Behavioral Neuroscience

Hormonal regulation of patch-leaving decision-making remains poorly understood. Here, young adult male and female Long-Evans rats were tested in a patch-leaving task before and after orchiectomy (ORCH), ovariectomy (OVX), or sham surgery, and were subsequently assessed in an impulsive-choice task. Patch leaving was measured under long- and short-travel conditions. Before surgery, males showed longer overstay than females during long-travel sessions, whereas no clear sex difference was detected during short-travel sessions. After surgery, orchiectomy did not produce a uniform shift in patch leaving but selectively disrupted the progressive reduction in overstay that normally emerged across repeated long-travel sessions. By contrast, ovariectomy produced weaker effects and did not reveal a comparably robust change in female patch leaving. Spatial and idle occupancy analyses showed that gonadectomy also altered within-patch behavior, with orchiectomy most strongly increasing idling-related measures in males, whereas ovariectomy more strongly redistributed female patch occupancy. Estrous stage did not significantly organize pre-surgical female overstay. Greater impulsive choice was associated with smaller post-surgical reductions in long-travel overstay in the unadjusted analysis. Together, these findings indicate that testicular hormones selectively support patch-leaving adaptation under high travel cost.

RationaleAppetitive Pavlovian cues can drive maladaptive reward seeking via stimulus-reward memories. Disrupting memory reconsolidation offers a potential strategy to reduce their influence, but evidence for {beta}-adrenergic blockade with propranolol is inconsistent across behavioural paradigms, particularly relative to NMDA receptor antagonism. ObjectivesWe tested whether propranolol disrupts reconsolidation of appetitive sucrose memories in a discriminative goal-tracking paradigm, and compared its effects with those of the most commonly used NMDA receptor antagonist, MK-801. MethodsAdult Lister hooded rats underwent discriminative Pavlovian conditioning. Thirty minutes before a brief memory reminder (non-reinforced or reinforced), rats received systemic drug treatment or saline control. In study 1, MK-801 (0.1 mg/kg) was administered to male rats. In study 2, propranolol (10 mg/kg) was administered to equal numbers of male and female rats. Goal-tracking was tested drug-free at 1 and 8 days. ResultsIn study 1, MK-801 impaired subsequent discriminated responding at test. These effects were observed not only when reminder was non-reinforced as in previous successful demonstrations, but also with reinforced reminder. In study 2, Propranolol also impaired subsequent goal-tracking, regardless of reminder type, and the effects were consistent across sexes. ConclusionsPropranolol can disrupt reconsolidation of appetitive goal-tracking memories to a similar extent as MK-801 under conditions that promote memory destabilisation. These findings demonstrate that {beta}-adrenergic blockade can impair appetitive memory reconsolidation in a goal-tracking paradigm, challenging prior null findings and revitalising the potential for propranolol-based interventions in maladaptive reward-seeking behaviours.

In animal learning research, learning is often represented by plotting a behavioral measure as a function of training trials. A particularly clear case is habituation, a basic form of learning in which repeated presentation of a stimulus produces a decrement in responding. Although retention tests provide the strongest basis for evaluating durable habituation once short-lived performance effects have dissipated, the pattern of response change across stimulus repetitions, or habituation curve, remains theoretically and empirically relevant because it is used to characterize determinants of habituation, individual and clinical profiles, and functional forms, including linear, curvilinear, asymptotic, and mixed incremental-decremental patterns of responding. However, group averaged curves may conceal substantial individual heterogeneity. Here, we analyzed archived human eyeblink habituation data from 157 participants to ask whether the curve shape selected for the group average reflects the curve shapes observed at the individual level. Five candidate functions were fitted separately to each participant and to the corresponding group average. No single function characterized most individuals. More importantly, the model selected for the group average differed from the most frequent individual model in all four groups. When data were pooled across groups, the average favored a dual-process form, a shape that matched the individual plurality in none of them. Simulation analyses showed that averaging heterogeneous individual trajectories can itself produce a group curve that favors a more complex model. Our findings show that group averaged habituation curves should not be treated as direct descriptions of the typical individual trajectory.

