Basolateral amygdala dopamine transmits emotional salience Authors: Megan A. Brickner, William E. Szot, Amy R. Wolff, Mark J. Thomas, Benjamin T. Saunders Published: 2026 ABSTRACT The basolateral amygdala (BLA) is important for assigning emotional weight to environmental stimuli to guide learning, discrimination, and adaptive vigilance. The contribution of neuromodulator inputs such as dopamine to this process is unclear. Here, we recorded in vivo BLA dopamine signals across emotional learning paradigms. We show that BLA dopamine scales with a stimulus’s emotional intensity, but not value or associative strength, and reward-predictive cues evoke BLA dopamine signals that diminish, rather than grow, with training. In a complex learning context, where rats discriminated between multiple valenced associations, the magnitude of cue-evoked BLA dopamine re-scaled to report the level of relative emotional saliency. Threat and safety cues prompted larger, sustained dopamine signals compared to reward and neutral cues. Together, our data reveal a pattern of dopamine encoding with a unique contribution to learning, for dynamic disambiguation of affective states. Introduction A foundation of behavior is the associative learning that cements relationships between environmental cues and biologically relevant outcomes. Disambiguation of positive and negative valenced sensory inputs is then imperative for directing appropriate decision making. The basolateral amygdala (BLA) is a key brain structure in this process, integrating sensory information from multiple brain regions and assigning emotional value to environmental cues and contexts1,2. The BLA is central to fear3,4 as well as reward learning5–13, and is involved in discrimination of fear and reward4,14–19 via dynamic valence encoding and memory representation14,20–23. The BLA’s role in emotional learning is partly driven by excitatory inputs from cortex, thalamus and other regions, but the contribution of neuromodulatory circuit inputs is less clear1,2,24–31. Notably, the ventral tegmental area (VTA), where dopamine neurons are classically known to regulate reward learning and incentive motivation via projections to the striatum32,33 also contains dense dopamine projections to the BLA34–38, and this system has received much less attention. On slow timescales, amygdala dopamine levels track reward discrimination learning39. Recent studies offer more insights, showing that the dopaminergic VTA-BLA system is rapidly activated during rewards and punishments16,36,40–42 and responds to predictive cues16,36,40,41,43,44. Further, BLA dopamine signaling encodes information relating to reward-identity memories16,36,40,41. Despite this growing focus, we still know very little about how the BLA dopamine system is engaged during learning. Delineating this aspect of heterogeneity is critical for understanding evolving complexities of the role of dopamine in valuation and salience encoding45–50. To investigate this, we used fiber photometry and optogenetics to characterize in vivo BLA dopamine signals as rats engaged in various learning tasks. We first show that appetitive and aversive stimuli evoke strong positive dopamine responses in the BLA, and reward-cue evoked responses diminish, rather than grow, with training. We then tracked the profile of BLA dopamine in response to four differently valenced cue types during a Pavlovian cue discrimination task. In this more complex learning context, where cues signaling threat, reward, safety, and neutral associations were intermingled within the same session, we found that the magnitude of cue-evoked BLA dopamine responses scaled with relative emotional saliency: fear and safety cues evoked strong signals early in conditioning that were larger than for reward and neutral cues, and these signals were resistant to extinction. In additional studies, we demonstrate optogenetic stimulation of dopamine release in the BLA does not directly drive learning via reinforcement, and finally, BLA dopamine signals do not track the value of preferred rewards but instead scale with stimulus uncertainty and threat intensity. Together, our data reveal a pattern of dopamine signaling with a unique contribution to learning compared to classic striatal systems, through encoding the emotional weight of sensory events, independent of associative strength or value. This state-guided affective signal is dynamically engaged when relative stimulus importance must be disambiguated. Results Optogenetic activation of VTA-BLA dopamine neurons evokes robust BLA dopamine release that does not support reinforcement To record amygdala dopamine signaling in freely behaving rats, we used fiber photometry with the dLight1.3b sensor (Fig. 1A). First, to validate this approach and determine how VTA TH-neuron stimulation influences dopamine dynamics in the BLA, a cre-driven Chrimson opsin was expressed into the VTA of TH-cre rats to specifically target the dopamine neurons (Fig. 1B,D), while dLight1.3b was expressed in the ipsilateral BLA (Figs. 1B,C; S1). Optogenetic activation of VTA dopamine neurons via 590 nm laser pulses evoked robust dopamine signals in the BLA, which positively scaled with stimulation frequency (Fig. 1E,F; t(7) = 5.031, p = 0.0015). We next targeted optogenetic activation to VTA dopamine terminals in the BLA to confirm locally evoked dopamine release. Terminal stimulation also resulted in robust dopamine signaling (Fig. 1G) that was comparable to that evoked by brief dopamine cell body stimulation (Fig. 1H; t(7) = 1.858, p = 0.1055).Fig. 1Optogenetic activation of VTA dopamine neurons evokes robust dopamine signaling in the BLA that does not support reinforcement.A Schematic of the photometry system to record BLA dopamine dynamics. B Viral approach for simultaneous in vivo optogenetic activation of VTA dopamine neuron cell bodies or terminals in the BLA paired with fiber photometry recording of dopamine transmission in the BLA. Dopamine neurons were targeted in TH-cre transgenic rats with DIO-Chrimson opsin. C Example of dLight1.3b viral expression in the BLA (n = 8; 4 F, 4 M), with tissue counterstained for visualization of overlapping tyrosine hydroxylase (TH). D Example of DIO-Chrimson expression in VTA dopamine neurons. E, F Optogenetic excitation of VTA dopamine cell bodies resulted in frequency-dependent dopamine release in the BLA, where 5 s stimulation at 20 Hz prompted a larger response compared to 5 s stimulation at 5 Hz (n = 8, paired t-test; p = 0.0015). G, H Optogenetic excitation of VTA dopamine neuron terminals in the BLA also resulted in robust BLA dopamine release, which was comparable to the signal evoked by cell body stimulation for 2 s at 20 Hz (n = 8, paired t-test; p = 0.1055). I Viral approach for intracranial self-stimulation (ICSS) of VTA-BLA dopamine terminals using DIO-Chrimson or DIO-ChR infused in the VTA of TH-cre rats and optic fibers implanted over the BLA. J Schematic of the ICSS behavioral paradigm, where rats could nose poke to activate VTA-BLA dopamine terminals to prompt dopamine transmission into the BLA. K Rats (n = 14; 6 F, 8 M) did not readily poke for BLA dopamine terminal activation during ICSS, making the same number of active nose pokes (ANPs) and inactive nose pokes (INPs) (n = 14, two-way RM ANOVA, main effect of nose poke, p = 0.1039). L Total number of ANPs and INPs during ICSS collapsed across sessions (n = 14, paired t-test; p = 0.1039). AUC = Area under curve. Data represent mean ± SEM. All tests are two-tailed. *p < 0.05, **p < 0.01. Optogenetic activation of VTA dopamine neurons evokes robust dopamine signaling in the BLA that does not support reinforcement. A Schematic of the photometry system to record BLA dopamine dynamics. B Viral approach for simultaneous in vivo optogenetic activation of VTA dopamine neuron cell bodies or terminals in the BLA paired with fiber photometry recording of dopamine transmission in the BLA. Dopamine neurons were targeted in TH-cre transgenic rats with DIO-Chrimson opsin. C Example of dLight1.3b viral expression in the BLA (n = 8; 4 F, 4 M), with tissue counterstained for visualization of overlapping tyrosine hydroxylase (TH). D Example of DIO-Chrimson expression in VTA dopamine neurons. E, F Optogenetic excitation of VTA dopamine cell bodies resulted in frequency-dependent dopamine release in the BLA, where 5 s stimulation at 20 Hz prompted a larger response compared to 5 s stimulation at 5 Hz (n = 8, paired t-test; p = 0.0015). G, H Optogenetic excitation of VTA dopamine neuron terminals in the BLA also resulted in robust BLA dopamine release, which was comparable to the signal evoked by cell body stimulation for 2 s at 20 Hz (n = 8, paired t-test; p = 0.1055). I Viral approach for intracranial self-stimulation (ICSS) of VTA-BLA dopamine terminals using DIO-Chrimson or DIO-ChR infused in the VTA of TH-cre rats and optic fibers implanted over the BLA. J Schematic of the ICSS behavioral paradigm, where rats could nose poke to activate VTA-BLA dopamine terminals to prompt dopamine transmission into the BLA. K Rats (n = 14; 6 F, 8 M) did not readily poke for BLA dopamine terminal activation during ICSS, making the same number of active nose pokes (ANPs) and inactive nose pokes (INPs) (n = 14, two-way RM ANOVA, main effect of nose poke, p = 0.1039). L Total number of ANPs and INPs during ICSS collapsed across sessions (n = 14, paired t-test; p = 0.1039). AUC = Area under curve. Data represent mean ± SEM. All tests are two-tailed. *p < 0.05, **p < 0.01. VTA dopamine projections to the striatum are known to support behavioral reinforcement51–54. Given the strong evoked dopamine signal we observed in the BLA, we next tested whether VTA-BLA dopamine transmission also promotes reinforcement (Fig. 1I). Rats were given the opportunity to nose poke for optogenetic activation of VTA dopamine terminals in the BLA, in an Intracranial Self Stimulation (ICSS) procedure (Fig. 1J). Overall, rats did not readily nose poke for stimulation (Fig. 1K; F(1,13) = 3.059, p = 0.1039), and active nose pokes were not significantly different from inactive nose pokes (Fig. 1L; t(13) = 1.749, p = 0.1039). In vivo BLA dopamine signals track stimulus emotional intensity but not value We next recorded BLA dopamine dynamics in response to differently valenced unconditioned stimuli - food reward or footshock. Rats were first given the opportunity to drink sucrose, delivered into a magazine port, which prompted a moderate increase in dopamine (Fig. 2A). Delivery of moderate (0.4 mA, 0.5 s) footshocks elicited a strong phasic increase in dopamine (Fig. 2B). Comparing the signals directly, footshock-evoked signals were significantly larger than sucrose evoked signals (Fig. 2C; t(18) = 4.191, p = 0.0005). Broadly, these results show that both positive and negative events prompt release of BLA dopamine, with a bias of stronger responses for aversive stimuli.Fig. 2In vivo BLA dopamine signals track stimulus emotional intensity but not value.A Rats (n = 21) were given unsignalled deliveries of a sucrose reward. Sucrose consumption prompted modest increases of in vivo BLA dopamine. Photometry traces were time locked to the first reward port entry after sucrose pump onset. B