A Quarter Century of Research on Infant Contingency Learning: Current and Future Directions

Abstract

Although traditional learning paradigms provided a substantial base for the emergence of the field of infant studies from the 1960s through the 1990s, research on contingency (operant) learning in infancy has not attracted much attention over the last 25 years. While the reasons for such neglect are unclear, learning protocols offer valuable contributions to the field of infant studies, spanning basic research, translational work, and application. An examination of the literature over the last quarter century shows operant learning concepts in use with respect to the development of agency, goal blockage reactivity, clinical cross-group comparisons, and developmental interventions. Building upon the foundation that infants are capable of contingency learning, research has explored underlying mechanisms, including coordinated movement dynamics and psychobiological correlates. Methodological innovations—such as novel paradigms and cutting-edge techniques like motion capture, eye-tracking, and computational modeling—have further refined our understanding of these processes. Efforts have also focused on identifying conditions that promote learning and factors contributing to data loss. An overarching question remains whether infants demonstrate agency during contingency learning. Additionally, recent research has shifted from a primarily experimental group approach to considering individual differences in early learning. However, it is unclear whether traditional learning metrics effectively capture nonmonotonic behavioral change and variability in learning patterns. The review offers cogent rationales for reintegrating these paradigms into the field of infant studies, discusses gaps in the literature that should be addressed for this goal to be realized, and proposes future directions for advancing the field.

1.0. An Introduction and a Brief History of the Study of Infant Contingency Learning

The field of infant studies emerged from a synthesis of developmental and comparative psychology in the 1960s. At the time of this emergence, the field of psychology was still heavily invested in traditional learning paradigms, owing to the vestigial dominance of behaviorism, learning theory, and behavioral approaches from the 1950s and 1960s. In this context, a robust literature on infant learning emerged over that time (Horowitz, 1968; Kessen, 1963; Lipsitt, 1964, 1966), which was largely inspired by questions about the fundamental behavioral competencies of the human infant (Friedman & Vietze, 1972; Stone, Smith, & Murphy, 1973).

Some of this early literature was devoted to classical conditioning (Brackbill, Fitzgerald, & Lintz, 1967), again motivated by questions around whether such learning was within the capacity of the neonate (Connolly & Stratton, 1969). However, much of the attention in this area became quickly focused to the study of paradigms involving contingencies (i.e., consequences for behavior), mostly conceptualized around operant conditioning (Watson, 1966, 1967). While a number of papers were devoted to basic parameters of infant learning (e.g., Siqueland, 1968), the early field of operant/contingency learning in infancy coalesced around two points.

First, having grown out of the functional school of psychology, the field turned to applied problems, such as the conditioning of social signals in the caregiving environment (Baer & Wolf, 1969; Brackbill, 1958; Etzel & Gewirtz, 1967; Rheingold, Gewirtz, & Ross, 1959; Siegel, 1969). The second involved the use of conditioning paradigms and protocols (e.g., Friedlander, 1966; Simmons & Lipsitt, 1961) to control infant behavior for the study of other behavioral phenomena. This included the use of contingencies to control infants’ visual fixation (Watson, 1969), head-turns (Siqueland, 1964), heart rate (Shearn, 1962), non-nutritive sucking (Haye, 1967), and motor behaviors, such as foot-kicking (Rovee & Rovee, 1969).

This latter line of work proved especially impactful, as it effectively launched numerous other lines of research. For example, the high-amplitude sucking paradigm was developed to study infant speech perception (Eimas, Siqueland, Jusczyk, & Vigorito, 1971; Eimas, 1985). The head-turn paradigm was also used for studying basic auditory processes in infancy (Schneider, Trehub, & Bull, 1979), music perception (Trehub, Thorpe & Trainor, 1990), and early language (Kuhl, 1979, 1983; Werker, Polka, & Pegg, 1997; Werker & Tees, 1984) development, and its legacy lives on as a clinical tool (visually reinforced audiometry) for screening infants’ hearing (Thompson & Wilson, 1984). The reinforcement of visual fixations was used in the assessment of infant auditory preferences (Colombo & Bundy, 1981; DeCasper & Fifer, 1980) and was eventually expanded for use in the synchronous reinforcement paradigm, a specific schedule in which reinforcement continues for the duration of the operant; this procedure was used for assessing discrimination learning in preverbal infants (Coldren & Colombo, 1994; Colombo, Mitchell, Coldren, & Atwater, 1990). Lastly, research on conjugate reinforcement—in which the intensity of the reinforcer is proportional to response strength (Lipsitt, Pederson, & DeLucia, 1966; McKirdy & Rovee-Collier, 1978; Rovee-Collier & Capatides, 1979; Rovee-Collier & Gekoski, 1979; Sullivan, Rovee-Collier, & Tynes, 1979))—formed the basis for Rovee-Collier’s seminal research program on infant long-term memory (Hartshorn & Rovee-Collier, 1997; Mast, Fagen, Rovee-Collier, & Sullivan, 1980; Rovee-Collier & Cuevas, 2008, 2009).