Voluntary physical activity is a highly motivated behavior with important implications for physical and mental health, yet the neural and endocrine mechanisms underlying motivation to exercise remain poorly understood. In contrast, motivation for sugar/palatable foods, drugs, and sex has been extensively characterized using operant paradigms. Here, we describe a novel progressive ratio operant task to measure motivation to run, independent of running ability. Using female Long Evans rats, which exhibit robust voluntary running behavior, we validated this paradigm by applying a manipulation well known to enhance the motivation to run: calorie restriction. Calorie-restricted animals exhibited increased operant responding to gain access to a running wheel, thus demonstrating heightened motivation for exercise. More specifically, calorie-restricted rats completed more ratios, reached a higher breakpoint in the progressive ratio task, ran more, and spent more time in the operant chamber. We did not observe any effects of calorie restriction on the estrous cycle or steroids (e.g. corticosterone, testosterone) in the blood or brain. Importantly, our task dissociates the motivational drive for physical activity from the ability to perform the physical activity itself, providing a new paradigm for studying the neural and endocrine mechanisms that regulate exercise motivation.

Social isolation has profound effects on behavior and cognition, but these effects can differ across sex and behavioral domains. We asked how social isolation shapes foraging decisions in male and female rats using a patch-leaving task that combines spatial navigation with sequential stay-or-go choices across varying travel costs and reward depletion rates. All groups scaled patch residence times with travel cost in a manner consistent with the marginal value theorem yet consistently overstayed beyond the optimal leaving time. The magnitude of overstaying was shaped by a strong interaction between sex and social status, with socially isolated females leaving patches closest to optimal and consuming food at the highest rate. Their foraging was accompanied by a coherent spatial and behavioral profile where the socially isolated females spent more time idle, preferentially occupying protected regions near the door and corridor, and avoiding the exposed patch center during both active foraging and periods of idling. These patterns are consistent with a conservative, safety-oriented strategy that simultaneously minimizes exposure and maximizes caloric return. Social isolation does not uniformly impair cognition but can selectively bias female rats toward efficiency-maximizing foraging decisions, consistent with the ecological pressures faced by outcast females in wild rat colonies.

The locus coeruleus noradrenaline (LC-NA) system is a key regulator of arousal, attention, and reward learning. Noradrenaline plays a critical role in impulse control, and recent evidence indicates the importance of noradrenaline signaling in cost-benefit decision making once choice strategies are established. However, whether the LC causally shapes the acquisition of decision strategies, and how this contribution may differ across sexes, remains unclear. We addressed these questions by chemogenetically inhibiting catecholaminergic neurons within the LC of adult tyrosine-hydroxylase Cre (TH::Cre) rats (n=69; 35 females) throughout acquisition of the cued rat gambling task (crGT), a probabilistic decision making paradigm that incorporates salient audiovisual reward-paired cues and simultaneously measures motor impulsivity. LC inhibition accelerated the development of risky choice strategies early in training in both males and females, reflected by impaired adoption of the most advantageous option and increased preference for risky options. Trial-by-trial analyses reveal that LC inhibition promoted switches in choice strategy following safe wins, while reducing switches away from risky options after both wins and losses. LC inhibition therefore seemed to encourage the repetition of actions that resulted in more uncertain outcomes. LC inhibition also selectively enhanced motor impulsivity in females, particularly early in training. These results provide causal evidence that the LC system guides the formation of optimal decisional strategies, while exerting sex-specific control over impulsive action.

Background: Dopamine (DA) is a neurotransmitter critically involved in food-related reinforcement learning. While mesolimbic DA reward-associated signaling in the nucleus accumbens has been widely investigated, far less is known about DA function in the hippocampus (HPC), a brain region traditionally known for its role in episodic and spatial memory processes that has recently been associated with appetite and food intake control. Methods: Here we investigated dorsal HPC DA signaling dynamics in rats using fiber photometry to detect changes in DA binding (via GRAB-DA sensors) before, during, and after a meal consumption in food-restricted rats. Pharmacological studies targeting HPC dopamine 2 receptors (D2R) assessed the functional role of HPC DA signaling in food intake and meal-related memory processes. Results: HPC DA binding was significantly elevated in the post-meal relative to the pre-meal state following standard chow consumption. This effect was replicated after consuming a high fat diet or liquid sucrose, but not a low-calorie sweetener. These post-meal DA signaling elevations are dependent on nutrient consumption, as HPC DA binding levels were unaffected by intraperitoneal administration of glucose or the satiation hormone, cholecystokinin, in otherwise fasted rats. Direct HPC D2R agonists administration reduced food intake, whereas HPC D2R blockade after a meal reduced the latency to the next meal and impaired spatial memory for meal location without affecting spatial memory for object location. Conclusions: Collective results identify HPC DA-D2R signaling as a candidate neurobiological mechanism through which nutrient consumption promotes meal-related episodic memory formation, and by extension, reduces subsequent food intake.

