Peak alignment in Estonian

Earlier studies of Estonian showed that vowel quantity words (i.e.
words dif- fering only in vowel quantity) produced with an H*+L
pitch accent differed in their peak alignment: While words with a
short and a long vowel had a peak late in the vowel of the stressed
syllable, words with an overlong vowel were characterised by a peak
earlier in the vowel (e.g. Asu et al., 2009; Lippus et al., 2013).
The main aim of this dissertation is to shed light on these peak
alignment differences: firstly, whether these alignment differences
can be ex- plained with the help of a segmental anchor; secondly,
whether alignment is similarily affected by quantity differences in
consonants and vowels; thirdly, whether such alignment differences
are stable in regard to the prosodic con- text, more precisely in
regard to the number of post-focal unstressed syllables (i.e. the
vicinity to the sentence boundary) and a variation of the speaking
rate. Additionally, not only the peak alignment in regard to the
vicinity of an upcoming sentence boundary was investigated, but
also the influence of the sentence boundary on segment durations
(phrase-final lengthening - PFL). Previous studies (e.g. Krull,
1997; Asu et al., 2009) showed that PFL occurs in Estonian, but it
was not studied yet whether PFL affects vowel and consonant
quantity words differently. Furthermore no attempt made to explain
PFL in Estonian with the help of abstract phonological models. The
purpose of this dissertation is to fill this gap. This dissertation
contains three different experiments which are presented in one
chapter each. The first experiment (chapter 2) explored the
influence of the upcoming sentence boundary and its interaction
with vowel (VQ) and consonant (CQ) quantity on the peak alignment
of falling nuclear H*+L pitch accents. Disyllabic target words (C 1
V 1 C 2 V 2 ) only differing in either the quantity of V 1
(VQ-words) or C 2 (CQ-words) were embedded in two different carrier
sentences: in one carrier sentence the target word was followed by
two unstressed syllables (long tail context) and in the other by
none (short tail ixcontext). All target words occured in three
quantity degrees: short (Q1), long (Q2) and overlong (Q3). There
were two main results: (1) In the short tail context the peak was
aligned earlier. (2) The peak alignment of VQ- and CQ-words was
similar. Quantity degree differences of both VQ- and CQ- words were
cued by the peak alignment in proportion to the V 1 C 2 -duration.
The proportional peak alignment had the order Q3 < Q2 < Q1,
where < denotes that the peak of Q3-words was proportionally
timed earlier than the peak of Q2-words and so on. The second
experiment (chapter 3) analysed the influence of the sentence
boundary, i.e. phrase-final lengthening (PFL), on the segment
durations of VQ- and CQ-words. The data used for the analysis was
the same as in the first experiment. There were two main results:
(1) The domain of PFL in Estonian was the main bearer of the
quantity contrast, i.e. V 1 in VQ-words and C 2 in CQ-words and can
be best accounted for in terms of a Structure- based model for
explaining PFL (Turk and Shattuck-Hufnagel, 2007). (2) Progressive
lengthening, i.e. the nearer a segment is to the final boundary the
more it is lengthened, occured in the data if the lengthened
segments were not in adjacent word-final position. The third
experiment (chapter 4) investigated whether speaking rate in-
fluences the alignment of the peak. VQ- and CQ-words were embedded
in carrier sentences with one unstressed syllable following the
target word. They were read in normal and fast speaking rate. There
were two main results: (1) In both VQ- and CQ-words the peak
alignment in proportion to the V 1 C 2 -duration had the order Q3
< Q2 < Q1, where < denotes that the peak of Q3-words was
aligned earlier than the peak of Q2-words and so on. (2) Speaking
rate did not influence the peak alignment in proportion to the V 1
C 2 -duration. The results of this dissertation favour in
interpretation in the sense of the segmental anchoring hypothesis
(see e.g. Ladd et al., 1999, 2000; Schepman et al., 2006) that
tonal targets are anchored with specific points of the seg- mental
string. The results of the current dissertation created the
impression that in Estonian the offset of the first mora could be
the anchorpoint for the peak - regardless of quantity degree and
type. Differences in the proportional peak alignment emerge because
the anchorpoint interacts with the temporal correlates of the
quantity contrast. Compatibly with Ladd (2008), the results of the
dissertation also show that peak alignment in Estonian is
influenced by phonologically-induced (an increase in the number of
post-focal syllables) but not phonetically-induced (faster speaking
rate) time pressure.