1.1. Summary and Outline

Here, we present a review of the last 25 years of infant contingency learning. The study of contingency has a long history within the context of infant-caregiver relationships, with the contingent nature of the caregivers’ responses being a critical factor in the judged quality of interaction, and the effect of that quality on infant outcomes (see Masek et al., 2021; Northup, 2017, for review). Our review is focused on infant contingency learning, rather than the contingent nature of the infant’s caregiving environment, and as such, coverage of caregivers’ contingent responding falls beyond its scope. As should be evident from our brief historical review of infant learning, researchers have been quick to recognize the potential and power of operant/contingency learning paradigms for the investigation of other behavioral processes, with large-scale research programs in (for example) memory, representation, and language generated on the back of learning protocols. Our review will be configured in terms of the topics that investigators have addressed with respect to infant learning over this period.

2.0. State of the Infant Continency Learning Literature: 2000-2025

Over four decades of empirical research have established that even very young infants exhibit learning across various response-reinforcement contingencies. This body of work has also identified ontogenetic changes in how quickly infants learn, with older infants acquiring operant contingencies more rapidly than younger infants (Davis & Rovee-Collier, 1983; Hill, Borovsky, & Rovee-Collier, 1988)—a pattern observed across other learning paradigms as well (e.g., Barr, Dowden, & Hayne, 1996; Morgan & Hayne, 2006). Building on these foundations, research in the past quarter century has explored underlying mechanisms (e.g., coordinated movement dynamics, psychobiological correlates; Kelso & Fuchs, 2016; Tummeltshammer, Feldman, & Amso, 2019); how infants respond to disruptions of learned contingencies, including extinction (e.g., Cuevas, Learmonth, & Rovee-Collier, 2016; Sullivan, 2018); and whether infants demonstrate evidence of agency during contingency learning (e.g., Fujihira & Taga, 2023; Zaadnoordijk et al., 2018). Additionally, contingency learning continues to be used to compare performance across clinical groups (e.g., Bhat, Galloway, & Landa, 2010; Sargent, Kubo, & Fetters, 2018) and as an intervention to promote development (e.g., Campbell et al., 2015; Chorna et al., 2014). There is also growing interest in individual differences in early learning and their connections to broader developmental markers (e.g., Lewis et al., 2015; Needham, Joh, Wiesen, & Williams, 2014). In the following sections, we review these key areas of inquiry within the infancy contingency learning literature. We begin with an overview of contingency learning that focuses on recent methodological advancements and ongoing challenges.

2.1. Methodological Challenges and Innovations

Since 2000, research has focused on optimizing methods of testing contingency learning in infants (e.g., Kraebel, Fable, & Gerhardstein, 2004; Merz et al., 2017), a topic explored in greater detail throughout our review. In 2020, Jacquey and colleagues published a comprehensive review examining factors that influence contingency learning during infancy, including contingency parameters, response and outcome characteristics, and individual differences (see also Pelaez & Monlux, 2017). Although we highlight only a few examples within the last quarter century, these questions have long been central for the field. Research comparing different types of reinforcers has found that conjugate reinforcement elicits higher peak responding than continuous reinforcement (Voltaire et al., 2005). Additionally, similar level of learning across reinforcers of various modalities has been found (i.e., auditory, visual, and audiovisual), while there is also evidence that audiovisual reinforcers enhance learning and retention at 3-4 months (Kraebel et al., 2004; Voltaire et al., 2005). Further, studies have shown that by 3 months, infants can learn highly specific movement patterns (e.g., 85-degree flexion right knee) in response to contingencies, rather than only displaying generalized behavioral changes (Angulo-Kinzler, & Horn, 2001).

