Article
To read the full-text of this research, you can request a copy directly from the authors.

Abstract

It is well-established that toddlers can correctly select a novel referent from an ambiguous array in response to a novel label. There is also a growing consensus that robust word learning requires repeated label-object encounters. However, the effect of the context in which a novel object is encountered is less wellunderstood. We present two embodied neural network replications of recent empirical tasks, which demonstrated that the context in which a target object is encountered is fundamental to referent selection and word learning. Our model offers an explicit account of the bottom-up associative and embodied mechanisms which could support children's early word learning and emphasises the importance of viewing behaviour as the interaction of learning at multiple timescales.

No full-text available

Request Full-text Paper PDF

To read the full-text of this research,
you can request a copy directly from the authors.

... An extended version of this model has already been used to replicate a range of other language acquisition phenomena (Morse and Cangelosi 2017;Cangelosi and Schlesinger 2018). For example, Twomey et al. (2016) used the ERA architecture to model mutual exclusivity-that is, the developmental phenomenon in which a child can learn the name of a new object if they hear a new label and are presented with an unseen (unlabeled) object among other objects with a known label. Other developmental language models have looked at the learning of both object and action labels, moving toward the first examples of syntax learning. ...
Chapter
Full-text available
The current state of the art in cognitive robotics, covering the challenges of building AI-powered intelligent robots inspired by natural cognitive systems. A novel approach to building AI-powered intelligent robots takes inspiration from the way natural cognitive systems—in humans, animals, and biological systems—develop intelligence by exploiting the full power of interactions between body and brain, the physical and social environment in which they live, and phylogenetic, developmental, and learning dynamics. This volume reports on the current state of the art in cognitive robotics, offering the first comprehensive coverage of building robots inspired by natural cognitive systems. Contributors first provide a systematic definition of cognitive robotics and a history of developments in the field. They describe in detail five main approaches: developmental, neuro, evolutionary, swarm, and soft robotics. They go on to consider methodologies and concepts, treating topics that include commonly used cognitive robotics platforms and robot simulators, biomimetic skin as an example of a hardware-based approach, machine-learning methods, and cognitive architecture. Finally, they cover the behavioral and cognitive capabilities of a variety of models, experiments, and applications, looking at issues that range from intrinsic motivation and perception to robot consciousness. Cognitive Robotics is aimed at an interdisciplinary audience, balancing technical details and examples for the computational reader with theoretical and experimental findings for the empirical scientist.
... Also, normative Bayesian models become computationally intractable if one aims to scale to real world high-dimensional data (Bossaerts & Murawksi, 2017). For these reasons, another very large family of models relies on heuristic models of learning, ranging from connectionist approaches Mareschal, 2014, Cangelosi andSchlesinger, 2015;Twomey et al., 2016) to heuristic statistical learning (Mangin et al., 2015) or symbolic learning (Mealier et al., 2017;Spranger and Steels, 2012). The advantage of these models is their very large expressivity, their capacity to address the problem of representation learning (especially in connectionist approaches), and their capacity to combine different kinds of learning mechanisms in the same model. ...
... Also, normative Bayesian models become computationally intractable if one aims to scale to real world high-dimensional data (Bossaerts & Murawksi, 2017). For these reasons, another very large family of models relies on heuristic models of learning, ranging from connectionist approaches Mareschal, 2014, Cangelosi andSchlesinger, 2015;Twomey et al., 2016) to heuristic statistical learning (Mangin et al., 2015) or symbolic learning (Mealier et al., 2017;Spranger and Steels, 2012). The advantage of these models is their very large expressivity, their capacity to address the problem of representation learning (especially in connectionist approaches), and their capacity to combine different kinds of learning mechanisms in the same model. ...
Preprint
Full-text available
We review computational and robotics models of early language learning and development. We first explain why and how these models are used to understand better how children learn language. We argue that they provide concrete theories of language learning as a complex dynamic system, complementing traditional methods in psychology and linguistics. We review different modeling formalisms, grounded in techniques from machine learning and artificial intelligence such as Bayesian and neural network approaches. We then discuss their role in understanding several key mechanisms of language development: cross-situational statistical learning, embodiment, situated social interaction, intrinsically motivated learning, and cultural evolution. We conclude by discussing future challenges for research, including modeling of large-scale empirical data about language acquisition in real-world environments. Keywords: Early language learning, Computational and robotic models, machine learning, development, embodiment, social interaction, intrinsic motivation, self-organization, dynamical systems, complexity.
... Another hypothesis regarding the lexical acquisition by an infant was mutual exclusivity bias (constraint) (Markman and Wachtel, 1988). In studies on lexical acquisition, this hypothesis was considered to be particularly important for CSL (Twomey et al., 2016). Mutual exclusivity bias assumes that the infant considers the name of an object to correspond to one particular category only. ...
Article
Full-text available
In this paper, we propose a Bayesian generative model that can form multiple categories based on each sensory-channel and can associate words with any of four sensory-channels (action, position, object, and color). This paper focuses on cross-situational learning using the co-occurrence between words and information of sensory-channels in complex situations rather than conventional situations of cross-situational learning. We conducted a learning scenario using a simulator and a real humanoid iCub robot. In the scenario, a human tutor provided a sentence that describes an object of visual attention and an accompanying action to the robot. The scenario was set as follows: the number of words per sensory-channel was three or four, and the number of trials for learning was 20 and 40 for the simulator and 25 and 40 for the real robot. The experimental results showed that the proposed method was able to estimate the multiple categorizations and to learn the relationships between multiple sensory-channels and words accurately. In addition, we conducted an action generation task and an action description task based on word meanings learned in the cross-situational learning scenario. The experimental results showed the robot could successfully use the word meanings learned by using the proposed method.
