Turk, R. , Jones, G. , Guest, D. , Young, A. & Andrews, M.
Nottingham Trent University
It has been demonstrated that both adults and children are capable of tracking statistical distributions within the environment. However, the exact nature of this ability is not fully understood, particularly in relation to natural language. Transitional probabilities are often used to describe the statistical relationship between items within a language but calculating these requires the continual tracking of both item frequency and item diversity (the number of co-occurring items); it follows therefore that these metrics may also be predictive of language proficiency when considered in isolation. The effects of these simpler metrics on language processing were examined at word level using a suite of primed word recognition studies (prime = first word of bigram, target = second word) where bigram frequency and bigram diversity were independently manipulated. High frequency bigrams and high diversity bigrams both led to faster response times to target words, over and above other effects such as the frequency and concreteness of the individual target words. Moreover, regression analyses showed that bigram frequency was more predictive of response times than bigram transitional probability. The results suggest that transitional probabilities may be masking a more fundamental role of frequency and diversity in language processing.