There have been efforts to identify factors contributing to data loss, such high baseline response levels and infant irritability (Lemelin, Tarabulsy, & Provost, 2002; Millar & Weir, 2015 Watanabe & Taga, 2011). Additionally, it remains unclear whether traditional learning metrics (e.g., percentage change response frequency learning criterion) effectively capture nonmonotonic behavioral change and variability in learning patterns (e.g., Popescu et al., 2021; Sloan et al., 2023). Finer-grained temporal analyses have detected evidence of learning in 4- to 8-month-olds sooner than when analyzed at 1-min or longer intervals (Popescu et al., 2021). Coarse temporal resolutions may overlook instances of rapid learning followed by fussiness or disengagement. Recent efforts have considered probability-based measures within specific time intervals to revisit Watson’s (1979, 1984, 1985) proposal that contingency detection is reflected in repeated response patterns to “test” the contingency (Popescu et al., 2021). These methodological advancements are valuable because they allow for more inclusive analysis of infants with variable response profiles while also providing deeper insight into real-time learning processes—patterns that traditional threshold-based learning criterion measures may overlook (see also Sen & Gredebäck, 2021).

Despite significant progress, the underlying mechanisms of infant contingency learning remain largely unknown. Research on associative learning with 7-month-olds (Tummeltshammer et al., 2019) has identified spontaneous eye-blink rate as a potential indirect marker of dopamine-based reward learning, a measure previously used in adults and nonhuman animals (see Eckstein et al., 2017, for review). Differences in such physiological markers could provide predictive error data for reinforcement learning (Van Slooten, Jahfari, & Theeuwes, 2019), an area yet to be explored in infancy. Incorporating additional physiological measures, such as pupil dilation, could offer insights into arousal and cognitive processing linked with noradrenergic activity—factors that contribute to variability in infants’ and adults’ reinforcement learning (Tummeltshammer et al., 2019; Van Slooten, Jahfari, Knapen, & Theeuwes, 2018).

Similarly, open questions remain regarding the neural correlates of contingency learning during infancy. Recent fMRI studies of awake infants have shown hippocampal activation during statistical learning and frontal activity related to attentional processes (Ellis et al., 2021a, 2021b). Advancements in gaze-contingent paradigms (Colombo et al., 1990; Wang et al., 2012) provide opportunities to investigate cortical and subcortical engagement during contingency learning while minimizing motor artifacts associated with traditional limb-based response measures. Despite the potential of psychophysiological approaches, only a few studies have examined EEG—focusing on event-related potentials to explore agency either following contingency learning (Zaadornojick et al., 2020) or during contingent versus noncontingent stimulation (Meyer & Hunnius, 2021). Notably, oscillatory EEG theta activity has been proposed as a neural correlate of learning (Begus & Bonawtiz, 2020; Clarke, Roberts, & Ranganath 2018), yet its role in contingency learning during infancy remains largely unexplored.

2.2. Extinction and Goal Blockage Reactivity

Contingency learning has been used to investigate the interplay between motoric, affective, physiological, and behavioral systems. A particularly informative context for studying infants’ emotional expressions and behavioral responses involves extinction (i.e., disruptions of acquired contingencies — when a previously rewarded response no longer yields a desirable outcome). Extinction has been conceptualized both as “goal blockage” and a form of “reversal learning,” depending on whether the focus is on affective or cognitive processes (e.g., Cuevas et al., 2016; Sullivan, 2018).

Although extinction has been used as an experimental manipulation in infant contingency research, little attention has been given to underlying learning processes. Pavlov (1927) considered extinction to involve the acquisition of a new association (R → no O) rather than simply erasing the original response-outcome (R → O) link. Research shows that even with extended extinction exposure beyond typical contingency learning investigations, 3-month-olds do not decrease their kicking rates during nonreinforcement (Cuevas et al., 2016; Shafer, 2008). However, they exhibit a strong extinction effect when tested the following day. Furthermore, when acquisition and extinction occur in different incidental contexts (i.e., colorful crib liners), 3-month-olds’ response levels depend on the testing context (Cuevas et al., 2016). When tested outside of the extinction context—either in the acquisition context or a novel one—infants show evidence of prior learning. These findings suggest that original learning is preserved after extinction and that infants not only encode details of their environment but also use this information to guide their future actions. At the same time, however, numerous questions remain about the developmental time course and individual differences in extinction learning.