... Recent years have seen the emergence of a new view of word learning as a low-level phenomenon that can proceed without invoking complex, metalinguistic awareness. These theories, often based on computational models, argue that low level associative processes of excitation and inhibition can give rise to the apparently complex behaviors children demonstrate during referent selection (Horst, Samuelson, Kucker, & McMurray, 2011;McMurray, Horst, & Samuelson, 2012;Samuelson, Kucker, & Spencer, 2016;Samuelson, Smith, Perry, & Spencer, 2011;Smith, 2000;Twomey, Morse, Cangelosi, & Horst, 2016). ...
Article
Full-text available
Variability is prevalent in early language acquisition, however whether it supports or hinders learning is unclear: while target variability has been shown to facilitate word learning, variability in competitor items has been shown to make the task harder. Here we tested whether background variability could boost learning in a referent selection task. Two groups of two-year-old children saw arrays of one novel and two known objects on a screen, and heard a novel or known label. Stimuli were identical across conditions, with the exception that in the constant color condition objects appeared on a uniform white background, and in the variable color condition backgrounds were different, uniform colors. At test, only children in the variable condition showed evidence of retaining label-object associations. These data support findings from the adult memory literature, which suggest that variability supports learning by decontextualizing representations. We argue that these data are consistent with dynamic systems accounts of learning in which low-level entropy adds sufficient noise to the developmental system to precipitate a change in behavior.
Article
Children encounter new words across variable and noisy contexts. This variability may affect word learning, but the literature includes discrepant findings. The current experiment investigated one source of these discrepant findings: whether contexts with familiar, nameable objects are associated with less robust label learning. Two year olds were exposed to word‐object pairings on variable contexts that either included nameable objects or did not. Target selection was more robust when exposure occurred without other nameable objects. The difference was present immediately, but not after a delay. This study provides the evidence that context effects are context‐bound.
Article
Learning words from ambiguous naming events is difficult. In such situations, children struggle with not attending to task irrelevant information when learning object names. The current study reduces the problem space of learning names for object categories by holding color constant between the target and other extraneous objects. We examine how this influences two types of word learning (retention and generalization) in both 30-month-old children (Experiment 1) and the iCub humanoid robot (Experiment 2). Overall, all children and iCub performed well on the retention trials, but they were only able to generalize the novel names to new exemplars of the target categories if the objects were originally encountered in sets with objects of the same colors, not if the objects were originally encountered in sets with objects of different colors. These data demonstrate that less information presented during the learning phase narrows the problem space and leads to better word learning success for both children and iCub. Findings are discussed in terms of cognitive load and desirable difficulties.
Article
The latest developments in artificial intelligence (AI) and machine learning, and the parallel advances in robotics, have contributed recently to a shift in the scientific approach to modeling human intelligence. These innovations, accompanied by the new emphasis on embodied and grounded cognition in AI and psychology, have led to the establishment of the field of developmental robotics. This field features an interdisciplinary approach, built on collaboration between cognitive robotics and child psychology, to the autonomous design of behavioral and cognitive capabilities in artificial cognitive agents, such as robots, which is inspired by developmental principles and mechanisms observed in children. In this article, we illustrate the benefits of this approach by presenting a case study of a baby robot with a focus on the role of embodiment during early word learning, as well as an overview of several developmental robotics model of perceptual, social, and language development.
Conference Paper
Full-text available
Determining in advance all objects that a robot will interact with in an open environment is very challenging, if not impossible. It makes difficult the development of models that will allow to perceive and recognize objects, to interact with them and to predict how these objects will react to interactions with other objects or with the robot. Developmental robotics proposes to make robots learn by themselves such models through a dedicated exploration step. It raises a chicken-and-egg problem: the robot needs to learn about objects to discover how to interact with them and, to this end, it needs to interact with them. In this work, we propose Novelty-driven Evolutionary Babbling (NovEB), an approach enabling to bootstrap this process and to acquire knowledge about objects in the surrounding environment without requiring to include a priori knowledge about the environment, including objects, or about the means to interact with them. Our approach consists in using an evolutionary algorithm driven by a novelty criterion defined in the raw sensorimotor flow: behaviours, described by a trajectory of the robot end effector, are generated with the goal to maximize the novelty of raw perceptions. The approach is tested on a simulated PR2 robot and is compared to a random motor babbling.
Article
Full-text available
For infants, the first problem in learning a word is to map the word to its referent; a second problem is to remember that mapping when the word and/or referent are again encountered. Recent infant studies suggest that spatial location plays a key role in how infants solve both problems. Here we provide a new theoretical model and new empirical evidence on how the body - and its momentary posture - may be central to these processes. The present study uses a name-object mapping task in which names are either encountered in the absence of their target (experiments 1-3, 6 & 7), or when their target is present but in a location previously associated with a foil (experiments 4, 5, 8 & 9). A humanoid robot model (experiments 1-5) is used to instantiate and test the hypothesis that body-centric spatial location, and thus the bodies' momentary posture, is used to centrally bind the multimodal features of heard names and visual objects. The robot model is shown to replicate existing infant data and then to generate novel predictions, which are tested in new infant studies (experiments 6-9). Despite spatial location being task-irrelevant in this second set of experiments, infants use body-centric spatial contingency over temporal contingency to map the name to object. Both infants and the robot remember the name-object mapping even in new spatial locations. However, the robot model shows how this memory can emerge -not from separating bodily information from the word-object mapping as proposed in previous models of the role of space in word-object mapping - but through the body's momentary disposition in space.
