RESEARCH DOSSIER
Pavlovian–Instrumental Transfer
Abstract. Pavlovian–instrumental transfer (PIT) describes how a cue learned to predict an outcome changes performance of an independently acquired action. This fundamental research monograph synthesizes the experimental origins of the paradigm, the distinction between general response invigoration and outcome-specific action selection, the effects of outcome devaluation and motivational state, and competing associative, inferential, and computational explanations. It assesses rodent neural interventions in amygdala and striatal systems alongside human behavioral, imaging, cognitive-load, and methodological evidence. A preregistered meta-analysis, tests of individual-difference reliability, recent working-memory experiments, aversive variants, and translational addiction studies are critically examined. The synthesis identifies what the principal contrasts establish, where mechanisms remain underdetermined, and how future experiments can discriminate causal accounts. This is a critical narrative synthesis of published research, with no original experiments or new clinical recommendations.
Keywords: Pavlovian–instrumental transfer; PIT; Pavlovian conditioning; instrumental learning; outcome-specific transfer; general transfer; incentive motivation; cue reactivity; reward prediction; outcome devaluation; goal-directed action; associative learning; amygdala; nucleus accumbens; ventral striatum; working memory; Bayesian inference; reproducibility; addiction psychology
The Scientific Object: Cues Changing Independently Learned Actions
Pavlovian–instrumental transfer (PIT) is the change in the probability, selection, intensity, or timing of an instrumentally acquired action caused by presenting a stimulus previously associated with an outcome through Pavlovian learning. A canonical experiment establishes the cue–outcome relationship separately from the action–outcome relationship and then measures responding in the cue's presence. This separation is indispensable: if the cue was also an instruction to respond during training, its control over behavior could reflect ordinary discriminative learning. The defining empirical problem is how knowledge acquired from relationships among environmental events influences action policies acquired from their consequences. The synthesis by Cartoni et al. (2016) locates this problem at the intersection of associative learning, motivation, action choice, and neurobiology.
A common observation is that a cue predicting food increases an animal's lever pressing or biases pressing toward the lever earning that very food. Yet these observations answer distinct questions. An increase in total responding could represent generalized motivational activation; redistribution between actions could reflect the retrieval of an outcome representation and its relationship to an action. Humans can additionally infer that cues provide relevant information about currently available outcomes, even when an experimenter intends those cues to be irrelevant. The conceptual distinction matters because a behavioral transfer score identifies modulation by a learned cue, whereas the mechanism responsible for modulation requires further manipulations. The animal and human literature supports multiple routes from predictive information to action Mahlberg et al. (2021).
This monograph reviews the historical construction of the transfer paradigm, outcome-specific and general effects, motivational and representational accounts, neural interventions, human replication, aversive variants, and applications to addictions. Its method is a critical narrative synthesis of identifiable primary experiments, methodological studies, meta-analysis, and contemporary theory; it reports no new data or quantitative synthesis. Evidence is assessed according to whether designs separate learning from performance, identify cue–response confounds, distinguish correlational signatures from causal mechanisms, and reproduce effects across tasks and populations. The resulting account describes PIT as a family of experimentally defined interactions whose functional interpretation depends on the outcome structure, motivation, instructions, and learning history Badioli et al. (2024).
Historical Origins and the Two-Process Problem
Early instrumental-learning experiments documented how stimuli that predicted reinforcement could affect responding under different experimental contingencies. Contemporary reviews trace the first transfer experiments to the 1940s, when behavioral psychologists began to analyze influences of conditioned signals upon independently trained instrumental acts. A decisive theoretical consolidation came from Rescorla and Solomon's account of two-process learning, which proposed that Pavlovian motivational states could affect instrumental performance even though the two learning procedures were established independently Rescorla & Solomon (1967). Its historical importance lies in identifying a cross-system explanatory problem; its motivational terminology should be interpreted in light of subsequent outcome-specific experiments and more detailed representations.
As research acquired more discriminating designs, the same physical outcome could be made accessible through different stimulus and action contingencies. Colwill and Rescorla analyzed associations among discriminative stimuli, reinforcers, and actions; Rescorla later demonstrated that transfer could depend on a shared outcome between a stimulus and an action Colwill & Rescorla (1988) Rescorla (1994). This work made response-specific changes scientifically consequential: predictive cues did more than simply elevate a generalized level of excitation. With multiple outcomes and response options, a test could ask whether a cue strengthened the action that produced its predicted outcome more than an action associated with another outcome.
The historical sequence was not a replacement of one theory by a single successor. Experimental programs progressively sharpened the separation between broad motivational influence and sensory- or identity-specific influence. The development of separate lesion, motivational, and imaging tests created an operational distinction between general and specific transfer, elaborated in the modern review Cartoni et al. (2016). Historical attribution must recognize that a term's original theoretical meaning and a contemporary assay's operational definition may differ. The field's durable achievement is a set of comparison procedures: separate acquisition histories, controlled cue presentation, action competition, and counterbalanced outcomes provide conditions under which a strong claim about cross-learning influence can be tested.
The Triadic Architecture: Stimulus, Response, and Outcome
The minimal vocabulary of PIT comprises a conditioned stimulus S, an instrumental response R, and an outcome O. Pavlovian training establishes an S→O relation, typically with outcome delivery independent of the participant's behavior; instrumental training establishes an R→O contingency, meaning action changes the probability of an outcome. During the critical transfer test, the stimulus is presented while instrumental responding is available, often without delivering the instrumental outcomes. A cue's effect can then be assessed against a neutral-cue baseline or a pre-cue interval. This architecture supports a test of influence from learned outcome prediction to independently trained behavior, as illustrated by the distinct procedures reviewed in Cartoni et al. (2016).
A two-outcome arrangement may train R1→O1 and R2→O2 while separately pairing S1→O1 and S2→O2. Presenting S1 during choice between R1 and R2 provides an outcome-identity contrast. A third cue S3 predicting O3, an outcome never earned through the tested actions but of comparable motivational value, can provide a general-transfer probe. This design matters because S1 can produce both a general increase and a selective bias, while S3 is structured to reveal a broader effect. The use of three outcomes to distinguish these forms was central to the animal dissociations developed by Corbit & Balleine (2005) and Corbit & Balleine (2011).
Scientific interpretation depends on preserving independence across the phases. Action identities, cue modalities, and outcome identities should be counterbalanced; otherwise stimulus salience or a preferred response can masquerade as transfer. The test should measure the same responses under cue-present and appropriate comparison conditions, with the schedule designed to reduce new learning during testing. Extinction transfer tests are common because deliveries during testing could reinforce newly emerging stimulus–response associations. Even in extinction, however, declining responses, temporal expectation, and baseline suppression influence estimates. A valid result requires appropriate control for these alternative sources of change, alongside evidence that both Pavlovian and instrumental contingencies were learned Belanger et al. (2022).