A relevant area of inquiry has examined infants’ coordinated behavioral, physiological, and affective responses during contingency learning and its disruption. For example, during arm pulling contingency acquisition, 4-month-olds exhibit decreased heart rate (HR) in conjunction with increased respiratory sinus arrythmia (RSA), autonomic markers interpreted as indicators of attention supporting learning (Lewis, Hitchcock, & Sullivan, 2004). In contrast, during a brief 2-min extinction phase, infants typically maintain or increase their responses while displaying more anger or sadness facial expressions. Psychobiological responses to extinction also vary as a function of facial expressions: anger is linked to increased HR, while sadness is associated with elevated cortisol levels, potentially signaling a withdrawal response (Lewis, Ramsay, & Sullivan, 2006; see also Lewis & Ramsay, 2005). Anger expressions during extinction at 4-5 months have been proposed to reflect approach behaviors toward obstacles; they are not tied to presumed traits of emotional reactivity, such as concurrent temperament or future tantrum behaviors (Sullivan, 2016; Sullivan & Lewis, 2012). Notably, these anger responses predict greater behavioral persistence in the face of goal blockage at 20 months (Lewis, Sullivan, & Kim, 2015).

Furthermore, infants display anger expressions when learned contingencies are disrupted in other real-world contexts, such as partial reinforcement (when the frequency or intensity of the reinforcer decreases) or noncontingent reinforcement (when the reinforcer occurs independently of their actions; Sullivan & Lewis, 2003). Interestingly, while 4- to 5-month-olds continue responding during partial reinforcement, their responding declines within 2 min of noncontingent reinforcement. Broader research on contingency learning has also shown that 3- and 4-month-olds decrease pulling or kicking movements in response to noncontingent stimulation (Sloan et al., 2023; Watanabe et al., 2011). In contrast, 2-month-olds do not exhibit this response differentiation, instead increasing movement across all limbs regardless of whether reinforcement is contingent or not (Watanabe et al., 2011). Recent dynamic systems modeling has proposed that these age-related changes in noncontingent responding reflect the development of action differentiation in contingency learning, which requires the ability to both enhance and inhibit movements (bifurcation dynamics). This plays a crucial role in constraining the formation of spurious action-outcome associations, in conditions such noncontingent stimulation (Fujihara & Taga, 2023). Taken together, these findings suggest that behavioral persistence during brief extinction and partial reinforcement is unlikely attributed to a lack of inhibitory control in infants older than 2 months. Whether these patterns relate to self-agency has been a recurring theme in infancy contingency research over the past quarter century, as discussed in more detail in the following sections.

2.3. Contingency Learning Characteristics: Examinations of Agency

The past quarter century has also seen more granular analyses of behavioral patterns and the integration of computational modeling and neural measures to further characterize the nature of early learning processes and underlying mechanisms. Advances such as 3D motion capture—where sensors are placed on limbs and joints—allow researchers to analyze whole-body movement dynamics (e.g., movement velocity, coordination) across various contexts and time scales. For instance, this approach has revealed that while 2-month-olds increase responses across all limbs in an arm pulling-contingency task, 3- and 4-month-olds display more topographically specific bilateral and unilateral arm movements, respectively (Watanabe & Taga, 2006). These distinct behavioral profiles have been interpreted as potential evidence of agency—a construct that is conceptualized in various ways, including having a sense that our behavior has consequences or forming action-outcome mental representations and/or expectations. In considering more parsimonious explanations of infants’ contingency responses, a simple computational learning model—lacking representational mechanisms such as memory or agency and relying solely on operant conditioning—was able to replicate the observed pattern of single-limb activation during training (Zaadnoordijk et al., 2018).

A simple, “nonrepresentational” model of infant contingency learning (Zaadnoordijk et al., 2018), however, failed to reproduce the “extinction burst” that infants typically display shortly after reinforcement is removed (e.g., Cuevas et al., 2016, Sloan et al., 2023). Heightened responding during extinction has been considered as potential evidence of agency, reflecting infants’ anticipation of the consequences of their actions. Accordingly, one might therefore expect that EEG correlates of violation of expectation, such as mismatch negativity (MMN) during extinction would offer additional insights into processes underlying infants’ behavioral responses. Three- to 4.5-month-olds who exhibited MMN during extinction of arm pulling a computerized mobile also showed limb-specific increases in responding, while those without a clear MMN response did not (Zaadnoordijk et al., 2020). Thus, young infants display behavioral, neural, and psychobiological responses when action-outcome associations are disrupted; however, there are individual differences in these responses which have been interpreted as reflecting variations in emerging agency as well as approach-withdrawal reactions (e.g., Lewis et al., 2006; Sloan et al., 2023).