Delivery of moderate 0.4 mA footshocks evoked large BLA dopamine responses in the same rats. C The footshock-evoked signal was larger than the BLA dopamine response to sucrose consumption when comparing AUC values (n = 21, paired t-test; p = 0.0005). D Rats spend significantly more time in the ensure reward port compared to the sucrose reward port (n = 8, paired t-test; p = 0.0093). E Sucrose consumption and ensure consumption elicit moderate increases of BLA dopamine, F that are equal in AUC magnitude of the first 5 s after rewarded port entry onset (n = 8, paired t-test; p = 0.8212). G Rats (n = 10) were exposed to uncued footshocks. Footshocks of both mild (0.2 mA) and moderate (0.4 mA) currents prompt robust BLA dopamine transients. H BLA dopamine responses were significantly larger to the higher magnitude shocks (n = 10, paired t-test; p = 0.0086). Data represent mean ± SEM. All tests are two-tailed. *p < 0.05, **p < 0.01, ***p < 0.001. In vivo BLA dopamine signals track stimulus emotional intensity but not value. A Rats (n = 21) were given unsignalled deliveries of a sucrose reward. Sucrose consumption prompted modest increases of in vivo BLA dopamine. Photometry traces were time locked to the first reward port entry after sucrose pump onset. B Delivery of moderate 0.4 mA footshocks evoked large BLA dopamine responses in the same rats. C The footshock-evoked signal was larger than the BLA dopamine response to sucrose consumption when comparing AUC values (n = 21, paired t-test; p = 0.0005). D Rats spend significantly more time in the ensure reward port compared to the sucrose reward port (n = 8, paired t-test; p = 0.0093). E Sucrose consumption and ensure consumption elicit moderate increases of BLA dopamine, F that are equal in AUC magnitude of the first 5 s after rewarded port entry onset (n = 8, paired t-test; p = 0.8212). G Rats (n = 10) were exposed to uncued footshocks. Footshocks of both mild (0.2 mA) and moderate (0.4 mA) currents prompt robust BLA dopamine transients. H BLA dopamine responses were significantly larger to the higher magnitude shocks (n = 10, paired t-test; p = 0.0086). Data represent mean ± SEM. All tests are two-tailed. *p < 0.05, **p < 0.01, ***p < 0.001. To dissociate how BLA dopamine signals are engaged by stimuli of different values versus different intensities, we compared responses to sucrose versus a highly palatable Ensure solution. In consumption tests, rats preferred the Ensure reward, spending more time in the Ensure magazine compared to the sucrose magazine (Fig. 2D; t(7) = 3.551, p = 0.0093). However, both reward types prompted modest increases in BLA dopamine (Fig. 2E) that were equal in magnitude (Fig. 2F; t(7) = 0.2346, p = 0.8212). We next tracked BLA dopamine responses to random, brief footshocks of either 0.2 mA or 0.4 mA (Fig. 2F). Each footshock intensities evoked large dopamine transients, but the stronger footshock level evoked greater dopamine release in the BLA (Fig. 2F,G; t(9) = 3.345, p = 0.0086). Overall, these patterns suggest that BLA dopamine release scales with relative stimulus intensity (Fig. 2F, G) but does not scale with relative stimulus value (Fig. 2D,E). Reward learning promotes diminishing cue-evoked BLA dopamine signals across training To explore how BLA dopamine dynamics track associative learning and valence discrimination, rats were prepared for BLA dopamine photometry and then trained on a multi-phase Pavlovian cue discrimination paradigm adapted from previous studies14 (Fig. 3A). Rats first underwent reward-only conditioning (RC), followed by a compound reward and fear conditioning phase (CC). This was then followed by the full cue discrimination task (CD), where rats had to distinguish between four different, interleaved Pavlovian cue associations—reward, fear, safety, and neutral. Finally, reward and fear cues underwent extinction (Fig. 3A). Throughout this, we measured two main conditioned responses: sucrose port entries and freezing, which were largely similar across males and females (Fig. S2). Further, in our assessment of the photometry data, we did not see any consistent sex-related biases in BLA dopamine signals nor qualitatively different patterns of activity (Fig. S3), in various task phases. These results are interesting in light of previous data showing a wide range of sex differences in behavioral threat and safety responses55–59, suggesting complex and distinct fear response strategies between female and male rodents. While we found no obvious connection between cue-evoked dopamine and behavior in our studies, our data do not rule out possible sex-biased dopaminergic mechanisms of behavioral vigilance and decision making within the amygdala60.Fig. 3BLA dopamine signals preferentially represent conditioned threat over reward expectation.A Schematic of the multi-phase Pavlovian cue discrimination paradigm (n = 24; 11 F, 13 M). Behavior and photometry data for the first two phases: reward conditioning and compound (reward and fear) conditioning are shown. B Throughout training, time spent in the reward port increased during reward cue presentations (n = 24, two-way RM ANOVA, main effect of session, p < 0.0001), while fear cue presentations did not prompt port entry behaviors (n = 17, paired t-test; p < 0.0001). C Cue-induced freezing was low for the reward cue during both reward (n = 24, one-way ANOVA, p = 0.6271) as well as compound conditioning but was higher for the fear cue in compound conditioning (n = 17, paired t-test; p = 0.0004). D During reward conditioning, a 20 s auditory cue was paired with a 3 s delivery of sucrose into the reward port. E Reward cue evoked BLA dopamine responses were largest initially and decreased with training, F as shown by AUC comparisons of the first 2 s after reward cue onset (n = 24, one-way RM ANOVA, p = 0.0002). G There was no correlation between port entry behavior and reward cue peak dopamine signals on the final day of reward conditioning (n = 24, Pearson correlation, p = 0.2702). H During compound conditioning (n = 17; 7 F, 10 M), trials containing a new 20 s auditory cue that predicted a footshock (0.5 s 0.5 mA) were intermingled with reward cue trials. I Rats exhibited large dopamine dynamics to the fear cue and moderate dopamine dynamics to the reward cue. J The fear cue-evoked signal was greater than the reward cue when comparing AUC values (n = 17, paired t-test; p = 0.0032). K The reward cue during both reward conditioning day 1 and compound conditioning produced similarly sized dopamine events that did not show different AUC comparisons (n = 17, paired t-test; p = 0.3832). L There is no correlation between conditioned freezing behaviors and the magnitude of the fear cue signal (n = 24, Pearson correlation, p = 0.3881), M or between conditioned port entry behaviors and the magnitude of the reward cue signal (n = 24, Pearson correlation, p = 0.8830). CC = compound conditioning. RC = reward conditioning. Data represent mean ± SEM. All tests are two-tailed. *p < 0.05, **p < 0.01, ****p < 0.0001. BLA dopamine signals preferentially represent conditioned threat over reward expectation. A Schematic of the multi-phase Pavlovian cue discrimination paradigm (n = 24; 11 F, 13 M). Behavior and photometry data for the first two phases: reward conditioning and compound (reward and fear) conditioning are shown. B Throughout training, time spent in the reward port increased during reward cue presentations (n = 24, two-way RM ANOVA, main effect of session, p < 0.0001), while fear cue presentations did not prompt port entry behaviors (n = 17, paired t-test; p < 0.0001). C Cue-induced freezing was low for the reward cue during both reward (n = 24, one-way ANOVA, p = 0.6271) as well as compound conditioning but was higher for the fear cue in compound conditioning (n = 17, paired t-test; p = 0.0004). D During reward conditioning, a 20 s auditory cue was paired with a 3 s delivery of sucrose into the reward port. E Reward cue evoked BLA dopamine responses were largest initially and decreased with training, F as shown by AUC comparisons of the first 2 s after reward cue onset (n = 24, one-way RM ANOVA, p = 0.0002). G There was no correlation between port entry behavior and reward cue peak dopamine signals on the final day of reward conditioning (n = 24, Pearson correlation, p = 0.2702). H During compound conditioning (n = 17; 7 F, 10 M), trials containing a new 20 s auditory cue that predicted a footshock (0.5 s 0.5 mA) were intermingled with reward cue trials. I Rats exhibited large dopamine dynamics to the fear cue and moderate dopamine dynamics to the reward cue. J The fear cue-evoked signal was greater than the reward cue when comparing AUC values (n = 17, paired t-test; p = 0.0032). K The reward cue during both reward conditioning day 1 and compound conditioning produced similarly sized dopamine events that did not show different AUC comparisons (n = 17, paired t-test; p = 0.3832). L There is no correlation between conditioned freezing behaviors and the magnitude of the fear cue signal (n = 24, Pearson correlation, p = 0.3881), M or between conditioned port entry behaviors and the magnitude of the reward cue signal (n = 24, Pearson correlation, p = 0.8830). CC = compound conditioning. RC = reward conditioning. Data represent mean ± SEM. All tests are two-tailed. *p < 0.05, **p < 0.01, ****p < 0.0001. In the simple reward conditioning phase, an auditory cue predicted sucrose reward delivery (Fig. 3D). Rats learned this association quickly and increased their sucrose port entry behaviors across training (Fig. 3B; F(6,135) = 5.187, p < 0.0001), while conditioned freezing behaviors were minimal (Fig. 3C; F(2,69) = 0.4699, p = 0.6271). Females and males responded with similar levels of locomotion and port entry behavior during reward conditioning. While there were no significant differences in freezing during reward cue presentations for females and males, the trajectory of response patterns was different. Freezing behaviors decreased for the females, but stayed slightly elevated for the males (Fig. S2; session x sex interaction, F(2,44) = 4.566, p = 0.0158). Reward cue presentations evoked dopamine increases on all days of training (Fig. 3E). Notably, cue-evoked dopamine was the greatest on the first day of conditioning and decreased significantly as reward conditioning progressed (Fig. 3F; F(2,46) = 10.73, p = 0.0002; post hoc comparisons for session 4 vs 1, p = 0.0017; session 7 vs 1, p = 0.0003). There was no relationship between cue-evoked dopamine signals and the intensity of port entry behavior (Fig. 3G; Pearson correlation, p = 0.2702). BLA dopamine signals preferentially represent conditioned threats over reward expectation Following simple reward conditioning, a cue predicting footshock was added, interleaved with reward cue trials in a compound conditioning session (Fig. 3H). On this session, rats maintained their sucrose port entry behaviors during the reward cue and spent significantly more time in the sucrose port during the reward cue compared to the fear cue (Fig. 3B; t(16) = 6.383, p < 0.0001). Freezing behavior rapidly emerged to the fear cue but not the reward cue (Fig. 3C; t(16) = 4.455, p = 0.0004), and reward cue freezing did not differ from simple reward conditioning (Fig. 3C; t(16) = 1.394, p = 0.1824). This indicates rats were able to behaviorally discriminate the cues during compound conditioning, with similar levels of conditioned port entry behaviors and cued locomotion between females and males (Fig. S2). We examined BLA dopamine signals on this session, when both reward and fear learning were engaged. The fear cue elicited a significantly larger and more sustained dopamine peak than the reward cue signal (Fig. 3I,J; t(16) = 3.465, p = 0.0032). Notably, on this compound conditioning session, the reward cue evoked dopamine signal rebounded relative to the final day of simple reward conditioning (Fig. 3E,I), returning to the same magnitude it showed during the first day of reward conditioning (Fig. 3K; t(16) = 0.8966, p = 0.3832). Despite these large signal differences for fear versus reward cues, there were no significant correlations between fear or reward cue-evoked dopamine and freezing (Fig. 3L; Pearson correlation, p = 0.3881) or port entry behavior (Fig. 3M; Pearson correlation, p = 0.8830). Rats discriminate fear, safety, and reward predictive cues Previous in vivo single unit recordings have demonstrated that a population of BLA neurons becomes responsive during cue discrimination training, including to a safety cue that designates relief from shock14. We hypothesized that BLA dopamine transmission may be an important contributor to the disambiguation of differently valenced, salient events. Following fear plus reward compound conditioning rats were trained to distinguish between 4 different cues predicting various outcomes presented pseudorandomly during each cue discrimination behavioral session: the previously experienced (1) reward cue (paired with sucrose) and (2) fear cue (paired with a footshock), as well as a (3) neutral cue with no paired outcome and a (4) compound cue (fear+neutral cue played concurrently) to signal a relief or “safety” from shock (Fig. 4A, B).Fig. 4Cue-evoked BLA dopamine scales with relative emotional saliency during discrimination learning.A Behavioral paradigm (n = 24; 11 F, 13 M). B In the cue discrimination phase, four cues were presented randomly in each session. Reward and fear cues were intermingled with a neutral cue (visual cue with no outcome) and a safety cue (fear and neutral cues presented together to signify relief from shock). C Rats spent the most time in the reward port during the reward cue and almost no time in the reward port during other cue presentations. This discrimination grew across training days as rats increased their time in port during reward cues (n = 24, two-way mixed model, p < 0.0001). D Cue-induced freezing remained high for the fear and safety cue and was higher than for the reward and neutral cues by the final session (n = 24, two-way RM ANOVA, p = 0.0048). E All cues elicited a positive peak on the first discrimination session, with similar dynamics for the fear and safety cues and similar dynamics for the reward and neutral cues. F By the last session, the fear cue produced the largest signal, while the safety, reward, and neutral cues produced modest and transient increases in BLA dopamine. G AUC values of the BLA dopamine signals comparing the first and last session (n = 24, two-way RM ANOVA, p = 0.0495). H In the first discrimination session, the peaks varied based on cue type, and by the final cue block, the fear and safety cues elicited higher peak signals compared to the reward and neutral cues (n = 24, two-way mixed model, p = 0.0373). I By the last session, the fear cues maintained sustained BLA dopamine responses and were larger than the safety, reward, and neutral cues (n = 24, two-way mixed model, p < 0.0001). J There was no correlation between freezing behavior and the safety cue evoked dopamine peaks, while the fear cue dopamine signal negatively correlated with fear cue-induced freezing behavior by the end of training (n = 24, Pearson, safety: p = 0.9615, fear: p = 0.0102). K The dopamine signal for the reward and neutral cues was not correlated with port entries (n = 24, Pearson, reward: p = 0.2234, neutral: p = 0.2918). Data represent mean ± SEM. All tests are two-tailed. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Cue-evoked BLA dopamine scales with relative emotional saliency during discrimination learning. A Behavioral paradigm (n = 24; 11 F, 13 M). B In the cue discrimination phase, four cues were presented randomly in each session. Reward and fear cues were intermingled with a neutral cue (visual cue with no outcome) and a safety cue (fear and neutral cues presented together to signify relief from shock). C Rats spent the most time in the reward port during the reward cue and almost no time in the reward port during other cue presentations. This discrimination grew across training days as rats increased their time in port during reward cues (n = 24, two-way mixed model, p < 0.0001). D Cue-induced freezing remained high for the fear and safety cue and was higher than for the reward and neutral cues by the final session (n = 24, two-way RM ANOVA, p = 0.0048). E All cues elicited a positive peak on the first discrimination session, with similar dynamics for the fear and safety cues and similar dynamics for the reward and neutral cues. F By the last session, the fear cue produced the largest signal, while the safety, reward, and neutral cues produced modest and transient increases in BLA dopamine. G AUC values of the BLA dopamine signals comparing the first and last session (n = 24, two-way RM ANOVA, p = 0.0495). H In the first discrimination session, the peaks varied based on cue type, and by the final cue block, the fear and safety cues elicited higher peak signals compared to the reward and neutral cues (n = 24, two-way mixed model, p = 0.0373). I By the last session, the fear cues maintained sustained BLA dopamine responses and were larger than the safety, reward, and neutral cues (n = 24, two-way mixed model, p < 0.0001). J There was no correlation between freezing behavior and the safety cue evoked dopamine peaks, while the fear cue dopamine signal negatively correlated with fear cue-induced freezing behavior by the end of training (n = 24, Pearson, safety: p = 0.9615, fear: p = 0.0102). K The dopamine signal for the reward and neutral cues was not correlated with port entries (n = 24, Pearson, reward: p = 0.2234, neutral: p = 0.2918). Data represent mean ± SEM. All tests are two-tailed. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. During this experimental phase, rats spent more time in the sucrose port during reward cue presentation periods than any other cue period (Fig. 4C; F(3,69) = 18.24, p < 0.0001). They quickly discriminated cue types and this discrimination grew across sessions (Fig. 4C; session x cue type interaction F(12,258) = 7.410, p < 0.0001). Additionally, cue-induced freezing was modulated by cue type (Fig. 4D; F(3,69) = 35.46, p < 0.0001), with a graded response in freezing across fear, safety, reward, and neutral cues. Further, fear cue discrimination grew across sessions (Fig. 4D; session x cue type interaction F(3,69) = 4.694, p = 0.0048). Compared to the reward and neutral cues, the fear cue elicited more freezing behaviors in both the first (p = 0.0025; p < 0.0001) and last training session (p < 0.0001; p < 0.0001). The safety cue also evoked significantly more freezing than the reward cue (p < 0.0001) by the last cue discrimination day and the neutral cue presentations on both the first and last cue discrimination days (p < 0.0001; p < 0.0001). There were no significant changes to reward or neutral cue freezing across training, which remained low. When comparing males and females, there were no significant differences in port entry behaviors to any of the cue types. There were also no sex differences in cued freezing behaviors during the cue discrimination phase. However, females exhibited significantly more locomotion than males during the neutral cue presentations (Fig. S2). BLA dopamine responds preferentially to fear and safety cues throughout discrimination learning Prominent increases in BLA dopamine were seen in response to all cues during the first (Fig. 4E) and last (Fig. 4F) discrimination training sessions, with different cues prompting different dopamine levels as the sessions progressed (Fig. 4G, AUC session x cue interaction, F(3,69) = 2.746, p = 0.0495). The fear and safety cues evoked larger responses than the reward and neutral cues (Fig. 4E-G), but there were no differences between the AUC values for the fear and safety signals (p = 0.9993) nor the reward and neutral signals (p = 0.3787) in the first discrimination session. Early in discrimination training, the dopamine responses to the fear and safety cues were characterized by large peaks followed by a sustained elevation, while the reward and neutral cues prompted smaller peaks that more readily returned to baseline (Figs. 4E; S4). In contrast, by the last discrimination session (Figs. 4F; S4), the reward and neutral cues maintained a similar evoked peak but returned to baseline more quickly. Additionally, within the last discrimination session, the fear cue elicited a similar high peak and sustained signal, but the safety cue signal diminished relative to session 1 (Fig. 4G; AUC: p < 0.0001), reaching an intermediate level interposed between the fear cue trace and the reward and neutral cue traces (Fig. 4F). The reward cue response also decreased relative to session 1 (Fig. 4G; AUC p = 0.0325), mirroring the pattern seen during reward conditioning training (Fig. 3F). On the last discrimination day, the fear cue presentations elicited larger and more sustained dopamine dynamics than the safety cue (p = 0.0231), reward cue (p < 0.0001), and neutral cue (p < 0.0001; Fig. 4G). Throughout the first cue discrimination session, cued dopamine levels were rapidly updated as rats began to discriminate the different cue types (Fig. 4H; trial x cue type interaction F(12,270) = 1.876, p = 0.0373), as demonstrated during the final block of cue presentations, where both the fear and safety cues elicited signals with higher peak values compared to reward (fear: p = 0.0001; safety: p = 0.0467) and neutral cues (fear: p < 0.0001; safety: p = 0.0018). Fear cues also continued to elicit strong and sustained BLA dopamine responses, which were generally greater than those evoked by the safety, reward, or neutral cues (Fig. 4I; F(3,69) = 15.18, p < 0.0001). During the final cue discrimination session, there was no relationship between safety cue-evoked dopamine and freezing behavior (4 J; Pearson correlation, p = 0.9615). However, after discrimination learning, there was a mild negative relationship between the peak dopamine to the fear cue and conditioned freezing behavior (Fig. 4J; Pearson correlation, p = 0.0102). There were no relationships found between evoked dopamine and conditioned port entry behaviors for the reward cue (Pearson correlation, p = 0.2234) or the neutral cue (Pearson correlation, p = 0.2918) on the last cue discrimination session (Fig. 4K). Overall, in various conditioning contexts, we saw positive-going in vivo BLA dopamine signals during rewarding and aversive events and their predictive