Student or educator? Sign up with your school email and get 6 months of Pro free.

The DOI matches its Crossref record; 102 of the 102 references it cites were found in a registry or an index; 1 of the 61 sources its claims cite were checked for support.

Revealing abrupt transitions from goal-directed to habitual behavior

Open access

Abstract

The speed of goal-directed to habit transitions has been debated since Clark Hull asked in 1943: is habit formation slow or sudden? To address this, male mice were given home-cage access to citric-acid water that reduced—without eliminating—reward-seeking for plain water in an auditory go/no-go task. Animals learned to discriminate quickly but exhibited ongoing state-like fluctuations in engagement. Strikingly, these fluctuations abruptly ceased (transition) long after discrimination stabilized, with HMM-GLM modeling pinpointing a ~3-trial transition. We confirmed this as a goal-directed to habit transition using sensory-specific outcome devaluation, DLS lesions, motor stereotypy, and pupillary responses. Dual-site fiber photometry showed equally abrupt DLS dynamics at the transition: outcome-related activity dropped and stimulus-response activity sharpened, suggesting a switch-like mechanism that recruits a readily available habit circuit rather than gradual changes across a threshold. Thus, habits can emerge suddenly, mediated by an abrupt dorsostriatal shift from outcome- to stimulus-driven processing. Whether habits emerge gradually or suddenly remains an open question. Here the authors show that mice abruptly switch from goal-directed to habitual control, marked by a rapid dorsal striatal shift from outcome- to stimulus-driven processing.

Overview

CoChat Summary

TL;DR

In male mice trained on an auditory go/no-go task, reduced reward-seeking from home-cage citric-acid water was associated with an abrupt, not gradual, transition from goal-directed to habitual control. Multiple behavioral, lesion, pupillometry, and dorsal striatal photometry measures converged on a switch-like habit transition occurring in about 3 trials.

Key findings

  • Self-restricted CA mice lost less weight than standard water-restricted mice before training: 8.9% ± 1.9% vs 17.8% ± 2.3% (p = 5.5×10-5), and they obtained fewer rewards during lick training (p = 0.00079).
  • In the go/no-go task, CA mice showed 10/12 (83.3%) high-variability engagement phases, and 8/10 (80%) of those transitioned to low variability; these transitions happened >1000 trials after expert discrimination (d’ > 2 expert = 1436 ± 454 trials vs transition = 3832 ± 385 trials).
  • Sensory-specific outcome devaluation showed that pre-transition behavior was goal-directed (devalued hit rates lower than valued; two-way repeated-measures ANOVA X2 = 5.61, p = 0.018), but post-transition the devaluation effect disappeared; the devaluation index decreased across the transition (t(6) = 2.42, p = 0.026).
  • DLS lesions slowed or prevented the transition: 60% of sham CA mice vs 20% of DLS-lesioned mice transitioned to low variability, without deficits in contingency learning or lick-related motor behavior.
  • Lick microstructures became more automatic after transition, with increased consummatory licks (t(7) = -4.87, p = 0.0018) and shorter reaction time (W = 28, p = 0.016); pupillary responses also became less outcome-sensitive, with post-transition effects visible within 10 trials of the transition (z = -2.78, p = 0.0054).

Methods

  • Male C57BL/6J mice (11–15 weeks old at training start) were assigned to standard water restriction or self-restriction via ad libitum citric-acid (CA) water in the home cage; 45/48 mice with d’ > 2 were included.
  • Animals were trained on a head-fixed auditory go/no-go task with target and foil tones; the study used behavioral analysis, sensory-specific outcome devaluation, bilateral DLS lesions, pupillometry, and dual-site fiber photometry in DLS and DMS.
  • Main analyses included blockwise behavior metrics, sensory-specific devaluation tests, HMM-GLM modeling of trial-by-trial choice, and statistical testing with t-tests, Wilcoxon tests, ANOVA/ANCOVA, and mixed-effects models.