Neural activity in frontal-striatal circuits underlies reinforcement learning. Traditional theories suggest that reinforcement signals, which drive learning, strengthen connections within the basal ganglia. This strengthening is believed to shift information processing from cortical regions to subcortical regions as learning becomes established over time. To examine this hypothesis, we trained macaques to associate multiple sets of images with their values. Selecting different images led to either an increase (+2, +1) or a decrease (-1, -2) in the number of tokens, which subsequently determined the amount of juice reward the macaques received. We simultaneously recorded neuronal activity from orbitofrontal cortex, ventral striatum, amygdala, and dorsomedial thalamic nucleus, analyzing the dynamic changes in these brain regions during both the initial learning and overlearned stages. The results indicated that as learning progressed from the initial stage to the overlearned stage, information processing shifted from the ventral striatum to the orbitofrontal cortex, corresponding to the abstraction from stimulus value to state value. This finding challenges traditional theories and provides a new perspective on the neural circuit mechanisms of learning.

Despite the well-established health benefits of exercise, adherence to physical activity remains low, highlighting the need to identify biological factors that regulate the initiation and persistence of exercise behavior. Here, we tested whether ovarian hormone state at the onset of voluntary wheel running (VWR) influences both immediate and long-term exercise behavior in female rats. Females that initiated VWR during proestrus (Pro) ran farther, spent more time running, and ran at higher speeds on the first day of wheel access than females initiating VWR outside of Pro. Remarkably, initiating VWR during Pro also produced persistent increases in running distance, duration, speed, and escalation across subsequent weeks, despite normal cycling through other estrous phases. In contrast, exogenous estradiol (E2) administered at VWR initiation did not alter day-1 behavior, but increased running distance and duration across subsequent weeks without affecting running speed or escalation. To determine whether dorsal striatal dopamine contributes to these effects, we inhibited the substantia nigra (SN) to dorsolateral striatum (DLS) pathway on the first day of VWR. This manipulation reduced the immediate and long-term effects of initiating VWR during Pro on running duration and distance but not speed or escalation. These findings identify behavioral initiation as a critical window during which hormones and nigrostriatal signaling influence future engagement in physical activity. Furthermore, analysis of individual components of VWR architecture reveals that distinct features of VWR behavior can be dissociated mechanistically and thus could be used to investigate separate motivational processes underlying physical activity.

Although female rats are typically less aggressive than male rats, lactating females will vigorously defend their nests and pups against an intruder. Much attention has been directed at the consequences of this aggression on the intruder and less on the consequences for the mother and her subsequent interactions with her pups. Here, we exposed resident Sprague-Dawley dams to the resident-intruder paradigm twice daily for five consecutive days, beginning when the dam's (RES) pups were 7 days old, to assess social stress effects on maternal behavior and neurobiology. Controls were dams that had time-matched (TMC) separation from their pups but were not exposed to intruders, and naive moms which were never separated nor exposed to an intruder (CTL). We assessed the dam's subsequent behavior and interactions with her pups on Day 1 and Day 5, and Fos expression after Day 5 in select regions of the prefrontal cortex, amygdala, hypothalamus and periaqueductal gray of the midbrain. In separate cohorts, after pups were weaned, the dams underwent restraint stress and plasma corticosterone assayed. PCA analysis of the dam's behaviors identified three components: normal self-focused behaviors; nurturing behaviors and rough non-nurturing behaviors. Relative to CTL, RES dams exhibited more disrupted behaviors towards their pups, including, rough transport, stepping on pups, and flinging/tossing pups around the cage. In contrast, TMC Dams showed some, but fewer changes relative to CTL, suggesting that separation from pups alone does not account for all disrupted behavior in RES dams. The bulk of these behavioral effects occurred in the first 5-10 min after reunion with the pups and were seen on both the first and fifth day of testing. Of the brain regions examined, the prefrontal cortex was activated by both the defeat/intruder stress (RES) and separation stress (TMC), whereas the dorsal PAG was activated specifically by the defeat/intruder stress. The medial and basolateral amygdala exhibited differential neuronal activity between the RES defeat/intruder-exposed dams and the other two groups. The RES moms exhibited an insufficient adrenocortical response to acute restraint stress. The results suggest that amygdala-dPAG activity is important for dissociating disrupted maternal care in RES (due to defense of the nest against an intruder) from simple pup separation, both of which activate the mPFC. The experience of repeatedly defending the nest may induce subsequent disruptions in HPA responses. The amygdala-dPAG pathway may regulate aspects of stress and emotional regulation exhibited by mothers who defend their offspring against intruders.