Outcome-Specific and General Transfer
Outcome-specific PIT is demonstrated when a Pavlovian cue preferentially elevates or selects an instrumental response that has produced the same outcome. Its critical observable is a differential action effect, not merely an increase in all responding. The outcome can have multiple properties—sensory features, identity, value, and inferred availability—and experiments must establish which information participants use. The general form is inferred when a reward-predicting stimulus enhances responding across actions without a shared specific outcome, under comparisons that rule out simple response bias. The two forms can coexist within one training design and need not correspond to two mutually exclusive psychological faculties Corbit & Balleine (2011).
General transfer is often described as motivational invigoration because appetitive predictive cues can increase response rate even for instrumental actions earning another appetitive reward. Outcome-specific transfer is commonly described as selective retrieval because a cue may bring to mind the identity of its associated outcome and thereby facilitate the matching action. These descriptions guide measurement but remain theoretical interpretations. General increases may depend on estimated reward availability; selective choices may depend on conscious expectation, acquired association, or both. Interactions between outcome identity and value require independent interventions to discriminate theories. The methodological discussion in Badioli et al. (2024) emphasizes how task variants complicate direct comparison across published human experiments.
Two operational distinctions protect the construct. First, specific transfer requires within-cue comparison between matched and unmatched responses, ideally supplemented by neutral-cue controls. Second, a cue predicting a novel reward cannot automatically be assumed motivationally equivalent to a cue predicting a tested reward: attractiveness, satiety, probability, and experience may vary. Neural experiments showing dissociations across structures provide compelling support for differentiated mechanisms under specified conditions Corbit & Balleine (2005). Yet the number of outcomes, pretraining phase order, species, and outcome modality affect which processes the task recruits. Outcome-specific and general PIT therefore name reproducible families of effects that demand explicit protocols.
Operationalization, Contrasts, and Inferential Validity
A PIT result should be defined before analysis as a contrast among conditions. In an outcome-specific task, a common contrast is responding for the matched outcome during a cue minus responding for a mismatched outcome during that cue, with response preference and baseline differences modeled appropriately. General PIT may be estimated as responding during an outcome-predictive cue minus neutral or pre-cue responding for actions lacking a matching outcome. These contrasts need counterbalancing and, in multi-option choice paradigms, attention to the fact that increasing one response proportion necessarily reduces another. Researchers should report raw response trajectories as well as modeled effects Badioli et al. (2024).
Instrumental acquisition tests whether participants learned how responding changes outcome delivery; Pavlovian acquisition tests whether predictive stimuli acquired significance. In animals, magazine entries, conditioned orienting, or cue-elicited approaches can provide behavioral evidence of Pavlovian acquisition, but are not interchangeable with instrumental modulation. In humans, knowledge questions and outcome-expectancy ratings can establish explicit comprehension, while also creating demand or prompting strategies. Most transfer tests withhold outcomes, which reduces new reinforcement but generates extinction and changes perceived reward availability. The temporal arrangement of cues, baseline intervals, and response opportunities accordingly belongs in the construct's operational definition Belanger et al. (2022).
A statistical association between cue presentation and response frequency establishes a task effect when controlled conditions are adequate; it does not by itself identify a habit, unconscious impulse, unique neurotransmitter, or prediction error. Interpretation is strengthened by selective devaluation, outcome identity changes, contingency reversals, motivational shifts, trial-level modeling, and independently replicated manipulations. Human task versions also differ in whether actions are rewarded with points, money, food, or hypothetical currency, and whether cue meanings are implicit or explicitly instructed. These differences may explain heterogeneous effect estimates without requiring that one laboratory has discovered an entirely different phenomenon Badioli et al. (2024).
Incentive Motivation and Reward-Predictive Cues
The incentive interpretation of transfer begins with a conditional fact: a stimulus that forecasts an appetitive event can change how vigorously an individual performs an independently acquired response. A cue may acquire the capacity to activate appetitive approach, increase effort, or bias selection among possible actions. The motivational effect need not be a direct expression of cue–response training, because the instrumental response may have been absent during Pavlovian acquisition. Rescorla and Solomon's two-process formulation Rescorla & Solomon (1967) supplied a historically influential explanation of this cross-system interaction. Later behavioral experiments established that different cue effects require different response comparisons and can depend on the architecture of the outcome relationships Cartoni et al. (2016).
General motivational invigoration is often interpreted through changes in arousal, incentive salience, or the momentary willingness to act for a reward. Operationally, however, an increased response rate could reflect a change in action initiation, persistence, estimated opportunity, or the vigor with which a selected action is executed. These potential mediators are experimentally separable. Researchers can compare choice proportions and absolute rates, introduce response costs, manipulate deprivation, or compare rewarded and unrewarded test periods. A cue may boost response vigor while leaving the relative allocation of responses largely unchanged; conversely, a cue may redistribute choices without changing total output Corbit & Balleine (2003).
The word 'motivation' must therefore earn an explanatory role through independent predictions. If stimulus-driven response energization varies with hunger or thirst, the experiment provides information about motivational modulation; it does not resolve whether a cue conveys explicit knowledge of reward availability. An account that simply renames the measured response increase as 'motivation' does not specify a mechanism. A more informative hypothesis predicts that motivational state changes should affect the cue–response interaction under controlled outcome identities, while separate tests identify cue content and action contingencies Corbit et al. (2007). Motivational interpretation remains strongest when state manipulations and behavior jointly support it.
Outcome Identity and Selective Action Retrieval
Outcome-specific transfer is anchored in a finer informational claim: a learned cue can retrieve a representation of a particular outcome and thereby favor the action associated with producing that outcome. In a two-action experiment, a cue for a food pellet might selectively increase the response previously earning the pellet rather than the response producing sucrose. The crucial contrast holds broad appetitive value relatively constant while changing outcome identity. Experiments demonstrating selective transfer support the involvement of sensory-specific or identity-specific information in behavioral control Rescorla (1994). The term 'representation' here refers to information required to explain the experimentally observed discrimination; it does not specify a conscious image or particular neural code.
Selective retrieval can operate through several pathways. A cue might activate a representation of its outcome, which then activates a response historically associated with the outcome; alternatively, participants might infer that the cue reveals which reward is currently available and select an action they believe obtains it. Both pathways produce matching-response preferences. Human studies have provided evidence consistent with active use of expected outcomes and task reasoning, and recent cognitive-load manipulations sharpen these interpretations Mahlberg et al. (2021) Huseynzade et al. (2026). Distinguishing pathways requires changing outcome knowledge or instruction while measuring whether residual selective transfer persists.
