Shifting the focus from what to why

Sensorimotor Contingencies as a Key Drive of Development: From Babies to Robots

Much current work in robotics focuses on the development of robots capable of autonomous unsupervised learning. An essential prerequisite for such learning to be possible is that the agent should be sensitive to the link between its actions and the consequences of its actions, called sensorimotor contingencies. This sensitivity, and more particularly its role as a key drive of development, has been widely studied by developmental psychologists. However, the results of these studies may not necessarily be accessible or intelligible to roboticians. In this paper, we review the main experimental data demonstrating the role of sensitivity to sensorimotor contingencies in infants' acquisition of four fundamental motor and cognitive abilities: body knowledge, memory, generalization, and goal-directedness. We relate this data from developmental psychology to work in robotics, highlighting the links between these two domains of research. In the last part of the article we present a blueprint architecture demonstrating how exploitation of sensitivity to sensorimotor contingencies, combined with the notion of "goal," allows an agent to develop new sensorimotor skills. This architecture can be used to guide the design of specific computational models, and also to possibly envisage new empirical experiments.

A Motor Learning Paradigm Combining Technology and Associative Learning to Assess Prone Motor Learning in Infants

BACKGROUND

Associative learning is the ability to discover a relationship between two or more events. We combined principles of learning and technology to develop a paradigm to assess associative learning in prone.

PURPOSE

The purpose of this study was to determine whether 3- to 6-month-old infants can demonstrate: (1) short-term learning of an association between their upper body movements in prone and activation of a toy, and (2) retention of the association learned on day 1, 24 hours later.

METHODS

Twenty-eight infants who were 3 to 6 months of age and who were typically developing were tested for 2 consecutive days in an instrumented play gym. Both days of testing had a baseline and 4 acquisition phases (2 minutes each). During the acquisition phase, the toy activated for a maximum of 10 seconds when the infant's head was above a threshold. A criterion was set a priori to distinguish infants as short-term learners and retainers of the association learned on day 1.

RESULTS

Of 28 infants, 22 and 14 infants completed all phases of the testing on day 1 and day 2, respectively. Fourteen (50%) of the infants met the criteria for short-term learners. On day 2, there was an analyzable sample of 9 short-term learners. Three of the 12 short-term learners (25%) demonstrated retention on day 2.

CONCLUSION

Consistent with prior infant motor learning research, half of the infants demonstrated associative learning in this novel assessment in prone; however, based on study criteria, the infants had limited retention of the association on day 2.

Ontogenesis of learning and memory: Biopsychosocial and dynamical systems perspectives

In this article, we review recent empirical and theoretical work on infant memory development, highlighting future directions for the field. We consider the state of the field since Carolyn Rovee-Collier's call for developmental scientists to "shift the focus from what to why," emphasizing the function of infant behavior and the value of integrating fractionized, highly specialized subfields. We discuss functional approaches of early learning and memory, including ecological models of memory development and relevant empirical work in human and non-human organisms. Ontogenetic changes in learning and memory occur in developing biological systems, which are embedded in broader socio-cultural contexts with shifting ecological demands that are in part determined by the infants themselves. We incorporate biopsychosocial and dynamical systems perspectives as we analyze the state of the field's integration of multiple areas of specialization to provide more holistic understanding of the contributing factors and underlying mechanisms of the development of memory.

The impact of memory load and perceptual cues on puzzle learning by 24-month olds

