Gradience in an abrupt change
Two seemingly abrupt shifts affecting disyllabic noun/verb pairs are currently in progress in English. One is the diatonic stress shift, whereby noun/verb pairs originally stressed on the final syllable develop into pairs in which the noun is stressed on the first syllable while the verb remains stressed on the final syllable (as in PDE <i>COMpress </i>n. vs<i>.</i> <i>comPRESS</i> v.). This shift has a long history, going back to at least the sixteenth century (Minkova 1997, 160), but is still seen in variability in the stress placement in some words, such as <i>address</i>, with some speakers preferring initial stress on the noun and others final stress. The second shift moves the stress of the verbs in diatonic pairs to the first syllable as well, as in <i>CONtrast </i>n. or v. Whereas the first shift may have its roots in the overwhelming type frequency in English of two-syllable forestressed nouns and backstressed verbs (Sereno 1986), such cannot be the basis of the second shift. Yet both would seem to be abrupt shifts of stress. But that too is questionable. Sereno and Jongman (1995) have found gradient stress in bisyllabic words such as <i>poison </i>and <i>debate</i>, which function as either nouns or verbs, but which have only one stress pattern; that is, they found the ratio of noun to verb usage for such a word influences the amplitude of the stress in each of the word’s syllables. For example, a word such as <i>poison</i>, which is more often used as a noun, will have a higher amplitude on the first syllable when used as a verb than will a word which is more often used as a verb, such as <i>notice</i> (Sereno & Jongman 1995, 69). This paper investigates whether the diatonic stress shift in English and its further development in some words to intial stress for both noun and verb might display a similar gradient component. Specifically, this study uses current U.S. and British pronunciations (Wells 2008) and frequency counts from the spoken portion of the <i>Corpus of Contemporary American English</i> (COCA) to compare the relative effects of noun frequency, noun-plus-verb frequency, and noun/verb ratio on stress shifts in English bisyllabic noun/verb pairs. The importance of the noun/verb ratio on these shifts has been explained above. The importance of whether the noun-plus-verb frequency or noun frequency alone is a better predictor of the change rests on the question of whether English is a language for which homophones share frequency effects, and hence whether it is appropriate to expect homophones to behave differently in this and other changes, as implied by Labov (2010, 273) and Joseph (2012, 416). The following topics are therefore investigated: firstly, if the token noun-plus-verb frequency of bisyllabic pairs that have become diatonic is, on average, less than that of the noun-plus-verb frequency of those which have remained backstressed; secondly, if the token frequency of nouns in bisyllabic pairs that have become diatonic is, on average, less than that of the nouns which have remained backstressed; thirdly, if the bisyllabic pairs that have become diatonic exhibit a higher noun-verb ratio than those which have remained backstressed; and lastly, if the diatones which have, according to Wells (2008), become possible forestressed pairs (such as <i>‘REsearch, </i>n. or v.) have higher noun or noun-plus-verb frequencies and/or noun-verb ratios than diatones which have not changed.