On the Birth and Growth of Concepts

Abstract

This article describes what the earliest concepts are like and presents a theory of the spatial primitives from which they are formed. The earliest concepts tend to be global, like animal and container, and it is hypothesized that they consist of simplified redescriptions of innately salient spatial information. These redescriptions become associated with sensory and other bodily experiences that are not themselves redescribed, but that enrich conceptual thought. The initial conceptual base becomes expanded through subdivision, sometimes aided by language that points up these divisions or suggests new spatial analyses, and by the analogical extension of spatially derived concepts to nonspatial domains. This formulation is contrasted with Fodor's (1998) metaphysical theory of concept formation.

Keywords:

• Concept Formation
• Infant Concepts
• Innate Primitives
• Perceptual Meaning Analysis
• Spatial Representations

Acknowledgement

This article is based on a paper given at the 2005 Cognitive Science Society annual meeting in a symposium entitled ‘Solutions to Fodor's Puzzle of Concept Acquisition’. I thank Adele Abrahamsen, William Bechtel, and Daniel Weiskopf for helpful comments on an earlier version and David Kirsh and George Mandler for interesting discussion on the issues.

Notes

Jean Mandler is a Distinguished Research Professor in the Department of Cognitive Science at the University of California, San Diego.

Notes

[1]  The term ‘basic-level’ has never been given a satisfactory definition, but is usually used to refer to the level at which familiar mammals, household objects, vehicles, and sometimes plants are commonly labeled—hence, dog, cup, car, tree; see Mandler (1998, 2004) for discussion. The belief that such concepts are the first to be formed has generated a number of mistaken conclusions about early concept formation, for example, Fodor (1980, p. 311).

[2]  These results are not merely due to differences in inter- and intra-domain perceptual similarity; e.g., 9-month-olds categorize birds and airplanes as different in spite of great similarity of the models used to test them (Mandler & McDonough, 1993). This finding can be contrasted with their failure to categorize dogs and fish as different, in spite of their different physical appearance.

[3]  I am not discussing ‘true’ or ‘expert’ definitions, rather the properties of classes that hold in our experience and that we think of as necessary, whether or not that is the case—in short, the concepts infants and other laymen think with. No reference to experts is needed to understand concepts.

[4]  Conceptual primitives should be considered foundational or developmental primitives rather than traditional semantic primitives (see Carey, 1982 and Mandler, 2004).

[5]  For infants, the relevant distinction is almost certainly animal versus nonanimal rather than animate versus inanimate. Including plants with animals in a common animate domain is a late, verbally taught acquisition (Carey, 1985), and may not occur at all.

[6]  Differentiating kinds within domains may occur in more than one way. Some subdivisions can be made by combining primitives (e.g., land or ‘surface’ animals and air or ‘up’ animals), which is the approach of the Rogers and McClelland model. However, other subdivisions (e.g., various mammal kinds) may take place by associating unanalyzed perceptual appearances with conceptual descriptions. This issue is discussed later.

[7]  Although plants sometimes exhibit motion, ±PATH seems more important in this case than ±MOTION.

[8]  Current data suggest that support is not as straightforward a concept as containment (Casasola & Cohen, 2002; Choi, 2006), although this might be because attachment and support have not been separately studied.

[9]  Indoor things, outdoor things, taking in, and behind are given question marks because the evidence collected so far is indirect.

[10]  And of course, language can be used to teach metric and adventitious spatial relations such as ‘six feet tall’ or ‘slanted backward’.

[11]  Some of the apparent synonyms in English may be due to the two different linguistic streams—Romance and Germanic—that formed the modern language.

[12]  Of course, they also learn that words refer to actions, and there is evidence that spatial image-schemas are used in everyday verb understanding (Richardson, Spivey, Barsalou, & McRae, 2003)

Additional information

Notes on contributors

Jean M. Mandler

Jean Mandler is a Distinguished Research Professor in the Department of Cognitive Science at the University of California, San Diego.