Detailed analyses of infants’ movement patterns have revealed shifts in movement coordination dynamics throughout contingency learning. Three- to 4-month-olds exhibit highly coordinated connected limb-mobile movement during contingency learning, surpassing both interlimb and nonconnected limb-mobile coordination (Sloan et al., 2023). This pattern is absent during a pre-learning noncontingent stimulation phase, where interlimb movement correlations dominate. Similarly, by 10 months, infants demonstrate anticipatory looking with gaze shifts to the reinforcer location 0.5 s prior to operant button presses (Kenward, 2010). These looking-pressing correlations persist during extinction, providing another instance of system coordination during infancy, potentially indicative of learned action-outcome expectations.

The animal learning literature has explored the role of intention in goal-directed behavior by investigating outcome devaluation following contingency learning (Dickinson, 1985, 1989). This approach assesses whether responses are linked to a specific outcome rather than general positive consequences. A developmental study using a touch-screen display and video reinforcers found that 18- to 48-month-olds exhibited differential responding indicative of outcome devaluation during a reacquisition period (Klossek, Russell, & Dickinson, 2008). However, differential responding during extinction emerged only after 27 months. This finding was interpreted as evidence that goal-directed behavior is dissociable from contingency learning in young infants (Klossek et al., 2008). That said, age-related differences in amount of training with action-outcome contingencies may have also influenced these findings. This work provides a more nuanced perspective of ontogenetic changes in goal-directed contingency learning during early childhood and also offers an approach that could be applied across different paradigms and at earlier points in development.

Novel analytical approaches have also provided insights into individual differences in contingency learning. Traditional assessments rely on learning criteria based on whether the frequency of a behavior within predetermined periods of time surpasses a set threshold (e.g., 150% above baseline; see Cuevas et al., 2024 for review). However, to capture sudden changes in activity—an “aha moment” of contingency detection—recent work has analyzed peak increases in movement rate using sliding time windows. This method also allows for examining variations in coordinated responding before and after peak detection (Sloan et al., 2023), a point initially suggested in studies of infant discrimination learning (Colombo et al., 1990). Using this approach, distinct clusters of 3- to 4-month-olds were identified, exhibiting peak changes in mobile contingency responding early, middle, and late in the training session. Additionally, fine-grained analysis of infants’ movement patterns revealed alternating bouts of kicking and stillness. These moments of stillness, as opposed to continuous movement to maximize reinforcement, would be a requisite for recognizing self-generated actions as the cause of mobile movement. Sloan et al. (2023) propose that “emergence of agency can take the form of a punctuated self-organizing process, with meaning found both in movement and stillness.” Beyond agency, detailed characterizations of infants’ contingency learning profiles may enhance our understanding of underlying mechanisms and improve the identification of variations in learning—offering valuable translational applications.

2.4. Infant Contingency Learning and Clinical Populations

Over the last two decades, a small literature has emerged examining differences (or lack of differences) in infant contingency or operant learning across different populations. One study (Gerhardstein, Dickerson, Miller, & Hipp, 2012) conducted within the US examined infant performance on operant tasks as a function of differences in environment. Like many objective measures of infant cognition in the first year (Bornstein, Pecheux, & Lecuyer, 1988; Seitz et al., 2024), infant contingency learning was not found to vary as a function of socioeconomic status. Graf et al. (2012) used the conjugate reinforcement protocol to examine learning, immediate, and long-term retention in infants from both German and African cohorts. Infants from both groups learned the contingency, and retention measures were equivalent once differences in the level of baseline activity were statistically controlled.

2.4.1. Autism Spectrum Disorders.

Several studies have been published on infant contingency learning in clinical populations. Given widespread societal interest in autism spectrum disorders (ASD) and a targeted initiative to study infant siblings of children with ASD (Rogers, 2009), it would seem natural that studies of infant learning might have been in the ASD-sibling literature might have emerged, either as a marker of endophenotyping (Mosconi et al., 2023) or for early identification of infants at risk (Zwaigenbaum, Bryson, & Garon, 2013). However, over the last 25 years, few papers have been published in this realm (Bhat et al., 2010; Neimy et al., 2017), with the former mostly drawing conclusion about gaze to social targets as a potential marker for ASD, and the latter merely a call to emphasize the use of operant learning models as a basis for studying the development of social and communicative skills in this population.