RationaleAdolescence is a period of heightened vulnerability to nicotine reinforcement. While zebrafish are a valuable model for investigating drug reward, standard conditioned place preference (CPP) assays typically test subjects in isolation. In this highly social species, solitary testing may act as an environmental stressor that confounds behavioral readouts. ObjectivesThis study examined how social context during testing (isolated vs. grouped) affects experimental attrition, behavioral stability, and nicotine CPP expression in late juvenile zebrafish. MethodsZebrafish housed in groups of four were tested either individually (isolated) or in their housing groups (grouped) during daily 20-minute sessions. Following baseline preference assessments, subjects underwent six days of conditioning pairing their initially non-preferred compartment with fish water or nicotine (0.5, 1.6, or 5.0 {micro}mol/L). Place preference, locomotion, and thigmotaxis were assessed on a drug-free test day. ResultsIsolated testing reduced distance traveled, decreased swimming speed, and increased time spent near tank walls, indicating heightened anxiety-like behavior. Experimental attrition was significantly higher in isolated (38.9%) than grouped (2.5%) subjects. Grouped subjects developed significant place preference at 1.6 and 5.0 {micro} mol/L nicotine, whereas preference was not detectable in isolated subjects. ConclusionsSolitary testing acts as a stressor that increases experimental attrition and masks place preference. Conversely, testing in the presence of conspecifics stabilizes behavior and facilitates the detection of nicotine reward in late juvenile zebrafish.

20% of Americans are at risk for environmental circadian rhythm disruptions (CRD) due to shift work, leading to substantial negative health outcomes. However, females are especially affected with greater vulnerability for substance use (SU) and adverse outcomes associated with pregnancy, including for offspring at birth and later in life. In mice, prenatal CRD (pCRD) recapitulates these risks, but it is unknown whether pCRD affects SU in mature offspring. To investigate this, C57BL/6J dams were disrupted by reversing the light/dark cycle during gestation. Following pCRD, reward- (cocaine conditioned place preference, intravenous self-administration) and mood-related behaviors (open field, elevated plus maze, light/dark box, forced swim) were measured in adult offspring. Adult female offspring of dams exposed to CRD developed an anhedonic-like phenotype with decreased food self-administration, cocaine intake and reinforcing properties of cocaine. Opposingly, pCRD male offspring showed a SU-like phenotype with increased cocaine preference, higher order food self-administration and cocaine reinforcement. Interestingly, these divergent behavioral outcomes were not specific to reward. While female pCRD mice showed increased anxiety-like behavior, pCRD males showed decreased anxiety/increased risk-taking behavior, as well as decreased immobility in the forced swim test. Rhythms in corticosterone were also sex-specifically affected by pCRD. These results suggest that pCRD may predispose individuals to distinct psychiatric disorders based on sex with mood disorders developing in females and SU disorders developing in males. By better understanding how disrupted rhythms during pregnancy affect behavior in adulthood, we can develop novel therapeutic approaches for SU and mood disorders in adults.

Stress during critical developmental periods causes lasting neurobiological alterations. Rodent models like neonatal maternal separation (NMS) induce cognitive alterations, particularly spatial memory deficits. Oxytocin (OT) system has been suggested to underlie these consequences, as it is critical for neurodevelopment. This neuropeptide also promotes maternal nurturing, prevents neuroinflammation and displays anxiolytic properties. This study hypothesized that early postnatal OT administration could prevent NMS-induced memory alterations in adult rats. Sprague-Dawley rat pups (both sexes, n=8-12/group) underwent NMS with concomitant intraperitoneal OT injections. At adulthood, novel object recognition and object location tasks were performed. Further investigation was conducted through ex vivo electrophysiological recordings of functional plasticity at Schaffer collateral-CA1 synapses (male, n=7-12/group), alongside RT-qPCR of synaptic, GABAergic, neuro-inflammatory, and oxytocin receptor markers in dorsal CA1 (male, n=4-6/group). NMS induced male-specific spatial memory impairment without affecting recognition memory. Early OT completely prevented spatial memory deficits in NMS males. Electrophysiological recordings revealed that NMS suppressed CA1 long-term potentiation (LTP), and neonatal OT restored it. NMS induced transcript overexpression of neuro-inflammatory markers, GABAergic markers, and synaptic proteins in dorsal CA1. OT treatment normalized or reduced these mRNA expressions, consistent with restoration of CA1 synaptic function. Early postnatal OT prevents NMS-induced spatial memory deficits and hippocampal LTP impairments in male rats, which is associated with normalized or reduced neuro-inflammatory and GABAergic transcript expressions. These findings establish exogenous oxytocin administration during a critical neonatal window as sufficient to prevent male-specific hippocampal dysfunction and cognitive deficits induced by early-life stress, identifying the oxytocinergic system as a promising target for early neuroprotective interventions.

