Presentation on theme: "18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 1 Are there “Shapers” and “Aligners” ? Individual differences in signalling pitch accent category."— Presentation transcript:
18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 1 Are there “Shapers” and “Aligners” ? Individual differences in signalling pitch accent category Oliver Niebuhr Mariapaola D’Imperio Barbara Gili Fivela Francesco Cangemi Special Session “Shapes and tones – Towards a more holistic perspective in intonation research” 17th International Congress of Phonetic Science, Hong Kong, China, August 2011
18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 2 Individual speakers employ different production strategies towards a similar acoustic/perceptual goal. Talker variability is at least partly determined by anatomical features but linguistic factors such as dialectal and speech community membership give rise to noticeable variation. The relative strength of perceptually trading acoustic cues can differ from speaker to speaker, at least for segmental variability (Jonhnson, 1997). But what about prosodic variability? What about tonal alignment? Introduction
18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 3 Strict segmental anchoring (Ladd et al. 1999): tonal targets are timed to occur at a fixed distance from segmental boundaries irrespective of rate changes and/or syllabic structure differences for a specific pitch accent type. If tonal alignment is variable on an individual basis, can it be in a trade- off relationship with an alternative cue to the same pitch accent category? What about dynamic cues, such as “accent shape”? But individual variability in alignment production is rarely pointed out. Here the issue is investigated through a database for three language varieties (Standard German, Pisa and Neapolitan Italian) and three pairs of well studied pitch accent contrasts.
18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 4 Method 3 well-established PA contrasts: 10 short A-B dialogues H+L* vs. H* in Standard Northern German, L+H* vs. L*+H in Neapolitan Italian and H* vs. H*+L in Pisa Italian Recordings made in sound-treated rooms at the Universities of Kiel, Naples, and Pisa. The German dialogues were read by 35 speakers; the Italian recordings included 17 Neapolitan and 20 Pisa speakers.
18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 5 German intonation contrast Two types of nuclear pitch accents: ‘Early Peak’ vs. ‘Medial Peak’ or H+L* vs. H* according to Grice and Baumann (2002, GToBI) Well-known from the seminal experiments of Kohler (1987), and numerous other production and perception studies –H+L*= F0 peak reached before accented-vowel onset, F0 falls into vowel –H* = F0 peak reached 30-60 after accented vowel-onset, falls sets in before vowel offset Kohler (1991:341) Grice and Baumann (2002:25)
18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 6 Results for German I Intonation labels were cross-checked by two other trained prosodic labellers H+L* and H* productions were first analyzed in terms of overall mean values (x) and standard deviations (s) for –F0 peak range (semitones, st) –Shape index (duration rise / duration fall) –Peak alignment relative to accented vowel onset
18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 7 Results for German II t F0 Number of cases = 525 CV: x= -29 ms, s=55 ms x= 68 ms, s=78 ms Accented [CV:] syllable H+L* Shape index: x=1.31, s=0.33 H* Shape index: x=0.83, s=0.25 (nuclear, voiced initial syllable of disyllabic noun, followed by voiced CVC syllable + 2 unaccented syllables until end of phrase) x= 10.5 st, s= 1.8 st x= 10.6 st, s= 1.7 st Mean alignment values are congruent with those of previous studies.
18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 8 BUT: individual production patterns blur this picture –(1) Across all 35 speakers, peaks are not solidly anchored relative to segmental boundary. Rather they form two alignment continua. –(2) The alignment continua of H+L* and H* overlap substantially. H* peaks could be aligned before the accented-vowel onset (up to 100ms earlier than H+L* peaks) H+L* peaks could occur after the vowel onset. –Trade-off between peak alignment and shape index, reflected in a highly significantly negative correlation (r= -0.72; p<0.001) - align distance+ align distance
18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 9 Results for German IV –Based on this trade-off… –5 speakers showed clearly different alignment before (H+L*) or after (H*) the vowel onset. Shapers Aligners t test shape: p<0.001; t test alignment: p<0.001 –For 5 other speakers, alignment was completely overlapping but showed strong shape differences…
18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 10 Nuclear, focus interrogative Peak aligned around accented syllable offset Nuclear, narrow focus declarative Peak aligned around accented syllable onset Neapolitan Italian intonation contrast L+H* vs L*+H Their shape and function have been established by means of analyses in production and perception (e.g., D’Imperio 2000, D’Imperio, 2002, Petrone, 2008)
18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 11 Considerable inter-speaker differences in peak alignment. A set of 6 speakers made little or no use of systematic alignment greater shape differences Results for Neapolitan Italian Alignment differences between L+H* and L*+H by speaker
Discussion & Conclusion 18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 12 When viewed superficially in terms of overall mean values, our results replicate previous findings on peak alignment in differently timed pitch accent movements. However, when broken down into the individual patterns, the clearly bimodal production picture appears to be blurred. Peak alignment relative to vowel onset forms a continuum in all three varieties under investigation, if individual patterns are taken into account. Also, systematic alignment is either in a trade-off relationship with peak shape (as in German and in Neapolitan) or in an enhancing relationship (as in Pisa).
Discussion & Conclusion 18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 13 Future studies will determine, if the production strategies are correlated with indexical features, i.e. in our data female speakers are mostly “shapers”. Studying the production and perception of intonational cues must be complemented by a closer look at talker variability, in addition to variety induced variability. Only perception studies will be able to determine whether tonal target alignment can be compensated for or even substituted by more dynamic accentual features, such as peak and rise/fall shape.
18.08.2011 Niebuhr, D‘Imperio, Gili Fivela, Cangemi 14 Thank you for your attention