Joseph Tyler attended the Annual Meeting of the International Pragmatics Association, from July 3-8 at the University of Manchester, in Manchester UK.


Prosodic correlates of coherent discourse structure


Do speakers communicate discourse structure in their prosody? And if they do, what specific features of discourse do speakers encode prosodically? This study looked for prosodic correlates of participant readings of a newspaper article, whose discourse structure was annotated within a semantically-motivated discourse theory (Segmented Discourse Representation Theory (SDRT)), as a way of gaining independent, empirical access to how speakers represent and communicate the structure of discourse.

Three structural measures were of interest: how closely related adjacent discourse segments are, how embedded a discourse segment is in the overall discourse, and whether a rhetorical relation links discourse segments at the same hierarchical level (coordinating) or different levels (subordinating). Prosodic measures including pause duration, pitch, intensity and speech rate were correlated with those structural measures.

To exemplify, the following text is an excerpt from a Wall Street Journal article, segmented according to SDRT (Reese, Denis, Asher, Baldridge, & Hunter, 2007).


  40.   But the mainstream civil-rights leadership generally avoided the rhetoric of "law and order,"

  41.   regarding it as a code for keeping blacks back.

  42.   Law and order didn't mean justice,

  43.   Mr. Jackson used to say,

  44.   but "just us."

  45.   In the past, many were hesitant to speak about crime in public

  46.   because "the larger community would talk about 'lock them up and throw the key away' and hide behind black leaders in doing it,"

  47.   explains Rep. Craig Washington,

  48.   the Houston Democrat who led the caucus hearing.

  49.   Now there is escalating discourse within the black community about what it can and must do to stop crime.

  50.   Just after the new year, Mr. Jackson held the first of several conferences focusing on just that.


In this excerpt, segments 41 and 42 are directly linked by a discourse relation while segments 48 and 49 are not, indicating the former are more closely related than the latter. Segments 40 and 49 are less embedded, i.e. subordinated fewer times, than all of the other segments in this excerpt. And segments 40 and 49 are linked by a coordinating relation while segments 40 and 45 are linked by a subordinating relation. Do these structural features correlate with prosody?

         To answer this question, ten participants studied silently and then read aloud twice a 1218-word text divided into ninety discourse segments and relations, providing almost 900 data points for each phenomenon. Statistical analysis involved a linear mixed model with subject as a random effect and words per discourse segment, quotation and sentence-initiality as fixed effects.

         Results showed that more distantly related adjacent discourse segments had significantly longer intervening pause durations, and higher initial pitch and max intensity on the second segment. A discourse segment’s overall embeddedness correlated with f0min, intensity and speech rate. And coordinating and subordinating relations correlated differently with pause duration, f0min and mean pitch, intensity and speech rate. The findings for f0min are interesting because prior discourse prosody research has tended to exclude it from its analysis (den Ouden, Noordman, & Terken, 2009; Hirschberg & Grosz, 1992; Pierrehumbert & Hirschberg, 1990).