Early childhood is characterized by memory capacity limitations and rapid perceptual and motor development [Rovee-Collier (1996). Infant Behavior & Development, 19, 385-400]. The present study examined 2-year olds' reproduction of a sliding action to complete an abstract fish puzzle under different levels of memory load and perceptual feature support. Experimental groups were compared to baseline controls to assess spontaneous rates of production of the target actions; baseline production was low across all experiments. Memory load was manipulated in Exp. 1 by adding pieces to the puzzle, increasing sequence length from 2 to 3 items, and to 3 items plus a distractor. Although memory load did not influence how toddlers learned to manipulate the puzzle pieces, it did influence toddlers' achievement of the goal-constructing the fish. Overall, girls were better at constructing the puzzle than boys. In Exp. 2, the perceptual features of the puzzle were altered by changing shape boundaries to create a two-piece horizontally cut puzzle (displaying bilateral symmetry), and by adding a semantically supportive context to the vertically cut puzzle (iconic). Toddlers were able to achieve the goal of building the fish equally well across the 2-item puzzle types (bilateral symmetry, vertical, iconic), but how they learned to manipulate the puzzle pieces varied as a function of the perceptual features. Here, as in Exp. 1, girls showed a different pattern of performance from the boys. This study demonstrates that changes in memory capacity and perceptual processing influence both goal-directed imitation learning and motoric performance.

Transitions in the temporal parameters of sensory preconditioning during infancy

Sensory preconditioning (SPC) is a form of latent learning in which preexposure to co-occurring neutral stimuli (S1 -S2 ) permits subsequent learning to be transferred from one stimulus (S1 ) to the other (S2 ). We examined whether human infants exhibit developmental transitions in the temporal parameters of SPC by manipulating the preexposure regimen. Infants received simultaneous or sequential preexposure to puppets S1 and S2 (Days 1-2); saw target actions modeled on S1 (Day 3); and were tested for deferred imitation with S2 (Day 4). Although 6-, 9-, and 12-month-olds associated the puppets, there was a shift in the effective regimen from simultaneous to sequential preexposure-similar to prior findings with rat pups (Experiment 1). Experiment 2 revealed that human infants potentially exhibit another transition in SPC at 15 and 18 months of age. We consider the roles of ontogenetic shifts in infants' ecological niche, selective attention, and unitization in developmental transitions in SPC.

Contribution of Embodiment to Solving the Riddle of Infantile Amnesia

At least since the late nineteenth century, researchers have sought an explanation for infantile amnesia (IA)-the lack of autobiographical memories dating from early childhood-and childhood amnesia (CA), faster forgetting of events up until the age of about seven. Evidence suggests that IA occurs across altricial species, and a number of studies using animal models have converged on the hypothesis that maturation of the hippocampus is an important factor. But why does the hippocampus mature at one time and not another, and how does that maturation relate to memory? Our hypothesis is rooted in theories of embodied cognition, and it provides an explanation both for hippocampal development and the end of IA. Specifically, the onset of locomotion prompts the alignment of hippocampal place cells and grid cells to the environment, which in turn facilitates the ontogeny of long-term episodic memory and the end of IA. That is, because the animal can now reliably discriminate locations, location becomes a stable cue for memories. Furthermore, as the mode of human locomotion shifts from crawling to walking, there is an additional shift in the alignment of the hippocampus that marks the beginning of adult-like episodic memory and the end of CA. Finally, given a reduction in self-locomotion and exploration with aging, the hypothesis suggests a partial explanation for cognitive decline with aging.

Infant expressions in an approach/withdrawal framework

Since the introduction of empirical methods for studying facial expression, the interpretation of infant facial expressions has generated much debate. The premise of this article is that action tendencies of approach and withdrawal constitute a core organizational feature of emotion in humans, promoting coherence of behavior, facial signaling, and physiological responses. The approach/withdrawal framework can provide a taxonomy of contexts and the neurobehavioral framework for the systematic, empirical study of individual differences in expression, physiology, and behavior within individuals as well as across contexts over time. By adopting this framework in developmental work on basic emotion processes, it may be possible to better understand the behavioral principles governing facial displays, and how individual differences in them are related to physiology and behavioral function in context.

Ten-month-old infants use prior information to identify an actor's goal

For adults, prior information about an individual's likely goals, preferences or dispositions plays a powerful role in interpreting ambiguous behavior and predicting and interpreting behavior in novel contexts. Across two studies, we investigated whether 10-month-old infants' ability to identify the goal of an ambiguous action sequence was facilitated by seeing prior instances in which the actor directly pursued and obtained her goal, and whether infants could use this prior information to understand the actor's behavior in a new context. Experiment 1 demonstrated that the goal preview impacted infants' subsequent action understanding, but only if the preview was delivered in the same room as the subsequent action sequence. Experiment 2 demonstrated that infants' failure to transfer prior goal information across situations arose from a change in the room per se and not other features of the task. Our results suggest that infants may use their understanding of simple actions as a leverage point for understanding novel or ambiguous actions, but that their ability to do so is limited to certain types of contextual changes.