Specific transfer also depends on the granularity of outcomes. Two rewards described as 'points' may be experienced as one common abstract incentive unless the learning environment assigns distinct categories or uses distinguishable sensory properties. A matched-action advantage then may reflect verbal labels, symbol mapping, or learned outcome identity. Outcome identity should be validated directly, and the two actions should be equated on baseline strength wherever possible. Colwill and Rescorla's instrumental stimulus–reinforcer work Colwill & Rescorla (1988) provides a broader associative context for understanding why relationships to outcomes have explanatory value beyond a simple strength-of-response account.
Outcome Devaluation, Revaluation, and the Value–Identity Distinction
Outcome devaluation is a discriminating intervention because it changes what an outcome is currently worth while preserving, at least initially, the previously established history of action and cue association. Sensory-specific satiety can reduce the value of a particular food; conditioned taste aversion can create a different kind of motivational change. In instrumental tasks, devaluation-sensitive choices provide evidence that actions are controlled by their anticipated consequences. Transfer experiments ask a more specific question: will a cue continue to select the action related to an outcome whose value has changed? The answers depend on which PIT form is measured and on the devaluation procedure Holland (2004).
A durable response bias toward the action matching a devalued outcome would establish that outcome identity can continue to guide choice even when outcome value changes. It would not prove value-independent behavior across all test conditions. Devaluation might alter baseline action probability, retrieval of sensory attributes, and willingness to contact the outcome. In contrast, a decline in cue-induced response invigorating after satiety could imply stronger motivational sensitivity. The study by Holland (2004) compared transfer and reinforcer devaluation and illustrated why apparently similar effects cannot be assumed to depend on one representational dimension.
Investigators must also distinguish outcome devaluation from changes in the predictive validity of a cue. Changing the reward's desirability is not the same operation as informing a participant that the cue–outcome relationship is no longer reliable. An updated belief about contingency can suppress deliberate cue use while leaving learned associative responses partly intact. Tests combining independent value changes, contingency changes, and instructions are especially important for resolving the relation between incentive and inference. Behavioral dissociations should be reported at the level of the particular training and testing regime rather than generalized immediately to claims about conscious control or automatism Huseynzade et al. (2026).
Contingency, Context, Extinction, and Transfer Expression
A transfer test takes place in a context whose contingencies may differ from either training phase. During Pavlovian acquisition, the cue predicts an outcome independent of the focal action; during instrumental acquisition, an action earns an outcome; during test, the action often yields nothing. An animal or human may therefore discover that the task has changed, even as experimenters treat the test as a passive readout of previously acquired learning. Extinction of instrumental responding can reduce the window in which PIT is measurable. Transfer estimates are better understood as time-dependent expressions of acquired relationships under a changed contingency structure Cartoni et al. (2016).
Context can influence the accessibility and significance of cue–outcome and action–outcome memories. A predictive stimulus may carry more information in an ambiguous environment, while a distinctive test context may signal that previous reward rules are suspended. Cue-on versus cue-off contrasts cannot automatically remove such context dependence when cue blocks differ systematically in temporal position. Randomized and counterbalanced cue scheduling, analysis of early versus late trials, and clear reporting of extinction dynamics can help establish that the effect is tied to cue presentation rather than task drift Belanger et al. (2022).
An important experimental issue is whether transfer requires prior cue and instrumental responses to have been learned in the same motivational state or context. Some manipulations preserve outcome identity while shifting deprivation or pharmacological state. Corbit, Janak, and Balleine Corbit et al. (2007) showed that motivational shifts and interventions in reward-related circuitry can distinguish transfer processes under specific laboratory conditions. Such interactions help delineate conditions under which a cue's learned value contributes to response performance, while also warning against estimating a stable personal 'cue-reactivity trait' from one short and context-dependent task.
Appetitive and Aversive Transfer as Experimental Families
Most canonical PIT examples use appetitive outcomes such as food or monetary rewards. In these tasks, researchers ask how appetitive predictions invigorate reward seeking or selectively redirect reward-related choice. Aversive variants examine cues predicting threat, shock, loss, or unpleasant consequences and their influence on approach, avoidance, or defensive action. These paradigms can share the architecture of independently trained cue–outcome and action–outcome relations, yet the measured responses have different functional meanings. An increase in avoidance may reflect defensive preparation, learned safety seeking, or the expectation that an action prevents harm Eder & Mitschke (2025).
The distinctions between appetitive and aversive settings matter for both theory and measurement. Threat-related cues may globally suppress reward-oriented responding even while facilitating an avoidance action; a single overall response-rate score can conceal these concurrent effects. An aversive cue could selectively prime an action that had reduced a matching aversive outcome, although outcome identity is often harder to equate when threats differ in intensity or timing. Eder and Mitschke Eder & Mitschke (2025) investigated cue control over fight-or-flight decisions, providing an example of an aversive-oriented action-selection setting rather than a straightforward replication of a food-reward task.
Stress and affective state may further alter transfer, but results depend on how stress is induced and how subsequent action control is assessed. Tang, Price, and Bonardi Tang et al. (2026) studied stress induction in a quasi-avoidance transfer paradigm with online and in-person procedures. A single stress manipulation cannot establish a general causal law for every form of aversive PIT. A more rigorous comparative program defines outcome valence, response type, controllability, subjective threat, and baseline tendencies before comparing mechanisms. Appetitive and aversive transfer belong in one scientific discussion because both address cue influence across learning systems, while the concrete outcome–action structure remains decisive.
Amygdala Contributions and Causal Dissociations
The amygdala has supplied some of the clearest causal evidence that generalized and outcome-specific effects involve distinguishable circuitry. Corbit and Balleine Corbit & Balleine (2005) compared lesions of the basolateral complex and central nucleus in rats trained with distinct instrumental outcomes and Pavlovian cues. The basolateral lesions disrupted selective sensitivity to outcome devaluation and specific transfer, while lesions of the central nucleus disrupted general transfer in the relevant test. This is a double dissociation under defined conditions, because the two interventions produced different impairments in two behavioral operations. It supports a differentiation of functional contributions in appetitive PIT rather than establishing that each nucleus implements only one psychological computation.
The interpretation of an amygdala lesion requires understanding its timing and the sequence of learning operations. A deficit at transfer test can arise from impaired acquisition, retrieval, expression, altered orienting, or motivational responsiveness. Corbit and Balleine reported broadly comparable acquisition under their conditions, strengthening their inference about transfer expression Corbit & Balleine (2005). Holland and Gallagher Holland & Gallagher (2003) also compared basolateral and central effects across conditioned stimulus-potentiated feeding and PIT, showing the relevance of task-specific contrasts. Causal intervention is stronger than correlation, yet lesions can disrupt several processes and networks, so selective behavioral controls remain essential.
