Скачать презентацию Experiments concerning boundary tone perception in German 3 Скачать презентацию Experiments concerning boundary tone perception in German 3

a5036873849f79ffc85cd37f9b6a7c76.ppt

  • Количество слайдов: 26

Experiments concerning boundary tone perception in German 3 rd Workshop of the SPP-1234 Potsdam, Experiments concerning boundary tone perception in German 3 rd Workshop of the SPP-1234 Potsdam, 7 th January 2009 Presentation of the Stuttgart Project (Möbius & Dogil) Katrin Schneider

Outline n n n Categorical Perception (CP) Perceptual Magnet Effect (PME) Experimental designs Stimulus Outline n n n Categorical Perception (CP) Perceptual Magnet Effect (PME) Experimental designs Stimulus preparation Experiments: q q n Boundary tone perception in German without context information Boundary tone perception in German including context information (Identification only) Outlook

Categorical Perception (CP) n Test design developed according to the perception of plosives (Repp, Categorical Perception (CP) n Test design developed according to the perception of plosives (Repp, 1984) q n Perception is categorical if the peak in discrimination corresponds to the category crossover found in identification. Experiments in the prosodic research area: q q q House (1996) Kohler (1987, 1990) Ladd & Morton (1997) Remijsen & van Heuven (1999); van Heuven & Kirsner (2004) Schneider & Lintfert (2003) Falé & Hub Faria (2006)

Perceptual Magnet Effect (PME) n Kuhl (1991): q q Phonetic perception is influenced by Perceptual Magnet Effect (PME) n Kuhl (1991): q q Phonetic perception is influenced by language experience PME: discrimination ability differs inside a category n n prototype (P) attracts immediate neighbors – low discrimination performance around a non–prototype (NP) – better discrimination performance

Experimental designs n Testing for CP and PME simultaneously q q Identification: assign stimulus Experimental designs n Testing for CP and PME simultaneously q q Identification: assign stimulus to one of the given categories Goodness rating (only for PME): n n q separately for each category found in identification How well does the presented stimulus fits into the assigned category? scale given Discrimination: n n Does the presented stimulus pair consists of identical or of different stimuli? differences in the construction of stimulus pairs between CP and PME test design

Stimulus preparation n n Test for German boundary tones (BT) Test stimulus: q q Stimulus preparation n n Test for German boundary tones (BT) Test stimulus: q q q n ambiguous between statement/question interpretation no syntactic bias: PP n noun consisting of mostly sonorants; no /ə/ n pitch accent not on the last syllable n polysyllabic noun; no compound noun “ins kalte Panama” (in the cold Panama), embedded in carrier sentences male native German speaker

Stimulus preparation n Manipulation of fundamental frequency (F 0) of the last 2 syllables Stimulus preparation n Manipulation of fundamental frequency (F 0) of the last 2 syllables of the target PP: q q q n Calculation F 0 range: mean rise to H%: 90 Hz; mean fall to L% : 50 Hz ERB (Equivalent Rectangular Bandwidth) scale used PSOLA resynthesis Headphones used during each test

Additional stimuli for PME test H% H*L 0, 338 ERB L% Additional stimuli for Additional stimuli for PME test H% H*L 0, 338 ERB L% Additional stimuli for PME test

Perception of German boundary tones (without context) n Results (25 subjects (10 females)): q Perception of German boundary tones (without context) n Results (25 subjects (10 females)): q q s-shaped curves in identification; 2 categories CP and PME for the statement category: n n q unclear results for the question category: n n q clear prototype & non-prototype (goodness rating) disc. peak & warping of perceptual space around P better within-category discrimination performance than in statement category → no clear CP, but also no gradual perception individual prototypes differ → no PME problematic: combination of CP and PME test design in discrimination

Perceptual reference space Perceptual reference space

Perception of German boundary tones (including context information) n Why? q q n context Perception of German boundary tones (including context information) n Why? q q n context might have an influence on the location of the category boundary there is always context in normal conversation – no “out-ofthe-blue” sentences How? q 2 different context types we tested for: n n q BT height of the preceding sentence (influence of F 0)) syntactic structure of the preceding sentence (influence of syntax) same speaker as in previous experiment

