Disfluencies (``p'')

It is common in spontaneous speech for the speaker to hesitate, stop abruptly and restart, or produce other similar disfluencies. Since the aim of J_ToBI is to describe spontaneous as well as read lab speech, there must be a mechanism for marking such disfluent junctures. Following English ToBI, the diacritic ``p'' following a break index value is used to mark these cases. The use of this diacritic on the break index tier is a cue that the corresponding tones on the tone tier may be incomplete or ill-formed. Since this ``p'' label is reserved for disfluent junctures only, labellers should ask themselves whether the utterance might have been produced differently (more fluently) if the the speaker was given a second chance to produce it.

A 1p marking on the break index tier indicates cases of abrupt cut-off in which there is no sense of the L% boundary tone which accompanies an accentual phrase juncture (BI 2). Utterance heikoo shows an example of 1p marking. Here, the speaker stops abruptly after the words /ima no/ `livingroom-GEN' but then continues on with the following /ma'do to/ `window-with' as if no disfluency had occurred (without restart). Tonally, the string /ima no ma'do to/ constitutes a well-formed accentual phrase (which also happens to be a single intonation phrase). The break index value 1 marked after the /no/ reflects the fact that this juncture falls inside a larger unit (accentual phrase), and the ``p'' diacritic flags the disfluency (see also section 6 for discussion of the ``disfl'' label on the miscellaneous tier).

<<heikoo>> (part 1) ``Um, the one on top, the window on top, um, I will make it so that it lines up level with the livingroom window.''

<<heikoo>> (part 2) ``Um, the one on top, the window on top, um, I will make it so that it lines up level with the livingroom window.''

A 2p marking on the break index tier, on the other hand, marks a disfluent juncture which is accompanied by the sense of a L% accentual phrase final boundary tone. Example utterance shikakui shows 2p marking. In this utterance, the speaker hesitates after the word /sikakui/ `square', but then continues on after a moment with the rest of the phrase /tyairo no kami'/ `brown-GEN paper'. The downtrend of the words (not downstep here, since the words are unaccented) gives the sense that they are grouped into a single intonation phrase, and that no reset has occurred after the disfluent pause. The boundary after /sikakui/ is a medium disjuncture (BI 2), and indeed if the pause is cut out entirely the utterance sounds like a fluent intonation phrase, with no strong boundary intervening (i.e. BI 3). Therefore, the break index marked here reflects the medium disjuncture (BI 2), as well as the fact that there is a disfluency due to hesitation.

<<shikakui>> ``The square, brown paper ...''

The utterance heikoo also gives an example of break index 2p. There is a disfluent break after the first phrase /ue no hO' no/ `the one toward the top', and the speaker chooses to restart the utterance after this point. However, the break after /no/ of the first phrase does not have the sense of a strong disjuncture (BI 3), but rather, it sounds as if the speaker would continue on with the utterance, despite the disfluency. Thus, a medium disjuncture is marked (BI 2), along with the ``p'' diacritic.

Word-internal breaks such as in [tya-tyairo] `br-brown' should not be indicated on the break index tier, but only by a ``disfl'' label in the miscellaneous tier (see section 6 below). However, word-internal breaks followed by a restart, such as in [tya- ore'Nzi no kami'] `the br- orange paper' should be marked using 1p on the break index tier, as well as with a ``disfl'' label in the miscellaneous tier.