cues (Figs. 2A, B, D, G; 3E, H; 4E, F). Cue-evoked signals had clear qualitative distinction in the representation of different types of cues by the end of discrimination training (Fig. 4E, F). While there was a general bias toward the more fearful or ambiguous cues, dopamine responses were larger and more preferentially biased toward threat-related cues, in particular early in discrimination learning. Reward cue-evoked dopamine signals dynamically rescale with changing emotional uncertainty Presentations of a reward cue predicting a sucrose reward delivery (Fig. 5A) reliably elicited positive BLA dopamine dynamics during the multi-phase Pavlovian cue discrimination paradigm (Figs. 3E, I; 4E, F). When comparing the conditioned behaviors during the first and last reward conditioning sessions, compound conditioning, and the first and last cue discrimination sessions, we saw that the time spent in the sucrose port varied by phase (Fig. 5B; F(4,83) = 8.149, p < 0.0001). Specifically, reward seeking behaviors increased across reward conditioning (day 1 vs. day 7: p = 0.0025) and cue discrimination (first vs. last day: p = 0.0128), but the time spent in the reward port during reward cue presentations was blunted on the first cue discrimination session compared to the final reward conditioning session (p < 0.0001). Likewise, the reward cue evoked dopamine signals across the paradigm also varied by phase (Fig. 5C). In general, the cued dopamine responses were characterized by a quick rise to a peak of differing magnitudes followed by varying levels of maintained activity. The dopamine signals were biased in magnitude according to phase for AUC (Fig. 5D; F(4,85) = 9.091, p < 0.0001) and peak value measures (Fig. 5E; F(4,85) = 9.455, p < 0.0001),Fig. 5Reward-cue evoked dopamine signals dynamically rescale with changing emotional uncertainty.A Throughout the multi-phase Pavlovian cue discrimination paradigm, rats learned a 20 s auditory cue predicted a 3 s delivery of sucrose into the reward port. B In all phases, rats spent time in the reward port during reward cue presentations. Port entry behaviors increased during reward conditioning and cue discrimination, but generally transiently decreased as new cue types were introduced (n = 24, one-way mixed-effects model, p < 0.0001). C Here, the reward cue evoked BLA dopamine responses across the experiment phases are plotted together. These signals consisted of a mild to moderate peak with varying levels of sustained activity. D AUC comparisons of the first 2 s after reward cue onset (n = 24, one-way mixed-effects model, p < 0.0001) and E peak measures of the BLA dopamine signals (n = 24, one-way mixed-effects model, p < 0.0001) show the reward cue elicited the highest amount of dopamine during the first day of cue discrimination compared to all other phases. RC = reward conditioning. CC = compound conditioning. CD = cue discrimination. Data represent mean ± SEM. All tests are two-tailed. *p < 0.05, **p < 0.01, ****p < 0.0001. Reward-cue evoked dopamine signals dynamically rescale with changing emotional uncertainty. A Throughout the multi-phase Pavlovian cue discrimination paradigm, rats learned a 20 s auditory cue predicted a 3 s delivery of sucrose into the reward port. B In all phases, rats spent time in the reward port during reward cue presentations. Port entry behaviors increased during reward conditioning and cue discrimination, but generally transiently decreased as new cue types were introduced (n = 24, one-way mixed-effects model, p < 0.0001). C Here, the reward cue evoked BLA dopamine responses across the experiment phases are plotted together. These signals consisted of a mild to moderate peak with varying levels of sustained activity. D AUC comparisons of the first 2 s after reward cue onset (n = 24, one-way mixed-effects model, p < 0.0001) and E peak measures of the BLA dopamine signals (n = 24, one-way mixed-effects model, p < 0.0001) show the reward cue elicited the highest amount of dopamine during the first day of cue discrimination compared to all other phases. RC = reward conditioning. CC = compound conditioning. CD = cue discrimination. Data represent mean ± SEM. All tests are two-tailed. *p < 0.05, **p < 0.01, ****p < 0.0001. During the first discrimination session, the reward cue elicited a more sustained signal (Fig. 5D) and a higher peak value (Fig. 5E) compared to the reward cue signals on reward conditioning day 1 (AUC: p = 0.0034; peak value: p = 0.0014), reward conditioning day 7 (AUC: p < 0.0001; peak value: p < 0.0001), compound conditioning (AUC: p = 0.0167; peak value: p = 0.0157), and the final discrimination (AUC: p = 0.0408; peak value: p = 0.0357). Further, on the last discrimination day, the reward cue signal was greater than the reward conditioning day 7 signal in terms of AUC (p = 0.0234) and peak value (p = 0.0215). These data support the notion that BLA dopamine scales with emotional uncertainty, in this case, when new cues need to be disambiguated. The first day of the cue discrimination phase had the highest reward cue induced dopamine response (Fig. 5C), but also had the lowest port entry behaviors (Fig. 5B). Moreover, while time in the reward port increased across the cue discrimination phase (Fig. 5B), the reward cue traces diminished (Fig. 5C). These data reinforce the notion from our correlation data that BLA dopamine response does not encode behavioral expression, instead dynamically reflects the relative emotional intensity of sensory cues. Bias in extinction of reward versus fear cue-evoked BLA dopamine dynamics We next conducted extinction of cue discrimination (Fig. 6A), wherein the reward and fear cues were presented without sucrose or footshock (Fig. 6B). Conditioned behavioral responses extinguished quickly. When compared to the final day of cue discrimination, time spent in the reward port during reward cue presentations significantly decreased (Fig. 6C; t(19) = 4.134, p = 0.0006). Within extinction training, the time spent in the reward port during the reward cue presentations was generally greater than the fear cue presentations (Fig. 6C; F(1,46) = 8.674, p = 0.0050), but this difference was gone by the third extinction session (Fig. 6C; p = 0.2148). Concurrently, cue-induced freezing behavior to the fear cue was higher than the reward cue (Fig. 6D; F(1,46) = 18.16, p < 0.0001) throughout all of extinction training, but this cued freezing behavior discrimination decreased across sessions (Fig. 6D; session x cue type interaction F(2,78) = 5.882, p = 0.0042). This effect was especially blunted by the end of extinction as freezing to the fear cue was significantly lower on day 3 compared to day 1 (p = 0.0029). Females and males had similar cue-induced freezing during extinction training; however, different cue types prompted varying levels of freezing between the sexes, as reward cue induced freezing decreased in females, but was maintained in males during extinction (Fig. S2; session x sex interaction, F(2,37) = 4.852, p = 0.0135). Behavioral changes in cued locomotion levels were similar in both females and males (Fig. S2). While the qualitative signal patterns and dynamics of the cue-evoked BLA dopamine established during discrimination training were partially maintained throughout all three extinction sessions for the reward cue (Fig. 6E) and fear cue (Fig. 6F), there were significant changes in the magnitude of evoked signals and encoding of both cue types. Dopamine responses decreased to the reward cue throughout extinction training (Fig. 6E, G; F(2,39) = 12.35, p < 0.0001; day 1 vs day 2, p = 0.0139 and day 3, p < 0.0001), but not when comparing the first extinction session to the final discrimination session (t(23) = 1.403, p = 0.1740). Alternatively, the fear cue-evoked signal quickly decreased relative to cue discrimination (t(23) = 3.125, p = 0.0048) and continued to decrease as extinction training progressed (Fig. 6F, H; F(2,39) = 19.28, p < 0.0001; day 1 vs day 2, p = 0.0187 and day 3, p < 0.0001).Fig. 6Bias in extinction of reward versus fear cue-evoked BLA dopamine dynamics.A Schematic of the multi-phase Pavlovian cue discrimination paradigm the rats were trained on (n = 24; 11 F, 13 M). Behavior and photometry data from the extinction training phase are shown. B The reward cue and fear cue were presented without their previously paired outcomes (sucrose or footshock). C Compared to the last day of cue discrimination, port entry behaviors diminished during the reward cue when sucrose is omitted and remained absent during fear cue presentations (n = 24; paired t-test; p = 0.0006; two-way mixed-effects model, main effect of cue, p = 0.0050). D Cue induced freezing behavior decreased for the fear cue across extinction training, while freezing responses remained low for the reward cue (n = 24, two-way mixed-effects model, session by cue interaction, p = 0.0042; main effect of cue, p < 0.0001). E The reward cue continued to elicit moderate BLA dopamine dynamics during the first extinction with similar levels to the final discrimination session but began to lessen throughout extinction training. F The fear cue dopamine trace immediately decreased from the levels seen in the discrimination phase and continued to temper down until the final extinction training day. G Within extinction, AUC values decreased across extinction for the reward cue (n = 24, paired t-test; p = 0.1740; one-way mixed-effects model, p < 0.0001), H and also decreased for the fear cue (n = 24, one-way mixed-effects model, p < 0.0001). This pattern was also seen when comparing the fear cue AUC values from extinction day 1 to the final cue discrimination session (paired t-test; p = 0.0048). CD=cue discrimination. Data represent mean ± SEM. All tests are two-tailed. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. Bias in extinction of reward versus fear cue-evoked BLA dopamine dynamics. A Schematic of the multi-phase Pavlovian cue discrimination paradigm the rats were trained on (n = 24; 11 F, 13 M). Behavior and photometry data from the extinction training phase are shown. B The reward cue and fear cue were presented without their previously paired outcomes (sucrose or footshock). C Compared to the last day of cue discrimination, port entry behaviors diminished during the reward cue when sucrose is omitted and remained absent during fear cue presentations (n = 24; paired t-test; p = 0.0006; two-way mixed-effects model, main effect of cue, p = 0.0050). D Cue induced freezing behavior decreased for the fear cue across extinction training, while freezing responses remained low for the reward cue (n = 24, two-way mixed-effects model, session by cue interaction, p = 0.0042; main effect of cue, p < 0.0001). E The reward cue continued to elicit moderate BLA dopamine dynamics during the first extinction with similar levels to the final discrimination session but began to lessen throughout extinction training. F The fear cue dopamine trace immediately decreased from the levels seen in the discrimination phase and continued to temper down until the final extinction training day. G Within extinction, AUC values decreased across extinction for the reward cue (n = 24, paired t-test; p = 0.1740; one-way mixed-effects model, p < 0.0001), H and also decreased for the fear cue (n = 24, one-way mixed-effects model, p < 0.0001). This pattern was also seen when comparing the fear cue AUC values from extinction day 1 to the final cue discrimination session (paired t-test; p = 0.0048). CD=cue discrimination. Data represent mean ± SEM. All tests are two-tailed. *p < 0.05, **p < 0.01, ***p < 0.001, ****p < 0.0001. BLA dopamine scales with threat intensity but not fear expression Given that we saw larger BLA dopamine signals to fear-related cues during compound conditioning (Fig. 3H) and cue discrimination (Fig. 4E,F), we sought to understand if BLA dopamine would also scale when rats were tasked with discriminating cues associated with different levels of aversive valence. During a fear scale discrimination behavioral paradigm, rats were presented with intermingled fear cues within the same session (Fig. 7A): a low shock cue (20 s auditory tone paired with a 0.2 mA, 0.5 s footshock) and a higher, moderate shock cue (20 s auditory tone paired with a 0.4 mA, 0.5 s footshock), presented 5 times each per session. Rats froze to both cues, and this freezing grew across sessions (Fig. 7B; F(1,9) = 44.91, p < 0.0001). By the final day of fear scale discrimination, rats exhibited higher freezing levels compared to first day levels for the high shock cue (p = 0.0014) as well as the low shock cue (p = 0.0164). However, freezing was not significantly modulated by shock cue type (F(1,9) = 0.09096, p = 0.7698).Fig. 7BLA dopamine tracks threat intensity but not fear expression.A Rats (n = 10; 4 F, 6 M) underwent scaling fear discrimination, where two 20 s auditory cues were intermingled in each training session, one cue culminated with a 0.2 mA footshock (“Low Shock Cue”) and the other cue with a 0.4 mA footshock (“High Shock Cue”). B Cue-induced freezing behavior was high for both fear cue types (n = 10, two-way RM ANOVA, main effect of session, p < 0.0001). C Upon the first discrimination sessions, both fear cues evoked robust BLA dopamine responses that were sustained in nature. D By the last fear discrimination session, both fear cues maintained their robust and sustained BLA dopamine responses. E There were five cue trials for each fear cue type. In the first session, peak values of the dopamine traces began to diverge in magnitude, such that by the final cue presentation, the peak of dopamine trace for the high shock cue was larger than the peak of the low shock cue (n = 10, paired t-test; p = 0.0141). F By the last session, both fear cues maintained robust and sustained BLA dopamine responses (n = 10, two-way RM ANOVA, main effect of intensity, p = 0.0087). G Within the last session, the high shock cue presentations elicited higher dopamine dynamic peak values compared to the low shock cue presentations (n = 10, two-way mixed model, main effect of intensity, p = 0.0141). H BLA dopamine traces for the full 20 s cue presentations of the low and high shock cues. The high shock cue remained elevated compared to the low shock cue throughout the entire cue presentation as well as during the brief shock at 20 s. I On the final discrimination session, neither the low shock cue dopamine signals (orange) nor the high shock cue dopamine signals (brown) were correlated with cue-induced freezing behaviors (n = 10, Pearson correlation, low: p = 0.8169, high: p = 0.4391). J A separate group of TH-cre rats (n = 8; 3 F, 5 M) underwent fear conditioning paired with optogenetics. Dopamine terminals in the BLA were excited (5 sec, 20 Hz) at the onset 50% of fear cue presentations, which did not affect expression of freezing (n = 8, paired t-test; p = 0.3156). Data represent mean ± SEM. All tests are two-tailed. *p < 0.05, **p < 0.01. BLA dopamine tracks threat intensity but not fear expression. A Rats (n = 10; 4 F, 6 M) underwent scaling fear discrimination, where two 20 s auditory cues were intermingled in each training session, one cue culminated with a 0.2 mA footshock (“Low Shock Cue”) and the other cue with a 0.4 mA footshock (“High Shock Cue”). B Cue-induced freezing behavior was high for both fear cue types (n = 10, two-way RM ANOVA, main effect of session, p < 0.0001). C Upon the first discrimination sessions, both fear cues evoked robust BLA dopamine responses that were sustained in nature. D By the last fear discrimination session, both fear cues maintained their robust and sustained BLA dopamine responses. E There were five cue trials for each fear cue type. In the first session, peak values of the dopamine traces began to diverge in magnitude, such that by the final cue presentation, the peak of dopamine trace for the high shock cue was larger than the peak of the low shock cue (n = 10, paired t-test; p = 0.0141). F By the last session, both fear cues maintained robust and sustained BLA dopamine responses (n = 10, two-way RM ANOVA, main effect of intensity, p = 0.0087). G Within the last session, the high shock cue presentations elicited higher dopamine dynamic peak values compared to the low shock cue presentations (n = 10, two-way mixed model, main effect of intensity, p = 0.0141). H BLA dopamine traces for the full 20 s cue presentations of the low and high shock cues. The high shock cue remained elevated compared to the low shock cue throughout the entire cue presentation as well as during the brief shock at 20 s. I On the final discrimination session, neither the low shock cue dopamine signals (orange) nor the high shock cue dopamine signals (brown) were correlated with cue-induced freezing behaviors (n = 10, Pearson correlation, low: p = 0.8169, high: p = 0.4391). J A separate group of TH-cre rats (n = 8; 3 F, 5 M) underwent fear conditioning paired with optogenetics. Dopamine terminals in the BLA were excited (5 sec, 20 Hz) at the onset 50% of fear cue presentations, which did not affect expression of freezing (n = 8, paired t-test; p = 0.3156). Data represent mean ± SEM. All tests are two-tailed. *p < 0.05, **p < 0.01. Both fear cues evoked large and sustained BLA dopamine signals during the first fear discrimination session (Fig. 7C). When comparing the average cue peak dopamine response for each trial, we saw a rapid distinction between the higher shock and lower shock cues emerge by the end of the first session, where the final high shock cue was significantly larger than the final low shock cue (Fig. 7E; t(8) = 3.126, p = 0.0141). The robust and sustained BLA dopamine responses were preserved during the last fear discrimination session for both the low and high fear cues (Fig. 7D). The high shock versus low shock distinction persisted and was clearer across the last fear discrimination session. The high shock cue evoked higher dopamine peak values compared to the low shock cue presentations across trials in the final session (Fig. 7F; F(1,9) = 9.216, p = 0.0141). Overall across training sessions, the high shock cue dopamine AUC was greater than the low shock cue AUC (Fig. 7G; F(1,9) = 11.15, p = 0.0087). This was driven by a trend for a larger high shock cue AUC in the first discrimination session (p = 0.0637), which was significantly larger than the low shock cue by the final session (p = 0.0483), and the high shock cue signal on the first session (p = 0.0025). The low shock cue AUC did not significantly change between the first and last fear discrimination sessions (p = 0.8121). Looking at the entire cue window, we saw both shock cues elicited prolonged BLA dopamine signals that persisted throughout the entire 20 s cue presentations. The high shock cue remained elevated in comparison to the low shock cue during the cue period as well as the footshock (Fig. 7H). Despite this, cue evoked dopamine signals were not significantly correlated with freezing behavior for either the low or high shock cue (Fig. 7I,J; Pearson correlation, low: p = 0.8169, high: p = 0.4391). VTA dopamine neuron activation can facilitate fear learning61 and specific inhibition of VTA-BLA dopamine neuron projections can disrupt fear memory recall41. While we established fear cue presentations prompt strong BLA dopamine signals (Figs. 3I; 4E, F; 7C, D, H), across studies, we did not find consistent correlations between fear cue evoked dopamine and conditioned freezing (Figs. 3L; 4J; 7I). To further explore this, we tested the impact of exogenously increasing BLA dopamine levels, in TH-cre rats. We found that optogenetic excitation of VTA-BLA dopamine terminals during fear cue presentations did not affect freezing behavior (Fig. 7J; t(7) = 1.081, p = 0.3156). Together with our above results, these data suggest that elevated cue-evoked dopamine reports the state of emotional intensity but does not acutely drive behavioral expression of emotional responses. Discussion We investigated how the BLA dopamine system is engaged as rats learn appetitive, aversive, and ambiguous associations. First, we show that VTA dopamine neuron stimulation engages robust BLA dopamine signaling that is not intrinsically reinforcing. Second, appetitive and aversive events, and their predictive cues, evoke BLA dopamine in simple cue conditioning paradigms, where signals scale with stimulus intensity but not reward value. Critically, cues evoked dopamine in a pattern that did not match the strength of learning, as signals were larger early and diminished across training, but rebounded when the emotional uncertainty of the sensory environment changed. As we increased the complexity of the learning context, where cues signaling threat, reward, safety, and neutral associations were intermingled, the level of cue-evoked BLA dopamine reported the relative emotional salience: fear and safety cues elicited larger dopamine signals than either reward or neutral cues. Finally, we show that BLA dopamine signals scale with the intensity of conditioned threats, but not the expression of fear. Together, our results show that BLA dopamine responds to salient cues, regardless of valence, and the magnitude and persistence of these signals dynamically adjust according to the emotional relevance of the stimulus within a given learning context. These unique patterns highlight a dopamine system that encodes sensory-guided state transitions in a way that is independent of associative strength or value. Our data help expand conceptual frameworks of emotional learning and offer a number of insights into amygdala and dopamine functions9,22,62–70. BLA dopamine signals do not predominantly reflect value or cue associative strength We show that VTA dopamine neurons evoke strong dopamine release in the BLA34–36. While activation of VTA dopamine projections to the striatum is well documented to create associative learning and drive behavioral reinforcement51–54, we found that VTA-BLA dopamine signaling is not reinforcing. Further, BLA dopamine levels increased during both aversive and appetitive events16,71, and these signals scaled with stimulus intensity, but not value. Together, this suggests that dopamine release in the BLA does not directly drive learning, nor does it signal the positive or negative aspects of stimuli per se, but is instead involved broadly in the encoding sensory transitions. Given that dopamine inputs to the BLA do not predominantly signal valence or