Research should not map the human amygdala directly onto the rodent experiment without examining species, measurement, and outcome type. Human imaging resolves regional blood-oxygen responses rather than single-cell computations, and regional labels do not guarantee homologous functional tasks. A synthesis that pairs controlled rodent interventions with human imaging therefore identifies convergent patterns and unresolved translation questions. Prévost and colleagues Prévost et al. (2012) used high-resolution fMRI to assess neural correlates of specific and general transfer within human amygdalar subregions, creating a bridge between these literatures but not substituting for causal manipulation in people.
Nucleus Accumbens Core and Shell
The ventral striatum participates in the convergence of reward prediction and action regulation, making it an important candidate interface between Pavlovian and instrumental learning. A prominent rodent experiment distinguished the nucleus accumbens core and shell by examining the general and outcome-specific forms of PIT with controlled outcome structures Corbit & Balleine (2011). The study reported differential dependence of these forms on core and shell, respectively, strengthening the conclusion that general response invigoration and selective outcome retrieval can be experimentally and neurally dissociated. Its significance comes from the alignment between the behavioral contrast and the anatomical manipulation.
The term 'interface' should be treated as a systems hypothesis rather than a single anatomical relay. The accumbens receives multiple inputs, contributes to incentive processing and movement, and functions within loops linking limbic, cortical, and motor systems. An intervention altering the response during a transfer test might affect selection, energy, state representation, or motor output. Measuring baseline instrumental capacity, Pavlovian acquisition, and outcome sensitivity is necessary to interpret a transfer-specific impairment. A scientifically robust neural account states what was manipulated, what remained intact, and which transfer contrast was selectively changed Corbit & Balleine (2011).
Human imaging studies have frequently implicated ventral striatal activity during predictive-cue effects, but blood-oxygen-level-dependent responses represent population-level vascular signals and can covary with many aspects of motivation and task structure. Talmi and colleagues Talmi et al. (2008) examined human transfer with functional imaging, contributing evidence that conditioned incentive cues recruit reward-related systems while influencing action performance. When a neural signal predicts a larger PIT score, the result supports a relationship between task-evoked brain activity and behavior. Intervention and longitudinal replication remain necessary before an imaging correlation can be treated as a necessary causal mechanism.
Dopamine, Ventral Tegmental Area, and Incentive Processes
Dopamine is involved in learning from reward prediction and regulating behavioral vigor, but its role in PIT must be specified by which transfer component and experimental condition is being discussed. Corbit, Janak, and Balleine Corbit et al. (2007) investigated effects of motivational-state shifts and inactivation of the ventral tegmental area on general and specific transfer in rats. Such experiments directly probe a contributor to reward-related circuits while holding parts of the associative training history constant. Their results are relevant because they address whether the motivational component of cue influence depends upon neural systems associated with incentive-related processing.
An apparent dopamine–PIT association is underdetermined if instrumental responding, learning, movement initiation, and cue prediction are all measured by a single rate variable. Dopaminergic manipulations may change how quickly an organism responds even in the absence of the critical conditioned cue; this is why baseline responding and matched control intervals matter. Computational accounts often link dopamine to prediction error, but prediction error during learning and cue-dependent action invigoration at test are different operations. A neurotransmitter can influence both without those operations being identical Corbit et al. (2007) Cartoni et al. (2016).
A translational question is whether dopamine-sensitive cue effects in rodents explain cue-triggered decisions in everyday human settings. This requires convergence between behavioral task designs, neural measurements, and causal pharmacological or stimulation studies that isolate a specified mechanism. Clinical drugs and disease states produce broad changes, so effects on PIT cannot automatically be interpreted as pure dopamine measures. The appropriate conclusion from the literature is functional participation within circuit-level interactions, bounded by evidence on tasks and interventions. A single dopamine variable is insufficient as a universal explanation for general and outcome-specific effects.
Formal Associative and Reinforcement-Learning Models
Computational descriptions make the explanatory alternatives explicit. A traditional associative model assigns strengths to S–O and R–O relationships and predicts how activation of a shared outcome can facilitate a response. A reinforcement-learning model instead may represent learned action values, cue-dependent state values, or a change in policy temperature or response vigor. Either approach can generate a positive cue-on-minus-cue-off contrast under some assumptions. Model adequacy is therefore judged by out-of-sample predictions for outcome-specific choice, devaluation, changes in available actions, and transfer after contingency reversal, rather than by a good fit to one response curve Cartoni et al. (2016).
General PIT can be formalized as a cue-dependent increase in the effective value or vigor of multiple actions, whereas specific PIT may require an identity-indexed mapping between cue and outcome and between outcome and action. But such decomposition is only one family of possible models. A person may use a symbolic rule about which reward is available to choose a response; mathematically this can mimic the output of automatic associative retrieval. The challenge of identifiability is acute when measures are limited to one final choice. Investigators should vary incentives, instructions, and memory demands to generate predicted divergences Mahlberg et al. (2021).
Computational modeling can also reveal whether observed response changes are due to the probability of selecting an action or the speed of executing it. Drift-diffusion, value-learning, and hierarchical models can be useful when their parameters have explicit measurement interpretations and are identifiable with task data. Model comparison should report parameter recovery, predictive checks, sensitivity to priors, and test-retest stability. Such analyses are valuable tools for resolving alternative mechanisms; they cannot, without further intervention, establish that the fitted algorithm is the unique process implemented by the nervous system Degni et al. (2022).
Bayesian Inference, Beliefs, and Latent Task Structure
An inferential approach treats predictive cues as information about the current environment. A person who has learned that a color forecasts one reward may interpret that color as evidence that the reward is available during the instrumental test, even if actual delivery is suspended. The effect on action selection may then follow rational inference from a mistaken model of the test structure, not simple automatic activation. Degni and colleagues Degni et al. (2022) presented evidence and Bayesian interpretations relevant to general human transfer, providing a formal alternative to explanations that depend exclusively on reflexive motivation.
Beliefs can differ between participants and change within a task. Some people explicitly report using cue information strategically, while others try to disregard it and nevertheless show residual transfer. A theory that treats all selective responding as conscious strategy overlooks such residual effects; a theory that assumes conscious task inference never matters overlooks instruction and cognitive-load findings. The informative problem is how learned associations, expected reward availability, and task instructions jointly determine behavior. Mahlberg and colleagues Mahlberg et al. (2021) argued for a goal-directed interpretation of aspects of human appetitive transfer, and contemporary experiments provide direct tests of that proposal.
Investigators should measure belief states without making their measurement the cause of the effect. Asking repeated expectancy questions can focus attention on cue–outcome relations, change their salience, or supply an unintended instruction. Separate groups with and without explicit probes, trial-level confidence measures, and pre-registered mediation hypotheses can help. The Bayesian approach is strongest when it specifies what prior belief is updated, which environmental observations are available, and how posterior inferences translate into instrumental action. When these terms are free to vary arbitrarily, the model explains after the fact rather than predicts under controlled manipulations Degni et al. (2022).
