FACTORS OF VARIATION IN THE PRODUCTION OF THE GERMAN DORSAL FRICATIVE B. Pompino-Marschall and Ch. Mooshammer Allgemeine Sprachwissenschaft, Typologie und Universalienforschung, Geisteswissenschaftliche Zentren Berlin e.V. . 10/11, D-10117 Berlin, Germany. Tel.: +49 30 20 192 419, FAX: +49 30 20 192 402, E-mail: bobby@fas.ag-berlin.mpg.de ABSTRACT The articulatory variability of the Standard German (voiceless) dorsal fricative / / - surfacing as [ ], [ ] or [ ] depending on the position within the syllable and the vowel context (cf. [2, 3]) - is analysed by using elec- tropalatographic and electromagnetic articulographic as well as averaged spectral data. The results are compared to those for the coronal fricatives and the velar plosive in German. 1. INTRODUCTION As has been shown (cf. [1]) phonemic variation in place of articulation should be distinguished from coarticula- tory/assimilative allophonic variation (e.g. fronting of velar stops). In Standard German there is a complementary distri- bution of the allophones of the (voiceless) dorsal fricative / /: [ ] occurs after front vowels and consonants as well as morpheme initially whereas [ ] (and [ ]) occur after high (and nonhigh) back vowels. In the following study we want to compare this varia- tion in place of articulation from palatal to velar (and even uvular) (1) with the coarticulatory ef fect of vowels on following coronal fricatives as well as (2) with the fronting of the velar stop depeending on the identity of the following vowel. 2. METHOD 2.1. Material and procedure The recorded material consisted of German words embedded in the carrier phrase Ich habe __ er (‘I’ve mentioned __’). These words varied systematically with respect to (lax) vowel ([ , , , , ]) and target con- sonant ([ , , / / , ]); where [ ] and [ ] where only pro- duced directly following the vowel (e.g. fis [ ] (‘F sharp’), fesch [ ] (‘smart’)); [ / / ] (e.g. mich [ ] (‘me’), Bucht [ ] (‘bay’), Koch [ ] (‘cook’)) also after a preceding coronal [ , ] (e.g. Milch [ ] (‘milk’), manch [ ] (‘many a’)) or a preceding rhotic coda consonant (e.g. Storch [ ] (‘stork’)). For every vowel series one of the word items was chosen to start with [ ] (e.g. Kuß [ ] (‘kiss’)). In a pilot study this material was recorded with five repe- titions in randomized order by means of the Reading EPG.3 system both palatographically and acoustically by two native German speakers, BPM (male) and TMO (female). The same material was recorded with the male subject (BPM) of the pilot study and two further male native German subjects (JDR, PJA) with simultaneous recordings of tongue movements by a Carstens AG100 articulograph with 5 sensor coils mounted midsagittally on the tongue surface 1 cm from the tip (tongue blade - TB), at the point opposite to the border between hard and soft palate (tongue dorsum - TD) and halfways in between (predorsally - PD). Reference coils were placed at the upper incisors and at the bridge of the nose. 2.2. Data analysis The EPG data was analyzed with respect to the position of the centre of gravity of the area of linguopalatal con- tact at the point of maximal constriction at the temporal centre of the obstruent articulation calculated with refer- ence to row number (1-8, counting from front; cf. CGR in the tables) as well as to distance from the inner edge of the upper incisors (cf. fi g. 1; CGD). The second EPG parameter was the weight of these centres of gravity (expressed in numbers of electrodes contacted; WR/WD), the third maximal contacts within a single row of EPG- electrodes (MC). The EMMA data of the consonantal articulations were analyzed with respect to their extreme positions of the PD coil at the articulatory obstruent centre (defined as the minimum in the tangential velocity function of this coil; cf. MAXX/MAXY in table II). This coil was chosen since the articulation of the dorsal fricative was the main focus of this study . This choice will clearly somewhat distort the results for the coronal fricatives as well as for the velar stops since their extreme articulatory positions better should be registered with reference to the TB or TD coil respectively. Besides the measurement of consonantal duration, the fricative productions were acoustically analysed with respect to their average spectrum during the steady state phase with the help of the Signalyze software for Apple MacIntosh (300Hz broadband without preamplification). Statistical analysis was done for each subject individually with the help of the ANOVA procedure of the StatView software for Apple MacIntosh. Single ANOVAs were cal- culated for the EPG and EMMA parameters: (1) compar- This document was created with FrameMaker 4.0.4