Provided by Cognitive Sciences ePrint Archive
Secondary stress in Brazilian Portuguese: the interplay between production
and perception studies
Pablo Arantes and Plinio A. Barbosa
Speech Prosody Studies Group, State University of Campinas, Brazil
Abstract
This paper reports experiments on speech production show-
ing that secondary stress in Brazilian Portuguese (BP) can be
best described as phrase-initial prominence cued by greater du-
ration and pitch accent excursion in initial position. It also re-
ports a perception experiment in which clicks were associated
to consecutive V-to-V positions in stress groups. Mean click de-
tection RTs are gradient, but show no influence of initial length-
ening. RTs near the phrasally stressed position are shorter and
almost 60% of RT variance can be accounted for by produced
timing patterns.
1. Introduction
Recent phonological accounts of secondary stress in Brazil-
ian Portuguese (henceforth BP) [1] follow traditional ap-
proaches [2] in saying that secondary stress is assigned left-
wards to every even syllable counting from the lexically stressed
one. Most of the literature on secondary stress in other Ro-
mance languages agree on the claim about its binary nature.
They have in common the fact that they analyse isolated words,
compounds (and in some cases two-word noun-phrases) and
rely solely on impressionistic methodologies to base their pro-
posals.
Apart from intuitions based on linguists’ introspection, ex-
perimental studies have until the moment failed in providing a
sound empirical basis for the binary alternation claim in BP (see
[3] and others cited in [5]), as well as other Romance languages
([4] for instance). Experimental analysis seem to suggest that
some kind of initial prominence cued by greater duration or f0
excursion is a common feature in this language group. The
tendency toward initial prominence is also pointed out in the
phonological literature and can be formalized as an iambic re-
versal rule.
Earlier results [5] of a more comprehensive phonetic study
of stress groups containing polysyllabic words with a vary-
ing number of prestressed syllables in BP confirmed the ini-
tial prominence tendency. Normalized V-to-V duration patterns
were compared to those produced by a simulation with a dy-
namical model of rhythm production [6] and the simulated du-
rational contours mimicked the natural ones quite satisfactorily.
Since rhythm (and other prosodic traits) are traditionally
seen as an optimal solution between speakers’ and listeners’
oposing needs, a more complete account of secondary stress
and other rhythm-related phenomena requires a better under-
standing of the perceptual mechanisms involved in the on-line
processing of the patterned signal speakers provide listeners
with. Stating the problem in dinamical systems terms it is nec-
essary to uncover how production and perception, each one pos-
sessing it own intrinsic dynamics, couple to each other. One
way to do this is to determine what parameters listeners are sen-
sitive to when figuring it out what elements in the speech chain
stand out as prominent.
The successful interplay between the study of BP timing
and its modelling in a dynamical systems framework attained
in [5] encouraged us to experimentally investigate the possi-
bility that listerners actively evaluates V-to-V units duration as
a cue to detect an upcoming phrasal stress. If this hypothesis
comes to be proved true it will be an indication that the the un-
derlying dynamics governing rhythm production and listerners’
attention deployment rhythm are similar.
As far as secondary stress goes, the approach we are devel-
oping can help answering if the binarity embodied in most pho-
nologists’ analysis can be said to play any role in defining the
dynamics underlying rhythm perception. The experiment pre-
sented in section 3 helps answering this question showing how
listeners process the duration patterns elicited in the production
study reported in section 2 and in [5].
2. Production Study
Our corpus is composed of 17 penultimately-stressed target
words, with the number of prestressed syllables ranging from
two to five. An independent variable, dσ , was introduced to
control for the possible interplay between phrasal and hypothet-
ical secondary stresses. This variable measures the distance be-
tween the syllable bearing lexical stress in the target words and
the syllable bearing the main phrasal stress in the NP, in the fol-
lowing carrier sentences: “[A target ]NP parece menor hoje.”
(dσ = 0) and “[ A target rude/rural/bicolor ]NP parece menor”
(dσ = 2, 3, 4). A naive male speaker of the southeastern BP va-
riety read ten repetitions of the sentences in a sound-attenuated
booth. Duration and f0 were the main dependent variables.
2.1. Timing Patterns
Segmental duration was grouped in vowel, CV and V-to-V
frames (for a reference on the use of V-to-V units for charac-
terization of rhythmic patterns, see [7]) and then normalized
by means of z-score transformation to minimize intrinsic du-
ration effects. An ad-hoc reference corpus was used for this
purpose. No statistical evidence favoring either initial or alter-
nating prominence could be found if individual phone duration
was grouped in syllables or if vowel duration was taken alone.
V-to-V duration contours are shown in figures 1 and 2
alongside the first positions in the stress groups for the nine
words with four syllables (figure 1) and the six words with five
syllables (figure 2). Point markers stand for the different values
of dσ .
Taking consecutive positions in the stress group and dif-
ferent values of the distance parameter (dσ) as categori-