaker response

Figure 5C and 5D; Results, Test 8, paragraph 3

What the paper reports

• ratio of intermediate to previous cue response against zero: one sample t-test, t(6) = 6.64, p = 5.6 × 10⁻⁴ (n = 7 animals)

Per animal ratio of intermediate to previous cue response 0.12 to 0.68, mean 0.435; t(6) = 6.66, p = 5.6e-4 against 0 (paper t = 6.64), and below 1 (t = -8.66, p = 1.3e-4). 7 to 18 pairs per animal over 2 to 4 sessions. None of the 190 cues of the pairs has a NaN dF/F sample in its windows.

and reward response (One sample t test against a null of one; t(6) = 2.95; two-tailed p value = 0.0256; n=7 animals)

Figure 5D, right

What the paper reports

• ratio of intermediate to previous reward response against one: one sample t-test, t(6) = 2.95, p = 0.0256 (n = 7 animals)

Per animal ratio of intermediate to previous reward response 1.08 to 3.25, mean 1.85; t(6) = 2.96, p = 0.025 against 1 (paper t = 2.95, p = 0.0256).

Here, we measure baseline as the dopamine fluorescence immediately prior to the intermediate cue delivery, as opposed to the baseline prior to the first cue in Fig 5C, D. Cue response was defined as baseline subtracted peak response. The intermediate cue response was still positive.

Figure S11D and S11E; Methods, Experiment 6, paragraph 2

What the paper reports

• one sample t-test of the cue response ratio, t = 10.42, p = 4.6 × 10⁻⁵ (n = 7 animals)

Under all four readings (peak over 0 to 0.5 s or 0 to 1 s, baseline the last sample or the mean of the 0.5 s before the cue) the ratio of intermediate to previous peak response is positive in every animal (minimum 0.135), mean 0.25 to 0.45, t = 5.21 to 8.35, p <= 0.002 (paper t = 10.42).

How the test differed from the paper’s method

• The peak window and the baseline are not fixed by the paper; the four readings of the criterion are run and agree.

Our observations were not consistent with a backpropagating bump of activity and were instead consistent with an increase in cue response over trials of learning (Fig 6B)

Figure 6A and 6B; Results, Tests 9-11, paragraph 2; Supplementary Note 9

What the paper reports

• late minus early normalized response, first 100 trials: paired t-test, t = -0.16, p = 0.8771 (n = 7 animals)

Normalized late minus early response over the first 100 CS+ trials, -0.008 with the Methods' baseline (t = -0.23, p = 0.83) and -0.006 without (t = -0.16, p = 0.876; paper t = -0.16, p = 0.8771). 473 to 900 trials per animal, so all have the 400 the normalizer needs. Leaving out the 0 to 4 trials per animal with a NaN dF/F sample from -1 to 3 s gives t = -0.22, p = 0.84.

Consistent with ANCCR, we experimentally found that cue2 and cue1 responses increased together early in learning prior to separating later in learning (Fig 6D).

Figure 6C and 6D; Results, Tests 9-11, paragraph 2

What the paper reports

• CS2 minus CS1 normalized response, early in learning: paired t-test, t = -0.03, p = 0.9752 (n = 7 animals)

Normalized CS2 minus CS1 over the first 50 trials, mean -0.003, t = -0.03, p = 0.975 without a baseline (the paper's values), -0.002 (p = 0.98) and -0.001 (p = 0.99) with the 1 s and 1.5 s baselines. 225 to 350 trials per control animal.