Limitations

  • The authors note that pan-neuronal fiber photometry is a bulk measure and does not resolve cell types or projections; this limits mechanistic interpretation.
  • They state that calcium sensor kinetics are slow relative to task timing, making it hard to fully disentangle cue, action, outcome, and evaluative signals.
  • Evident from the text: the study is restricted to male mice and to a specific head-fixed go/no-go paradigm, so generalization beyond this preparation is not directly tested.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41467-026-71048-0

Cites

102 of 102 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Self-restriction reduces reward-seeking for plain water without impacting discrimination learning. a Protocol outline: after head-post implantation, mice underwent a habituation period and introduction to the water restriction paradigms. ‘Control’ mice underwent a common (maintenance at 80-85% original body weight) water restriction paradigm (WR85, blue). CA mice had ad libitum access to a water bottle with a low percentage of a less palatable hydration source (citric acid laced water) in their home cage (CA, yellow) to which they were progressively introduced (starting with 0.5%, reaching maximum 3%). After a few days, both cohorts underwent two days of lick training (lavender), followed by an auditory go/no-go training (black). b CA mice lost significantly less weight compared to WR85. Median diff. = 10.06, n = 11/12 mice, two-sided Wilcoxon rank-sum test, z = 4.03, p = 0.000055, HL = 8.78, 95% CI[7.37, 11.05]. c CA mice obtain significantly fewer rewards compared to WR85 in one lick training session. Median diff. = 75.00, n = 11/12 mice, two-sided Wilcoxon rank-sum test, z = 3.36, p = 0.000791, HL = 77.50, 95%CI [43.00, 110.00]. d (i) After lick-training, mice underwent auditory cued go/no-go training that consisted of ~300 trials per session. (ii) Mice learn to lick after a S+ tone to obtain a plain water reward (3 µl) and withhold licking to an S- tone to avoid a time-out. (iii) Correct responses are hits (licking to the S+ tone) and correct rejects (withhold licking to the S-), while incorrect responses are false alarms (licking to the S-) and misses (not licking to the S + ). e Accuracy comparison between WR85 (blue, n = 11) and CA (yellow, n = 12) mice on highly engaged trial blocks is similar. Two-way ANOVA revealed a significant effect of Group (F(1, 319) = 5.45, p = 0.020, ηp2 = 0.017), and a significant Trial_block main effect (F(25, 319) = 7.93, p = 1.47 × 10-21, ηp2 = 0.38), and no significant Group × Trial_block interaction effect (F(25, 319) = 0.80, p = 0.741, ηp2 = 0.059). Bonferroni-corrected post hoc pairwise t-test identified: Group=2,Trial_block=2 vs Group=1,Trial_block=9 (mean diff. = -0.33, 95%CI [-0.64,-0.010], p_adj = 0.025); Group=2,Trial_block=2 vs Group=1,Trial_block=10 (mean diff. = -0.32, 95%CI [-0.63,-0.020], p_adj = 0.015); Group=2,Trial_block=2 vs Group=1,Trial_block=11 (mean diff. = -0.33, 95%CI [-0.65,-0.00], p_adj = 0.047), plus 109 additional significant contrasts; full statistics are provided in Supplementary Table (‘SuppTable PostHoc Bonferroni Accuracy’). Expert accuracy is defined as 75% correct (gray horizontal line). f No differences were observed in the number of trials to reach expert accuracy between groups. Median diff. = 100.00, n = 11/12, two-sided Wilcoxon rank-sum test: z = -0.090, p = 0.93, HL = -50.00, 95%CI [-1004.00, 304.00]. Except a, b all data are presented as mean values ± SEM.
Fig. 1 Self-restriction reduces reward-seeking for plain water without impacting discrimination learning. a Protocol outline: after head-post implantation, mice underwent a habituation period and introduction to the water restriction paradigms. ‘Control’ mice underwent a common (maintenance at 80-85% original body weight) water restriction paradigm (WR85, blue). CA mice had ad libitum access to a water bottle with a low percentage of a less palatable hydration source (citric acid laced water) in their home cage (CA, yellow) to which they were progressively introduced (starting with 0.5%, reaching maximum 3%). After a few days, both cohorts underwent two days of lick training (lavender), followed by an auditory go/no-go training (black). b CA mice lost significantly less weight compared to WR85. Median diff. = 10.06, n = 11/12 mice, two-sided Wilcoxon rank-sum test, z = 4.03, p = 0.000055, HL = 8.78, 95% CI[7.37, 11.05]. c CA mice obtain