2.4.2. Preterm Infants.

Contingency-paradigm performance has also been examined in preterm infant groups, with full-term infants as controls. Heathcock et al. (2004), Haley et al. (2008), Lobo and Galloway (2013), and Sargent et al. (2018) all used the footkick-mobile/conjugate reinforcement technique. Most studies have concluded that preterms were shown to be consistently less active, and showed less robust evidence for learning and retention, although the focus for each study varied to some degree. Sargent et al. (2018) were largely interested in the topography of the footkick. Hayley et al. (2008) supplemented the learning assessment with psychophysiological and endocrine measures, and found evidence that preterms were less able to regulate parasympathetic activity during the administration of the tasks, and that arousal levels may mediate the underlying processes responsible for differences in the two groups’ performance. Finally, Lobo and Galloway (2013) reported that learning deficiencies during the first year persisted to 24 months, and that granular learning and problem-solving tasks might be more sensitive predictors of delay than global developmental assessments.

2.4.3. Other Populations and Conditions.

Several isolated studies of other clinical populations were also published. Taylor et al. (2013) examined learning in 6-month-old infants with spina bifida, relative to a group of typically-developing controls in a task where a mobile was tethered to infants’ wrists. Both groups showed increased hand-waving while tethered, thus signaling acquisition of the contingency, although infants with spina bifida were less likely to attain a predetermined criterion for learning or retain the contingency over time. Campbell et al. (2015) used the conjugate/mobile footkick paradigm with a small and heterogeneous sample of infants with periventricular damage to demonstrate the potential of the task for intervention purposes.

2.4.4. Use of Contingency as Intervention.

Some investigators have emphasized the possibility of using operant paradigms for widely divergent purposes, including the prevention of sudden infant death (Paluszynkska et al., 2004), although few have engaged programmatically in seeking these applied or translational goals. One promising trend, however, has been reference to the use of operant or contingency-based experiences as possible interventions for improving longer-term outcomes; this literature (see Inamdar, Khurana, & Dusing, 2022, for a systematic review) suggests that early engagement of infants in highly contingent experimental protocols might have positive effects across a broader realm of behavioral or biobehavioral domains. Given the consistent literature demonstrating positive effects of parental/caregiver responsiveness as a critical variable in cognitive and socioemotional development during infancy (e.g., Frenkel et al., 2024), the application of predictable consequences to endogenous responses would seem to be a natural and potentially well-controlled extension in the search for effective early interventions.

Toward that end, two studies (Chorna et al., 2014; Hamm et al., 2015) have been conducted with high-risk infants have shown that using a pacifier-activated music player (PAM), presenting the mother’s voice contingent upon oral feeding increases the rate, volume, and frequency of oral feeding, and also led to shorter hospital stays. The authors conceptualize this as an operant paradigm, although the control group was simply a standard-practice (no PAM) group rather than a yoked or random control group. As such, other potential explanations (e.g., increases in arousal) cannot be definitively eliminated as confounds.

3.0. Conclusions and Future Directions

As we bring this review of the last 25 years to a conclusion, a number of points stand out for summary and elaboration and encouragement for the next quarter century. We hope these are helpful in guiding the next generation of scientists working in the field of infant behavior and development.

3.1. Contingency Learning: A Means or an End for Inquiry?

First, while contingency and operant learning procedures in infancy continue as active options for some laboratories, the focus of their use remain (as they have since the 1960s) largely as a window to other behavioral or cognitive realms (e.g., Cuevas, Rovee-Collier, & Learmonth, 2006; Needham et al., 2014). Relatively little work has been done on infant contingency learning per se, even as many papers throughout the period covered here – in various disciplines – have called for a resurgence in its use and inquiry. Older infants learn more quickly than younger infants, and the standard effects of reinforcement schedules presumably hold for infants across the age range (e.g., Davis & Rovee-Collier, 1983; Hartshorn et al., 1998; Hill et al. 1988). Beyond that, however, we know relatively little about basic parameters of infants’ processing of contingencies, such as the breadth of the temporal windows that bind consequences to actions, and the developmental functions for those windows (see Jacquey et al., 2020: Pelaez & Monlux, 2017, for reviews). The majority of the contingency learning literature has focused on infants between 2 and 7 months of age, despite its relevance across the lifespan. In addition, although infant contingency learning seems readily conceptualized as a unitary function, we do not know whether the different forms of contingency learning cohere as a phenomenon or whether they correlate across contexts.