Human PIT: Experimental Translation and Initial Demonstrations
Human PIT experiments translate the logic of independent Pavlovian and instrumental training into laboratory decisions about money, points, food, or other incentives. The design must retain the separation between cue learning and action learning while accounting for instructions, verbal comprehension, and differences in baseline response speed. Talmi, Seymour, Dayan, and Dolan Talmi et al. (2008) provided an influential demonstration of human Pavlovian–instrumental transfer, examining how reward-related cues altered instrumental performance and how these effects related to neural activity. Its contribution was to show that experimental transfer procedures can be adapted to the human laboratory while retaining the core inferential structure.
The human participant also brings a task model. Instructions may explicitly explain the Pavlovian cue–outcome pairing or the action–outcome contingency, and the participant can treat transfer as a puzzle about reward availability. This capacity complicates simple generalization from nonverbal conditioning paradigms. Inferences about an experimenter's intentions can enhance or suppress specific transfer, especially where correct responding appears to depend on reading the cues. The contrast between automatic and goal-directed accounts has consequently become a productive human research program Seabrooke et al. (2017) Mahlberg et al. (2021).
A defensible human experiment should measure acquisition, describe the degree of instructed versus experiential learning, and identify whether the transfer test is performed under extinction. It should report both successful training exclusions and any selection bias those exclusions create. Participants who fail a learning criterion may be informative about boundary conditions rather than merely unsuitable observations. Generalizing laboratory findings to shopping, gambling, or substance-use decisions requires additional evidence because real-world environments often combine cue exposure with actual action reinforcement, social messaging, and changing incentives Belanger et al. (2022).
Goal-Directed Choice and Deliberate Cue Use
Human outcome-specific PIT poses a problem for simple labels such as 'automatic incentive response.' If an instructed participant expects a particular reward when a stimulus appears, the participant may choose the action previously associated with that reward as a deliberate strategy. Seabrooke and colleagues Seabrooke et al. (2017) examined evidence of goal-directed processes in human PIT, and their subsequent work Seabrooke et al. (2019) continued the analysis of goal-directed control. These studies create testable distinctions among the learned informational content of cues, current beliefs about the task, and response selection during transfer.
Goal-directed inference and learned associative activation need not be mutually exclusive at the level of individual behavior. A participant may first acquire a cue–outcome relationship through experience, later form a belief that the cue signals current availability, and then choose an action based on a remembered outcome relation. A separate component of the response may be expressed even when the participant regards the cue as irrelevant. Studies that include postexperimental awareness questions require careful interpretation because reported strategies are retrospective and may be incomplete. Behavioral effects under altered instructions provide an independent experimental test of the contribution of deliberate cue use Mahlberg et al. (2021).
Another source of ambiguity is that a participant's 'goal' may mean earning a target reward, complying with experimenter instructions, or performing accurately in a task described as a game. Each goal makes different predictions when outcomes are absent. Consequently, evidence for goal-directed transfer should specify the information available to participants and the outcome they believed their actions could obtain. The goal-directed account gains strength when changes in perceived cue relevance systematically change choice under preserved Pavlovian associations, and when residual effects are measured rather than assumed away Huseynzade et al. (2026).
Working Memory, Directed Forgetting, and Recent Human Evidence
An October 2026 experimental article by Huseynzade, Dignath, and Eder Huseynzade et al. (2026) directly addressed whether outcome-selective transfer in humans depends upon cognitive processing of cue information. Across three experiments using a monetary stock-market scenario, the investigators trained separate Pavlovian cue–outcome and instrumental action–outcome relationships. They then manipulated the credibility or accessibility of the Pavlovian information at transfer. This work is particularly informative because it manipulated possible cognitive mediators rather than deriving an unconscious-process interpretation from the mere presence of a cue-induced effect.
The first experiment established selective transfer and associated larger effects with strategic cue use. In a second experiment, participants were instructed to forget previously learned Pavlovian relationships after a purported software problem, which reduced selective transfer. In a third experiment, concurrent working-memory load also reduced the effect. The study reported a residual selective effect under combined manipulations, with no further attenuation from adding load to directed forgetting. The authors interpret the overall pattern as evidence that reward expectancies and beliefs concerning cue relevance influence human selective PIT Huseynzade et al. (2026).
These results narrow the space of plausible theories while retaining several unresolved questions. Directed-forgetting instructions alter perceived validity as well as memory accessibility; working-memory load may alter attention, response selection, or the maintenance of task rules. Residual transfer is consistent with a contribution beyond deliberate strategic use but does not uniquely identify an associative pathway. Generalization requires independent studies using distinct outcomes, instructions, and populations. The experiments advance an explicit mechanistic test by showing that changing cognitive task conditions changes the magnitude of cue-selective action bias, while leaving room for multiple contributing processes Huseynzade et al. (2026).
Replication, Meta-Analysis, and Between-Study Heterogeneity
Individual demonstrations are important for establishing that a transfer pattern can occur; general conclusions depend on independent replication and transparent synthesis. Badioli and colleagues Badioli et al. (2024) reported a preregistered meta-analysis on Pavlovian-to-instrumental transfer, explicitly examining methodological heterogeneity in human research. Its significance is methodological as well as substantive: experiments with different cues, outcomes, choice structures, test instructions, and scoring procedures cannot simply be treated as identical measurements of one constant effect. Preregistered synthesis can evaluate classes of effects while also identifying where design differences limit cross-study comparisons.
A meta-analysis can provide a weighted summary only under the assumptions of its inclusion rules and statistical model. Publication bias, selective reporting, nonindependence among effects from the same experiment, and variability in task definitions can alter conclusions. A pooled effect for appetitive transfer does not establish that aversive transfer, substance-specific cue reactivity, and human outcome-specific choice all share one common psychological mechanism. Researchers should therefore report the distribution of observed effects, prespecified moderators, heterogeneity, uncertainty, and sensitivity to inclusion criteria rather than using a single average as proof of universal stimulus control Badioli et al. (2024).
Experimental standardization also requires distinguishing a failed acquisition phase from a genuine absence of transfer. An unlearned cue has limited capacity to alter actions for its predicted outcome, whereas a learned cue may have no effect because the test context changes participants' expectations. Exclusion based on successful acquisition can improve internal validity but restrict generalizability and introduce selection effects. Belanger and colleagues Belanger et al. (2022) developed task variants that explicitly evaluate specific and general transfer in humans. The most useful replication strategy combines conceptually comparable tests with careful documentation of differences that may moderate expression.
Psychometrics: Internal Consistency and Longitudinal Stability
A robust average PIT effect does not guarantee that a task measures stable individual differences. Group-level experimental effects and between-person psychometric reliability answer distinct questions. A task may reliably show an average increase in responding during predictive cues while producing unstable rankings of individuals over time. Conversely, a consistent difference among participants could be driven by general motor speed, attention, or questionnaire familiarity instead of PIT-specific susceptibility. Separating these claims is especially important when the paradigm is proposed as a biomarker of substance-use vulnerability or a predictor of clinical outcomes Belanger et al. (2025).
