Embodied Identification of Objects

In this article by Smith and Gasser, one of the lessons concerning the development of embodied cognition that we can learn from babies is how language provides us with the opportunity for abstract thought.  Smith and Gasser believe language may be the basis for all symbolic reasoning, including mathematics.  It is argued this happens in four steps.  In the first step, the child learns to associate individual words with specific objects which they identify by shape.  The second step is when the child can recognize similarities between objects within the same category – again, distinguished by shape.  For example, this object is round like that ‘ball’ over there, so this must also be a ‘ball’.  This leads to the third step, or ‘second order generalization’ where the child realizes any novel object may belong to a category which contains similarly-shaped objects.  Finally, in the fourth step, the child learns to attend to the shape of an object in order to learn its name. 

It is interesting that Smith and Gasser’s account of language includes very little about the body.  In fact, despite including a previous section dedicated to multi-modality, Smith and Gasser’s argument seems to be that babies primarily learn language through visually recognizing shapes and associating them with sounds.  However, if we are to take previous sections or ‘lessons’ in the article as true (Be multi-modal, Be incremental, Be physical, Explore, and Be social), then in the first step the baby would not only identify a specific object by its shape.  By the time the child is physically and perceptually developed enough to understand language, he or she is already experiencing the world as an active and engaged, multi-modal subject.  A ball would not be identified through shape alone (which, it could be argued, is already a geometric or mathematic concept rather than the factor in identification which underpins language development which, in turn, is said to underpin mathematic concepts) but also colour, texture, taste, smell, etc.

However, I think it can alternatively be argued that objects are not necessarily identified through the collection of sensory ‘input’ they afford, but instead through their use or what they do.  Smith and Gasser claim on p. 6 “…visual object recognition appears to automatically activate the actions associated with the object.” Is it not possible this suggests a ball is identified as something that rolls when pushed across a surface or something that bounces rather than something which is ‘round’?  Could a cup be identified as an object which contains liquid from which people drink rather than something cup-shaped?  If this is correct, then babies would generalize across ‘the actions associated with the object’ – not merely shapes.

On p. 22 Smith and Gasser state, “animate categories are organized by many different kinds of similarities across many modalities; artefact categories are organized by shape; and substance categories by material.”  Why would artefacts and substances, unlike so-called ‘animate categories’ be organized by only one modality?  If artefact categories are organized by shape alone, how would a child learn to recognize differently-shaped balls as balls, for example?  If substance categories are organized by material alone, why would a baby be capable of identifying and naming toy, plastic foods and/or drinks?  Furthermore, if “animate categories are organized by many different kinds of similarities across many modalities”, how could a child recognize a toy dog as a ‘dog’ when presented less than the full range of modalities through which the child comes to identify a ‘dog’?  I believe the argument that babies learn to identify everything through multiple modalities based in experiences of interaction seems more in line with the previous claims made in the article.

Reference:

Smith, L. and Gasser, M. (2005) “The Development of embodied  cognition: Six lessons from babies", Artificial Life, 11 (1-2), pp. 13-29.