Article
Full-text available
Children develop in a real, messy world in which learning unfolds through time and space shared with others. Understanding how children develop in this complex environment will require a solid, theoretically-grounded understanding of how the child and environment interact-both within and beyond the laboratory. We as researchers understand the scientific value in testing children in carefully-controlled environments, but for our findings to have any impact on children's lives we must strive to understand how the processes we study in the lab operate in the real, busy environments in which children interact with peers and adults.
Article
Full-text available
From at least two months onwards, infants can form perceptual categories. During the first year of life, object knowledge develops from the ability to represent individual object features to representing correlations between attributes and to integrate information from different sources. At the end of the first year, these representations are shaped by labels, opening the way to conceptual knowledge. Here, we review the development of object knowledge and object categorization over the first year of life. We then present an artificial neural network model that models the transition from early perceptual categorization to categories mediated by labels. The model informs a current debate on the role of labels in object categorization by suggesting that although labels do not act as object features they nevertheless affect perceived similarity of perceptually distinct objects sharing the same label. The model presents the first step of an integrated account from early perceptual categorization to language-based concept learning.
Article
Full-text available
A critical question about the nature of human learning is whether it is an all-or-none or a gradual, accumulative process. Associative and statistical theories of word learning rely critically on the later assumption: that the process of learning a word's meaning unfolds over time. That is, learning the correct referent for a word involves the accumulation of partial knowledge across multiple instances. Some theories also make an even stronger claim: Partial knowledge of one word-object mapping can speed up the acquisition of other word-object mappings. We present three experiments that test and verify these claims by exposing learners to two consecutive blocks of cross-situational learning, in which half of the words and objects in the second block were those that participants failed to learn in Block 1. In line with an accumulative account, Re-exposure to these mis-mapped items accelerated the acquisition of both previously experienced mappings and wholly new word-object mappings. But how does partial knowledge of some words speed the acquisition of others? We consider two hypotheses. First, partial knowledge of a word could reduce the amount of information required for it to reach threshold, and the supra-threshold mapping could subsequently aid in the acquisition of new mappings. Alternatively, partial knowledge of a word's meaning could be useful for disambiguating the meanings of other words even before the threshold of learning is reached. We construct and compare computational models embodying each of these hypotheses and show that the latter provides a better explanation of the empirical data.
Article
Full-text available
The intermodal preferential looking paradigm (IPLP) has proven to be a revolutionary method for the examination of infants' emerging language knowledge. In the IPLP, infants' language comprehension is measured by their differential visual fixation to two images presented side-by-side when only one of the images matches an accompanying linguistic stimulus. Researchers can examine burgeoning knowledge in the areas of phonology, semantics, syntax, and morphology in infants not yet speaking. The IPLP enables the exploration of the underlying mechanisms involved in language learning and illuminates how infants identify the correspondences between language and referents in the world. It has also fostered the study of infants' conceptions of the dynamic events that language will express. Exemplifying translational science, the IPLP is now being investigated for its clinical and diagnostic value. © The Author(s) 2013.
Conference Paper
Full-text available
Digital control of color television monitors—in particular, via frame buffers—has added precise control of a large subset of human colorspace to the capabilities of computer graphics. This subset is the gamut of colors spanned by the red, green, and blue (RGB) electron guns exciting their respective phosphors. It is called the RGB monitor gamut. Full-blown color theory is a quite complex subject involving physics, psychology, and physiology, but restriction to the RGB monitor gamut simplifies matters substantially. It is linear, for example, and admits to familiar spatial representations. This paper presents a set of alternative models of the RGB monitor gamut based on the perceptual variables hue (H), saturation (S), and value (V) or brightness (L). Algorithms for transforming between these models are derived. Particular emphasis is placed on an RGB to HSV non-trigonometric pair of transforms which have been used successfully for about four years in frame buffer painting programs. These are fast, accurate, and adequate in many applications. Computationally more difficult transform pairs are sometimes necessary, however. Guidelines for choosing among the models are provided. Psychophysical corrections are described within the context of the definitions established by the NTSC (National Television Standards Committee).
Article
Full-text available
Classic approaches to word learning emphasize referential ambiguity: In naming situations, a novel word could refer to many possible objects, properties, actions, and so forth. To solve this, researchers have posited constraints, and inference strategies, but assume that determining the referent of a novel word is isomorphic to learning. We present an alternative in which referent selection is an online process and independent of long-term learning. We illustrate this theoretical approach with a dynamic associative model in which referent selection emerges from real-time competition between referents and learning is associative (Hebbian). This model accounts for a range of findings including the differences in expressive and receptive vocabulary, cross-situational learning under high degrees of ambiguity, accelerating (vocabulary explosion) and decelerating (power law) learning, fast mapping by mutual exclusivity (and differences in bilinguals), improvements in familiar word recognition with development, and correlations between speed of processing and learning. Together it suggests that (a) association learning buttressed by dynamic competition can account for much of the literature; (b) familiar word recognition is subserved by the same processes that identify the referents of novel words (fast mapping); (c) online competition may allow the children to leverage information available in the task to augment performance despite slow learning; (d) in complex systems, associative learning is highly multifaceted; and (e) learning and referent selection, though logically distinct, can be subtly related. It suggests more sophisticated ways of describing the interaction between situation- and developmental-time processes and points to the need for considering such interactions as a primary determinant of development. (PsycINFO Database Record (c) 2012 APA, all rights reserved).