Belanger and colleagues Belanger et al. (2025) evaluated behavioral and neural reliability in a PIT task. The published study reported high split-half reliability of behavioral interference and motivational contrasts, with coefficients spanning approximately .92 to .98 across its tabulated samples and time points. However, intraclass correlations for repeated measurements were around .53 in one clinical sample and approximately .26–.29 in a developmental sample measured across a longer interval; the authors also reported weak longitudinal reliability of the neural contrasts. The longer interval may include meaningful change in the underlying person as well as measurement instability, so its coefficient should not be interpreted as a pure repeated-session error estimate.
The psychometric lesson concerns unit of inference. A well-designed within-person cue manipulation can be sensitive to a reliable laboratory effect without supporting individual clinical classification from one session. Test–retest analyses should define intervals, populations, exclusion rules, motor and motivational covariates, and the precise PIT metric. Neuroimaging reliability must be assessed separately from behavioral reliability, and task contrasts with near-zero stable between-person variance may fail as trait markers even if mean activation exists. The path from PIT to a clinical instrument requires longitudinal calibration and predictive validation in independent samples Belanger et al. (2025).
Substance Use, Cue Reactivity, and Translational Boundaries
Substance-related cues can acquire predictive significance through repeated experience, making PIT an attractive experimental model for understanding when environmental signals change reward-seeking actions. The clinical question is narrower than the claim that all addiction is caused by Pavlovian cue control: do experimentally measured changes in action under reward-predictive cues predict meaningful substance-related behavior, over and above learning, motivation, stress, and other contributors? Garbusow and colleagues Garbusow et al. (2016) investigated PIT-related effects in the nucleus accumbens and their relationship to relapse in alcohol dependence. The study contributes longitudinally relevant evidence connecting a laboratory cue-action process with a clinically important outcome.
Translation from a food or monetary PIT experiment to substance use requires careful control of outcome specificity. A cue associated with alcohol may predict opportunity, relief from negative affect, social setting, or learned action sequences rather than one discrete sensory outcome. A person can respond to such a cue with deliberate drug seeking, automatic motor preparation, craving, or avoidance; these are related but operationally different processes. Hogarth Hogarth (2020) critically argued that goal-directed choice under negative affect deserves greater attention in models of addictive behavior. Taken together, these research lines caution against using a positive PIT contrast as a direct diagnostic measurement of compulsive habit.
Clinical correlations are vulnerable to factors that influence both PIT performance and outcomes: withdrawal state, medication, baseline attention, response control, learning history, and socioeconomic context. Predictive claims require independent longitudinal cohorts and a validated model with out-of-sample performance, while mechanistic causal claims require an intervention that specifically changes the relevant cue influence and then alters a clinical outcome. A laboratory association between neural activity and relapse does not by itself justify treatment selection. The research value of PIT lies in furnishing an experimentally manipulable process that can be linked to clinical questions without assuming that the link has a simple one-to-one mapping Garbusow et al. (2016) Belanger et al. (2025).
Individual Differences, Impulsivity, and Affective Context
Some studies ask whether personality dimensions, impulsivity, or affective traits modify the influence of conditioned cues on instrumentally learned actions. A plausible hypothesis is that people high in affect-linked impulsivity will express larger or more persistent cue effects. Yet the relevant construct must be measured independently: self-reported urgency, delay discounting, motor impulsivity, and substance-related impairment are not interchangeable. González and colleagues González et al. (2025) investigated relationships among affect-driven impulsivity, extinction of outcome-specific PIT, reward-related attentional capture, and outcome-devaluation sensitivity. The research places transfer within a broader set of reward-control processes while leaving questions about predictive specificity.
Associations among individual-difference measures do not establish that they share one latent mechanism. For instance, an association between transfer and attentional capture could arise from common sensitivity to reward cues, general task engagement, or correlated measurement errors. Stronger conclusions require multivariate modeling, reliability estimates, sufficient sample size, and replication with different tasks. A person can show robust outcome-specific transfer and remain sensitive to outcome devaluation, so transfer is not synonymous with behavioral inflexibility or a habitual personality. Findings should be interpreted in terms of the exact manipulation and score González et al. (2025).
Affective state can alter attention, perceived reward availability, and the relative value of actions, all of which may change transfer expression. Stress might increase defensive cue control in one task and suppress appetitive response rate in another. Experiments should distinguish state effects from stable traits by assessing participants repeatedly or randomizing affective manipulations. A useful hypothesis predicts which cue–outcome relations and response classes will be affected, rather than expecting a uniform increase in every transfer measure. The experimental comparison of stress induction procedures in Tang et al. (2026) underscores why state manipulation and task design must be considered jointly.
Human Neural Measurements and Motor Preparation
Functional neuroimaging and electrophysiological research extend PIT investigation from behavior to systems involved in evaluating cues, preparing responses, and selecting actions. Talmi and colleagues Talmi et al. (2008) connected human transfer to neural reward-processing activity, while Prévost and colleagues Prévost et al. (2012) investigated human amygdalar subregions during specific and general transfer contrasts. Such findings can test whether a proposed anatomical network has a signal related to the behavioral effect. They are most informative when neural contrasts are independently defined and movement, difficulty, and cue salience are controlled.
A 2026 human neuroimaging experiment by Huang and colleagues Huang et al. (2026) examined motor activation in cue-guided behavior during PIT. The work places response preparation and motor-system recruitment alongside reward-related explanation. This expansion matters because an increase in action rate may be partly implemented through preparatory processes rather than a change in consciously expressed preference. A neural association with response preparation is not sufficient to infer that the same process generates every cue-related choice effect; independent tasks should dissociate speed, action identity, and expected outcome value.
Neural studies also face the challenge of reliability. Group-average condition differences can be experimentally repeatable while person-level activation contrasts show poor temporal stability, limiting their use as predictive markers. Belanger and colleagues Belanger et al. (2025) reported weak longitudinal reliability for their neural PIT measures despite stronger within-session behavioral consistency. Responsible interpretation distinguishes anatomical localization, computational mediation, and the psychometric stability of an individual score. Strong cross-species inference calls for convergence between animal interventions, human behavior, and measurements with clearly characterized uncertainty.
Designing an Informative and Reproducible PIT Experiment
A rigorous experimental design starts with the scientific contrast. If the objective is outcome-specific transfer, two identifiable outcomes and independently learned matching actions provide a basis for selectivity; if the objective is general invigoration, a suitably controlled nonmatching outcome or neutral cue can make the motivational effect interpretable. Before collecting data, researchers should prespecify stimulus assignment, acquisition criteria, cue-order counterbalancing, response-window definition, primary dependent variable, and treatment of zero responses. The designs reviewed in Belanger et al. (2022) and the methodological heterogeneity synthesized by Badioli et al. (2024) provide concrete reasons to make such choices explicit.
