ICPhS XVII Regular Session Hong Kong, 17-21 August 2011 1590 PLOSIVE REDUCTION AT THE GROUP LEVEL AND IN THE INDIVIDUAL SPEAKER Nicolai Pharao The LANCHART Centre; the Dept. of Scandinavian Studies and Linguistics, University of Copenhagen, Denmark nicolaip@hum.ku.dk ABSTRACT This study presents results of phonetic factors on the reduction of plosives in spontaneously spoken Danish. Using mixed-effect logistic regression with speaker as random effect factor, it is shown that reduction of /d/ is distinct from reduction of /b/ and /g/. Speaker tendency for reduction relative to the group average is shown to be consistent across place of articulation for /b/ and /g/ for the majority of speakers sampled. Keywords: plosive reduction, Danish 1. INTRODUCTION The reduction of plosives is a well-studied phenomenon in laboratory speech and increasingly in spontaneous speech. But it is relatively understudied in contemporary Danish spontaneous speech. This paper presents the results of an investigation of the variation in the manifestation of the phonemes /b/, /d/ and /g/ in word internal position in spontaneous speech by using the DanPASS corpus [5]. Language internal factors known to affect plosive reduction in other languages [2, 4] are included in order to investigate their effect in Danish. The main question is whether the reduction of plosives can be seen as the result of a general process of reduction or as separate processes pertaining to the manifestation of each phoneme. In other words, is plosive reduction in Danish like spirantization in Spanish where all plosives are reduced to fricatives in particular segmental environments, albeit to different degrees [3]? Or has reduction become phonologized separately for one or more of the three plosives such that the situation is similar to that in varieties of English which show near- categorical flapping of coronals in intervocalic position [6]? In addition to studying plosive reduction in Danish by looking at the aggregate behavior of a group of individuals, the present study also investigates the behavior of individual speakers across the three phonemes. This is done to further study the extent to which reduction of each plosive should be viewed as a separate process, like flapping of coronals in American English, or as a general process of plosive reduction, like spirantization in Peninsular Spanish. Can we expect a speaker to be more or less prone to reduction in general, and therefore assume that if he or she has a relatively high tendency for reducing coronals compared to other speakers, he or she should also have a relatively high tendency to reduce labial and dorsal plosives in comparison with a group average? For plosives at different places of articulation that can be seen to be reduced in a similar way in the same phonetic contexts, it is hypothesized that individuals will show the same tendency for reduction relative to the group mean irrespective of place of articulation, which will support the interpretation of the process as a general phenomenon that applies to both or all three plosives. The probability for reduction may still vary between phonemes, but a speaker’s tendency for reduction relative to the group average should be the same. To investigate plosive reduction in groups and in individual speakers, multiple logistic regression with mixed-effects was used. Phonetic factors were incorporated as fixed effects in the regression models, and individual speakers were included as a random effect factor in the modelling of reduction for each of the three plosives /b d g/. The use of speaker as a random factor allows for control of heterogeneity in speaker behavior [1], but here they will also be examined in order to further understand the behavior of individuals. 2. DESIGN OF THE STUDY The study is based on quantitative analyses of the transcriptions of the dialogues included in the DanPASS Corpus [5].