Article
Full-text available
In 2 experiments, adults and children were tested in an object-selection task that examined whether Ss would (1) map a novel word onto a previously unnamed object and (2) extend the newly learned word to another exemplar. Exp 3 was a control study. Ss overwhelmingly selected the novel object as the referent for the novel term, even though the new label was never explicitly linked to the novel object. Ss also extended the new term and allowed it to preempt yet another novel label from applying to the just-named object. The existence of several lexical principles and the power of indirect word learning is supported. (PsycINFO Database Record (c) 2012 APA, all rights reserved)
Article
Full-text available
The acquisition of word meaning is often partially attributed to fast mapping. However, recent research suggests that fast mapping and word learning may represent distinct components of language acquisition. Here we examine their interaction with a Hebbian Normalized Recurrence Network, a connectionist architecture that captures both online processing and long-term statistical learning. After training on a small lexicon, the model performed well above chance on a fast mapping task. Careful analyses of the weight changes, however, suggest that the fast mapping task can be solved with minimal learning. Thus, this model not only captures both long-term learning and online processes, but also provides unique insights regarding the relationship between fast mapping and word learning and that the two should be carefully distinguished.
Article
Full-text available
Building intelligent systems with human level competence is the ultimate grand challenge for science and technology in general, and especially for cognitive developmental robotics. This paper proposes a new approach to the design of cognitive skills in a robot able to interact with, and communicate about, the surrounding physical world and manipulate objects in an adaptive manner. The work is based on robotic simulation experiments showing that a humanoid robot (iCub platform) is able to acquire behavioral, cognitive, and linguistic skills through individual and social learning. The robot is able to learn to handle and manipulate objects autonomously, to understand basic instructions, and to adapt its abilities to changes in internal and environmental conditions.
Article
Full-text available
In this paper, we discuss the requirements of cognitive architectures for epigenetic robotics, and highlight the wider role that they can play in the development of the cognitive sciences. We discuss the ambitious goals of ongoing development, scalability, concept use and transparency, and introduce the epigenetic robotics architecture (ERA) as a framework guiding modeling efforts. A formal implementation is provided, demonstrated, and discussed in terms of meeting these goals. Extensions of the architecture are also introduced and we show how the dynamics of resulting models can transparently account for a wide range of psychological phenomena, without task dependant tuning, thereby making progress in all of the goal areas we highlight.
Article
Full-text available
The current study examines how focusing children's attention immediately after fast mapping improves their ability to retain novel names. Previous research suggests that young children can only retain novel names presented via referent selection if ostensive naming is provided and that such explicit naming works by increasing children's attention to the target and decreasing their attention to the competitor objects (Horst and Samuelson, 2008). This explanation of the function of ostensive naming after referent selection trials was tested by drawing 24-month-old children's attention to the target either by illuminating the target, covering the competitors, or both. A control group was given a social pragmatic cue (pointing). Children given social pragmatic cue support did not demonstrate retention. However, children demonstrated retention if the target object was illuminated, and also when it was illuminated and the competitors simultaneously dampened. This suggests that drawing children's attention to the target object in a manner that helps focus children's attention is critical for word learning via referent selection. Directing attention away from competitors while also directing attention toward a target also aids in the retention of novel words.
Article
Full-text available
Upon fast mapping, children rarely retain new words even over intervals as short as 5 min. In this study, we asked whether the memory process of encoding or consolidation is the bottleneck to retention. Forty-nine children, mean age 33 months, were exposed to eight 2- or-3-syllable nonce neighbors of words in their existing lexicons. Didactic training consisted of six exposures to each word in the context of its referent, an unfamiliar toy. Productions were elicited four times: immediately following the examiner's model, and at 1-min-, 5-min-, and multiday retention intervals. At the final two intervals, the examiner said the first syllable and provided a beat gesture highlighting target word length in syllables as a cue following any erred production. The children were highly accurate at immediate posttest. Accuracy fell sharply over the 1-min retention interval and again after an additional 5 min. Performance then stabilized such that the 5-min and multiday posttests yielded comparable performance. Given this time course, we conclude that it was not the post-encoding process of consolidation but the process of encoding itself that presented the primary bottleneck to retention. Patterns of errors and responses to cueing upon error suggested that word forms were particularly vulnerable to partial decay during the time course of encoding.
Article
Most theories of learning would predict a gradual acquisition and refinement of skills as learning progresses, and while some highlight exponential growth, this fails to explain why natural cognitive development typically progresses in stages. Models that do span multiple developmental stages typically have parameters to “switch” between stages. We argue that by taking an embodied view, the interaction between learning mechanisms, the resulting behavior of the agent, and the opportunities for learning that the environment provides can account for the stage-wise development of cognitive abilities. We summarize work relevant to this hypothesis and suggest two simple mechanisms that account for some developmental transitions: neural readiness focuses on changes in the neural substrate resulting from ongoing learning, and perceptual readiness focuses on the perceptual requirements for learning new tasks. Previous work has demonstrated these mechanisms in replications of a wide variety of infant language experiments, spanning multiple developmental stages. Here we piece this work together as a single model of ongoing learning with no parameter changes at all. The model, an instance of the Epigenetic Robotics Architecture (Morse et al 2010) embodied on the iCub humanoid robot, exhibits ongoing multi-stage development while learning pre-linguistic and then basic language skills.