Procedural transparency matters because transfer scores are sensitive to baseline response rate, the number of available actions, cue duration, extinction, and the interpretation participants assign to rewards. A transparent report includes phase diagrams, training trial counts, instructions, exclusion criteria, acquisition analyses, exact comparison conditions, response distributions, and code or data when ethically permissible. Analyses should distinguish within-person cue effects from between-person trait estimates, correct for repeated observations, and present uncertainty intervals. Researchers can supplement standard contrasts with hierarchical models, but the model's assumptions and parameter-identifiability checks should be documented Badioli et al. (2024).
Convergent causal tests should target competing theories. Outcome devaluation tests whether current value affects transfer; changing outcome identity or availability tests the information used in action selection; directed forgetting and working-memory load test cognitive dependence; targeted neural manipulation tests circuit contributions. None of these alone answers every mechanistic question. Their joint value lies in patterns of predicted dissociation under well-defined conditions. Preregistration reduces undisclosed analytic flexibility, while replication across species, laboratories, and task formats establishes which findings survive changes in materials and implementation Huseynzade et al. (2026) Corbit & Balleine (2005).
Open Questions and the Artificial Era Research Horizon
The current literature leaves unresolved the relative contributions of automatic associative retrieval, deliberate inference, motivational state, and motor preparation to specific transfer outcomes. One research priority is factorial manipulation of cue relevance, working-memory demands, outcome valuation, and response cost in the same carefully powered design. Another is longitudinal measurement that distinguishes real psychological change from unstable scores. A third is computational adjudication: models should generate different quantitative predictions for new conditions rather than post hoc explanations of an already observed transfer effect. The 2024 meta-analysis Badioli et al. (2024) and the 2026 cognitive-intervention experiments Huseynzade et al. (2026) provide complementary starting points.
Generalization outside the laboratory is a separate problem. Everyday cues may predict incentives with changing reliability, and actions can be reinforced while cues are present, breaking the independent-training assumption that makes canonical PIT interpretable. In consumer environments, reward symbols, advertising, and digital feedback can be analyzed experimentally through similar cue–outcome and action–outcome arrangements, but field experiments must establish actual contingencies and measure behavior rather than assuming a laboratory effect directly predicts purchases. Substance-use, avoidance, and digital-behavior studies should distinguish their intervention targets and avoid treating a shared label as evidence of one universal mechanism Cartoni et al. (2016) Hogarth (2020).
In the Artificial Era, PIT also supplies a methodological comparison for studying cue-conditioned artificial decision policies and human interaction with artificial agents. Artificial reinforcement-learning systems can be trained under segregated predictive and action contingencies, then tested for cue-dependent changes in action. Such an artificial-system response is a functional analogue defined by the same observable experimental arrangement; claims about human subjective motivation, neural substrates, or experiential states require separate evidence. For human–artificial interaction, predictive cues emitted by artificial interfaces may shape human reward expectations and decisions. The appropriate research program preregisters cross-system behavioral contrasts and distinguishes engineered policy responses from empirically documented human psychological processes Cartoni et al. (2016).
Conclusion
Pavlovian–instrumental transfer identifies a causal interaction between two independently established kinds of learning: predictive relations between stimuli and outcomes, and contingency relations between actions and outcomes. The phenomenon is scientifically informative because a cue can alter an instrumental action even when that action was acquired elsewhere. Outcome-specific transfer shows selective control organized by outcome identity, whereas general transfer tests broader response invigoration. Both must be measured with designs that account for cue history, action strength, outcome identity, extinction, and concurrent motivation. The historical two-process problem remains recognizable within current behavioral and neural methodology Rescorla & Solomon (1967) Cartoni et al. (2016).
Experimental dissociations support differentiated circuit contributions in appetitive animal transfer, notably in amygdala and ventral-striatal systems Corbit & Balleine (2005) Corbit & Balleine (2011). Human evidence establishes transfer under several task implementations and shows that deliberate beliefs, working-memory demands, and instructions can modulate outcome-selective action bias Mahlberg et al. (2021) Huseynzade et al. (2026). The research also identifies limits: a positive transfer score does not uniquely reveal automatism, habit, consciousness, or a single neurotransmitter mechanism. Neural associations and clinical correlates require designs suited to causal and predictive inference. The distinction between robust group-level effects and stable person-level measurement remains indispensable Belanger et al. (2025).
The central synthesis is a testable framework for interpreting cross-learning influence: specify S–O and R–O histories, separate selective choice from general activation, manipulate outcome identity and value, determine the participant's informational model, and evaluate convergent behavioral, neural, and computational evidence. The strongest next advances will compare mechanistic predictions under interventions, harmonize reporting without erasing meaningful task differences, and validate generalization beyond controlled settings. PIT thus provides an enduring research instrument for studying how learned predictions organize action, as well as a precise comparative paradigm for new forms of human and artificial decision systems Badioli et al. (2024) Huseynzade et al. (2026).