value, the encoding of stimulus properties to represent valence instead is likely processed through different BLA efferent pathways to downstream regions and/or via different populations of BLA neurons28,72. The BLA contains valence-sensitive clusters of neurons, where positive and negative conditioned stimuli are preferentially encoded1,2,28,29,67–69,73,74. These ensembles are anatomically interspersed but are largely separate and similarly sized populations. While there is clear evidence that some BLA neurons have inherent bias for representation of appetitive or aversive stimuli, the differential encoding of valence information among BLA neurons is dynamic with learning14,20,21. Plasticity driving this discrimination is supported by a complex array of circuit inputs and local communication within the amygdalar complex9,30,75–77. Dopamine input also likely plays a key role, as recent studies and computational models suggest42,78–80. While additional experiments are needed to investigate this directly, we hypothesize that moment-to-moment, sensory-guided BLA dopamine fluctuations interface with distinct BLA neural ensembles for dynamic disambiguation of state. Notably, across our experiments, BLA dopamine signal patterns did not fit clearly with associative learning or prediction error frameworks - cue-evoked dopamine responses were generally larger at the beginning of a conditioning phase, before strongest cued behavioral performance. Further, these signals were engaged independent of valence and value, as neutral, threat, reward, and safety cues all evoked a positive excitation of dopamine release. We also found no consistent connection between cue-evoked BLA dopamine signals and behavioral expression, with the only significant relationship emerging as a negative correlation in fear cue-evoked dopamine signal and freezing during cue discrimination. This one correlation is difficult to interpret in isolation, but overall, these patterns are in stark contrast to classic notions of the striatal dopamine system, where learning rate and prediction are reflected in incrementally accrued cue responses, and value is represented by negatively biased signals to aversive stimuli45–48,81–86. This will require further study, but a lack of clear connection between BLA dopamine signals and behavioral expression in our data is perhaps consistent with studies showing that the BLA is not essential for expression of some forms of learning87–89. While there are other possible BLA dopamine input sources, such as from the substantia nigra, periaqueductal gray, and dorsal raphe nucleus37,90–93, we hypothesize that input from VTA dopamine neurons is responsible for signaling the cued state changes we see in the BLA. This notion is supported by previous studies. For example, lesions of VTA dopamine neurons impair BLA cue attentional signals94. VTA to BLA dopaminergic terminals are activated during learning36 and they are necessary for some aspects of reward and fear learning16,95. Further, there is heterogeneity of stimulus response profiles within populations of VTA dopamine neurons61,96–102 but a relatively anatomically uniform distribution of VTA to BLA projection neurons37,41. While it is unclear if they project to the BLA, a subset of VTA dopamine neurons are activated by both fear cues and non-predictive cues, suggesting the capacity to signal more general salience information61. Collectively, our data reveal a unique dopamine signaling profile, which adds insight to the growing understanding that dopamine neurons contribute heterogeneous information across projection sites45–50. BLA dopamine dynamically tracks cue salience and emotional ambiguity Our findings suggest that rather than reporting the strength of learning, valence, value, or behavioral vigor, BLA dopamine transmits information related to the current emotional salience and ambiguity of sensory transitions, with signals biased towards the most motivationally important stimulus in the environment. We found that dopamine signals in response to reward cues dynamically shrunk or grew not as a function of behavioral expression as learning progressed but instead tracked the emotional uncertainty of the learning context. Reward cue-evoked dopamine release attenuated over simple reward conditioning sessions, but when additional cues were added, increasing the emotional complexity of the environment and requiring discrimination, reward cue signals rebounded above those previously low levels. This qualitative pattern, which is again in contrast to classic associative dopamine patterns in the striatum, underscores the notion that BLA dopamine signals salient shifts in state and local stimulus context, rather than simply denoting either stimulus novelty or accrued learning. Consistent with this, past studies using microdialysis show that tonic dopamine levels in the BLA slowly increase while rats undergo discrimination learning39, and pharmacologically boosting BLA dopamine enhances the early acquisition of a stimulus-reward association103. Further, in humans, dopamine is released in the amygdala during shifts from rest to working memory104, suggesting that this system is re-engaged when memory of past associations are needed8,16. In our data, we also saw a clear change over time in BLA dopamine in response to the safety cue, where signals were initially larger and comparable to the fear cue, but decreased as discrimination learning progressed. These patterns support the notion that BLA dopamine is not purely biased towards negatively-valenced events but, more broadly, is invigorated and then reinvigorated when new stimuli in the environment are introduced. Collectively, our results and other important recent studies16, support a framework where BLA dopamine contributes more directly to flexible state or model-based learning, rather than reward prediction error, incentive motivation, or reinforcement105,106. We found that cues bearing greater emotional weight not only evoked larger peak BLA dopamine at cue onset, but also more sustained increases throughout the entire cue period. Fear cue-evoked signals were larger and broader, while reward and neutral cue signals were smaller and more transient. Interestingly, the safety cue evoked an intermediate response that was sustained but smaller than the fear cue signal. In line with this, when fear cues predicting different threat levels were intermixed, dopamine responses were more sustained to whichever threat was larger in the moment. This scaling of sustained dopamine matched the perceived emotional vigilance, a pattern that is again distinct from cue-evoked dopamine signals in striatum, which are typically composed of rapid phasic fluctuations45–48,81–86. Mechanistically, sustained BLA dopamine patterns may be supported by the relatively low levels of the dopamine transporter (DAT) and slower dopamine clearance rates within the amygdala, compared to the striatum107,108. Dopamine signaling in the BLA is known to affect long-term potentiation and engenders complex effects on local and projection neuron populations, partially via modulation of glutamatergic inputs109,110. Dopamine neurons projecting to the BLA exhibit unique physiological characteristics111,112, preferentially targeting BLA projection neurons containing the D1-type dopamine receptors (D1DR)35,90,113–115. Optogenetic activation of dopamine release increases cyclic AMP activity in BLA D1DR neurons and enhances their response to sensory inputs42, while pharmacological activation of BLA D1DR signaling facilitates contextual fear learning116, freezing behavior117, reward seeking and learning118, as well as safety cue-induced fear suppression19. Our results motivate future studies aimed at dissecting the contribution of stimulus-evoked BLA dopamine signals of different duration and intensity on plasticity in the BLA. One prediction, motivated by our data, is that sustained levels of dopamine during discrimination of different cues may support stimulus context encoding on longer timescales. Given that BLA dopamine levels are dynamically modulated by changes in affective and homeostatic state, these signals could contribute to scaling of value and memory representations among BLA neurons3,44,75,119. Here, we showed BLA dopamine release tracks reward, fear, discrimination, and extinction learning. Taking the classic view that the BLA processes sensory stimuli on both the proximal and environmental levels to support behavioral outputs, our results imply that BLA dopamine signals are involved in valence discrimination via a transmission bias that reports the relative emotional saliency and uncertainty of the moment. These signals are dynamically engaged in learning contexts where relative stimulus importance must be rapidly disambiguated. Our results support a framework where the BLA dopamine system, in contrast to classic striatal dopamine frameworks, serves as a general, sensory state-guided saliency signal to support emotional learning. Methods Subjects Male and female Long Evans rats (n = 41; 19 F, 22 M) were used. All rats weighed 250-500 g at the time of surgery and were 4-9 months old at the time of experimentation. Experimental procedures were approved by the Institutional Animal Care and Use Committee at the University of Minnesota and were carried out in accordance with the guidelines on animal care and use of the National Institutes of Health of the United States. Surgical procedures Rats were anesthetized with 5% isoflurane and placed in a stereotaxic frame, after which anesthesia was maintained at 1-3%. Saline, carprofen anesthetic (5 mg/kg), and cefazolin antibiotic (70 mg/kg) were administered subcutaneously prior to start of surgery. The top of the skull was exposed and holes were made for viral infusion needles, optic fiber implants, and 4 skull screws. A total of 0.5 µL of a virus coding for the dopamine biosensor dLight 1.3b (AAV5-hSyn-dLight1.3b, University of Minnesota Viral Vector and Cloning Core) was infused unilaterally into the BLA (2.7 AP, +/− 4.9 ML, −8.2 DV). In a separate cohort of rats, a total of 1.0 μL of virus (pAAV5-Syn-FLEX-Chrimson-tdTomato, Addgene) was infused unilaterally into the VTA ( − 5.6 AP, +/− 0.7 ML, −8.0 DV), while dLight1.3b injections were targeted to the ipsilateral BLA. In another cohort of rats, a total of 1.0 μL of virus (pAAV5-EF1a-double floxed-hChR2(H134R)-EYFP-WPRE-HGHpA, Addgene) was infused unilaterally into the VTA. Each virus was delivered at 0.1 µL/min. Immediately after injection, the needle was raised 100 µm and then left in place for an additional 10 min to allow for diffusion. Optical fibers (9 mm length, 400 µm diameter, Doric Lenses) for photometry recordings were chronically implanted into the BLA (2.7 AP, +/− 4.9 ML, −7.9 DV) in the same hemisphere as the viral infusion. For optogenetic stimulation, optical fibers (10 mm length, 300μm diameter, Thorlabs) were implanted above the VTA ( − 5.6 AP, +/− 0.7 ML, −7.9 DV) and/or the BLA (2.7 AP, +/− 4.9 ML, −8.0 DV) in the same hemisphere of viral injections. All coordinates are in mm relative to bregma and skull surface. Implants were secured to the skull with dental acrylic (Lang Dental) applied around the skull screws and