Article
Infants' own activities create and actively select their learning experiences. Here we review recent models of embodied information seeking and curiosity-driven learning and show that these mechanisms have deep implications for development and evolution. We discuss how these mechanisms yield self-organized epigenesis with emergent ordered behavioral and cognitive developmental stages. We describe a robotic experiment that explored the hypothesis that progress in learning, in and for itself, generates intrinsic rewards: The robot learners probabilistically selected experiences according to their potential for reducing uncertainty. In these experiments, curiosity-driven learning led the robot learner to successively discover object affordances and vocal interaction with its peers. We explain how a learning curriculum adapted to the current constraints of the learning system automatically formed, constraining learning and shaping the developmental trajectory. The observed trajectories in the robot experiment share many properties with those in infant development, including a mixture of regularities and diversities in the developmental patterns. Finally, we argue that such emergent developmental structures can guide and constrain evolution, in particular with regard to the origins of language.
Article
Although the brain mechanisms are extremely complex and contain many details, the theory of their cognitive abilities must be based on much fewer general principles. Traditionally it has been held that "intelligent" information processing ensues from signal processing in adaptive network structures. A new aspect emphasized in this paper is that the degree of adaptation itself should be made controllable in a particular way, which leads to a broad category of new phenomena. A successful implementation of this principle is the Self-Organizing Map (SOM), which can be discussed on different levels of abstraction: as a physiological model, or a simplified numerical algorithm. Accordingly, the SOM may be regarded as a paradigm in brain theory, but the simplified algorithm can also be used for many "intelligent" tasks in real-world operations.
Book
Developmental robotics is a collaborative and interdisciplinary approach to robotics that is directly inspired by the developmental principles and mechanisms observed in children’s cognitive development. It builds on the idea that the robot, using a set of intrinsic developmental principles regulating the real-time interaction of its body, brain, and environment, can autonomously acquire an increasingly complex set of sensorimotor and mental capabilities. This volume, drawing on insights from psychology, computer science, linguistics, neuroscience, and robotics, offers the first comprehensive overview of a rapidly growing field. After providing some essential background information on robotics and developmental psychology, the book looks in detail at how developmental robotics models and experiments have attempted to realize a range of behavioral and cognitive capabilities. The examples in these chapters were chosen because of their direct correspondence with specific issues in child psychology research; each chapter begins with a concise and accessible overview of relevant empirical and theoretical findings in developmental psychology. The chapters cover intrinsic motivation and curiosity; motor development, examining both manipulation and locomotion; perceptual development, including face recognition and perception of space; social learning, emphasizing such phenomena as joint attention and cooperation; language, from phonetic babbling to syntactic processing; and abstract knowledge, including models of number learning and reasoning strategies. Boxed text offers technical and methodological details for both psychology and robotics experiments.
Article
Recent research suggests that infants tend to add words to their vocabulary that are semantically related to other known words, though it is not clear why this pattern emerges. In this paper, we explore whether infants leverage their existing vocabulary and semantic knowledge when interpreting novel label-object mappings in real time. We initially identified categorical domains for which individual 24-month-old infants have relatively higher and lower levels of knowledge, irrespective of overall vocabulary size. Next, we taught infants novel words in these higher and lower knowledge domains and then asked if their subsequent real-time recognition of these items varied as a function of their category knowledge. While our participants successfully acquired the novel label-object mappings in our task, there were important differences in the way infants recognized these words in real time. Namely, infants showed more robust recognition of high (vs. low) domain knowledge words. These findings suggest that dense semantic structure facilitates early word learning and real-time novel word recognition.
Article
In this article, we review literature on word learning and propose a theoretical account of how lexical knowledge and word use emerge and develop over time. We contend that the developing lexical system is built on processes that support children's in-the-moment word usage interacting with processes that create long-term learning. We argue for a new characterization of word learning in which simple mechanisms like association and competition, and the interaction between the two, guide children's selection of referents and word use in the moment. This in turn strengthens and refines the network of relationships in the lexicon, improving referent selection and use in future encounters with words. By integrating in-the-moment word use with long-term learning through simple domain-general mechanisms, this account highlights the dynamic nature of word learning and creates a broader framework for understanding language and cognitive development more generally.
Article
From soon after their first birthdays young children are able to make inferences from a communicator's referential act (e.g., pointing to a container) to her overall social goal for communication (e.g., to inform that a searched-for toy is inside; see Behne, Carpenter, & Tomasello, 2005; Behne, Liszkowski, Carpenter, & Tomasello, 2012). But in such cases the inferential distance between referential act and communicative intention is still fairly close, as both container and searched-for toy lie in the direction of the pointing gesture. In the current study we tested 18- and 26-month-old children in a situation in which referential act and communicative goal were more distant: In the midst of a game, the child needed a certain toy. The experimenter then held up a key (that they knew in common ground could be used to open a container) to the child ostensively. In two control conditions the experimenter either inadvertently moved the key and so drew the child's attention to it non-ostensively or else held up the key for her own inspection intentionally but non-communicatively. Children of both ages took only the ostensive showing of the key, not the accidental moving or the non-ostensive but intentional inspection of the key, as an indirect request to take the key and open the container to retrieve the toy inside. From soon after they start acquiring language young children thus are able to infer a communicator's social goal for communication not only from directly-referential acts, but from more indirect communicative acts as well. Copyright © 2014 Elsevier B.V. All rights reserved.
Article
Infants can see someone pointing to one of two buckets and infer that the toy they are seeking is hidden inside. Great apes do not succeed in this task, but, surprisingly, domestic dogs do. However, whether children and dogs understand these communicative acts in the same way is not yet known. To test this possibility, an experimenter did not point, look, or extend any part of her body towards either bucket, but instead lifted and shook one via a centrally pulled rope. She did this either intentionally or accidentally, and did or did not address her act to the subject using ostensive cues. Young 2-year-old children but not dogs understood the experimenter's act in intentional conditions. While ostensive pulling of the rope made no difference to children's success, it actually hindered dogs' performance. We conclude that while human children may be capable of inferring communicative intent from a wide variety actions, so long as these actions are performed intentionally, dogs are likely to be less flexible in this respect. Their understanding of communicative intention may be more dependent upon bodily markers of communicative intent, including gaze, orientation, extended limbs, and vocalizations. This may be because humans have come under selective pressure to develop skills for communicating with absent interlocutors – where bodily co-presence is not possible.