Crawling is associated with more flexible memory retrieval by 9-month-old infants

In the present experiment, we used a deferred imitation paradigm to explore the effect of crawling on memory retrieval by 9-month-old human infants. Infants observed an experimenter demonstrate a single target action with a novel object and their ability to reproduce that action was assessed after a 24-hr delay. Some infants were tested with the demonstration stimulus in the demonstration context and some infants were tested with a different stimulus in a different context. Half of the infants in each test condition were crawling at the time of participation and half were not. Both crawling and non-crawling infants exhibited retention when tested with the demonstration stimulus in the demonstration context, but only infants who were crawling by 9 months of age exhibited retention when tested with a different stimulus in a different context. These findings demonstrate that the onset of independent locomotion is associated with more flexible memory retrieval during the first year of life.

The interactive development of social smiling

Infant smiles emerge even in the absence of visual feedback, but their interactive development and intensification appear to be dependent on experiences of visually mediated interaction. Although neonatal smiling has no clear emotional content, social smiling emerges out of attentive engagement with an interactive caregiver. This process illustrates the dynamic systems postulate that real-time interaction is a window on developmental process. On the one hand, specific dimensions of smiling may have qualitatively different psychologically meanings. On the other hand, different features of infant smiling may reflect linked indices of a single dimension of positive emotion that ebbs and flows in time. The resolution of this paradox will likely involve continued attention to the interactive flow of positive emotion communication. This will be facilitated by new methods for measuring smiling and positive emotion in time. Smiling may simultaneously index a desire to interact and the dissipation of arousal associated with that interaction. Infants' capacity to become actively and vigorously caught up in emotionally positive smile-mediated interaction is linked to their ability to regulate that emotion by gazing away from their interactive partners. Ultimately, this attentional control paves the way for infant's tendency to use smiles to initiate early referential communication with a partner. These anticipatory smiles may provide a developmental bridge between early emotionally positive dyadic responsivity and later patterns of social competence.

Relational information processing of novel unrelated actions by infants

Declarative memory in infants is often assessed via deferred imitation. Not much is known about the information processing basis of the memory effect found in these experiments. While in the typical deferred imitation study the order of actions remains the same during demonstration and retrieval, in two experiments with n=30 respective n=25, 10- and 11-month-old infants, the order of novel unrelated actions in demonstration and retrieval was varied (same, reversed, mixed). This allowed a separation of item-specific from item-relational information processing. In both experiments best memory performance was found when the order of target actions remained the same during encoding and recall, demonstrating that infants seem to rely on item-specific as well as item-relational information which has to be ad hoc constructed while encoding.

Object permanence after a 24-hr delay and leaving the locale of disappearance: the role of memory, space, and identity

Fourteen-month-old infants saw an object hidden inside a container and were removed from the disappearance locale for 24 hr. Upon their return, they searched correctly for the hidden object, demonstrating object permanence and long-term memory. Control infants who saw no disappearance did not search. In Experiment 2, infants returned to see the container either in the same or a different room. Performance by room-change infants dropped to baseline levels, suggesting that infant search for hidden objects is guided by numerical identity. Infants seek the individual object that disappeared, which exists in its original location, not in a different room. A new behavior, identity-verifying search, was discovered and quantified. Implications are drawn for memory, spatial understanding, object permanence, and object identity.

