This paper addresses the role of imitation as a means to enhance the learning of communication skills in autonomous robots. A series of robotic experiments are presented in which autonomous mobile robots are taught a synthetic proto-language. Learning of the language occurs through an imitative scenario where the robot replicates the teacher's movements. Imitation is here an implicit attentional mechanism which allows the robot imitator to share a similar set of proprio- and exteroceptions with the teacher. The robot grounds its understanding of the teacher's words, which describe the teacher's current observations, upon its own perceptions which are similar to those of the teacher. Learning of the robot is based on a dynamical recurrent associative memory architecture (DRAMA). Learning is unsupervised and results from the self-organization of the robot's connectionist architecture. Results show that the imitative behavior greatly improves the efficiency and speed of the learning. More...

Abstract not found

... this paper demonstrates scaling up of this movement imitative strategy for transmitting a vocabulary across a group of robotic agents, i.e. from a teacher agent to several learner agents. In particular, it shows that imitative behaviour is necessary for the grounding of the agents' proprioceptions and speeds up the grounding of exteroceptions. These studies stress the importance of behavioural social mechanisms in addition to general cognitive abilities of associativity for grounding communication in embodied agents. In particular, it shows that a simple movement imitation strategy is an interesting scenario for the transmission of a language, as it is an easy means of getting the agents to share a common context of perceptions, which is a prerequisite for a common understanding of the language to develop. It is thus suggested that a behaviour -oriented approach might be more appropriate than a pure cognitivist one which is dominating in related studies of the mechanisms involved in grounding communication.

Previous research has shown that nave participants display a high level of agreement when asked to choose or draw schematic representations, or image schemas, of concrete and abstract verbs (Richardson, Spivey, Edelman & Naples, 2001). For example, participants tended to ascribe a horizontal image schema to push, and a vertical image schema to respect. This consistency in offline data is preliminary evidence that language invokes spatial forms of representation. It also provided norms that were used in the present research to investigate the activation of spatial image schemas during online language comprehension. We predicted that if comprehending a verb activates a spatial representation that is extended along a particular horizontal or vertical axis, it will affect other forms of spatial processing along that axis. Participants listened to short sentences while engaged in a visual discrimination task (Experiment 1) and a picture memory task (Experiment 2). In both cases, reaction times showed an interaction between the horizontal/vertical nature of the verb's image schema, and the horizontal/vertical position of the visual stimuli. We argue that such spatial effects of verb comprehension provide evidence for the perceptual-motor character of linguistic representations. 1.

Human cognition is unique in the way in which it relies on combinatorial (or compositional) structures. Language provides ample evidence for the existence of combinatorial structures, but they can also be found in visual cognition. To understand the neural basis of human cognition, it is therefore essential to understand how combinatorial structures can be instantiated in neural terms. In his recent book on the foundations of language, Jackendoff formulated four fundamental problems for a neural instantiation of combinatorial structures: the massiveness of the binding problem, the problem of 2, the problem of variables and the transformation of combinatorial structures from working memory to long-term memory. This paper aims to show that these problems can be solved by means of neural ‘blackboard ’ architectures. For this purpose, a neural blackboard architecture for sentence structure is presented. In this architecture, neural structures that encode for words are temporarily bound in a manner that preserves the structure of the sentence. It is shown that the architecture solves the four problems presented by Jackendoff. The ability of the architecture to instantiate sentence structures is illustrated with examples of sentence complexity observed in human language performance. Similarities exist between the architecture for sentence structure and blackboard architectures for combinatorial structures in visual cognition, derived from the structure of the visual cortex. These architectures are briefly discussed, together with an example of a combinatorial structure in which the blackboard architectures for language and vision are combined. In this way, the architecture for language is grounded in perception. 2 Content

According to perceptual symbol systems (Barsalou, 1999), sensory-motor simulations underlie the representation of concepts. It follows that sensory-motor phenomena should arise in conceptual processing. Previous studies have shown that switching from one modality to another during perceptual processing incurs a processing cost. If perceptual simulation underlies conceptual processing, then verifying the properties of concepts should exhibit a switching cost as well. For example, verifying a property in the auditory modality (e.g., BLENDER-loud) should be slower after verifying a property in a different modality (e.g., CRANBERRIES-tart) than in the same modality (e.g., LEAVES-rustling). Only words were presented to subjects, and there were no instructions to use imagery. Nevertheless switching modalities incurred a cost, analogous to switching modalities in perception. A second experiment showed that this effect was not due to associative priming between properties in the same modality. These results support the hypothesis that perceptual simulation underlies conceptual processing. Modern psychology relies heavily on the digital computer as a metaphor for human cognition (e.g., Fodor, 1975; Pylyshyn, 1984). According to this view, the software of the mind can be distinguished from the hardware of the body, with mental representations being amodal redescriptions of sensory-motor experience. Increasingly, however, researchers argue that this approach is fundamentally wrong, suggesting instead that interactions between sensory-motor systems and the physical world underlie cognition. For example, Barsalou's (1999) theory of perceptual symbol systems proposes that conceptual knowledge is grounded in sensory-motor systems. To represent a concept, neural systems partially run a...