Article
Children overgeneralise verbs to ungrammatical structures early in acquisition, but retreat from these overgeneralisations as they learn semantic verb classes. In a large corpus of English locative utterances (e.g., the woman sprayed water onto the wall/wall with water), we found structural biases which changed over development and which could explain overgeneralisation behaviour. Children and adults had similar verb classes and a correspondence analysis suggested that lexical distributional regularities in the adult input could help to explain the acquisition of these classes. A connectionist model provided an explicit account of how structural biases could be learned over development and how these biases could be reduced by learning verb classes from distributional regularities. https://sites.google.com/site/sentenceproductionmodel/cv/twomey%2Cchang%2Cambridge%2Cinpress.pdf?attredirects=0&d=1
Article
Gaze control requires the coordination of movements of both eyes and head to fixate on a target. We present a biologically constrained architecture for gaze control and show how the relationships between the coupled sensorimotor systems can be learnt autonomously from scratch, allowing for adaptation as the system grows or changes. Infant studies suggest developmental learning strategies, which can be applied to sensorimotor learning in humanoid robots. We examine two strategies (sequential and synchronous) for the learning of eye and head coupled mappings, and give results from implementations on an iCub robot. The results show that the developmental approach can give fast, cumulative, on-line learning of coupled sensorimotor systems.
Article
Previous research indicates learning words facilitates categorisation. The current study explores how categorisation affects word learning. In the current study, we investigated whether learning about a category facilitates retention of newly learned words by presenting 2-year-old children with multiple referent selection trials to the same object category. In Experiment 1, children either encountered the same exemplar repeatedly or encountered multiple exemplars across trials. All children did very well on the initial task; however, only children who encountered multiple exemplars retained these mappings after a short delay. Experiment 2 replicated and extended this finding by exploring the effect of within-category variability on children's word retention. Children encountered either narrow or broad exemplars across trials. Again, all children did very well on the initial task; however, only children who encountered narrow exemplars retained mappings after a short delay. Overall, these data offer strong evidence that providing children with the opportunity to compare across exemplars during fast mapping facilitates retention.
Article
Computational models are tools for testing mechanistic theories of learning and development. Formal models allow us to instantiate theories of cognitive development in computer simulations. Model behavior can then be compared to real performance. Connectionist models, loosely based on neural information processing, have been successful in capturing a range of developmental phenomena, in particular on-line within-task category learning by young infants. Here we describe two new models. One demonstrates how age dependent changes in neural receptive field sizes can explain observed changes in on-line category learning between 3 and 10 months of age. The other aims to reconcile two conflicting views of infant categorization by focusing on the different task requirements of preferential looking and manual exploration studies. A dual-memory hypothesis posits that within-task category learning that drives looking time behaviors is based on a fast-learning memory system, whereas categorization based on background experience and assessed by paradigms requiring complex motor behavior relies on a second, slow-learning system. The models demonstrate how emphasizing the mechanistic causes of behaviors leads to discovery of deeper, more explanatory accounts of learning and development.
Article
When hearing a novel name, children tend to select a novel object rather than a familiar one, a bias known as disambiguation. Using online processing measures with 18-, 24-, and 30-month-olds, we investigate how the development of this bias relates to word learning. Children’s proportion of looking time to a novel object after hearing a novel name related to their success in retention of the novel word, and also to their vocabulary size. However, skill in disambiguation and retention of novel words developed gradually: 18-month-olds did not show a reliable preference for the novel object after labeling; 24-month-olds reliably looked at a novel object on Disambiguation trials but showed no evidence of retention; and 30-month-olds succeeded on Disambiguation trials and showed only fragile evidence of retention. We conclude that the ability to find the referent of a novel word in ambiguous contexts is a skill that improves from 18 to 30 months of age. Word learning is characterized as an incremental process that is related to – but not dependent on – the emergence of disambiguation biases.
Article
Although vocabulary acquisition requires children learn names for multiple things, many investigations of word learning mechanisms teach children the name for only one of the objects presented. This is problematic because it is unclear whether children's performance reflects recall of the correct name-object association or simply selection of the only object that was singled out by being the only object named. Children introduced to one novel name may perform at ceiling as they are not required to discriminate on the basis of the name per se, and appear to rapidly learn words following minimal exposure to a single word. We introduced children to four novel objects. For half the children, only one of the objects was named and for the other children, all four objects were named. Only children introduced to one word reliably selected the target object at test. This demonstration highlights the over-simplicity of one-word learning paradigms and the need for a shift in word learning paradigms where more than one word is taught to ensure children disambiguate objects on the basis of their names rather than their degree of salience.
Article
This paper views lexical acquisition as a problem of induction: Children must figure out the meaning of a given term, given the large number of possible meanings any term could have. If children had to consider, evaluate, and rule out an unlimited number of hypotheses about each word in order to figure out its meaning, learning word meanings would be hopeless. Children must, therefore, be limited in the kinds of hypotheses they consider as possible word meanings. This paper considers three possible constraints on word meanings: (1) The whole object assumption which leads children to interpret novel terms as labels for objects—not parts, substances, or other properties of objects; (2) The taxonomic assumption which leads children to consider labels as referring to objects of like kind, rather than to objects that are thematically related; and (3) The mutual exclusivity assumption which leads children to expect each object to have only one label. Some of the evidence for these constraints is reviewed.
