Range-estimation in learning word meanings:  a recipe for
	semantic change ?
	
	Daniel W. Smith
	Biology Dept., MS 34, Woods Hole Oceanographic Institution

	dwsmith@whoi.edu


	Like many other traits of human individuals, the meanings of
	words pass from one human to the next through cultural
	inheritance (e.g., learning and imitation).  Many of these
	meanings can be most realistically modelled in "semantic space"
	as extending over ranges (spans or volumes in the space of
	possible referents), rather than being pegged to single points
	within such a space (as e.g. in Fillenbaum & Rapoport 1971, K.
	Smith et al 2003).  Precisely because these meanings extend
	over ranges, while instances of their use typically correspond
	to points within those ranges, a language learner is likely to
	lack enough experience with others' uses of a given meaning to
	permit an exact determination of where its range should begin
	and end.  One might thus expect variation in individuals'
	meaning-ranges to arise as a byproduct of range "guesses"
	during cultural transmission, particularly in the case of rare
	words.  On a larger time-scale, one might expect that
	synchronic meaning variation arising from such guesses could
	foster change in average meanings over successive cultural
	"generations," particularly in the face of a changing
	environment.

	The poverty of data available to a range-estimating learner for
	any given meaning imposes a "transmission bottleneck" distinct
	from, and in addition to, the possibility that the learner
	lacks any experience at all with some subset of specific
	meanings (as treated by K. Smith et al 2003).  How does
	language squeeze through this bottleneck ?  Observing what
	kinds of variation and change in meanings actually do arise in
	human languages, among those theoretically possible given
	cultural transmission of meaning ranges, may provide insights
	into both cognitive psychology (how do cognitive/perceptual
	constraints limit shifts of meanings ?) and the cultural
	transmission process (how might language itself have adapted to
	facilitate its own transmission in the face of limited data?).
	At the same time, computer simulation experiments in a
	simplified, one-dimensional meaning space can elucidate the
	potential interplay of factors in the contribution of
	range-estimation to cultural evolution of meanings.



	REFERENCES

	Fillenbaum, S., and Rapoport, A.  1971.  Structures in the
	Subjective Lexicon.  New York:  Academic Press.
	
	Smith, K., Kirby, S., and Brighton, H.  2003.  Iterated
	learning: a framework for the emergence of language.  E-print
	from the Dept. of Theoretical and Applied Linguistics,
	University of edinburgh, available online at
	http://archive.ling.ed.ac.uk/documents/disk0/00/00/01/83/index.html