How the test differed from the paper’s method

• The Methods' baseline window is not given; both readings of the criterion are run and agree with the code's unsubtracted measure.

cue2 and cue1 responses increased together early in learning prior to separating later in learning

Figure 6D, right; Results, Tests 9-11, paragraph 2

CS2 over CS1 response over trials 151 to 200, mean 0.549 (0.30 to 0.71) without a baseline, t(6) = -8.01, p = 2.0e-4 against 1; 0.573 and 0.574 with the 1 s and 1.5 s baselines (p <= 1.1e-4).

we first verified robust inhibition of mesolimbic dopamine release during the cue2→reward delay in the experimental group (~0.6 times the reward response on day 1 of conditioning)

Figure 6F and 6G; Results, Tests 9-11, paragraph 2; Supplementary Note 10

What the paper reports

• dopamine in the last 0.5 s before reward, DAT-Cre against controls: two sample t-test, t = -5.74, p = 1.3 × 10⁻⁴ (n = 13 animals)
• reward response on day 1, DAT-Cre against controls: two sample t-test, t = -3.99, p = 0.0021 (n = 13 animals)

With the baseline before CS1, dopamine 1 to 1.5 s after CS2 over the reward response is lower in the 6 DAT-Cre mice than in the 7 controls under every reading (t = -8.62, p = 3.2e-6 for (a) and (c); t = -6.86, p = 2.7e-5 for (b)), and the day 1 reward response over the random reward session's is lower (0.53 against 1.48, t = -4.13, p = 0.0017 with the 1 s window; t = -3.76, p = 0.0032 with 0.5 s). The DAT-Cre mean inhibition is -0.565 under reading (b), within tolerance of -0.6; (a) -0.247 and (c) -0.494 are not. With the code's baseline over every code-15 event (CS1 and CS2), (b) gives -0.632, t = -5.74, p = 1.3e-4, and the day 1 reward test with the 0.5 s window t = -3.99, p = 0.0021, both the paper's values. The excluded DAT-Cre mouse M8 has a negative day 1 reward response (-0.43), as Fig S12D says.

How the test differed from the paper’s method

• The claim's baseline is the 0.5 s before CS1; measure_inhibition_sequential.py takes it before every code-15 event, CS1 and CS2 alike (reftimes). Both are shown; the code's reproduces the paper's t values exactly.
• For the 1 s reward window the 0.5 s baseline AUC is doubled, so that the baseline covers a window of equal length (claims header); a choice made here, which reading (b) does not involve.

Consistent with ANCCR, we observed that every experimental animal learned the task and that mesolimbic dopamine acquired positive responses to cue1 in all experimental animals (Fig 6I).

Figure 6I; Results, Tests 9-11, paragraph 2; Table S1

What the paper reports

• CS1 response across sessions, wild type: linear regression, t = 5.49, p = 4.3 × 10⁻⁶ (n = 35 sessions from 7 WT animals)
• CS1 response across sessions, DAT-Cre: linear regression, t = 2.55, p = 0.0166 (n = 30 sessions from 6 DAT-Cre animals)
• anticipatory licks across sessions, wild type: linear regression, t = 3.61, p = 9.9 × 10⁻⁴ (n = 35 sessions from 7 WT animals)
• anticipatory licks across sessions, DAT-Cre: linear regression, t = 4.92, p = 3.4 × 10⁻⁵ (n = 30 sessions from 6 DAT-Cre animals)

Over the first three and last two sessions, CS1 response slope 0.134 per session in controls (t = 4.17, p = 2.1e-4, n = 35) and 0.076 in DAT-Cre mice (t = 2.87, p = 0.0077, n = 30); lick slopes 1.53 (t = 3.67, p = 8.4e-4) and 1.28 (t = 4.22, p = 2.3e-4) (paper t = 5.49, 2.55, 3.61, 4.92). In the last session all 6 DAT-Cre animals have a positive CS1 response (0.12 to 0.65 of the day 1 reward response) and positive anticipatory licking (0.52 to 10.78 licks). HJ-FP-WT-stGtACR-M2's last session has no dF/F for its CS1 trials 1 to 17, and its mean is over the other 33.

Not quoted: Only Table S1 gives this comparison (Behavior (right), t = 0.88, p = 0.3999); no sentence states it. The Supplementary Note 10 says the experimental animals learned "albeit slower than control animals".

