english only
Visuo-motor imitation
Introduction
Humans' capacity to imitate has been extensively investigated through a wide-range of behavioral and developmental studies. Yet despite the huge amount of phenomenological evidence gathered, we are still unable to relate this behavioral data to any specific neural substrate. Specifically, we address the principle of ideomotor compatibility , by which observing the movements of others influences the quality of one's own performance and develop two neural models which account for a set of related behavioral studies (Sauser et al., 2006).
|
Related Experimental Study The study in which we are interested here is a stimulus-response experiment designed to verify two hypotheses of the ideomotor theory (Brass et al., 2000). These two hypotheses are based on the neural correlate that neural circuits devoted to the recognition of movements performed by others are likely to be shared by the motor preparation circuits (Decety et al., 2003). a) When a subject is requested to respond to the motion of a demonstrator then (s)he would experience a motor facilitation . b) the facilitatory effect is greater when the movements of the demonstrator and subject were very similar (ideomotor compatible) than when they were of a different type (ideomotor incompatible). The experimental stimuli consisted of a combination of a finger-lifting movement (either index or middle finger) and of a spatial cue consisting of a cross painted on the corresponding or opposite fingernail (see Figure. 1). |
Figure 1: Stimuli as used by Brass et al. (2000). Congruent condition: a left (right) finger movement with a cross on the left (right) fingernail. Incongruent condition: a left (right) finger movement with a cross on the right (left) fingernail. |
Models
We propose two similar models to replicate the cited behavioral study. The two networks consist of three major parts: the perceptual, decisional and motor preparatory layers. (see Figure 2). In our models, perception is only considered in its final stage, in that we assume visual information to have already been processed by highly specialized circuits and represented in a manner relevant for the task. Nevertheless, active processes such as top-down modulation of cortical activity are involved in the perceptual level of our models. Then, as the main task of the proposed models is to perform a selection among different sources of stimulus, the perceptual information is then fed to a decisional process . And as soon as it is performed, the network transforms information from stimulus space to motor space by means of stimulus-response associations. Considering now motor preparation, it consists of three areas, coding respectively for a) the motor plans, b) the shared representation between movement observation and motor execution (the ideomotor region ), and c) the final motor selection. The motor plans are fed either directly or indirectly through the ideomotor area to the final motor selection layer which is waiting for the execution signal coming from the decisional process. Indeed, the key difference that distinguishes between our two models is that one assumes shortcuts in the decisional process mediated by a direct-mapping between movement observation and execution (direct-matching model), whereas the other is designed using an information bottleneck in the decisional process (single route model).
|
Figure 2: Architecture of the two models. (Left) Bottleneck-like model with a single selection mechanism. (Right) The model is endowed with a direct-mapping from movement observation to motor execution. |
|
Implementation The neural implementation of the two models is inspired by neurophysiological evidence suggesting a continuous representation of stimuli in feature maps (Cisek et al., 2005). In such neural populations, neurons generally respond to external stimuli with broad tuning curves of activity. Therefore we adopted the dynamic neural field approach (Erlhagen et al., 2002) which integrates the principles of continuous representations endowed with a metric, and can account for temporal dynamics of stimuli interactions. Indeed, the important dynamical properties of such networks include abilities for stimulus enhancement, for cooperative and competitive interactions within and across neural representations. |
Figure 3: (Left) Canonical equation of a neural field. (Right) Illustration of the selection dynamics (up) when two identical inputs are applied (down). |
|
Results and Discussion The behavioral experiments were simulated by the two models using the same initial conditions as those used in the behavioral study (Brass et al., 2000). The results of the two models are in good agreement with the original data. Since the reported simulation results are barely distinguishable, we devised a method for determining which model best reflects the information pathway in the brain. To achieve this, we once again took inspiration from the literature on stimulus-response compatibility and decided to confront our models with an incompatible stimulus-response mapping, i.e. the subjects should respond to a left (right) cue with a right (left) finger movement. Under this experimental paradigm, the models now exhibit clearly different behaviors (see Fig. 4). To conclude, we present two biologically inspired computational models capable of reproducing the experimental results obtained by Brass et al. (2000). These models are in line with other computational models addressing imitation mechanisms in both humans and monkeys such as Arbib et al. (2000), in that they all assume a shared representation between movement observation and action execution which is mediated by competitive interactions. |
Figure 4: Original behavioral data, results and prediction of the models under different experimental conditions. |
Acknowledgments This work was supported by the Swiss National Science Foundation, through grant no 620-066127 of the SNF Professorships program.
References
E. L. Sauser and Aude G. Billard, Parallel and Distributed Neural Models of the Ideomotor Principle: An Investigation of Imitative Cortical Pathways, Neural Networks, Vol. 19(3), Special Issue on Brain Mechanisms of Imitation Learning, 2006.
M. Brass, H. Bekkering, A. Wohlschlager and W. Prinz, Compatibility between observed and executed finger movements: comparing symbolic, spatial and imitative cues, Brain and Cognition, Vol. 44, pp. 124-143, 2000.
J. Decety and J. A. Sommerville, Shared representations between self and others: a social cognitive neuroscience view, Trends in Cognitive Science, Vol. 7, pp. 527-533, 2003.
P. Cisek and J. F. Kalaska, Neural correlates of reaching decision in dorsal premotor cortex: specification of multiple direction choices and final selection of action, Neuron, Vol. 45, pp. 801-814, 2005.
W. Erlhagen and G. Schoner, Dynamics field theory of movement preparation, Psychological Review, Vol. 109, pp. 545-572, 2002.
M. Arbib, A. Billard, M. Iacoboni and E. Oztop, Mirror neurons, imitation and (synthetic) brain imaging, Neural Networks, Vol. 13, pp. 953-973, 2000.
Selected publications
People involved in this project
