The Situated Multimodal Facets of Human Communication - Coverbal Synchrony in Human-Machine Interaction - page 183

Empty speech pauses, defined as a silent interval longer than 0.150
s, were divided into three categories according to their duration: a)
short—from 0.150 up to 0.500 s long; b) medium—from 0.501 up to
0.900 s long; c) long—more than 0.900 s long. Added verbal material
was defined as any verbal material that produces a modification in the
listener's conscious knowledge , and given verbal material as intended
not to produce such a modification (Chafe, 1974). The label unclassified
was attributed to filled pauses (such as “ ehmm ”, “ uhh ”), and/or short
interruptions (such as “ ap* ”) following empty pauses. Empty pauses
identifying changes in scene, time and event structures were labeled
changes , a clause defined as “ a sequence of words grouped together on a
semantic or functional basis” , and a paragraph as “ a sequence of clauses
connected together by the same subject or scene” .
The analyses of the children's narrations showed that according to
their duration, each speech pause category plays a different role in the
discourse organization. Pauses of short and medium length generally
appear after utterances containing added information. Long pauses play
the functional role of delimiting paragraphs and identifying changes
in scene, time and event structures (Table 2).

Table 2. Percentage and absolute occurrences of pauses of short, medium and long length

associated with given , added , unclassifi ed information and changes of scene in the
discourse structure. The percentage is computed over the number of pauses in
each duration range.

10 Female Children Short EP Medium EP Long EP

Given 9 4% 2 3% 0

Added 225 88% 67 89% 36 77%

Unclassifi ed 21 8% 6 8% 11 23%

Changes 51 20% 61 81% 45 96%

4 Male children
Given

5 6% 0 0

Added 65 86% 37 88% 13 81%

Unclassifi ed 6 8% 5 12% 3 19%

Changes 6 8% 14 33% 15 94%

The data reported in Table 2 suggest that pauses belonging to
different duration ranges may play a different role in structuring
narrations, but more data and experiments are needed to confirm this
hypothesis. Most empty pauses are associated with the production of
added information, independently of the duration category to which
they belong. The majority of pauses of long (96% for female and 94%