significantly fewer rewards compared to WR85 in one lick training session. Median diff. = 75.00, n = 11/12 mice, two-sided Wilcoxon rank-sum test, z = 3.36, p = 0.000791, HL = 77.50, 95%CI [43.00, 110.00]. d (i) After lick-training, mice underwent auditory cued go/no-go training that consisted of ~300 trials per session. (ii) Mice learn to lick after a S+ tone to obtain a plain water reward (3 µl) and withhold licking to an S- tone to avoid a time-out. (iii) Correct responses are hits (licking to the S+ tone) and correct rejects (withhold licking to the S-), while incorrect responses are false alarms (licking to the S-) and misses (not licking to the S + ). e Accuracy comparison between WR85 (blue, n = 11) and CA (yellow, n = 12) mice on highly engaged trial blocks is similar. Two-way ANOVA revealed a significant effect of Group (F(1, 319) = 5.45, p = 0.020, ηp2 = 0.017), and a significant Trial_block main effect (F(25, 319) = 7.93, p = 1.47 × 10-21, ηp2 = 0.38), and no significant Group × Trial_block interaction effect (F(25, 319) = 0.80, p = 0.741, ηp2 = 0.059). Bonferroni-corrected post hoc pairwise t-test identified: Group=2,Trial_block=2 vs Group=1,Trial_block=9 (mean diff. = -0.33, 95%CI [-0.64,-0.010], p_adj = 0.025); Group=2,Trial_block=2 vs Group=1,Trial_block=10 (mean diff. = -0.32, 95%CI [-0.63,-0.020], p_adj = 0.015); Group=2,Trial_block=2 vs Group=1,Trial_block=11 (mean diff. = -0.33, 95%CI [-0.65,-0.00], p_adj = 0.047), plus 109 additional significant contrasts; full statistics are provided in Supplementary Table (‘SuppTable PostHoc Bonferroni Accuracy’). Expert accuracy is defined as 75% correct (gray horizontal line). f No differences were observed in the number of trials to reach expert accuracy between groups. Median diff. = 100.00, n = 11/12, two-sided Wilcoxon rank-sum test: z = -0.090, p = 0.93, HL = -50.00, 95%CI [-1004.00, 304.00]. Except a, b all data are presented as mean values ± SEM.CC BY-NC-ND · PubMed Central
Fig. 2 Abrupt state-like transitions from goal-directed to habitual behavior appear spontaneously within individual animals. a Hit (color) and FA (black) action rates for example WR85 (left) and a CA (right) animals. b Hit rate of CA example animal (in a) showing periods of high (gray shadow) and low engagement, followed by a spontaneous transition (red vertical line) to low hit rate variability. c CA mice have significantly less engaged blocks compared to WR85. Median diff. = 13.92, n = 11/12 mice, two-sided Wilcoxon rank-sum test: z = 2.98, p = 0.0028, HL = 19.27, 95%CI [8.90,37.32]. d (i) Most CA (8/12, 66.7%) showed hit-rate variability and the presence of a transition. Some CA mice that transitioned to low variability hit rate, seemed to transition to high variability after a while (asterisks). (ii) Only a few CA mice showed no hit rate variability (2/12, 16.7%). (iii) Few CA mice showed initial hit rate variability, but never transitioned to low variability (2/12, 16.7%). e An HMM-GLM was used to model behavioral data, which allows us to analyze the state-like nature of transitions. f An HMM-GLM with two states provides the best fit for most of the CA animals based on a BIC analysis (n = 8 CA mice). g Both states are equally stimulus-driven, but state 2 is characterized by a no-go, or disengaged bias (n = 8 CA mice). h State 1 (pink) is highly stimulus selective between target and foil trials with high engagement, while State 2 (green) shows overall task disengagement (n = 8 CA mice). i A CA exemplar shows that the two-state HMM-GLM model accurately recapitulates the behavior (top), and the two states govern the pre-transition phase, while only one state becomes explanatory of the post-transition phase. j (i) The GLM states reflect transitions between an engaged state (state 1, pink, hit rate = 0.98) and a disengaged state (state2, olive, hit rate = 0.35). Across all mice, the HMM-GLM model predicted that the probability of staying in engaged or disengaged state (trial-by-trial) is 0.995 and 0.975 respectively, whereas the transition probability between states is 0.005 (engaged to disengaged) and 0.025 (disengaged to engaged). (ii) State occupancy pre-transition (black) is approximately divided 50%–50% between State 1 and State 2, while post-transition (red), State 1 dominates (n = 8 CA mice). Two-way ANOVA revealed a significant State main effect (F(1, 28) = 62.20, p = 1.37×10-8, ηp2 = 0.69), and no significant Transition main effect (F(1, 28) = 0.020, p = 0.90, ηp2 = 0.0010), and