3.2. Use of Contingency Learning in Translational and Applied Contexts.

As we have suggested throughout the review, infant learning can be harnessed for impact in the field beyond as a simple methodological tool. One of the clearer rationales for the study of early behavioral measures lies with their potential for use in translational contexts. In large part, the relatively sparse and fragmented nature of work in this area suggests that researchers may have missed an opportunity over the last 25 years to exploit operant/contingency learning for these purposes. One challenge noted in some recent learning research is the limited use of noncontingent reinforcement comparisons and baseline activity prior to contingency manipulations (e.g., Tripathi et al., 2019; Wang et al., 2012). Without these controls, it can be difficult to determine whether observed changes in behavior result from contingency learning itself or are instead driven by arousal, general motor activity, or other factors.

Given the fundamental nature of learning in the human behavioral sciences, measures of learning can be used as biobehavioral markers for intellectual or developmental disorders, or as indicators of potential delay in clinical populations. It is heartening to see some scattered work on this topic over the last two decades, but discouraging that there no sustained programs of research have sought to address this potential. The small but extant literature on preterms and at-risk samples suggests less robust learning profiles in those populations, but in many cases, the data suggest that even these infants are capable of basic acquisition (e.g., Hayley et al., 2008; Sargent et al., 2018). It seems clear that work in this area would necessitate a more comprehensive inquiry toward the development of more challenging configurations of contingency learning to develop more sensitive indices of later risk.

3.2.1. Individual Differences.

A corollary to this proposal is the study of individual differences in infant learning. It is relatively surprising that, aside from some work with small samples, we know relatively nothing of the intraindividual reliability of learning indices, nor do we know whether the concept of learning itself represents a strong unitary latent factor or whether infants’ performance across tasks featuring different conditioning parameters is dissociable. Such work would also be important toward the inclusion of contingency learning measures in longitudinal clinical trials, as learning protocols can be readily adapted for administration across different age groups and (when appropriately configured) can remain engaging and yield data with lower attrition rates.

3.3. Moving Beyond Traditional Models of Learning and Development

One exciting aspect of the current literature on infant learning is the inclusion of ancillary measures of cognition, psychophysiology, and psychobiology during contingency tasks. While this approach would represent a clear departure from the traditional Skinnerian design of operant conditioning tasks, it would be resonant with efforts to identify underlying mechanisms, co-active processes, and the emergence of developmental cognitive neuroscience over the last several decades. Developmental cascades approaches (Oakes & Rakison, 2019) emphasize more holistic examinations of developmental change, considering the profound behavioral, neural, and biological changes across multiple domains of function and their interrelations across time. For instance, even within the realm of infant cognition—the examination of learning, memory, and attention processes are largely studied in isolation in well-controlled investigations. However, in our daily lives these processes affect one another; what we focus on affects what we perceive, learn, and retain and vice versa (see Cuevas, Learmonth, & Davinson, in press, for review). Characterizing development change in contingency learning within broader contexts is prime area for basic and applied research during the next quarter century,

3.4. Contingency Learning as Experience/Intervention

Finally, we have seen the suggestion and several small-sample demonstrations of the potential for infant contingency learning to serve as intervention techniques (e.g., Chorna et al., 2014; Keren-Portnoy et al., 2021). Indeed, much of the research on the quality of caregiving has focused on contingency/responsiveness as a critical factor (e.g., Masek et al., 2021, for review); it seems logical that the use of repeated exposure to laboratory experiences featuring highly contingent parameters might work as a means of providing stimulative and engaging environments in addition to the promotion of a sense of agency early in life.

Highlights:

• Infant contingency learning protocols are tools for basic and translational science

• Agency and goal blockage reactivity have been studied using contingency paradigms

• Recent efforts have explored underlying mechanisms of infant contingency learning

• Infant operant learning metrics can yield both clinical markers and interventions

• Variations in infant operant learning are explored by recent innovative approaches