Article
Does making an inference lead to better learning than being instructed directly? Two experiments evaluated preschoolers' ability to learn new words, comparing their memory for words learned via inference or instruction. On Inference trials, one familiar and one novel object was presented and children were asked to Point at the [object name (i.e., pizer)]. These trials required the child to infer that the novel label referred to the novel object and not to the familiar object. On Instruction trials, a novel object label directly referred to a novel object (e.g., This is a glark) and no familiar distracter object was shown. We found that although children looked longer at the novel target on Instruction trials, they showed poorer retention of the newly learned label compared to words learned on Inference trials. Hence, we found that inferential learning was superior to instruction. Relevance for optimal learning contexts and education are discussed.
Chapter
Children learn words with remarkable speed and flexibility. However, the cognitive basis of young children’s word learning is disputed. Further, although research demonstrates that children’s categories and category labels are interdependent, how children learn category labels is also a matter of debate. Recently, biologically plausible, computational simulations of children’s behavior in experimental tasks have investigated the cognitive processes that underlie learning. The ecological validity of such models has been successfully tested by deploying them in robotic systems (Morse, Belpaeme, Cangelosi, & Smith, 2010). The authors present a simulation of children’s behavior in a word learning task (Twomey & Horst, 2011) via an embodied system (iCub; Metta, et al., 2010), which points to associative learning and dynamic systems accounts of children’s categorization. Finally, the authors discuss the benefits of integrating computational and robotic approaches with developmental science for a deeper understanding of cognition.
Article
During the second year of life, infants develop a preference to attach novel labels to novel objects. This behavior is commonly known as “mutual exclusivity” (Markman, 198914. Markman , E. M. 1989 . Categorization and naming in children: Problems of induction. , Cambridge, MA : MIT Press . View all references). In an intermodal preferential looking experiment with 19.5- and 22.5-month-olds, stimulus repetition was critical for observing mutual exclusivity. On the first occasion that a novel label was presented with 1 familiar object and 1 novel object, looking behavior was unsystematic. However, on reexposure to the same stimuli, 22.5-month-olds looked preferentially at the novel object prior to the re-presentation of the novel label. These findings suggest a powerful memory mechanism for novel labels and objects, enabling mutual exclusivity to emerge across repeated exposures to potential referents.
Article
Four experiments explored the processes that bridge between referent selection and word learning. Twenty-four-month-old infants were presented with several novel names during a referent selection task that included both familiar and novel objects and tested for retention after a 5-min delay. The 5-min delay ensured that word learning was based on retrieval from long-term memory. Moreover, the relative familiarity of objects used during the retention test was explicitly controlled. Across experiments, infants were excellent at referent selection, but very poor at retention. Although the highly controlled retention test was clearly challenging, infants were able to demonstrate retention of the first 4 novel names presented in the session when referent selection was augmented with ostensive naming. These results suggest that fast mapping is robust for reference selection but might be more transient than previously reported for lexical retention. The relations between reference selection and retention are discussed in terms of competitive processes on 2 timescales: competition among objects on individual referent selection trials and competition among multiple novel name–object mappings made across an experimental session.
Article
Toddlers' acquisition of the Novel Name–Nameless Category (N3C) principle was examined to investigate the developmental lexical principles framework and the applicability of the specificity hypothesis to relations involving lexical principles. In Study 1, we assessed the ability of 32 children between the ages of 16 and 20 months to use the N3C principle (operationally defined as the ability to fast map). As predicted, only some of the children could fast map. This finding provided evidence for a crucial tenet of the developmental lexical principles framewor: Some lexical principles are not available at the start of language acquisition. Children who had acquired the N3C principle also had significantly larger vocabularies and were significantly more likely to demonstrate 2-category exhaustive sorting abilities than children who had not acquired the principle. The 2 groups of children did not differ in either age or object permanence abilities. The 16 children who could not fast map were followed longitudinally until they attained a vocabulary spurt; at that time, their ability to fast map was retested (Study 2). Results provided a longitudinal replication of the findings of Study 1. Implications of these findings for both the developmental lexical principles framework and the specificity hypothesis are discussed.
Article
I consider the role of cognitive modeling in cognitive science. Modeling, and the computers that enable it, are central to the field, but the role of modeling is often misunderstood. Models are not intended to capture fully the processes they attempt to elucidate. Rather, they are explorations of ideas about the nature of cognitive processes. In these explorations, simplification is essential—through simplification, the implications of the central ideas become more transparent. This is not to say that simplification has no downsides; it does, and these are discussed. I then consider several contemporary frameworks for cognitive modeling, stressing the idea that each framework is useful in its own particular ways. Increases in computer power (by a factor of about 4 million) since 1958 have enabled new modeling paradigms to emerge, but these also depend on new ways of thinking. Will new paradigms emerge again with the next 1,000-fold increase?
Article
This paper considers possible problems researchers might face when interpreting the results of studies that employ variants of the preference procedure. Infants show a tendency to shift their preference from familiar to novel stimuli with increasing exposure to the familiar stimulus, a behaviour that is exploited by the habituation paradigm. This change in attentional preference with exposure leads us to suggest that researchers interested in infants' pre-experimental or spontaneous preferences should beware of the potentially confounding effects of exposing infants to familiarization trials prior to employing the preference procedure. The notion that infant attentional preference is dynamic also calls into question the use of the direction of post-familiarization preference per se when interpreting the knowledge or strategies available to infants. We look into the results of a cross-modal word learning study to show how the interpretation of results may be difficult when infants exhibit a significant preference in an unexpected direction. As a possible solution to this problem we propose that significant preferences in both directions should be sought at multiple intervals over time. Copyright © 2004 John Wiley & Sons, Ltd.