a significant State × Transition interaction effect (F(1, 28) = 0.99, p = 8.7×10-5, ηp2 = 0.43). Bonferroni-corrected post hoc pairwise comparisons across State × Transition cell means identified: State=State1,Transition=Pre vs State=State1,Transition=Post (mean diff. = -0.30, 95%CI [-0.56,-0.040], p_adj = 0.015); State=State1,Transition=Pre vs State=State2, Transition=Post (mean diff. = 0.50, 95%CI [0.24,0.76], p_adj = 4.38×10-5); State=State2, Transition=Pre vs State=State1, Transition=Post (mean diff. = -0.51, 95% CI [-0.77, -0.26], p_adj = 2.72×10-5), plus 2 additional significant contrasts; full statistics are provided in Supplementary Table (‘SuppTable PostHoc Bonferroni Occupancy’). k, No-go to go (State 2 → State 1) transitions and go to no-go (State 1 → State 2) transitions happening in the goal-directed phase, are abrupt (n = 8 CA mice). l, No differences between the first (blue, belonging to the goal-directed phase), and last (red, belonging to goal-directed to habitual) transitions in abruptness. Both happen within 1-4 trials (n = 8 CA mice). Two-sided Wilcoxon rank-sum test: U = 30, p = 0.89, mean diff. = 0.054, median diff. = 0.50, 95%CI [-2.00, 2.00], HL = 0.00, rrb = 0.071. In g, h, j-ii, k, and l all data are presented as mean values ± SEM.
Fig. 2 Abrupt state-like transitions from goal-directed to habitual behavior appear spontaneously within individual animals. a Hit (color) and FA (black) action rates for example WR85 (left) and a CA (right) animals. b Hit rate of CA example animal (in a) showing periods of high (gray shadow) and low engagement, followed by a spontaneous transition (red vertical line) to low hit rate variability. c CA mice have significantly less engaged blocks compared to WR85. Median diff. = 13.92, n = 11/12 mice, two-sided Wilcoxon rank-sum test: z = 2.98, p = 0.0028, HL = 19.27, 95%CI [8.90,37.32]. d (i) Most CA (8/12, 66.7%) showed hit-rate variability and the presence of a transition. Some CA mice that transitioned to low variability hit rate, seemed to transition to high variability after a while (asterisks). (ii) Only a few CA mice showed no hit rate variability (2/12, 16.7%). (iii) Few CA mice showed initial hit rate variability, but never transitioned to low variability (2/12, 16.7%). e An HMM-GLM was used to model behavioral data, which allows us to analyze the state-like nature of transitions. f An HMM-GLM with two states provides the best fit for most of the CA animals based on a BIC analysis (n = 8 CA mice). g Both states are equally stimulus-driven, but state 2 is characterized by a no-go, or disengaged bias (n = 8 CA mice). h State 1 (pink) is highly stimulus selective between target and foil trials with high engagement, while State 2 (green) shows overall task disengagement (n = 8 CA mice). i A CA exemplar shows that the two-state HMM-GLM model accurately recapitulates the behavior (top), and the two states govern the pre-transition phase, while only one state becomes explanatory of the post-transition phase. j (i) The GLM states reflect transitions between an engaged state (state 1, pink, hit rate = 0.98) and a disengaged state (state2, olive, hit rate = 0.35). Across all mice, the HMM-GLM model predicted that the probability of staying in engaged or disengaged state (trial-by-trial) is 0.995 and 0.975 respectively, whereas the transition probability between states is 0.005 (engaged to disengaged) and 0.025 (disengaged to engaged). (ii) State occupancy pre-transition (black) is approximately divided 50%–50% between State 1 and State 2, while post-transition (red), State 1 dominates (n = 8 CA mice). Two-way ANOVA revealed a significant State main effect (F(1, 28) = 62.20, p = 1.37×10-8, ηp2 = 0.69), and no significant Transition main effect (F(1, 28) = 0.020, p = 0.90, ηp2 = 0.0010), and a significant State × Transition interaction effect (F(1, 28) = 0.99, p = 8.7×10-5, ηp2 = 0.43). Bonferroni-corrected post hoc pairwise comparisons across State × Transition cell means identified: State=State1,Transition=Pre vs State=State1,Transition=Post (mean diff. = -0.30, 95%CI [-0.56,-0.040], p_adj = 0.015); State=State1,Transition=Pre vs State=State2, Transition=Post (mean diff. = 0.50, 95%CI [0.24,0.76], p_adj = 4.38×10-5); State=State2, Transition=Pre vs State=State1, Transition=Post (mean diff. = -0.51, 95% CI [-0.77, -0.26], p_adj = 2.72×10-5), plus 2 additional significant contrasts; full statistics are provided in Supplementary Table (‘SuppTable PostHoc Bonferroni Occupancy’). k, No-go to go (State 2 → State 1) transitions and go to no-go (State 1 → State 2) transitions happening in the goal-directed phase, are abrupt (n = 8 CA mice). l, No differences between the first (blue, belonging to the goal-directed phase), and last (red, belonging to goal-directed to habitual) transitions in abruptness. Both happen within 1-4 trials (n = 8 CA mice). Two-sided Wilcoxon rank-sum test: U = 30, p = 0.89, mean diff. = 0.054, median diff. = 0.50, 95%CI [-2.00, 2.00], HL = 0.00, rrb = 0.071. In g, h, j-ii, k, and l all data are presented as mean values ± SEM.CC BY-NC-ND · PubMed Central