What the paper reports

• two sample t-test, t = 0.88, p = 0.3999 (n = 13 animals)

Last-session anticipatory licks 7.25 in controls against 4.84 in DAT-Cre mice, difference 2.41, t(11) = 1.17, p = 0.27 (paper t = 0.88, p = 0.40). No difference is found.

Quantification of pre-CS2 anticipatory licks showing reward-seeking prior to CS2 (onset of dopamine neuron inhibition).

Figure S12F; Table S1

What the paper reports

• session effect: two-way mixed ANOVA, F = 9.07, p = 0.0118 (n = 13 animals)
• genotype effect: two-way mixed ANOVA, F = 0.25, p = 0.6241 (n = 13 animals)
• genotype by session interaction: two-way mixed ANOVA, F = 0.07, p = 0.7915 (n = 13 animals)

Pre-CS2 licks rise from 0.33 to 1.62 between the first and last sessions; session F(1, 11) = 7.39, p = 0.020 (paper F = 9.07, p = 0.0118), genotype F = 0.57, p = 0.47, interaction F = 0.0003, p = 0.99. With the code's baseline window (1500 s before CS1), session F = 5.23, p = 0.043, genotype p = 0.57, interaction p = 0.60.

All 28 subjects of the code's lists are in the dandiset, matched with _ read as -. Of the 119 listed days (startday, endday, learnedday and the single days of omission.m), 117 are sessions with the condition the script names; HJ-FP-M4 Day38, the first background reward day in cumsum_pavlovian.m (condition 4) and cumsum_bgdrw_vs_extinction.m, is not in the dandiset. All 320 sessions inside the ranges have the condition (3 s or 9 s cue-reward delay, background rewards, no delivered reward, trial-less task). Counted literally, 320 of the 331 integer days in the ranges are sessions; the 7 missing days inside the Pavlovian ranges are M2 Day23, M3 Day16 and Day31, M4 Day30 and Day38, F2 Day16 and M6 Day49. The subjects match and more than 90% of the listed days do, so the link holds in part. The dependent claims are tested on the sessions present; the missing sessions bear on fig4p-background-faster-decay (M4's first background session) and on the session counts. The spreadsheet named in the dandiset's description was not used, as it has no version to pin. The notebook reads the six scripts at the pinned commit and finds that every mouse and day list in jeong.py equals the script's, and that cue_acquisition.m and backpropagation_pavlovian.m use the initial learning days of cumsum_pavlovian.m.

How the test differed from the paper’s method

• Day numbers are read from the file names as load_dandi_url does; the Matlab scripts read them from folder names of the authors' local copies, which the dandiset does not hold.
• The criterion's every Day number from startday to endday is reported both for the listed days and literally for every integer day; the outcome uses the listed days, since integer days without a session can be days without training.

The dopamine cue response remained significantly positive but decayed across trials faster than during extinction (Fig 4M-P).

Figure 4M to 4P; Results, Tests 3-7, paragraph 4

What the paper reports

• background rewards against extinction: paired t-test, t(6) = -3.51, p = 0.0126 (n = 7 animals)

The background reward cue response stays positive after the change (mean of the 90 post-change trials over animals, t(6) = 3.40, p = 0.015). It decays further than in extinction on average under every reading, but not significantly; (a) normalized cumulative sum at trial 140, 1.60 against 2.13, paired t = -2.40, p = 0.053; (b) mean over the 90 trials, t = -1.97, p = 0.097; (c) post over pre mean, t = -2.40, p = 0.053 (paper t = -3.51, p = 0.0126). Background is below extinction in 5 of 7 animals. The first background reward session of HJ-FP-M4 (Day38) is not in the dandiset, so that animal's 90 post-change trials start a session later; without M4, reading (a) gives t(5) = -1.85, p = 0.12. Readings (a) and (c) are one affine function of the other ((a) = 1 + 1.8 (c)), so only two independent readings were tested. The criterion asks for background below extinction at p < 0.05 under every reading, which none meets, so the main part of the claim, the faster decay, is not reproduced. Leaving out the 0 to 4 trials per animal and condition with a NaN dF/F sample gives p = 0.048 for (a), 0.047 for (c) and 0.088 for (b). With the sessions of cumsum_pavlovian.m (conditions 4 and 3, one session before each) no reading shows a difference (t = 0.69 and 0.72); there HJ-FP-M6's background sum over the 50 trials before the change is 37.3 (288.1 with the sessions of cumsum_bgdrw_vs_extinction.m), giving it a reading (a) of 12.15, and HJ-FP-M2's is -492.9.

