Segment durations in 20 minimal-paired words in Japanese were examined
to identify regular patterns of acoustic manifestations of mora timin
g in Japanese, The results indicate that segments consistently stretch
or compress, providing temporal compensation within a mora and betwee
n moras, to attain targeted length at the word level. A segment is 11%
longer in the pre-geminate-stop position than in the pre-single-stop
position; in the word final position, a segment is 9% shorter followin
g a geminate stop than a segment following a single stop; and [t] in t
he same phonotactic environment shows varied durations according to di
fferent moraic conditions. Fricatives followed by a devoiced vowel are
considerably longer than any other inherently long, nonmoraic segment
s, and are comparable to moraic segments and the average mora duration
. Syllable durations are not isochronous, varying by as much as 50%. T
he word duration is affected more by the number of component sounds an
d their inherent durations in two-mora than in three-or four-mora word
s. When the word durations are converted into ratios, they are markedl
y similar to the mora values. The results suggest a timing control mec
hanism adjusting segments to obtain targeted word durations. These phe
nomena describe mora timing in Japanese.