Fig. 3 Sensory-specific devaluation and bilateral DLS lesions confirm a transition to habit. a Sensory-specific devaluation protocol. b Normalized hit rate across condition and transition. 2-way repeated-measure ANOVA revealed a significant condition x transition interaction effect (X2(1) = 5.61, p = 0.018, partial R2 = 0.39). Two-sided post hoc pairwise t-test of estimated marginal means with Tukey HSD correction and Kenward-Roger degree of freedom method: transition=Pre, condition=Valued vs transition=Pre, condition=Devalued (mean diff. = 0.25, 95%CI [0.032,0.47], p = 0.027); condition=Devalued, transition=Pre vs condition=Devalued, transition=Post (mean diff. = -0.47, 95%CI [-0.69,-0.25], p = 0.00020). n = 7 CA mice. For extended statistical analysis see Methods. c Devaluation index, pre-transition vs. post-transition. Mean diff. = 0.39, n = 7 mice. one-tailed paired-samples t-test: t(6) = 2.42, p = 0.026, two-sided 90%CI [0.078,0.71], Cohen’s dz = 0.92. Pre-transition: median = 0.37, Q1 = -0.018, Q3 = 0.63, lower bound = -0.99, upper bound = 1.60, lower whisker (min) = -0.29, upper whisker (max) = 0.77. Post-transition: median = 0.027, Q1 = -0.30, Q3 = 0.084, lower bound = -0.87, upper bound = 0.66, lower whisker (min) = -0.41, upper whisker (max) =0.095. H=habitual, GD=goal-directed. d Lesion protocol. e Lesion exemplar. Coronal view of the bilateral lesion sites at the DLS. Structural illustration by the Paxinos and Franklin’s Mouse Brain Atlas99. f Lesion map of all animals (n = 11 mice). g Action rate exemplars for sham (top, gray) and lesioned (bottom, purple) animals. h Cumulative distribution of transition trial for animals that showed behavioral variability (n = 10 sham and n = 9 lesioned mice). i ROC curve built from transition probabilities in (g). j AUC values for shuffled labels in our experimental groups. The difference in AUC between sham and lesioned mice (purple line) falls into the 95th significance percentile (red dotted line), compared to the difference between control animals (gray line). In b, c data are presented as mean values ± SEM.
Fig. 3 Sensory-specific devaluation and bilateral DLS lesions confirm a transition to habit. a Sensory-specific devaluation protocol. b Normalized hit rate across condition and transition. 2-way repeated-measure ANOVA revealed a significant condition x transition interaction effect (X2(1) = 5.61, p = 0.018, partial R2 = 0.39). Two-sided post hoc pairwise t-test of estimated marginal means with Tukey HSD correction and Kenward-Roger degree of freedom method: transition=Pre, condition=Valued vs transition=Pre, condition=Devalued (mean diff. = 0.25, 95%CI [0.032,0.47], p = 0.027); condition=Devalued, transition=Pre vs condition=Devalued, transition=Post (mean diff. = -0.47, 95%CI [-0.69,-0.25], p = 0.00020). n = 7 CA mice. For extended statistical analysis see Methods. c Devaluation index, pre-transition vs. post-transition. Mean diff. = 0.39, n = 7 mice. one-tailed paired-samples t-test: t(6) = 2.42, p = 0.026, two-sided 90%CI [0.078,0.71], Cohen’s dz = 0.92. Pre-transition: median = 0.37, Q1 = -0.018, Q3 = 0.63, lower bound = -0.99, upper bound = 1.60, lower whisker (min) = -0.29, upper whisker (max) = 0.77. Post-transition: median = 0.027, Q1 = -0.30, Q3 = 0.084, lower bound = -0.87, upper bound = 0.66, lower whisker (min) = -0.41, upper whisker (max) =0.095. H=habitual, GD=goal-directed. d Lesion protocol. e Lesion exemplar. Coronal view of the bilateral lesion sites at the DLS. Structural illustration by the Paxinos and Franklin’s Mouse Brain Atlas99. f Lesion map of all animals (n = 11 mice). g Action rate exemplars for sham (top, gray) and lesioned (bottom, purple) animals. h Cumulative distribution of transition trial for animals that showed behavioral variability (n = 10 sham and n = 9 lesioned mice). i ROC curve built from transition probabilities in (g). j AUC values for shuffled labels in our experimental groups. The difference in AUC between sham and lesioned mice (purple line) falls into the 95th significance percentile (red dotted line), compared to the difference between control animals (gray line). In b, c data are presented as mean values ± SEM.CC BY-NC-ND · PubMed Central