Sleep is thought to promote memory consolidation through the offline reactivation and reorganization of newly acquired information. However, most studies assess memory shortly after sleep, leaving unresolved whether an initial post-learning sleep episode produces enduring modifications that influence how memories respond to later reactivation. Importantly, the absence of behavioral differences after prolonged retention intervals does not necessarily imply that sleep failed to modify the original memory. Instead, sleep-dependent changes may persist in latent forms that are not readily captured by conventional memory assessments. Here, we investigated whether post-learning sleep produces lasting changes in declarative memories that influence their subsequent response to reactivation. In Study 1, participants learned a declarative memory task and were assigned to either a short nap, a wake condition, or an exploratory long-nap condition that included both NREM and REM sleep. Memory was assessed one week later. Despite substantial forgetting across the retention interval, no significant differences in memory performance were observed between groups. In Study 2, participants learned the same task and subsequently underwent either a short nap or wakefulness. Memory was reactivated six days after learning using an incomplete reminder previously shown to induce memory updating in human declarative memory, and memory was tested one day later. Under these conditions, participants who slept after learning showed better memory performance than wake controls. Moreover, sleep physiological measures predicted the magnitude of the post-reactivation memory benefit. These findings suggest that post-learning sleep induces enduring modifications in declarative memories that are not readily detectable through delayed memory testing alone. Instead, these sleep-dependent changes become evident when memories are challenged through subsequent reactivation. Our results indicate that sleep-dependent consolidation influences the future expression of memory, shaping how memories respond to later reactivation experiences and providing new insight into the relationship between consolidation and reconsolidation.

Social hierarchies are a nearly universal feature of animal groups, but whether dominance reflects a single generalized trait or a collection of context-specific competitive abilities remains unclear. Here, we assess social hierarchy in three strains of laboratory mice (BALB/c, C57BL/6, and Shank3B knockouts) of both sexes using three established paradigms: the tube test, the warm spot assay, and the void spot assay. Hierarchies emerged in all strains and both sexes across all three assays, but how animals established rank differed markedly by strain and sex. In the tube test, Shank3b-/- knockout females, but not males, lacked the winner effects seen in wild-type mice, indicating that the ability to build a winning streak depends on social recognition in a sex-specific manner. In the warm spot assay, females formed stronger hierarchies than males, particularly among mice on a C57BL/6 background, with high-ranking females actively displacing others from the warm platform. In the void spot assay, BALB/c mice of both sexes frequently displayed territory-marking behavior, a pattern that was less common in the other strains. Overall, individual rank rarely generalized across domains, despite high trial-to-trial repeatability for individuals within each assay. Together, these findings indicate that mice behave as dominance specialists rather than generalists, with strain- and sex-specific strategies for establishing rank in different competitive contexts, suggesting that distinct neural circuits likely underlie these separable components of competitive ability.

Animals navigating dynamic environments must transition between behavioral states dominated by exploitation of known rewards or exploration of alternatives. Understanding this flexibility requires characterizing not only choices and outcomes but also how reward information updates strategy. Common marmosets offer a unique combination of genetic tractability and frontal cortical complexity, yet scalable behavioral platforms for studying flexible decision making in this species remain limited. Here, we developed an automated home-cage touchscreen platform that enables voluntary behavioral testing across extended periods without water restriction. We trained marmosets on a series of reversal learning tasks of increasing complexity, culminating in a fully uncued probabilistic two-armed bandit task, and applied reinforcement learning models to examine how reward predictability shapes adaptive choice in marmosets. A Q-learning model with choice perseveration captured trial-by-trial dynamics, with choice variability consistent with stochastic sampling from the fitted policy. A trial-level behavioral state classifier identified distinct and persistent modes of exploitation and exploration that differed in their sensitivity to reward history and their transition structure. Comparing behavioral patterns within probabilistic and deterministic reward contingencies revealed that reward predictability reorganized state occupancy and sharpened error correction within predominately exploitative states, with animals responding to single negative outcomes more rapidly under deterministic feedback. These findings establish both a behavioral and computational framework to dissect the computations and circuits underlying adaptive decision making in marmosets.

