In lieu of an abstract, here is a brief excerpt of the content:

LENGTH AS A SUPRASEGMENTAL: EVIDENCE FROM SPEECH ERRORS Joseph Paul Stemberger Carnegie-Mellon University Segmental length can be analysed in one ofthree possible ways: as a segmental feature, as gemination, or as a suprasegmental. These three analyses make different predictions about how length should behave in language production. Treating length as a suprasegmental predicts that it will frequently be dissociated from a segment, while the other analyses predict it will usually not be. Speech error corpora in German, Swedish, and English are examined. The data suggest a suprasegmental analysis, most probably along the lines of recent autosegmental descriptions: long segments are associated with two positions in the syllable structure. A vowel and its associated structure show different degrees of cohesiveness in different languages, so that they behave quite differently in errors in the different languages.* Until recently, a long segment such as [i:] or [p:] was interpreted in one of two ways: as a geminate cluster /ii/ or /pp/, or as a single segment with the feature [ + long]. Leben 1980 and McCarthy 1981 have added a third possibility: it is a single segment whose length is represented suprasegmentally (in particular , encoded in the syllable structure). This paper will present psycholinguistic evidence from errors in normal speech that length for both vowels and consonants is represented in some suprasegmental fashion, most probably in the fashion assumed in recent autosegmental work. 1. Suprasegmental phenomena. Whenever a phenomenon is treated suprasegmentally , one ofthe strongest predictions made is that it will be relatively independent of the segments with which it is associated; Goldsmith 1976 refers to this phenomenon as the 'stability' of suprasegmental units. Segments may be deleted or moved without any effect on the suprasegmental, which will simply become associated with some nearby segment(s). Thus when a suprasegmental such as tone is linked to a segment, the segment may be deleted, but the tone remains and is realized on a neighboring segment. This is one of the arguments of Williams 1976 that tone is suprasegmental in Margi: LHLH I I V (1)tla-wa F tlwa Within psycholinguistics, this dissociation of segments and suprasegmentale provides a common argument for the suprasegmentalhood of certain phenomena . Hombert 1973 points out that tone and syllables are completely independent in Bakwiri word-play, where whole syllables are moved with no effect on tones: HLHL I I Il (2)likwe f kweli * Thanks go to Bruce Hayes and to anonymous readers of this article in an earlier form. 895 896LANGUAGE, VOLUME 60, NUMBER 4 (1984) In the study of speech errors, Fromkin 1971 notes that words can be reversed in order without changing the stress pattern of the sentence:1 (3)That's 'cause it soés fëel — feels so much nicer. On this evidence, she argues that English sentential stress is indeed supresegmental ; and Becker 1979 argues specifically for an autosegmental representation of stress. Such behavioral evidence reinforces the conclusion, reached on the basis of more traditional linguistic evidence, that stress and tone cannot be parts of a vowel, on a par with features of place or rounding; rather, they must be semi-independent suprasegmental units that are linked to segments. 1.1. Evidence used in this paper comes from errors occurring in natural speech situations. Speech errors have a number of characteristics that make them useful in addressing the question of how to represent length. First, one common error is the misordering of a segment—realizing it too early (anticipation , 4a), too late (perseveration, 4b), or reversed in order with another segment (transposition, 4c): (4)a. I'd AGROO with you. (agree) b.all over the ploe----- place c.soo-FEED (sea-food) Such errors usually involve single segments—or less frequently, clusters, rimes, or syllables. As in 4a-c, interacting segments are usually very similar not only in which features are present, but also in which part of the syllable they are found, and whether the syllable is stressed (Nooteboom 1969). Second, most errors can be described in terms of whole segments; relatively few errors must be explained as, e.g., anticipation of a feature, and almost none must be explained as transposition of features (Shattuck-Hufnagel & Klatt 1979). Further characteristics of errors will...

pdf

Share