Fig. 4 Motor automaticity signatures and pupillary response changes appear concomitant with habit transitions. a Exemplar lick raster plots of a CA animal, showing individual licks (green lines) to target tones throughout training. Tone onset (0, green note) is followed by a dead period (100 ms, white area) and the presence of operant licks (gray rectangle). b CA mice show a strong increase in post-transition number of consummatory licks. c Average number of licks per trial is significantly higher in CA mice post-transition compared to pre-transition (Mean diff. = -2.60, n = 8/8 mice, paired t-test: (t(7.00) = -4.87, p = 0.0018), Mean diff. = -2.60, 95%CI[-3.86,-1.34], Cohens dz = -1.72). d A significant reduction in the operant lick latency is observed post-transition (Median diff. = 75.00, n = 8/8 mice, two sided paired Wilcoxon signed-rank test (W = 28, p = 0.016), HL = 50.00 (95% CI[50.00, 150.00]). e Evoked pupil dilation separated by trial type (engaged only): black=pre-transition rewarded; red=post-transition rewarded; gray=pre-transition no-reward; pink=post-transition no-reward. f Significant increase in evoked pupil response in no-reward trials post-transition. Two-sided post hoc pairwise t-test with LSD test: Trial=Non-consumed,Transition=Pre vs Trial=Non-consumed,Transition=Post (mean diff. = -0.20, 95%CI [-0.65,-1.23], p = 0.019) for 3 days pre- and 3 days post-transition. The difference in pupil response between rewarded and no-reward trials decreased across transition, revealed by permutation-based two-way ANOVA, significant transition x trial interaction effect (F(2, 204) = 3.95, p = 0.017, ηp2 = 0.037). Data compiled from n = 7 CA mice, n = 35 trials per group. For extended statistical analysis see Supplementary Table 1. g Significant increase in evoked pupil response for 10 engaged, no-reward trials pre- and post-transition. Median diff. = -1.64, n = 45/35 trials, two-sided Wilcoxon rank-sum test: z = -2.78, p = 0.0054, HL = -1.36, 95%CI [-3.08, -0.34], data compiled from n = 7 mice. Except a all data are presented as mean values ± SEM.
Fig. 4 Motor automaticity signatures and pupillary response changes appear concomitant with habit transitions. a Exemplar lick raster plots of a CA animal, showing individual licks (green lines) to target tones throughout training. Tone onset (0, green note) is followed by a dead period (100 ms, white area) and the presence of operant licks (gray rectangle). b CA mice show a strong increase in post-transition number of consummatory licks. c Average number of licks per trial is significantly higher in CA mice post-transition compared to pre-transition (Mean diff. = -2.60, n = 8/8 mice, paired t-test: (t(7.00) = -4.87, p = 0.0018), Mean diff. = -2.60, 95%CI[-3.86,-1.34], Cohens dz = -1.72). d A significant reduction in the operant lick latency is observed post-transition (Median diff. = 75.00, n = 8/8 mice, two sided paired Wilcoxon signed-rank test (W = 28, p = 0.016), HL = 50.00 (95% CI[50.00, 150.00]). e Evoked pupil dilation separated by trial type (engaged only): black=pre-transition rewarded; red=post-transition rewarded; gray=pre-transition no-reward; pink=post-transition no-reward. f Significant increase in evoked pupil response in no-reward trials post-transition. Two-sided post hoc pairwise t-test with LSD test: Trial=Non-consumed,Transition=Pre vs Trial=Non-consumed,Transition=Post (mean diff. = -0.20, 95%CI [-0.65,-1.23], p = 0.019) for 3 days pre- and 3 days post-transition. The difference in pupil response between rewarded and no-reward trials decreased across transition, revealed by permutation-based two-way ANOVA, significant transition x trial interaction effect (F(2, 204) = 3.95, p = 0.017, ηp2 = 0.037). Data compiled from n = 7 CA mice, n = 35 trials per group. For extended statistical analysis see Supplementary Table 1. g Significant increase in evoked pupil response for 10 engaged, no-reward trials pre- and post-transition. Median diff. = -1.64, n = 45/35 trials, two-sided Wilcoxon rank-sum test: z = -2.78, p = 0.0054, HL = -1.36, 95%CI [-3.08, -0.34], data compiled from n = 7 mice. Except a all data are presented as mean values ± SEM.CC BY-NC-ND · PubMed Central
