Lexical priming and register variation

Lexical priming predicts that repeated encounters with lexical patterns will prime users for register awareness (Hoey 2013: 3344). To verify this prediction, this chapter reports on a study that determined the dimensions of collocation in American English, which are the parameters underlying the use of collocations in spoken and written text. The method was inspired by the multidimensional framework for register variation analysis introduced by Biber in the 1980s. The corpus used was the 450-million-word Corpus of Contemporary American English (COCA, 1990–2012 version). The most characteristic collocations of each register in COCA (spoken [American radio and television programs], magazine, newspaper, academic, and fiction) were computed using the logDice coefficient (Rychly 2008). These were then entered in a factor analysis, which yielded the statistical groupings of collocation across the registers. Nine dimensions were identified and are described in this chapter. The relationship between collocation and register was tested statistically through the dimensions, and the results suggested that register could predict the collocations (via the dimensions) between 39% and 67% of the time, which seems to lend support to the hypothesis that users are primed for register, as far as AmE collocations are concerned.