The Effect of Prior Practice on Memory Reactivation and Generalization

Three experiments examined the effect of practice on memory performance by 18-month-old infants. Infants were tested using an imitation paradigm; an adult demonstrated a series of actions with objects and infants were given the opportunity to reproduce those actions following a delay. Some infants practiced the target actions before the retention interval (practice) and some did not (no practice). In Experiment 1, a reminder treatment alleviated forgetting by infants who practiced but failed to alleviate forgetting by infants who did not practice. In Experiments 2A and 2B, infants who practiced generalized to novel test stimuli after a 24-hr delay, whereas infants without practice did not. Results suggest practice influences the accessibility and generality of infants' memories.

Does infant memory expression reflect age at encoding or age at retrieval?

Do human infants express a memory acquired earlier in ontogeny in a manner appropriate to their age at encoding or their age at the time of retrieval? To answer this, we exploited the fact that retention is highly context dependent at 6 months but not at 8-9 months of age. Six-month-olds learned an operant response in one context, and their memory was maintained by monthly reinstatements in the original context. At 8 or 9 months of age, 1 month after the last (or only) reinstatement, infants were tested in either the same or a different context. During testing, infants' retention was no longer context dependent; rather, they responded robustly in both test contexts. These results revealed that infants expressed a memory acquired when they were younger in a manner appropriate to their test age. They were interpreted in terms of changes in the functional significance of context before and after infants self-locomote.

Information pick-up by infants: what is it, and how can we tell?

Young infants spend the better portion of their 1st year merely looking around, and what they notice can have critical cognitive consequences. Although looking measures are commonly used to predict what information infants pick up about objects and/or their relationship to other objects, these predictions are often ambiguous. Ultimately, that knowledge can be unambiguously revealed only by giving infants an opportunity to directly act on or use it.

The development of declarative memory in human infants: age-related changes in deferred imitation

In 2 experiments, deferred imitation procedures were used to trace age-related changes in declarative memory by human infants over the first 2 years of life. An adult modeled 3 actions with an object, and infants' ability to reproduce those actions was assessed 24 hr later. Some infants were tested with a new object or in a new context relative to the original demonstration. Changes in the context disrupted the performance of 6-month-olds but had no effect on the performance of 12- and 18-month-olds. Changes in the object disrupted the performance of 6- and 12-month-olds but had no effect on the performance of 18-month-olds. This age-related increase in representational flexibility may account for the decline of childhood amnesia during the 3rd year of life.

Developmental changes in the specificity of memory over the first year of life

In two experiments with 260 infants between 2 and 12 months of age, we examined how differences between the conditions of encoding and retrieval affect retention. Initially, 9- and 12-month-olds were tested with a different cue (Experiment 1) or in a different context (Experiment 2) after delays spanning their respective forgetting functions. These data were then combined with corresponding data previously collected from 2-to 6-month-olds trained and tested in an equivalent task. The resulting analyses revealed that the specificity constraints on memory retrieval become progressively looser at the extremes of the forgetting function with age. With increasing age, retention was less affected by cue changes after shorter absolute delays and, except at 6 months, by context changes after longer absolute delays. This pattern dovetails with evidence of decreasing specificity in the retrieval cues required for deferred imitation during infants' 2nd year and reveals that the memory abilities of older children evolve gradually from early in infancy.

OBJECT REPRESENTATION, IDENTITY, AND THE PARADOX OF EARLY PERMANENCE: Steps Toward a New Framework

The sensorimotor theory of infancy has been overthrown, but there is little consensus on a replacement. We hypothesize that a capacity for representation is the starting point for infant development, not its culmination. Logical distinctions are drawn between object representation, identity, and permanence. Modern experiments on early object permanence and deferred imitation suggest: (a) even for young infants, representations persist over breaks in sensory contact, (b) numerical identity of objects (Os) is initially specified by spatiotemporal criteria (place and trajectory), (c) featural and functional identity criteria develop, (d) events are analyzed by comparing representations to current perception, and (e) representation operates both prospectively, anticipating future contacts with an O, and retrospectively, reidentifying an O as the "same one again." A model of the architecture and functioning of the early representational system is proposed. It accounts for young infants' behavior toward absent people and things in terms of their efforts to determine the identity of objects. Our proposal is developmental without denying innate structure and elevates the power of perception and representation while being cautious about attributing complex concepts to young infants.