Fig. 5 Switch-like reduction in DLS outcome-related signaling at the moment of a habit transition. a Bilateral simultaneous DLS and DMS fiber photometry recordings in behaving mice across training. Left: mice were injected with hSyn-GCaMP8m in both structures. Right: green rectangles show successful fiber placement and viral expression. b Analysis window ( ± 50 trials around the transition). c DLS (top, 50 trials per group from n = 7 mice, total n = 350 data points) and DMS (bottom, 50 trials per group from n = 6 mice, total n = 300 data points) activity pre (black) and post-transition (red). Data are shown as mean ± SEM. d Area under-the-curve (AUC) of the full trace in (c) for the DLS (top) (Median diff.=70.49, n = 350/350 trials from n = 7 mice, two-sided paired Wilcoxon signed-rank test (z = 7.53, p = 5.04×10-14), HL = 63.66 (95% CI[49.01, 77.52])) and DMS (bottom) overall trace (Median diff.= 33.64, n = 300/300 trials from n = 6 mice, two-sided paired Wilcoxon signed-rank test (z = 3.37, p = 0.00074), HL = 31.65 (95% CI[6.58, 54.32])). Data are shown as mean ± SEM. e First peak latency post-transition is significantly different for the DLS (top) (two-sample F-test, (F(349,349) = 1.87, p = 7.41×10−9)) but not the DMS (bottom) (two-sample F-test, F(294,291) = 0.94, p = 0.59). n = 350 trials DLS pre; n = 330 DLS trials post from n = 7 mice; n = 295 trials DMS pre; n = 292 trials DMS post from n = 6 mice. f Jitter quantification of first-peak latencies in DLS (top) and DMS (bottom), showing a significant reduction in DLS jitter post-transition (Levene’s test F(1,698) = 34.21, p = 7.62 × 10−9), but not in the DMS (Levene’s test F(1,585) = 2.94, p = 0.087). n = 350 trials DLS pre; n = 350 DLS trials post from n = 7 mice; n = 295 trials DMS pre; n = 292 trials DMS post from n = 6 mice. Data are shown as mean ± SEM. g Single animal exemplar of 5 trials immediately pre- (black) and post- (red) transition. Individual lines represent individual trials. h Logistic regression fit. Inflection point (in trials): full trace (top) DLS = 4.8, DMS = 9.4; ‘Cue’ epoch (middle) DLS none, DMS = 9.2; ‘Outcome’ epoch DLS = 3.3, DMS = 9.4. Red vertical line represents the transition point.
Fig. 5 Switch-like reduction in DLS outcome-related signaling at the moment of a habit transition. a Bilateral simultaneous DLS and DMS fiber photometry recordings in behaving mice across training. Left: mice were injected with hSyn-GCaMP8m in both structures. Right: green rectangles show successful fiber placement and viral expression. b Analysis window ( ± 50 trials around the transition). c DLS (top, 50 trials per group from n = 7 mice, total n = 350 data points) and DMS (bottom, 50 trials per group from n = 6 mice, total n = 300 data points) activity pre (black) and post-transition (red). Data are shown as mean ± SEM. d Area under-the-curve (AUC) of the full trace in (c) for the DLS (top) (Median diff.=70.49, n = 350/350 trials from n = 7 mice, two-sided paired Wilcoxon signed-rank test (z = 7.53, p = 5.04×10-14), HL = 63.66 (95% CI[49.01, 77.52])) and DMS (bottom) overall trace (Median diff.= 33.64, n = 300/300 trials from n = 6 mice, two-sided paired Wilcoxon signed-rank test (z = 3.37, p = 0.00074), HL = 31.65 (95% CI[6.58, 54.32])). Data are shown as mean ± SEM. e First peak latency post-transition is significantly different for the DLS (top) (two-sample F-test, (F(349,349) = 1.87, p = 7.41×10−9)) but not the DMS (bottom) (two-sample F-test, F(294,291) = 0.94, p = 0.59). n = 350 trials DLS pre; n = 330 DLS trials post from n = 7 mice; n = 295 trials DMS pre; n = 292 trials DMS post from n = 6 mice. f Jitter quantification of first-peak latencies in DLS (top) and DMS (bottom), showing a significant reduction in DLS jitter post-transition (Levene’s test F(1,698) = 34.21, p = 7.62 × 10−9), but not in the DMS (Levene’s test F(1,585) = 2.94, p = 0.087). n = 350 trials DLS pre; n = 350 DLS trials post from n = 7 mice; n = 295 trials DMS pre; n = 292 trials DMS post from n = 6 mice. Data are shown as mean ± SEM. g Single animal exemplar of 5 trials immediately pre- (black) and post- (red) transition. Individual lines represent individual trials. h Logistic regression fit. Inflection point (in trials): full trace (top) DLS = 4.8, DMS = 9.4; ‘Cue’ epoch (middle) DLS none, DMS = 9.2; ‘Outcome’ epoch DLS = 3.3, DMS = 9.4. Red vertical line represents the transition point.CC BY-NC-ND · PubMed Central

