1 Social and phonological dimensions of /l/‑ vocalization in West Australian English 1 Simon Gonzalez, Nathaniel Mitchell & Gerard Docherty Griffith University, Australia s.gonzalez@griffith.edu.au, n.mitchell@griffith.edu.au, gerry.docherty@griffith.edu.au /l/‑vocalization is commonly reported to be a prevalent characteristic of varieties of English spoken in Australia [1][2]. To date, the analysis of the /l/ and /l/‑vocalization in Australian English has tended to be auditory or articulatory in nature [3, 4]. A significant challenge posed by the automatic acoustic analysis of /l/‑vocalizations is their intermediate vowel‑consonant nature [5, 6]. As a result, we know relatively little about the acoustic properties of these realizational variants, and, because of the necessary limitations on sample size imposed by articulatory techniques, we know little beyond that which is available through auditory analysis about how variability in /l/‑vocalization is distributed across a larger sample of speakers of a particular variety of English. This paper reports findings from an acoustic analysis of the realization of post‑ vocalic /l/ in the conversational speech of young speakers of West Australian English (from the Perth metropolitan area – a location which has not previously been the object of sociophonetic investigation). The results are derived from our analysis of six pairs of same‑sex speakers recorded in unscripted conversations of around 30 minutes in duration. Our acoustic analysis sets out to classify /l/ variants along a consonantal‑vocalic continuum. Following [7], we base our acoustic characterization of /l/ realizations on statistical modeling of formant trajectories in vowel‑/l/ rhymes, using a dynamic approach to data collection coupled with a cluster analysis statistical model. Recordings were segmented in Elan [8] and subsequently force‑aligned within LaBB‑CAT [9], using HTK [10] for correcting misalignments in Praat [11]. Using the R package [12] PraatR [13], formants tracks were extracted. We included tokens of /l/ preceded by monophthongs before pauses or followed by consonants. We excluded contexts with following approximants in order to avoid formant dynamic trajectories beyond the offset of /l/. Adapting the methods described in [7] and [14], we tracked formant trajectories from the onset of the preceding vowel to the offset of the /l/. A total of 15 equidistant points were extracted for each interval. We analyzed the middle 80% (12 points) of the trajectory. We subsequently carried out a cluster analysis of formant trajectories, thereby allowing us to organize data into groups clustered around intrinsic data patterns, independent of categorical classifications. In this case, the analysis divides 1 Gonzalez, S., Mitchell, N. & Docherty, G. (2015). Social and phonological dimensions of /l/‑ vocalization in West Australian English. Paper presented at NWAV 44 New Ways of Analyzing Variation, 22‑25 October, Toronto, Canada.