Prosody Marks Different Kinds of Informativity: Interactions Between Frequency, Probability and Focus

Iris Chuoying Ouyang and Elsi Kaiser University of Southern California [email protected] Prosody has been found to encode the extent to which linguistic elements are informative. Prior work has approached informativity from two angles: (i) information theory/statistical probability and (ii) information structure/pragmatics. Prior work from the information-theory side has found that increased statistical probability (e.g. being high-frequency or contextually predictable) reduces words’ prosodic prominence (e.g.[1,6]). Prior work from the information-structure side has found that pragmatic, information-structural salience increases prosodic prominence (e.g., words in narrow focus are more prominent than words in wide focus, e.g.[3]). However, little attention has been paid to the potential interaction between these two kinds of factors (though see [2] on givenness+frequency). This was the focus of our study. We investigated whether word frequency and contextual probability modulate the prosodic effects of information structure, and whether they impact different categories of information structure in different ways. In a production study, participants (N=16) worked with a lab assistant as their partner: the partner (speaker A) asked questions and participants (speaker B) provided answers. Three question types were used to elicit focus: new-information narrow focus, corrective narrow focus, and VP (wide) focus (ex.1-3). Target objects differed in (a) word frequency (high, e.g. cars/books, or low, e.g. cans/shells) and (b) contextual probability (high, e.g. cars/cans, or low, e.g. books/shells). Contextual probability was estimated using a norming study; word frequencies are from SUBTLEXus [4].