Contexts: fundamental frequency (F 0) q Does the F 0 of the preceding sentence Contexts: fundamental frequency (F 0) q Does the F 0 of the preceding sentence influences perception? n preceding sentences: statement vs. question; only H% difference in BT height a) L% condition “Er will verreisen. b) L% Nach Panama. /? ” L% H% H% H% condition “Er will verreisen? L% Nach Panama. /? ”

Contexts: syntax q Does the syntactic structure of the preceding sentence influences perception? n Contexts: syntax q Does the syntactic structure of the preceding sentence influences perception? n preceding sentences: synt. statement vs. synt. question; differ in their syntactical constructions H% a) L% condition L% “Er will verreisen. c) L% Nach Panama. /? ” H% Wh_L% condition “Was liegt da? L% L% Ein Ticket nach Panama. /? ”

Identification n n 20 stimuli in each condition; 10 repetitions = 600 stimuli; randomized Identification n n 20 stimuli in each condition; 10 repetitions = 600 stimuli; randomized order 3 subtests: each incl. 6 training stimuli & 200 test stimuli 36 participants (23 females, 13 males) Exclusion of high reaction times (RT): q outliers > 2*sdev+mean n n RT outliers > 2, 374614 2, 5% of the data excluded

Identification: general results n Results: q clear s-shaped curves for all contexts L% H% Identification: general results n Results: q clear s-shaped curves for all contexts L% H% Wh_L%

Identification: general results n Results: q clear s-shaped curves for all contexts q significant Identification: general results n Results: q clear s-shaped curves for all contexts q significant differences n n n inside the BT-height condition as well as inside the syntax condition there seems to be an influence on the location of the category boundary in each context condition in the intended direction

Identification: gender differences n Gender differences: q in each single context condition as well Identification: gender differences n Gender differences: q in each single context condition as well as pooled over all data, females: n show an earlier crossover than males condition Wh_L% pooled over all H% condition contexts L%

Identification: gender differences n Gender differences: q in each single context condition as well Identification: gender differences n Gender differences: q in each single context condition as well as pooled over all data, females: n show an earlier crossover than males n are significantly faster than males pooled over all H% condition contexts L% condition Wh_L%

Identification: gender differences n Gender differences: q in each single context condition as well Identification: gender differences n Gender differences: q in each single context condition as well as pooled over all data, females: n show an earlier crossover than males n are significantly faster than males n no sign. differences inside each category (‘statement’ vs. ‘question’)

Identification: gender differences n Differences in context conditions: q males: n only BT height Identification: gender differences n Differences in context conditions: q males: n only BT height influences location of category boundary L% H% L% Wh_L%

Identification: gender differences n Differences in context conditions: q females: n only syntax influences Identification: gender differences n Differences in context conditions: q females: n only syntax influences location of category boundary L% H% L% Wh_L%

Identification: RT n RT as indicator for category boundary? q males: nearly perfect match Identification: RT n RT as indicator for category boundary? q males: nearly perfect match in all contexts H% condition L% condition Wh_L% condition

Identification n RT as indicator for category boundary? q females: nearly perfect match (except Identification n RT as indicator for category boundary? q females: nearly perfect match (except H% condition) Wh_L% condition L% H% condition

Summary Identification n significant gender differences with respect to the q q n category Summary Identification n significant gender differences with respect to the q q n category boundary location reaction times significant differences between presented context conditions, however q gender of the participant has to be taken into account

Outlook n finish the analyses of this experiment: q q q n Goodness rating Outlook n finish the analyses of this experiment: q q q n Goodness rating & CP discrimination: on the poster PME discrimination analysis: currently running Perceptual reference maps & comparison to withoutcontext results further experiments: q q Finish experiment with female voice & compare possible gender-specific behavior to male-voice experiments Experiments concerning pitch accents (preparation finished) & analyze the results

Thank you! Questions? Comments? Suggestions? … Thank you! Questions? Comments? Suggestions? …