The energy with which traditional ideas have been knocked down has not however been matched by equivalent energy in rebuilding theories, and where theories have been proposed, they have often been incomplete or unconvincing. Although, for example, Sinclair’s idiom principle articulates an important truth about the way we construct our utterances, it leaves unarticulated the mechanism whereby one might fall back onto grammar when idioms fail to serve the speaker’s communicative process; nor is it convincing about why we should, as children, develop a grammatical system as a security mechanism for when our idioms fail. (Sinclair & Mauranen, 2006, admittedly offer a subtler picture.) On the whole, corpus linguistics has either ignored theoretical questions or (less commonly) related itself to pre-existing theoretical perspectives that appeared capable of accommodating corpus linguistic findings, e.g. cognitive linguistics and connectionism (Ellis, 2005; Gries & Stefanowitsch, 2006). There is of course nothing wrong with the latter course of action, but it may be premature until we have explored further what a corpus-based theory might look like.

Much of the psycholinguistic literature used by linguists is arguably more linguistic than psychological. But there are two research threads in psycholinguistics that may be of relevance, that are perhaps more psychological than linguistic: work on semantic priming and work on repetition priming.¹

In semantic priming experiments, informants are shown a word or image (referred to as the prime) and then shown a second word or image (referred to as the target word). The speed with which the target word is recognized is measured. Some primes appear to retard informants’ recognition of the target and others appear to accelerate informants’ recognition of the target. For example, the prime word wing may be shown to inhibit the recognition of the word pig but to accelerate the recognition of the word swan.

As Pace-Sigge (2013) shows, the semantic priming literature is extensive and has an important pre-history; his book should be consulted by those interested in the psycholinguistic background. What is offered in this chapter is simply an outline account designed to explain the psycholinguistic underpinning of lexical priming theory before we consider its applicability to Chinese. The pioneering work on semantic priming was conducted by Meyer and Schvaneveldt (1971); they were the first to show that priming was scientifically demonstrable. Their work was followed through by Shelton and Martin (1992) (among others), who appear to show that semantic priming only works when the priming word and the target are associated in the informant’s mind. McRae and Boisvert (1998) argue, however, that if the words in question have closely related meanings there will be a priming effect even without such association. It will be noted that the earliest of the papers we have just cited is now over 40 years old and the phenomenon has not been refuted, even if the constraints upon its operation are still under discussion. While semantic priming does not explain the phenomenon of collocation, and is not necessarily the product of collocation, it is reasonable to look for a common explanation of both phenomena. This we find in repetition priming.

Repetition priming is rather different from semantic priming, in that the prime and the target are identical. Experiments with repetition priming centre on exposing informants to word combinations and then, sometimes after a considerable amount of time and after they have seen or heard a great deal of intervening material, measuring how quickly or accurately the informants recognize the combination when they finally see or hear it again. For example, a listener may be shown the word ALARMING followed by the word SUNSHINE. A day later, if s/he is shown the word ALARMING again, the informant may be found to recognize SUNSHINE more quickly than other words equivalently positioned after ALARMING. The assumption in such a case must be that s/he would be remembering the combination from the first time, since a combination of words such as ALARMING SUNSHINE will only rarely have occurred before, and quite probably not in the earshot or reading of the informant, and is therefore unlikely to comprise a learnt expression.

Again, we refer the reader to Pace-Sigge (2013) for a fuller account of the relevant literature. Key papers on these facets of repetition priming, though, are those of Jacoby and Dallas (1981), who observed greater accuracy in the identification of the tar...