V-V units
Figure 1: Normalized duration in stress groups containing four
syllable words. Point markers stand for different values of dσ.
◦ is 0, □ is 2, O is 3 and △ is 4.
V-V units
Figure 2: Normalized duration in stress groups containing five
syllable words. Point markers stand for different values of dσ .
◦ is 0, □ is 2, O is 3 and △ is 4
cal factors and mean normalized duration as dependent vari-
able, an ANOVA carried on the group of four-syllable words
yields significance only for the position factor (F (2, 1305) =
456.06, p < 10-4), although positions 1 and 2 are not statis-
tically different as pointed out by a Scheffe post-hoc test. A α
level of5% was adopted.
As for the five-syllable words group, only the position fac-
tor yields significance (F (3, 1144) = 124.3, p < 10-4).
Here, positions 1 and 2 are statistically different, according to
Scheffe test (p < 0.003 for dσ = 2 and p < 0.04 for dσ = 4).
These results are evidence for a gradient lengthening of the first
V-to-V unit as the stress group gets longer, due to longer target
words or the presence of an adjective (dσ = 4). No evidence
for binary alternations shows up.
2.2. Intonational Patterns
A set of f0 contour samples from our corpus was examined
and labeled with the help of three phoneticians. In the agreed
transcription, a H* tone is associated to the first syllable of the
target words, irrespective of its length. A complex tone H*+L
orL*+H is associated with the lexically stressed syllable of the
target.
Future work on BP intonation will show if pitch accents
should be expected in initial position in circumstances other
than those involving stress groups starting with polysyllabic
words.
2.3. Summing up
It’s difficult to see how the results, taken together, can be ac-
counted for by metrical-like representations. Timing and into-
national patterns suggest that the initial prominence is related
to prosodic phrasing rather than to a hierarchical relation estab-
lished with the lexically stressed syllable. It seems more appro-
priate to consider this subordinate prominence as a stress group
initial strengthening. Support for this interpretation comes from
the fact that initial lengthening can be generated by the rhythm
production model [6]. Follow-up studies should investigate how
the pattern of prominence(s) along prestressed syllables is af-
fected by (a) target word position within sentence and (b) se-
mantic factors like referential status (i.e., if target word is given
or new information).
3. Perception Study
Duration contours elicited in our production study are highly
constrained. The shape of change in duration that culminates
in a phrasal stress follows a pattern that can be successfully ac-
counted for by the model described in [6]. Given this fact, the
following questions can be raised: (1) since prosody is a trade-
off between speakers’ and listeners’ needs, then listeners’ per-
ception of such stimuli are in some way constrained? (2) If so,
is this pattern related to the one we find in production?
There has been some positive answers to question (1). It
has been found that when a word carries sentence stress its per-
ception is somehow facilitated [8] [9]. This finding has been
brought to light by having subjects responding to word-initial
phoneme targets on stressed and unstressed words and it came
out that reaction time (henceforth RT) speeds up when the target
phoneme is in a stressed position. The claim is that in every sen-
tence there are points which catch listeners’ attention. There-
fore, words in those spots are more accurately processed. It has
been also demonstrated [10] that in word lists where timing was
carefully manipulated so that stressed syllables seemed to occur
at periodic points in time, RT to target phonemes were shorter
when compared to a situation where some jitter was introduced
in the timing of stressed syllables. It seems thus clear that lis-
teners can benefit from rhythmically patterned stimuli.
These experiments cannot state, however, if the effects
of timing are local to the rhythmically crucial spots or if
these spots are actively exploited all over the stimuli. Studies
by experimental psychologists on how people attend to time-
changing stimulus [11] suggest the latter option is likely to be
true. To put forward the idea that speech perception and pro-
duction constrain each other is a way to start answering ques-
tion (2). The hypothesis stated here is that it can be expected
that listeners’ attention is actively entrained by speakers’ ac-