Social decision-making depends on inferring others hidden preferences from observable behavior. Yet it remains unclear how humans combine choices with process cues such as response times and gaze when learning about others in real-time interaction. Here we combine a novel multi-attribute bargaining task with eye-tracking and show that multiple decision-process cues support preference inference. Across 75 buyer-seller dyads, buyers acceptance rates tracked offer utility, rejection speed reflected decision confidence, and first fixations preferentially targeted the highest-weighted attribute. Sellers adapted subsequent offers using choices, response times, and, when available, gaze cues. A hierarchical inference and choice model suggested that sellers balanced expected utility with expected information gain and updated their beliefs in a Bayesian manner. Although gaze access did not improve overall performance, it changed how sellers used attentional information. These findings shed light on how humans infer others hidden preferences from decision dynamics in real-time social interaction.

Cognitive flexibility is the ability to change the way of responding when the demands of the environment change. This study tested how cognitive flexibility develops across the lifespan. We used a new computerized task that gives a continuous score instead of just right or wrong answers. 221 healthy adults aged 18 to 71 completed the Continuous-score Probabilistic Reversal Learning Test (CPRLT). We calculated mean absolute error and adjusted error for rule-based learning, and fitted a Rescorla-Wagner model to estimate each persons learning rate (alpha) for reward-based learning. All three scores have one breakpoint, performance improved rapidly from childhood to young adulthood, then declined slowly. Rule-based learning peaked around age 20. Reward-based learning peaked earlier, around age 18. This suggests that reward-based learning matures before rule-based learning. The pattern fits with brain development: reward circuits mature earlier, while prefrontal regions for rule-based learning develop later. Our continuous measure captured this difference, which binary tasks would miss.

Children often outperform adults in probabilistic statistical learning tasks, yet the mechanisms underlying this developmental advantage remain poorly understood. Here, we used eye-tracking measures of belief updating to examine how children and adults acquire and update predictions in a probabilistic sequence-learning task. Using the standard (oculomotor) reaction time measure, children showed stronger statistical learning than adults, replicating previous behavioral findings while revealing a more detailed profile of developmental differences in statistical learning. Critically, children updated their predictions more frequently: they were less likely to repeat previous predictions and more likely to shift their expectations in response to new input. Adults, in contrast, showed greater persistence, tending to maintain prior predictions even when those predictions were inconsistent with the underlying statistical structure. Despite these pronounced differences in updating behavior, the processing and use of prediction errors were remarkably similar across age groups. These findings indicate that developmental differences in statistical learning do not primarily arise from how prediction errors are computed, but rather from how prior beliefs and incoming information are weighted during belief updating. Children's enhanced learning may therefore reflect reduced reliance on stable priors and greater sensitivity to current sensory evidence, supporting a more exploratory learning strategy. Adults, by contrast, appear to favor an exploitative strategy that stabilizes existing predictions but reduces flexibility in probabilistic environments. More broadly, the results suggest that developmental changes in statistical learning may reflect age-related differences in how readily learners revise their predictions in response to incoming evidence. By integrating sensitive oculomotor measures with analyses that probe the mechanisms underlying belief updating, the present study provides a more fine-grained account of how predictive learning changes across development and offers a framework for reconciling previously inconsistent developmental findings in statistical learning.

Past sensory experience shapes our perceptual decision-making in the now. Popular models frame perceptual decisions as either attracted towards or repelled away from recent sensory information, but it is unclear when and why these distinct effects emerge. We here ask whether effects turn from attractive to repulsive depending on the level of surprise elicited by the precision-weighted discrepancy between past and present sensory states. This model is based upon the idea that attraction is adaptive for optimizing efficiency and accuracy when discrepancies are small, because they likely reflect sensory noise rather than real change in the environment. In contrast, repulsion may reflect the upweighting of counterfactual evidence when discrepancies are large because they more likely signal the need for model updating. We test this model on a large amount of recently-collated trial-by-trial serial dependence data and consistently find support for it across the dataset, participant, and trial-by-trial level. Specifically, serial dependence effects are attractive at low discrepancies between past and current sensory states but turn repulsive when discrepancies are larger. Higher sensory precision is found to accelerate this flip by reducing the modal discrepancy threshold required to trigger repulsion effects. We discuss how these findings necessitate extending existing theories of serial dependence, and how they may resolve conflicts in the broader predictive processing, learning and perception literatures.