Article
Recent research demonstrated that although twenty-four month-old infants do well on the initial pairing of a novel word and novel object in fast-mapping tasks, they are unable to retain the mapping after a five-minute delay. The current study examines the role of familiarity with the objects and words on infants' ability to bridge between the initial fast mapping of a name and object, and later retention in the service of slow mapping. Twenty-four-month-old infants were familiarized with either novel objects or novel names prior to the referent selection portion of a fast-mapping task. When familiarized with the novel objects, infants retained the novel mapping after a delay, but not when familiarized with the novel words. This suggests familiarity with the object versus the word form leads to differential encoding of the name-object link. We discuss the implications of this finding for subsequent slow mapping.
Article
Two-year-olds use the sentence structures verbs appear in--subcategorization frames--to guide verb learning. This is syntactic bootstrapping. This study probed the developmental origins of this ability. The structure-mapping account proposes that children begin with a bias toward one-to-one mapping between nouns in sentences and participant roles in events. This account predicts that subcategorization frames should guide very early verb learning, if the number of nouns in the sentences is informative. In 3 experiments, one hundred and thirty-six 21- and 19-month-olds assigned appropriately different interpretations to novel verbs in transitive ("He's gorping him!") versus intransitive sentences ("He's gorping!") differing in their number of nouns. Thus, subcategorization frames guide verb interpretation in very young children. These findings constrain theoretical proposals about mechanisms for syntactic bootstrapping.
Article
A major problem in language learning is to figure out the meaning of a word given the enormous number of possible meanings for any particular word. This problem is exacerbated for children because they often find thematic relations between objects to be more salient than the objects' taxonomic category. Yet most single nouns refer to object categories and not to thematic relations. How do children learn words referring to categories when they find thematic relations so salient? We propose that children limit the possible meanings of nouns to refer mainly to categorical relations. This hypothesis was tested in four studies. In each study, preschool children saw a series of target objects (e.g., dog), each followed by a thematic associate (e.g., bone) and a taxonomic associate (e.g., cat). When children were told to choose another object that was similar to the target (“See this? Find another one.”), they as usual often selected the thematic associate. In contrast, when the instructions included an unknown word for the target (“See this fep? Find another fep.”), children now preferred the taxonomic associate. This finding held up for 2- and 3-year-olds at the basic level of categorization, for 4- and 5-year-olds at the superordinate level of categorization, and 4- and 5-year-olds who were taught new taxonomic and new thematic relations for unfamiliar objects. In each case, children constrained the meaning of new nouns to refer mainly to categorical relations. By limiting the hypotheses that children need to consider, this constraint tremendously simplifies the problem of language learning.
Article
This paper views lexical acquisition as a problem of induction: Children must figure out the meaning of a given term, given the large number of possible meanings any term could have. If children had to consider, evaluate, and rule out an unlimited number of hypotheses about each word in order to figure out its meaning, learning word meanings would be hopeless. Children must, therefore, be limited in the kinds of hypotheses they consider as possible word meanings. This paper considers three possible constraints on word meanings: (1) The whole object assumption which leads children to interpret novel terms as labels for objects—not parts, substances, or other properties of objects; (2) The taxonomic assumption which leads children to consider labels as referring to objects of like kind, rather than to objects that are thematically related; and (3) The mutual exclusivity assumption which leads children to expect each object to have only one label. Some of the evidence for these constraints is reviewed.
Article
Previous studies have found that children can use social-pragmatic cues to determine “which one” of several objects or “which one’ of several actions an adult intends to indicate with a novel word. The current studies attempted to determine whether children can also use such cues to determine “what kind” of referent, object, or action, an adult intends to indicate. In the first study, 27-month-old children heard an adult use a nonce word in conjunction with a nameless object while it was engaged in a nameless action. The discourse situation leading into this naming event was manipulated so that in one condition the target action was the one new element in the discourse context at the time of the naming event, and in another condition the target object was the one new element. Results showed that children learned the new word for whichever element was new to the discourse context. The second study followed this same general method, but in this case children in one condition watched as an adult engaged in preparatory behaviors that indicated her desire that the child perform the action before she produced the novel word, whereas children in another condition saw no such preparation. Results showed that children who saw the action preparation learned the new word for the action, whereas children who saw no preparation learned the new word for the object. These two studies demonstrate the important role of social-pragmatic information in early word learning, and suggest that if there is a Whole Object assumption in early lexical acquisition, it is an assumption that may be very easily overridden.
Article
Very young children successfully acquire the vocabulary of their native language despite their limited information processing abilities. One partial explanation for children's success at the inductive problem word learning presents is that children are constrained in the kinds of hypotheses they consider as potential meanings of novel words. Three such constraints are discussed: (1) the whole-object assumption which leads children to infer that terms refer to objects as a whole rather than to their parts, substance, color, or other properties; (2) the taxonomic assumption which leads children to extend words to objects or entities of like kind; and (3) the mutual exclusivity assumption which leads children to avoid two labels for the same object. Recent evidence is reviewed suggesting that all three constraints are available to babies by the time of the naming explosion. Given the importance of word learning, children might be expected to recruit whatever sources of information they can to narrow down a word's meaning, including information provided by grammatical form class and the pragmatics of the situation. Word-learning constraints interact with these other sources of information but are also argued to be an especially useful source of information for children who have not yet mastered grammatical form class in that constraints should function as an entering wedge into language acquisition.