Fast Words, Slow Words

Mood. Good. Blood. Why is it that the vowels of these words are all spelled the same but pronounced differently? Historically, all go back to the same vowel in Renaissance English, namely, which mood retains. But the other two words have changed into and respectively. Often such sound changes travel different routes in different varieties of a language, whether regional or social. And some -oo- words do vary in that way: roof, for instance, has regional variants, /ruf/ and And we would probably make social judgments about a person who said I tuck it for 'I took it'. Even within the same variety, however, some words may be affected sooner than others. For instance, often the neighboring sounds play a role, so words that end in a sound--tooth, (for)sooth--are more likely to have kept the older /u/ (Ogura 1987). Yet another factor, however, in determining which words change first is the frequency with which a word is used. The theory that tries to explain why some words "hurry ahead" while others "drag behind" (as Schuchardt noted in 1885) has become known as the theory of lexical diffusion (Wang 1969).

Some variation in modern American English is due to changes affecting the least frequent words first. Hooper (1976, 99-100), for instance, found that more frequent verbs like kept, left, and slept resist being reshaped with an -ed suffix, whereas less frequent verbs like crept, leapt, and wept have developed the new forms creeped, leaped, and weeped. Phillips (1984) found that noun/verb pairs which retain the same stress pattern for both (e.g., expréss) tend to be used more frequently than pairs which have developed different stress patterns to indicate noun versus verb (e.g., éxploit n vs. explóit v). And Phillips (1981) uncovered the same pattern when the language realigns which consonants may occur next to each other. For example, the word nude, in most American dialects, is now pronounced /nud/, having lost the y-glide still extant in British Received Pronunciation and Southern American English: /njud/. In the latter variety, infrequent words such as tuber, duly, and nude are more likely to have lost the glide; more frequent words such as tube, during, and new are more likely to have retained it.

Many sound changes evince the opposite pattern, however, with the most frequent words changing first, as in the foot-boot example. These tend to be changes influenced by surrounding sounds, that is, assimilations. The shift in -oo- words, for instance, started in the words blood and flood probably because /l/s are produced with a tongue shape that lowers and centralizes the vowel, and following /d/s have similar effects (Ohala 1974, 265; Ladefoged and Maddieson 1996, 184, 303). Other changes affecting the most frequent words first are those that reduce vowels in unstressed syllables, as in mistáke, where the first syllable is often pronounced with a vowel more like that of but than of miss, versus mistóok, where the first syllable still sounds like miss (Fidelholz 1975). Even stress shifts, however, can affect the most frequent words first, if they involve ignoring the part of speech label rather than accessing it. The verb frústrate, therefore, is being pronounced frustráte more and more often in both British and American English, as are many other verbs ending in -ate--a pronunciation which follows the stress patterns of other English words. Less frequent verbs such as láctate, however, are less likely to alter their stress (Phillips 1998).

Phillips's (forthcoming) explanation for the connection between word frequency and lexical diffusion emphasizes the degree of attention or lexical analysis required of the speaker. That is, if the features of individual segments are blurred (as in assimilations) or ignored (as in certain stress shifts or segment reductions), the most frequent words are affected first. If speakers have to access specific information about a word, however, in order to implement a change...