the base of the ferrules containing the optic fibers. At the end of all surgeries, topical anesthetic and antibiotic ointment were applied to the surgical site, rats were placed on a heating pad and monitored until they were ambulatory. After surgery, rats were individually housed with ad libitum access to food and water on a 0800 to 2000 light/dark cycle (lights on at 0800), given carprofen (5 mg/kg) and cefazolin (70 mg/kg) for the first 3 days following surgery, and weighed and monitored for 6 days. Photometry recordings began at least 3 weeks after surgery. Optogenetic manipulations started at least 6 weeks after surgery. Fiber photometry recordings Behavioral testing days were paired with fiber photometry recording to measure dopamine signals (n = 32; 15 F, 17 M) in the BLA. The fiber photometry system was outfitted with optical components from Doric Lenses (Québec, Canada). This included fluorescence mini-cubes to transmit concurrent light streams from LEDs, 465 nm and 405/415 nm, that were sinusoidally modulated at 211 Hz and 330 Hz, respectively. The 465 nm channel stimulated fluorescence from the dopamine-dependent dLight, while the 405/415 nm served as an isosbestic reference that excites dLight in a dopamine-independent manner. Demodulating the brightness produced by both channels allows for reliable correction of movement or photobleaching artifacts. Before each photometry recording both LED currents were set to ~50μW. Fluorescence was transmitted from the fiber tip over the BLA back to the mini-cube via an optic cable, where it was passed through a GFP emission filter, amplified, and directed into a high sensitivity photoreceiver (Newport, Model 2151) for signal processing. Meanwhile, the session was controlled, and data were acquired via Synapse software running on a Tucker-Davis Technologies (TDT, Alachua, FL) real-time processor (RZ5P or RZ10X), as reported in previous studies47,120. Synapse software on the processor calibrated the output of each LED channel and simultaneously recorded photometry signals, which were being sampled from the photodetector at 6.1 kHz. To align the real-time neural signals to behavioral (e.g., port entries) and environmental (e.g., cue presentations) occurrences, TTL signals from the operant chambers (Med Associates) were sent to the RZ5P processor and timestamped for post-hoc analysis of the photometry data. Video recordings were also taken and integrated with the photometry and behavior data via the Synapse software. Optogenetic stimulation and behavior Optogenetic stimulation These studies used 589 nm lasers (OptoEngine and Dragon Lasers) and 473 nm lasers (OptoEngine). Light output included individual 5 ms light pulses for either 5 Hz or 20 Hz stimulation frequency. Light intensity was adjusted to be ~2 mW/mm2 at the tip of the intracranial fiber during individual pulses (~10 mW constant power). The optic tethers connected to the rats were sheathed in a lightweight armored jacket to block visible light transmission and reinforce the cable to prevent breakage. Laser pulses were initiated from Med-PC TTL pulses and patterned via Synapse software that also controlled photometry and video data acquisition. Fiber photometry recordings were concurrently collected on every optogenetic training day. Fiber photometry and optogenetics To understand how VTA TH-neuron activity influences dopamine dynamics in the BLA, in vivo optogenetics was paired with fiber photometry in TH-Cre rats (n = 8; 4 F, 4 M). Rats were first acclimated to the behavioral chambers (Med Associates) and optic cable tethering in a ~ 30 min habituation session. During VTA TH-cell body stimulation, an optic tether was attached over the VTA implant to deliver optogenetic stimulation, while another was attached to the BLA implant for dLight fiber photometry data collections. In separate sessions, rats were tethered to the optic cable and presented with 8-10 laser stimulations on a 120 s fixed time schedule. The stimulation parameters were (1) 20 Hz stimulation for 2 s, (2) 5 Hz stimulation for 5 s, and (3) 20 Hz stimulation for 5 s. These stimulation presentations were un-cued and not paired with any external stimulus. The same stimulation parameters and sessions were repeated for VTA TH-neuron terminal stimulation over the BLA. Here, the optic tether connected to the BLA implant concurrently delivered the laser stimulation, while collecting dLight fiber photometry fluorescence signals. Intracranial self-stimulation (ICSS) We expressed an excitatory opsin VTA dopamine neurons in TH-Cre rats (n = 14; 6 F, 8 M) and inserted an optic fiber in the BLA. Rats were placed in the testing chamber (Med Associates) and connected to an optic cable tether to receive optogenetic stimulation to target VTA dopamine terminals in the BLA. Two nose poke ports were positioned on one side of the chamber. Nose pokes in the active port resulted in a 2 s laser train (20 Hz, 5-ms pulses, fixed-ratio 1 schedule) that coincided with the delivery of a tri-color light illumination cue. Laser stimulation parameters were chosen to align with the results from the fiber photometry studies above where 2 s, 20 Hz VTA-BLA dopamine terminal stimulation prompted reliable dopamine dynamics in the BLA. Optogenetic stimulation during fear conditioning To investigate whether exogenously increasing BLA dopamine levels during a fear cue would impact freezing behaviors, Chrimson was virally expressed in VTA dopamine neurons in TH-cre rats (n = 8; 3 F, 5 M) and an optic fiber was implanted over the BLA. Rats then underwent two days of fear conditioning, where a 20 s auditory cue was paired with a moderate footshock (0.4 mA) delivered from 20 to 20.5 s. There were 14 cue-shock presentations each session with each cue trial separated by an intertrial interval that averaged 115 s (range: 100 s–130 s). Half of cue trials were paired with a 5 s, 20 Hz laser delivery concurrent with the fear cue onset to prompt VTA-BLA dopamine terminal activation. The stimulation trials were pseudorandomly delivered interwoven with the non-stimulation trials. Each session lasted ~30 min. Pavlovian behavioral assays All Pavlovian behavior training began at least 3–4 weeks after surgery. Med Associates chambers were outfitted with a syringe pump to deliver liquid sucrose reward, a magazine port equipped with an infrared beam to sense port entries, floor bars for the delivery of electrical shocks, speakers for the delivery of tone (high: 4500 Hz, low: 2900 Hz), and house lights for illumination. Cameras were positioned above each chamber to record movement throughout each session for post-hoc behavioral analysis. Port entry instances and durations were recorded by Med Associates software throughout the sessions. Chambers were cleaned with 70% ethanol and bedding was replaced after each session. Cue discrimination paradigm This Pavlovian cue discrimination task (n = 24; 11 F, 13 M) was adapted from previous studies (Fig. 4A)14. Habituation Training began with acclimation to the behavioral chambers and optic cable tethering in two ~30-minute habituation sessions. During these sessions, rats were tethered to the fiber photometry cable and allowed to freely move and explore around the chamber. Reward conditioning For the 7 following sessions, rats were presented with 25 auditory cue presentations lasting 20 s each on a 110 s VT schedule (range: 90–130 s). The auditory cue was either a pulsing low tone (repeating cycle of 0.2 s on and 0.2 s off) or a continuous high tone counterbalanced across rats. Each reward cue presentation was paired with a liquid reward (10% sucrose) delivered pseudorandomly between 10 s and 20 s after the reward cue onset. Each session lasted ~55 min. Cue habituation To acclimate the rats to the remaining cues, during an 8th reward conditioning session (25 auditory reward cue+sucrose presentations), the rats were pseudorandomly presented with noncontingent cues (5 house lights illuminations, 5 auditory (counterbalanced high or low) tones) lasting 20 s each. These cue presentations were not associated with any outcome. This session lasted ~75 min. Compound conditioning The next training session introduced the fear cue: 5 presentations of a 20 s auditory cue paired with a mild footshock (0.4 mA) delivered from 20 to 20.5 s. The fear cue presentations were pseudorandomly delivered intermixed with 25 reward cue presentations. This session lasted ~65 min. Cue discrimination After initial learning, the rats underwent the cue discrimination phase of the paradigm. Here, the rats had to distinguish between 4 different cues predicting various outcomes: the previously introduced (1) reward cue (paired with sucrose) and (2) fear cue (paired with a footshock), as well as a (3) neutral cue with no paired outcome and a (4) compound cue (fear+neutral cue played concurrently) to signal a relief from shock. The high and low tones, as well as pulsatile versus continuous tones, were counterbalanced across rats for the reward and fear cues. House light illumination was always used for the neutral cue. The compound cue signaling relief from shock was always the fear cue (counterbalanced auditory cue) played concurrently with the neutral cue (visual cue) for 20 s. There were 45 total cue presentations in each training session (15 reward cues, 5 fear cues, 10 neutral cues, and 15 compound cues). Each cue lasted 20 s and was delivered in pseudorandom orders on a 110 s average VT schedule (range 90-130 s). Each animal had 4-6 cue discrimination sessions, each lasting ~100 min. Extinction Behavioral testing concluded with 2–3 extinction sessions. Here, only the reward and fear cues were presented (20 presentations each; pseudorandom order) and both the reward and shock deliveries were omitted. These sessions lasted ~85 min each. Scaling valenced stimuli presentations Reward Scaling Rats (n = 8; 3 F, 5 M) were placed in Med Associates chambers while tethered to a fiber photometry cable. In one session, a 10% sucrose liquid reward was delivered via a syringe pump into the magazine port on the right side of the chamber. In a separate session, an Ensure liquid reward was delivered via a syringe pump into the magazine port on the left side of the chamber. For both reward types, there were 20 uncued, unsignaled reward presentations into the respective magazine port, which rats could enter at any time to consume the reward throughout the session. Both the left and right magazines were open during both reward sessions. Reward pumps were on for 3 s and delivery occurred on a 105 s average VT schedule (range: 90–120 s). Each session lasted ~38 min. Footshock scaling Rats (n = 10; 3 F, 7 M) were placed in Med Associates chambers while tethered to a fiber photometry cable. Footshocks were delivered via electric currents through the floorbars. In one session, 0.2 mA uncued, unsignaled 0.5 s footshocks were delivered. In a separate session, 0.4 mA uncued, unsignaled 0.5 s footshocks were delivered. Sessions were conducted in a counterbalanced manner. In both sessions, there were 6 footshock presentations that were delivered on a 210 s average VT schedule (range: 180–240 s). Each session lasted ~18 min. Fear scaling discrimination paradigm Single stimulus fear