How the test differed from the paper’s method

• What the paired t-test compares is not in the Methods, legend or code; the three readings of the criterion are run and agree that background is not below extinction at p < 0.05.
• HJ-FP-M4's first Experiment 5 session is missing from the dandiset (session-day-link).

mesolimbic dopamine acquired positive responses to cue1 in all experimental animals, albeit at a lower asymptote relative to the control group (Fig 6I)

Figure 6I; Supplementary Note 10, paragraph 2; Table S1

What the paper reports

• two sample t-test, t = 2.80, p = 0.0171 (n = 13 animals)

Last-session normalized CS1 response, controls 0.690 (n 7) against DAT-Cre 0.359 (n 6), lower on average but t(11) = 1.70, p = 0.117 (paper t = 2.80, p = 0.0171). One control, WT-M3, is at 0.077, below every DAT-Cre value; its last session has 25 trials, where every other animal's has 50. The outcome turns on this session and on the normalisation; with WT-M3's previous session in its place t = 2.43 (p = 0.033), normalised by the random reward session's reward response t = 2.84 (p = 0.016), and not normalised t = 2.85 (p = 0.016). Supplementary Note 10 and learnign_curve_sequential.py normalise by the reward response on day 1 of conditioning, as the criterion does, and no passage excludes a short session, so the outcome is kept on the criterion's measure. HJ-FP-WT-stGtACR-M2's last session (Day6) has no dF/F for its first 1117 s, so its CS1 trials 1 to 17 have no value and its mean is over the other 33; with its previous session in place of it, its value is 0.462 against 0.459 and t = 1.70, p = 0.117. No other Experiment 7 CS1 trial lacks a value.

How the test differed from the paper’s method

• The measure follows learnign_curve_sequential.py (session means of the CS1 AUC over 0 to 1.5 s minus the 1.5 s before, over the first conditioning session's mean reward response).

The average lick bout onset response was positive for consummatory lick bouts (during sucrose consumption) but was slightly negative for dry lick bout onsets away from reward consumption.

Figure S8E; Table S1

What the paper reports

• consummatory against non-consummatory bout onset response: paired t-test, t = 3.28, p = 0.0134 (n = 8 animals)
• consummatory bout onset response against zero: one sample t-test, t = 3.03, p = 0.0191 (n = 8 animals)
• non-consummatory bout onset response against zero: one sample t-test, t = -2.55, p = 0.0379 (n = 8 animals)

Under every reading the consummatory onset response is above 0 (mean z 1.76 to 3.07, t = 6.30 to 6.49, p <= 4.0e-4) and above the non-consummatory one (paired t = 7.17 to 8.58, p <= 1.8e-4). The non-consummatory response is negative on average under all readings but significant only under (c), 0 to 0.5 s (z = -0.35, t = -2.56, p = 0.0376, against the paper's t = -2.55, p = 0.0379); under (a) 0 to 1 s z = -0.15, t = -1.65, p = 0.14, and under (b) -0.5 to 1 s z = -0.02, p = 0.80. The window is open (Methods, legend and code do not give it) and the readings give different outcomes, so under the claims file's rule for readings of an open method detail the claim is blocked; the question is put to the authors under findings. The consummatory t values are about twice the paper's (3.03), so the paper's response measure differs from these readings.

How the test differed from the paper’s method

• The window of the tests is given by neither the Methods, the legend nor the code; the three readings of the criterion are run, and they give different outcomes for the non-consummatory test.