According to perceptual symbol systems, sensorimotor simulations underlie the representation of concepts. It follows that sensorimotor phenomena should arise in conceptual processing. Previous studies have shown that switching from one modality to another during perceptual processing incurs a processing cost. If perceptual simulation underlies conceptual processing, then verifying the properties of concepts should exhibit a switching cost as well. For example, verifying a property in the auditory modality (e.g., BLENDER-loud) should be slower after verifying a property in a different modality (e.g., CRANBERRIES-tart) than after verifying a property in the same modality (e.g., LEAVES-rustling). Only words were presented to subjects, and there were no instructions to use imagery. Nevertheless, switching modalities incurred a cost, analogous to the cost of switching modalities in perception. A second experiment showed that this effect was not due to associative priming between properties in the same modality. These results support the hypothesis that perceptual simulation underlies conceptual processing.

The authors identify 2 major types of statistical data from which semantic representations can be learned. These are denoted as experiential data and distributional data. Experiential data are derived by way of experience with the physical world and comprise the sensory-motor data obtained through sense receptors. Distributional data, by contrast, describe the statistical distribution of words across spoken and written language. The authors claim that experiential and distributional data represent distinct data types and that each is a nontrivial source of semantic information. Their theoretical proposal is that human semantic representations are derived from an optimal statistical combination of these 2 data types. Using a Bayesian probabilistic model, they demonstrate how word meanings can be learned by treating experiential and distributional data as a single joint distribution and learning the statistical structure that underlies it. The semantic representations that are learned in this manner are measurably more realistic—as verified by comparison to a set of human-based measures of semantic representation—than those available from either data type individually or from both sources independently. This is not a result of merely using quantitatively more data, but rather it is because experiential and distributional data are qualitatively distinct, yet intercorrelated, types of data. The semantic representations that are learned are based on statistical structures that exist both within and between the experiential and distributional data types.

& A recently emerging view sees language understanding as closely linked to sensory and motor processes. The present study investigates this issue by examining the influence of processing action verbs and concrete nouns on the execution of a reaching movement. Fine-grained analyses of movement kinematics revealed that relative to nouns, processing action verbs significantly affects overt motor performance. Within 200 msec after onset, processing action verbs interferes with a concurrent reaching movement. By contrast, the same words assist reaching movement when processed before movement onset. The cross-talk between language processes and overt motor behavior provides unambiguous evidence that action words and motor action share common cortical representations and could thus suggest that cortical motor regions are indeed involved in action word retrieval. &

minicolumns, forgetting, incremental learning, reaction time Summary: During the last few decades we have seen a convergence among ideas and hypotheses regarding functional principles underlying human memory. Hebb’s now more than fifty years old conjecture concerning synaptic plasticity and cell assemblies, formalized mathematically as attractor neural networks, has remained among the most viable and productive theoretical frameworks. It suggests plausible explanations for Gestalt aspects of active memory like perceptual completion, reconstruction and rivalry. We review the biological plausibility of these theories and discuss some critical issues concerning their associative memory functionality in the light of simulation studies of models with palimpsest memory properties. The focus is on memory properties and dynamics of networks modularized in terms of cortical minicolumns and hypercolumns. Biophysical compartmental models demonstrate attractor dynamics that support cell assembly operations with fast convergence and low firing rates. Using a scaling model we obtain reasonable relative connection densities and amplitudes. An abstract attractor network model reproduces systems level psychological phenomena seen in human memory experiments as the Sternberg and von Restorff effects. We conclude that there is today considerable substance in Hebb's theory of cell assemblies and its attractor network formulations, and that they have contributed to increasing our understanding of cortical associative memory function. The criticism raised with regard to biological and psychological plausibility as well as low storage capacity, slow retrieval etc has largely been disproved. Rather, this paradigm has gained further support from new experimental data as well as computational modeling.