References
- Badioli, M., Degni, L. A. E., Dalbagno, D., Danti, C., Starita, F., di Pellegrino, G., Benassi, M., & Garofalo, S. (2024). Unraveling the influence of Pavlovian cues on decision-making: A pre-registered meta-analysis on Pavlovian-to-instrumental transfer. Neuroscience & Biobehavioral Reviews, 164, 105829. https://doi.org/10.1016/j.neubiorev.2024.105829
- Belanger, M. J., Chen, H., Hentschel, A., Garbusow, M., Ebrahimi, C., Knorr, F. G., Zech, H. G., Pilhatsch, M., Heinz, A., & Smolka, M. N. (2022). Development of novel tasks to assess outcome-specific and general Pavlovian-to-instrumental transfer in humans. Neuropsychobiology, 81(5), 370–386. https://doi.org/10.1159/000526774
- Belanger, M. J., Chen, H., Fröhner, J. H., Garbusow, M., Heinz, A., & Smolka, M. N. (2025). Behavioural and neural reliability of a Pavlovian-to-instrumental transfer task. Addiction Biology, 30(12), e70112. https://doi.org/10.1111/adb.70112
- Cartoni, E., Balleine, B., & Baldassarre, G. (2016). Appetitive Pavlovian-instrumental transfer: A review. Neuroscience & Biobehavioral Reviews, 71, 829–848. https://doi.org/10.1016/j.neubiorev.2016.09.020
- Colwill, R. M., & Rescorla, R. A. (1988). Associations between the discriminative stimulus and the reinforcer in instrumental learning. Journal of Experimental Psychology: Animal Behavior Processes, 14(2), 155–164. https://doi.org/10.1037/0097-7403.14.2.155
- Corbit, L. H., & Balleine, B. W. (2003). Instrumental and Pavlovian incentive processes have dissociable effects on components of a heterogeneous instrumental chain. Journal of Experimental Psychology: Animal Behavior Processes, 29(2), 99–106. https://doi.org/10.1037/0097-7403.29.2.99
- Corbit, L. H., & Balleine, B. W. (2005). Double dissociation of basolateral and central amygdala lesions on the general and outcome-specific forms of Pavlovian-instrumental transfer. Journal of Neuroscience, 25(4), 962–970. https://doi.org/10.1523/JNEUROSCI.4507-04.2005
- Corbit, L. H., Janak, P. H., & Balleine, B. W. (2007). General and outcome-specific forms of Pavlovian-instrumental transfer: The effect of shifts in motivational state and inactivation of the ventral tegmental area. European Journal of Neuroscience, 26(11), 3141–3149. https://doi.org/10.1111/j.1460-9568.2007.05934.x
- Corbit, L. H., & Balleine, B. W. (2011). The general and outcome-specific forms of Pavlovian-instrumental transfer are differentially mediated by the nucleus accumbens core and shell. Journal of Neuroscience, 31(33), 11786–11794. https://doi.org/10.1523/JNEUROSCI.2711-11.2011
- Degni, L. A. E., Dalbagno, D., Starita, F., Benassi, M., di Pellegrino, G., & Garofalo, S. (2022). General Pavlovian-to-instrumental transfer in humans: Evidence from Bayesian inference. Frontiers in Behavioral Neuroscience, 16, 945503. https://doi.org/10.3389/fnbeh.2022.945503
- Eder, A. B., & Mitschke, V. (2025). Pavlovian to instrumental transfer of control over fight or flight decisions. npj Science of Learning, 10, 34. https://doi.org/10.1038/s41539-025-00331-4
- Garbusow, M., Schad, D. J., Sebold, M., Friedel, E., Bernhardt, N., Koch, S. P., Steinacher, B., Kathmann, N., Geurts, D. E. M., Sommer, C., Müller, D. K., Nebe, S., Paul, S., Wittchen, H.-U., Zimmermann, U. S., Walter, H., Smolka, M. N., Sterzer, P., Rapp, M. A., ... Heinz, A. (2016). Pavlovian-to-instrumental transfer effects in the nucleus accumbens relate to relapse in alcohol dependence. Addiction Biology, 21(3), 719–731. https://doi.org/10.1111/adb.12243
- González, F., Garre-Frutos, F., Hinojosa-Aguayo, I., & Hall, G. (2025). Extinction of outcome-specific Pavlovian-to-instrumental transfer (PIT), instrumental outcome devaluation, and reward-related attentional capture are predicted by affect-driven impulsivity. Learning & Behavior, 53(4), 335–354. https://doi.org/10.3758/s13420-025-00676-1
- Hogarth, L. (2020). Addiction is driven by excessive goal-directed drug choice under negative affect: Translational critique of habit and compulsion theory. Neuropsychopharmacology, 45(5), 720–735. https://doi.org/10.1038/s41386-020-0600-8
- Holland, P. C., & Gallagher, M. (2003). Double dissociation of the effects of lesions of basolateral and central amygdala on conditioned stimulus-potentiated feeding and Pavlovian-instrumental transfer. European Journal of Neuroscience, 17(8), 1680–1694. https://doi.org/10.1046/j.1460-9568.2003.02585.x
- Holland, P. C. (2004). Relations between Pavlovian-instrumental transfer and reinforcer devaluation. Journal of Experimental Psychology: Animal Behavior Processes, 30(2), 104–117. https://doi.org/10.1037/0097-7403.30.2.104
- Huang, Y., Qu, C., Gazzola, V., Garofalo, S., Starita, F., Krebs, R. M., Degni, L. A. E., Wei, J., Finotti, G., di Pellegrino, G., Desmurget, M., Sirigu, A., & Bardi, L. (2026). Motor activation in cue-guided behavior: Neural evidence from human Pavlovian-to-instrumental transfer (PIT). NeuroImage, 331, 121874. https://doi.org/10.1016/j.neuroimage.2026.121874
- Huseynzade, U., Dignath, D., & Eder, A. B. (2026). Directed forgetting and working-memory load reduce outcome-selective Pavlovian-to-instrumental transfer in humans. Scientific Reports, 16, 32070. https://doi.org/10.1038/s41598-026-75419-x
- Mahlberg, J., Seabrooke, T., Weidemann, G., Hogarth, L., Mitchell, C. J., & Moustafa, A. A. (2021). Human appetitive Pavlovian-to-instrumental transfer: A goal-directed account. Psychological Research, 85(2), 449–463. https://doi.org/10.1007/s00426-019-01266-3
- Prévost, C., Liljeholm, M., Tyszka, J. M., & O'Doherty, J. P. (2012). Neural correlates of specific and general Pavlovian-to-instrumental transfer within human amygdalar subregions: A high-resolution fMRI study. Journal of Neuroscience, 32(24), 8383–8390. https://doi.org/10.1523/JNEUROSCI.6237-11.2012
- Rescorla, R. A., & Solomon, R. L. (1967). Two-process learning theory: Relationships between Pavlovian conditioning and instrumental learning. Psychological Review, 74(3), 151–182. https://doi.org/10.1037/h0024475
- Rescorla, R. A. (1994). Control of instrumental performance by Pavlovian and instrumental stimuli. Journal of Experimental Psychology: Animal Behavior Processes, 20(1), 44–50. https://pubmed.ncbi.nlm.nih.gov/8308492/
- Seabrooke, T., Le Pelley, M. E., Hogarth, L., & Mitchell, C. J. (2017). Evidence of a goal-directed process in human Pavlovian-instrumental transfer. Journal of Experimental Psychology: Animal Learning and Cognition, 43(4), 377–387. https://doi.org/10.1037/xan0000147
- Seabrooke, T., Hogarth, L., Edmunds, C. E. R., & Mitchell, C. J. (2019). Goal-directed control in Pavlovian-instrumental transfer. Journal of Experimental Psychology: Animal Learning and Cognition, 45(1), 95–101. https://doi.org/10.1037/xan0000191
- Talmi, D., Seymour, B., Dayan, P., & Dolan, R. J. (2008). Human Pavlovian–instrumental transfer. Journal of Neuroscience, 28(2), 360–368. https://doi.org/10.1523/JNEUROSCI.4028-07.2008
- Tang, R., Price, J. M., & Bonardi, C. (2026). Online and in-person stress induction in a quasi-avoidance-based Pavlovian-to-instrumental transfer task. Learning & Behavior, 54, 302–318. https://doi.org/10.3758/s13420-026-00707-5