Verification

Claims the paper makes

1 Valid · 0 Not valid · 23 Not checked · 120 Own results and interpretations

81 more claims

Create a free account to see all 144 claims and to verify the 60 references still open.

See all claims — free account
  • It may state two separate claims as one. 44
  • The quote only partly states this claim. 2
  • Probably not a substantial claim. 25

Create a free account to see these claims.

Statistics recomputed

Every reported test result, recomputed from its test statistic and degrees of freedom. A result shows as not adding up only once an independent check has confirmed it.

  • Consistentt(6) = 2.42reported p = .026 · recomputed p = .052

    “t(6) = 2.42, p = .026”

    The paper states a one-tailed test

  • Consistentt(7) = -4.87reported p = .0018 · recomputed p = .002

    “t(7) = -4.87, p = .0018”

  • Consistentz = -2.78reported p = .0054 · recomputed p = .005

    “z = -2.78, p = .0054”

  • Consistentχ²(1) = 5.56reported p = .018 · recomputed p = .018

    “X2(1) = 5.56, p = .018”

  • Consistentχ²(1) = 8reported p = .0047 · recomputed p = .005

    “X2(1) = 8.00, p = .0047”

5 of 13 · the rest aren’t loaded

8 more recomputed tests

Create a free account to see every test the Stats check recomputed.

See all tests — free account

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

Self-check results appear here once an independent check has confirmed them.

Citations

Cites

Show all 91 references

Research works printed but not linked: 3

  1. Bouton, M. E. Learning and behavior: a contemporary synthesis. 2nd ed. J. Undergrad. Neurosci. Educ.15, R3 (2016).
  2. Lashley, K. S. et al. The Problem of Serial Order in Behavior. Jeffress, L. A. (Ed), Cerebral Mechanisms in Behavior: The Hixon Symposium. 112–136 (1951).
  3. Paxinos, G. & Franklin, K. B. J. The Mouse Brain In Stereotaxic Coordinates. 5th ed. London: Academic Press; (2019).

Cited by

See all 3

Metadata

Authors
Moore, S., Wang, Z., Zhu, Z., Wang, J., Sun, R., Lee, Y. A., Charles, A., & Kuchibhotla, K. V.
Affiliations
Johns Hopkins University (US)
Corresponding author
Kuchibhotla, Kishore V.
Venue
Nature Communications
Published
2026-03-31
Volume
17
Issue
1
ISSN
2041-1723 (ISSN-L)
OpenAlex
W7147635450
Cited by
3
References
91
Citations by year
2026: 3
FWCI
12.39
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
Funders
  • U.S. Department of Health & Human Services | NIH | National Institute on Drug Abuse (R01DA062689)
  • U.S. Department of Health & Human Services | NIH | National Institute on Deafness and Other Communication Disorders (R01DC018650)
Topics
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neural and Behavioral Psychology Studies
  • Life Sciences › Neuroscience › Cellular and Molecular Neuroscience › Neurotransmitter Receptor Influence on Behavior
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Memory and Neural Mechanisms
Keywords
goal-directed behavior, habit formation, dorsal striatum, fiber photometry, stimulus-response learning
MeSH terms
Animals, Behavior, Animal, Goals, Habits, Male, Mice, Inbred C57BL, Reward, Mice

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

A language model drew the claims from the paper on 2026-10-11, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-11 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Revealing abrupt transitions from goal-directed to habitual behavior. https://app.cochat.ai/papers/10.1038/s41467-026-71048-0
The paper
Moore, S., Wang, Z., Zhu, Z., Wang, J., Sun, R., Lee, Y. A., Charles, A., & Kuchibhotla, K. V. (2026). Revealing abrupt transitions from goal-directed to habitual behavior. Nature Communications. https://doi.org/10.1038/s41467-026-71048-0

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.