@misc{cogprints2132, editor = {A Cangelosi and D Parisi}, title = {Symbol Grounding and the Symbolic Theft Hypothesis}, author = {Angelo Cangelosi and Alberto Greco and Stevan Harnad}, publisher = {Springer, London}, year = {2002}, journal = {Simulating the Evolution of Language}, keywords = {perceptual learning, symbol grounding, categorical perception, categorization, evolution of language, neural nets}, url = {http://cogprints.org/2132/}, abstract = {Computational simulations are used to model the following: (1) category learning through sensorimotor trial and error ("sensorimotor toil") and how it generates categorical perception (decreased between-category similarity and increased within-category similarity); (2) symbol grounding (the connection between symbols and the sensorimotor categories that they name); (3) the origins of language as the capacity to acquire categories indirectly, by definition alone ("symbolic theft"); and (4) the evolutionary advantage of acquiring categories by this symbolic theft instead of sensorimotor toil.} }