VARIATION IN PHONETIC REALISATION OR IN PHONOLOGICAL CATEGORIES? INTONATIONAL PITCH ACCENTS IN EGYPTIAN COLLOQUIAL ARABIC AND EGYPTIAN FORMAL ARABIC Sam Hellmuth & Dina El Zarka Universität Potsdam, Universität Graz [email protected], [email protected] This paper uses qualitative and quantitative methods to compare the intonation of formal and colloquial varieties of Egyptian Arabic in a corpus of elicited read speech, to explore the widely held assumption that spoken formal Arabic will have the intonational characteristics of the speaker’s colloquial variety. Speakers are found to use broadly parallel phonological systems in each register, reflected in parallel distribution and type of pitch accents. A quantitative analysis of the pitch target alignment to the segmental string reveals only minor differences in the phonetic realisation of pitch accents across registers. Keywords: Arabic, intonation, sociolinguistic variation, pitch accent alignment. 1. INTRODUCTION This paper explores the hypothesis that the intonation of spoken formal Arabic (‘fusha’) will display the same intonational properties as those of the mother tongue dialect of the speaker (‘aammiyya’). This hypothesis arises from the commonly held assumption that, in general, the phonological (and phonetic) properties of spoken formal Arabic are transferred from the speaker’s colloquial dialect. For example, although all spoken Arabic dialects display quantity sensitive stress assignment, the exact rules for assigning primary stress vary from dialect to dialect [12]; in spoken formal Arabic speakers are expected to apply the particular stress assignment rules of their own mother tongue dialect [4]; studies have shown this to be the case for Egyptian Arabic [e.g. 19]. The distinction between standard and spoken varieties of Arabic is complex, and has been the subject of much attention in the literature [3, 7]. We adopt here Mitchell’s three way classification [20] whereby the primary distinction is between formal and informal Arabic, with a sub-division of informal Arabic into ‘careful’ and ‘casual’ registers. Our study addresses possible intonational differences between Egyptian Formal Arabic (EFA), defined as the rendition of formal Arabic (‘fusha’) by Egyptians, and Egyptian Colloquial Arabic (ECA, ‘aammiyya’). Cross-linguistic intonational variation is itself a relatively new area of research, although a number of possible parameters of intonational variation have been suggested [8-11, 15, 16]. Ladd [16:119] proposes a taxonomy of intonational variation in four categories: semantic (different meaning/use of phonologically identical contours), systemic (different inventory of phonologically distinct contours), realisational (differences of detail in the phonetic realisation of what is phonologically the same contour) and phonotactic (differences in contour-text association). Variation between EFA and ECA could occur in any of these four areas, however semantic and/or systemic differences require investigation in a large corpus of (preferably) spontaneous speech, which is beyond the scope of the present study. We thus focus here on the search for possible variation in just two categories: realisational differences, which we suggest would constitute evidence of phonetic variation between the two varieties, and phonotactic differences of contour-text association, to yield evidence of potential phonological variation. To this end our study comprises both qualitative and quantitative analysis. In particular we explore potential differences in the alignment of tonal events (L and H pitch targets) to the segmental string, since earlier studies suggest H peaks may align earlier in EFA (within the stressed syllable [24]) than in ECA (early in the second mora of the stress foot [13]) as illustrated in Figure 1. Figure 1: Schematised peak alignment in CV syllables in EFA and ECA, as observed in prior studies. EFA ECA stressed syllableAn extensive body of research on pitch target alignment indicates broad cross-linguistic similarities, such as a tendency towards more consistent alignment to the segmental string at the beginning of the syllable than at the end [22], but also extensive cross-linguistic variation in the fine detail of alignment patterns [1, 17]. For example, in a study of alignment in varieties of German, Atterer & Ladd [2] found consistent differences of alignment between Northern and Southern German, which they analyse as cross-dialectal variation in the phonetic realisation of a single phonological object (a pre-nuclear rising pitch accent), both of which are however quantitatively different from alignment patterns in, say, English. Indeed, in a corpus of speech in English produced by the same two sets of German speakers, German-like alignment patterns were found, and speakers did not produce English-like alignment patterns. The fine detail of alignment patterns can thus in principle serve as an indicator of realisational differences within a single phonological category (as the German North/South distinction) or of the transfer of phonotactic contour-text association patterns from one variety to another (as in the survival of dialectal alignment differences in speakers’ L2 productions). We therefore take the fine detail of tonal alignment in EFA and ECA to be a good potential source of evidence of variation in phonetic and/or phonotactic realisation across the two varieties, and the body of research into alignment variation allows us to investigate the question using established methodology. In the remainder of the paper we set out the details of the materials and analysis employed (§2), a survey of the qualitative and quantitative results (§3), a discussion (§4) and a brief conclusion (§5). 2. METHODOLOGY To test for variation in phonetic realisation between EFA and ECA (with alignment of pitch targets to the segmental string as our dependent variable), pairs of target words were sought containing parallel segmental content in the stressed syllable in each variety. Target stressed syllables elicited were of three types: short open (CV), short heavy (CVC) and long heavy (CVV). All target words contained the vowel [a] as the stressed vowel. To facilitate location of F0 events in the pitch track, we sought target words with sonorant consonants around the stressed vowel, but due to the limited number of lexical items meeting these