What the tip-of-the-tongue (TOT) says about homophone frequency inheritance.

The present study uses tip-of-the-tongue (TOT) states as a unique source of evidence to test the hypothesis of lexical access benefits for homophones--that is, whether low-frequency homophones, such as tee, inherit the lexical access benefits of their high-frequency homophonic counterparts, such as tea. We compared retrieval success rates for low-frequency homophones, for matched low-frequency controls, and for high-frequency controls with the combined frequency of the homophone set. In correct retrievals, low-frequency homophones behaved according to their specific frequency, not differing from the low-frequency controls. However, retrieval failures revealed a different kind of homophone effect. When retrieval failed for targets with a homophone partner, access difficulties tended to be less profound than for low-frequency controls, ending closer to target retrieval more often than low-frequency controls (at Step 2; in a self-resolved TOT or in a TOT with a strong feeling of knowing), and ending far away from target retrieval less often than low-frequency controls (at Step 1; in a notGOT). These results provide evidence against the notion of shared word-form representations for homophonic targets but leave open a door for a weaker form of homophone effects, possibly arising from feedback activation that influences retrieval only when access is sufficiently slowed (as when retrieval fails).

Linguistic evidence and grammatical theory.

This article surveys the major kinds of empirical evidence used by linguists, with a particular focus on the relevance of the evidence to the goals of generative grammar. After a background section overviewing the objectives and assumptions of that framework, three broad kinds of data are considered in the three subsequent sections: corpus data, judgment data, and (other) experimental data. The perspective adopted is that all three have their place in the linguist's toolbox: they have relative advantages and disadvantages that often complement one another, so converging evidence of more than one kind can reasonably be sought in many instances. Points are illustrated mainly with examples from syntax, but often can be easily translated to other levels (e.g., phonology, morphology, semantics, and pragmatics). WIREs Cogni Sci 2011 2 206-221 DOI: 10.1002/wcs.102 For further resources related to this article, please visit the WIREs website.