conditioning Rats (n = 10, 4 F, 6 M) underwent 2 distinct Pavlovian fear conditioning sessions. One session had 5, 20 s auditory cue presentations paired with a lower footshock (0.2 mA, 0.5 s) delivered at the end of the cue. A separate session had 5, 20 s auditory cue presentations paired with a higher footshock (0.4 mA, 0.5 s) delivered at the end of the cue. The tones used were counterbalanced across rats. In one group, a continuous low tone was used for the 0.2 mA cue, while a pulsing low tone (repeating cycle of 0.2 s on and 0.2 s off) was used for the 0.4 mA cue. In the other group, a pulsing high tone (repeating cycle of 0.2 s on and 0.2 s off) was used for the 0.2 mA cue, while a continuous high tone was used for the 0.4 mA cue. In both sessions, the cues were delivered in pseudorandom order on a 110 s average VT schedule (range 90-130 s) and each session lasted ~12 min. Fear discrimination After initial learning, the rats underwent the cue discrimination phase of the paradigm. Here, the rats had to distinguish between the 2 different cues predicting the footshock outcomes: the previously introduced (1) low fear cue (paired with a 0.2 mA footshock) and (2) high fear cue (paired with a 0.4 mA footshock). There were 10 total cue presentations in each training session (5 low fear cues and 5 high fear cues). Each cue lasted 20 s and was delivered in a pseudorandom order on a 110 s average VT schedule (range 90–130 s). Each animal had 3 fear discrimination sessions, each lasting ~22 min. DeepLabCut-based pose tracking Markerless tracking of animal body parts was conducted using the DeepLabCut Toolbox (DLC version 2.2.3 or 3.3.0)121 and analysis of movement features based on these tracked coordinates was conducted in Matlab R2024b (Mathworks). Videos (944 × 480 resolution) were recorded with a sampling frequency of 10–20 frames per second using TDT Synapse software with overhead cameras (Vanxse CCTV 960H 1000TVL HD Mini Spy Security Camera 2.8-12 mm Varifocal Lens Indoor Surveillance Camera). DeepLabCut model Videos from this experiment were analyzed through a refined DLC network47,54,120. To create the network, the body parts labeled included the nose, eyes, ears, fiber optic implant, shoulders, tail base, and an additional three points along the spine. Features of the environment were also labeled, including the 4 corners of the apparatus floor and the two magazine ports. Briefly, 2090 frames from 35 videos (32 different animals, 3 experiments) were labeled and 807 outlier frames were relabeled to refine the network for the current study. Labeled frames were split into a training set (95% of frames) and a test set (5% of frames). A ResNet-50 based neural network was used for 1,030,000 training iterations. After final refinement, we used a p-cutoff of 0.85 resulting in training set error of 2.99 pixels and test set error of 3.68 pixels. This model was then used to analyze videos from the rats (n = 26; 13 F, 13 M) who underwent the cue discrimination behavioral assay and the fear scale discrimination behavioral assay for all training sessions. Data analysis Behavioral analysis Port entries, locomotion bouts, and freezing bouts were the primary behavioral output measures. Port entry frequency and duration was analyzed from MedPC software outputs and the generated TTL signals from Synapse. Video Scoring: Freezing data was first analyzed with hand-scoring of recorded videos from the cue discrimination training sessions. For each animal, freezing duration for each cue presentation was calculated during at least the first and last cue discrimination training session. The experimenter was blinded for both rat and session information. Cue-evoked freezing was timed for each cue three times and the average was calculated. DeepLabCut Analysis: Freezing bouts and locomotion bouts were also calculated from positional information tracked using DLC. Bodypart coordinates and confidence values for each video frame were determined by the DLC network and were analyzed with Matlab. Broadly, any frames with body part coordinates where confidence values were below 0.7 were excluded from further data processing. The average coordinates of fixed features of the operant chamber were used for analysis. A pixel-to-cm ratio was calculated for each processed video in order to transform pixel distances into physical chamber dimensions for subsequent distance traveled and speed analyses. The pixel distance of all 4 perimeter sides and both diagonal cross-sections was determined, divided by the actual distance in cm, then that quotient value is averaged to give the pixel-to-cm conversion for each video. Movement speed was calculated from the coordinates of the mid-back, for each video frame, the formula [distance moved (pix per cm) * framerate] was used to provide movement speed in cm/s. For analyzing data across recordings with different sampling rates, missing data points for the lower sampling rate sessions were filled in using linear interpolation with the interpol1 function (MATLAB). Interpolated data points were not used to calculate peak or mean speeds for analysis. Freezing and locomotion detection was uniquely computed for each animal using a movement threshold determined by animal body size (distance between shoulder and lower back) and the predetermined optimal threshold for movement detection based on our operant chamber configuration47,54,120,122. This threshold value was used to detect fine movement in the face and head; for movement detection in the body, the limit was multiplied by 2 to account for larger body parts. Freezing bouts were defined as all visible body parts being below their corresponding movement threshold for any single video frame, while the transitions in and out of a movement bout were demarcated by a sliding window when the speed of 2 or more body parts exceeded the detection threshold for 0.3 s. For all bouts, video frames where less than 2 body parts were visible were excluded. For freezing, bouts lasting under 1 s were excluded from bout analysis. Locomotion was detected using the speed of the implant and 4 locations on the spine. Locomotion bouts were determined if all visible body parts were above their corresponding movement threshold for any single video frame, while the transitions in and out of a movement bout were demarcated by a sliding window when the speed of 2 or more body parts exceeded the detection threshold for 0.3 s. For locomotion, bouts lasting less than 0.5 s were excluded from bout analysis. For a subset of videos, the filtered predictions were used and the average frame x frame movement across the fixed features of the environment was subtracted to remove jitter caused by camera movement. This restored freezing detection to similar levels as videos where camera movement was not a factor. Fiber photometry analysis Photometry data were analyzed using MATLAB and GraphPad Prism. First, the photometry signals (both 405 nm/415 nm isosbestic and 465 nm) were low pass filtered and downsampled to 40 Hz. The isosbestic signal was then aligned to the 465 nm signal by applying a least-squares linear fit and this fitted isosbestic signal was used to normalize the 465 nm signal, where ΔF/F = (465 nm signal – fitted 415 nm signal)/(fitted 415 nm signal). Timestamps for each video recording (recorded at 10–20 FPS) were recorded in the Synapse photometry data file. Cue presentations, port entries, sucrose pumps, and footshocks were assigned a TTL signal from Med-PC that was used to timestamp the photometry data file in the Synapse software. The normalized ΔF/F signals were extracted 5 s before and 25 s following each cue presentation and Z-scored to the 5 s pre-cue period for each trial to minimize the effects of drift in the signal across experiment duration. Normalized signals for reward consumption were extracted in the window from 2 s preceding to 4 s following each port entry and Z-scored to the 2 s pre-port entry. Port entry occurrences were binned into rewarded and unrewarded port entries. The first port entry after reward delivery with a duration of >0.5 s was designated a rewarded port entry. Normalized signals for shock responses were extracted in the window from 2 s preceding to 4 s following each shock delivery and Z-scored to the 2 s pre-shock delivery. Cue responses were detected in the average waveform for each animal in the 5 s window beginning at cue onset, while reward and shock responses used a 2 s window beginning at event onset. The maximum and minimum responses in each window were detected, and latency to the maximum or minimum was calculated relative to the start of the detection window. Area under the curve (AUC) values were calculated from the Z-scored traces by numerical integration via the trapezoidal method using the trapz function (Matlab). Statistics Behavior and photometry signal data were analyzed with a combination of ANOVA models (one-way, two-way, and mixed effects), t-tests, and Pearson correlations. Post-hoc comparisons and planned t-tests were used to clarify main effects and interactions, with a Bonferroni correction applied. Photometry traces represent average values of sites/subjects, not individual trials. The data are expressed as the mean ± s.e.m. All tests were two-sided, and the statistical significance was set at p < 0.05. Histology Rats were deeply anesthetized via i.p. injections of Fatal-Plus (2 ml/kg; Paterson Veterinary) and were transcardially perfused with cold phosphate buffered saline (PBS) and subsequent 4% paraformaldehyde (PFA). Brains were extracted and continued to be fixed in 4% PFA overnight (~24 h) then were transferred to a cryoprotectant solution of 30% sucrose in PBS for at least 48 h. The tissue was cut into 50 μm sections on a freezing microtome (Leica Biosystems) and brain sections containing the BLA were mounted on microscope slides and coverslipped with Vectashield containing DAPI counterstain to validate viral expression and proper optic fiber placements. Images of the microscope slides were taken on a fluorescent microscope (Keyence BZ-X710) with a 4x and a 10x air immersion objective. Virus spreads and fiber positions were compared to the Rat Brain Atlas (Paxinos and Watson, 2007) and plotted with Adobe Illustrator. Immunohistochemistry To visualize dopamine projections in the amygdala, we performed immunohistochemistry for tyrosine hydroxylase. Brain sections were washed in PBS and incubated with bovine serum albumin (BSA) and Triton X-100 (each 0.2%) for 20 min. 10% normal donkey serum (NDS) was added for a 30 min incubation, before primary antibody incubation (rabbit anti-TH, 1:500, Invitrogen) overnight at 4 °C in PBS with BSA and Triton X-100 (each 0.2%). Sections were then washed and incubated with 2% NDS in PBS for 10 min and secondary antibodies were added (1:200 594 donkey anti-rabbit) for 2 h at room temperature. Sections were washed 2 times in PBS and mounted with Vectashield containing DAPI. Reporting summary Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article. Supplementary information Supplementary Information Reporting Summary Transparent Peer Review file Supplementary Information Reporting Summary Transparent Peer Review file Source data Source Data Source Data