All of the example utterances in sections 4.3 and 4.4 above show cases in which the units defined by break indices 2 and 3 corresponded to the prosodic units accentual phrase and intonation phrase, respectively. It was noted that while these perceptually-defined units and tonally-defined units coincide in most cases, there may be instances of mismatch between them, and thus they are not totally redundant to one another. In such cases of mismatch, the labeller should mark the break index according to her/his evaluation of the degree of disjuncture (exactly as in non-mismatch cases), but should then also add the diacritic ``m'' following the break index value.
Example utterance 
nibanme
shows the
marking of 2m. Here, the speaker has produced a H% boundary tone at
the edge of the first phrase /nibaNme' no/ `second-GEN'. As mentioned
above in section 4.4, H% and HL% tones are found at the edges
of intonation phrases, whose boundaries are most commonly
characterized by a strong degree of disjuncture (i.e. BI 3). Yet the
disjuncture between the /no/ of the first phrase and the following
/siNsitu/ `bedroom' is clearly not a strong one, but more like a
medium disjuncture. The 2m label reflects the fact that the perceived
juncture is similar to that commonly associated with an accentual
phrase (i.e. BI 2), but that there is a mismatch with the tonal
pattern.
The utterance 
sankaku
also shows an
example of a mismatch. In this utterance there is a pause between the
phrase /maNnaka ni/ `middle-LOC' and the verb /okima'su/ `put'. This
gives the sense of a strong disjuncture (BI 3). However, the pitch
range on the following verb seems reduced, as if the verb has been
downstepped. If this is the case, the string /maNnaka ni okima'su/
would form one intonation phrase, within which downstep applies. The
mismatch arises from the fact that there appears to be a strong
disjuncture marked with BI 3 (which is most often associated with an
intonation phrase) within an intonation phrase. Therefore, the
break has been labelled by 3m to mark this mismatch.
With instances of mismatch flagged using the ``m'' diacritic, it will enable researchers to search through large labelled databases and investigate these cases further. We will then be in a better position to say how common mismatches are, and whether they are in any way related to certain speaking styles.