The effects of intonation on acoustic properties of fricatives Marzena Żygis 1 , Daniel Pape 2 , Luis M.T. Jesus 2,3 , Marek Jaskuła 4 1 Centre for General Linguistics (ZAS) & Humboldt University, Berlin, Germany 2 IEETA, University of Aveiro, Portugal 3 ESSUA, University of Aveiro, Portugal 4 West Pomeranian University of Technology, Szczecin, Poland zygis@zas.gwz-berlin.de,danielpape@ua.pt,lmtj@ua.pt,Marek.Jaskula@zut.edu.pl ABSTRACT This study investigates acoustic properties of the Polish retroflex and alveolo-palatal fricatives and affricates /ʂ/, /ʈ͡ʂ/, /ɕ/ and /t͡ɕ/. The sibilants were produced in (i) yes/no questions with rising intonation and (ii) answers with falling intonation. Multitaper spectra were used to compare the two places of articulation. Results show that the Centre of Gravity (COG) did not differentiate retroflexes from alveolo-palatals in the question condition. The same consonants, however, display significantly different COG values if they are produced in the statement condition. Other spectral measurements (highest spectral peak, standard deviation, kurtosis) were also significantly different with regard to the two places of articulation and intonation types. The results indicate that special attention should be given to intonational differences when acoustic properties of sibilants are investigated. Keywords: segment-prosody interaction, sibilants, intonation 1. INTRODUCTION Much attention has been devoted to the investigation of spectral properties of fricatives in general, and sibilants in particular. Among various parameters, the spectral centre of gravity (COG) is probably the most frequent parameter used in the acoustic analysis of fricatives [1], [2], and successfully distinguishes sibilants produced at different places of articulation [3]. At the same time, studies which focus on spectral properties of fricatives have not taken into consideration intonational properties of the words/sentences in which the segments were embedded, following a still common tendency to investigate segments and intonation separately. More recently, however, some evidence was provided from the perspective of intonational research, according to which intonation patterns are not only encoded in vocalic but also in consonantal elements of speech. In particular, it was shown that sibilants contribute to differences in intonation by changing their spectral properties, e.g. higher COG was found for rising and lower COG for falling intonation [4], [5], [6]. The present paper seeks to provide an answer to the question of how spectral properties of sibilants are affected by intonational properties. Although linked in a sense to previous studies, the present investigation has a different aim. It does not analyse the contribution of sibilants to intonation patterns, but rather makes a point about the role of intonation on the spectral properties of sibilants. The main aim of the study is, thus, to show that spectral properties of sibilants – classically seen as invariant properties – might be seriously affected, and to some extent even blurred if intonation is not controlled for. 2.METHODS 2.1. Corpus In order to test the influence of intonation on spectral properties of sibilants, we investigated Polish phonemic retroflex and alveolo-palatal sibilant clusters consisting of (i) retroflex fricatives and affricates /ʂʈ͡ʂ/ and (ii) alveolo-palatal fricatives and affricates /ɕt͡ɕ/. The clusters always appeared in word-final position in sixteen words (eight words for each cluster). The words were monosyllabic and occurred in the sentence-final position which in turn differed in intonation contour: (a) rising intonation in yes/no questions and (b) falling intonation in statements: a) Widzi tę kość [kɔɕt͡ɕ]? Does he see this bone? b) Widzi tę kość [kɔɕt͡ɕ]. He sees this bone. Eight male native speakers of Standard Polish read all sentences three times. Sentences with both cluster types (divided into two sets) were randomised together with sentences containing bisyllabic words. In total we obtained 1536 items to analyse (16 words  2 sentence types [statement, question]  2 sibilant types [fricative, affricate]  3 repetitions x 8 speakers).