The Situated Multimodal Facets of Human Communication - Coverbal Synchrony in Human-Machine Interaction - page 182

of the glottis and/or the vocal tract in producing phonemes such as
stop consonants. Communicative and socio-psychological pauses are
meant to build up tension or generate the listener's expectations about
the rest of the story, assist the listener in his/her understanding of
the speaker, or to interrupt for questions and comments, as well as to
signal anxiety, emphasis, syntactic complexity, degree of spontaneity,
gender, and educational and socio-economical information (Bernestein,
1962; Goldman-Eisler, 1968; Abrams and Bever, 1969; Kowal et al.,
1975; Green, 1977; O'Connel and Kowal, 1983). Linguistic pauses are
more likely to coincide with discourse boundaries, realized as a silent
interval of varying length, at clause and paragraph level (Brotherton,
1979; Gee and Grosjen, 1984; Rosenfied, 1987; Grosz and Hirshberg,
1992). This is particularly true for narrative structures where it has
been shown that pausing marks the boundaries of narrative units
(Chafe, 1980; Rosenfield, 1987; O'Shaughnessy, 1995; Oliveria, 2000,
2002). Lastly, several cognitive psychologists have suggested that
pausing strategies will surface in the speech stream as the end product
of a “planning” cognitive process that reflects the complexity of the
neural information processing. The length of the pause is a sign of the
cognitive effort related to lexical choices and semantic difficulties for
generating new meanings (Goldman-Eisler, 1968; Butterworth, 1980;
Chafe, 1987; Brennan, 2001).
Investigations into the role of speech pauses have been neglected
by the recent literature leaving open several scientific questions.
These include, how children exploit pausing to shape their discourse
structure, and to signal the cognitive effort associated with the amount
of “ given ” and “ added ” information they are conveying. Shedding light
on this topic would help in the comprehension of the role that this
paralinguistic information is supposed to play, assisting humans in
building up meanings from them, suggesting appropriate mathematical
models for implementing more friendly human-machine interfaces,
and in particular, more efficient automatic speech recognition systems.
Here, I report the results of a series of analyses on narrations made
by 10 female and 4 male Italian children (mean age = 9 ± 3 months).
These children were asked to narrate episodes of a 7-minute
animated color cartoon “Sylvester and Tweety” they had just seen
(Esposito, 2006; Esposito and Esposito, 2011 for procedural details). The
cartoon is of a familiar type to Italian children, involving a cat and a
bird. The listener was familiar to the child (either the child's teacher or
other children). The children were recorded while narrating the cartoon
and the videos were analyzed frame by frame in slow motion to assess
the association between utterance meaning and empty speech pauses.