130 likes | 231 Views
Word order and tonal shape in the production of focus in short Finnish utterances. Martti Vainio 1 , Juhani Järvikivi 2 and Stefan Werner 3 1 University of Helsinki, 2 University of Turku, and 3 University of Joensuu, Finland. INTERSPEECH 2006, September 18. 2006 – Pittsburgh. Background.
E N D
Word order and tonal shape in the production of focus in short Finnish utterances • Martti Vainio1, Juhani Järvikivi2 and Stefan Werner3 • 1University of Helsinki, 2University of Turku, and 3University of Joensuu, Finland INTERSPEECH 2006, September 18. 2006 – Pittsburgh
Background • A former study on perception of prominence in Finnish showed that the perceived prominence is modulated by word order. • With regard to tonal shape the perception followed a flat hat-pattern: • The most important factor being the absolute difference between two f0 peaks, • With regard to a non-final word the size of the f0 rise was important – not the fall, • Respectively, the fall was more important than the rise with regard to a final word. • A production experiment was designed with similar materials to test hypotheses formed according to the findings in the perception test.
Word order • Changing the word order in an utterance “Menemme laivalla Lemille” (We go by boat to Lemi) from an unmarked to marked “Menemme Lemille laivalla” by itself can be used to focus the word “laivalla” by forming a presupposition that we did go by boat rather than, say, by car. • Additionally, prosody can be used to focus any constituent even in the unmarked case and regardless of word order. • Thus a Finnish speaker can say “Manne meni Lemille” as well as “Manne meni Lemille”. • Important question is, then, whether these two means – syntactic and prosodic – interact in production.
Tonal shape • Rather than having a “sagging” dip we predicted the production of tonal shape would follow a “flat hat-pattern”.
Materials • The sentence “Menemme laivalla Jimille” (We go by boat to Jimi’s) allows for three three different focus conditions with regard to the two nouns: narrow focus on either or broad focus, where neither of the nouns is more prominent than the other. • With three focus and two word order conditions, a set of 36 different sentences were created. • The intended focus condition was elicited from the speakers by providing them with three different spoken prompts that they had to reply: • Broad: “What do you do today?” • Narrow on “laivalla”: “How do you go to Jimi’s?” • Narrow on “Jimille”: Where do you go by boat?” • Eight speakers read the replies recorded by a female speaker from a sheet of paper with no additional marking for focus. The materials were recorded twice: with and without emphasis on the question prompt. • The materials were recorded in a sound proof recording studio with a high quality microphone and ADC.
Results • Produced items were labeled and f0 turning points corresponding to the ones used in the perception experiment were marked manually. 6
Measured differences • Differences between points marked with letters a,b, and c were used for analyses (semitones).
Tonal shape • Logistic regression was used to test the three hypotheses (non-linearities were modeled with restricted-cubic-splines): • Global peak height difference the most important, • The rise of the first peak more important than other local excursions, • The fall of the last peak more important than other local excursions. • Results 2: • Peak difference: Χ2(1)=63.94, p < .00001 • First peak rise: Χ2(1)=7.76, p < .0053 • First peak fall: Χ2 (1)=2,27, p < .109 • Results 3: • Peak difference: Χ2(1)=77.47, p < .0001 • Last peak rise: Χ2(4)=19.27, p < .0007, non-linearity: Χ2(3)=8.87, p < .03 • Last peak fall: Χ2(4)=12.49, p < .014, non-linearity: Χ2(3)=11.54, p < .009
Word order • Word order and emphasis on questions were investigated with 2x2x3 ANOVAs using peak difference in semitone as the dependent measure: averaged over subjects (F1) and items (F2). • Clear main effect of FOCUS: F1(2,14)=70.28,p<.001;F2(2,20)=521,22,p<.001. • Main effect of EMPHASIS: F1(1,7)=8.97,p<.001;F2 non-significant. No interactions (p’s>.09). • Similar results for separate analyses for emphasis conditions; except for a significant interaction between word order and focus: F1(2,14)=3,48,p<.059;F2(2,20)=4.17,p<.031. • Pairwise t-test within the broad condition revealed that the difference of the marked and unmarked word order condition was significant: t1(7)=2.51,p<.05;t2(10)=1.78,p<.05. • The top-line difference is raised or lowered to compensate for the extra focus from syntax in broad focus condition.
Conclusions • f0 is modulated by syntax and prosody in Finnish in a fairly complex and subtle way. • The production of prosody follows the non-obvious pattern with regard to perception of prominence and forms a flat hat pattern which is, however, not realized as such due to reasons beyond this study.