Psycholinguistic Probes of Native Speakers' Phonological Knowledge

It has been one of the tasks of phonologists since the time of Panini or Plato to demonstrate relationships between words based on their sound. In the past 200 years this has meant establishing the historical connection between words, i.e., to show that due to sound change, the same morpheme could show divergent forms in different phonological, morphological, or dialectal environments. More recently, especially since generative phonology, a new goal has been adopted, namely, to show how native speakers represent the phonological relationship between words in their language, e.g., musicjmusician, right/righteous. Most of the problems that have occupied phonologists within the past few decades have arisen in the course of this task, e.g. the character of the underlying forms common to morphological variants and of the rules that convert them to surface forms. But what does it take for native speakers to even conceive that two (or more) words are related, such that they are motivated to try to deduce an underlying form and the phonological rules which will make the relationship a regular one? It is this question we address here. First of all, for words to appear related they should presumably show some phonetic similarity, otherwise there would be little reason for speakers to try to work out phonological rules which relate the words. Second, they should be related semantically. Without this there would be as much reason to relate lathe/lather as there is Kate/Katherine, which is counterintuitive. Derwing and Baker (1977) conducted an interesting experiment which showed that similarity in meaning is a more important determinant than phonetic similarity of speakers' judgements that pairs of words are derivationally related. What else might influence these judgements? A very important factor should be the number of word pairs which show the same or similar phonological and semantic relationships. Thus, a pair which exemplifies a very common phonological pattern, e.g. that in extreme/extremity, should be a more likely candidate for being considered derivationally related than would pope/papal. Alongside the former, there are a host of examples showing the same vowel alternation, e.g. serene/serenity, obscene/obscenity, compete/competitive, obsolete/obsolescence, etc. The latter, however, has only (to our knowledge) one similar pair, nose/nasal. Thus the rules that would have to