138 Malcah Yaeger-Dror, Tania Granadillo, Shoji Takano, and Lauren Hall-Lew
2.5 Social situation and prosodic salience
Sociolinguists have shown that vowel positions, consonant realizations, and
even intonational contours vary with social situation (Labov 1972; Yaeger
1974; Yaeger-Dror 2001; Eckert and Rickford 2001; Tucker 2007). Social situ-
ation had initially been shown to inÀ uence intonation contours for quite styl-
ized genres of English such as story-telling, sports reporting, and political or
religious speeches, or direction-giving in a narrowly de¿ ned “game” setting.
(See, for example, Levin, Schaeffer, and Snow 1982; Grosz and Sidner 1986;
Liberman 1992; Nevalainen 1992; Blaauw 1995; Hirschberg and Nakatani
1996; Hirschberg 2000.) More recent studies have begun to look at less styl-
ized interactive situations (e.g., Bunnel and Idsardi 1996; IEEE 1997ff; Sagi-
saka et al. 1997; Chu-Carroll and Green 1998; COLING-ACL 1998ff), but the
vast majority of prosodic studies are still carried out on de-contextualized,
read sentences, or, at best, on newscasts, such as those in the “BUR” corpus
discussed earlier. This study will contrast the results of analysis of news-
broadcast data with results of a study using conversational speech.
2.6 Social situation, prosody and NEG:
The Social Agreement Principle
Both the Cogntive Prominence Principle and Cutler’s Corollary claim that
prominence is directly correlated with the importance of the information
being conveyed; all the read negatives which have been analyzed acoustically
support that claim. However, negatives must also be considered from another
point of view. One situational variable quite important to their analysis is the
distinction between informative and socially interactive situations (Yaeger-
Dror 1985, 1996, 2002a; Yaeger-Dror et al. 2002). That distinction will be
implicated in the study reported here.
We have already seen that both in isolated read sentences (O’Shaughnessy
and Allen 1983) and in informative readings (Hirschberg 1990) NEG carry impor-
tant information, and (therefore) are pitch prominent; however, conversation
analysts have shown that “preference for agreement” characterizes the conver-
sations they have analyzed (Schegloff, Jefferson, and Sacks 1977; Sacks 1992).
We will refer to that claim here as the Social Agreement Principle; NEG should
be prosodically reduced or deleted if they carry new information which might
be inferred as disagreeing with—or nonsupportive of—an earlier speaker.
“Preference for agreement” is obviously irrelevant for newscasts, or even
for read materials in general, but it is instructive to consider read dialogue:
when one reads from books, the F0 on NEG tokens is generally prominent in