Speech Units Workshop

It has long been known that despite their classification into several unrelated families, the languages of South Asia share many features, spanning syntax, morphology, and (segmental) phonology (Emeneau 1956). It has also been documented that in all of these features, we find regional variation across family lines (Subbārāo 2012). For example, while South Asian languages (SALs) generally use non-nominative case marking for subjects of experiencer verbs (e.g. be hungry), the choice of case varies by region, with most languages (spanning Indo-European, Dravidian, and Austroasiatic) using the dative, and languages in the Eastern Zone (spanning Indo-European and Tibeto-Burman) using the genitive. The current study extends this perspective to intonation, using a novel approach to the prosodic transcription of socially-distantly-gathered recordings of SALs ranging from fairly well-documented Bengali and Urdu to prosodically-undocumented Kannada and Kalasha. The main findings reveal a number of shared features across the region, as well as substantial crosslinguistic variation. The division of utterances into roughly content-word-sized elements like accentual phrases (APs) is universal across the region, and these APs canonically have two tones: one head-marking (on the stress) and the other edge-marking (on the final boundary). In most – but not nearly all – cases, the head-marking tone is centered around a L*, while the edge-marking tone is centered around a Ha. By looking carefully at the resulting AP contour, however, we find clear evidence of variation from a simplistic L* Ha model based on well-researched SALs: some SALs vary in how the AP's rise is realized while others have completely different AP patterns available, often forming regional groupings that cross family lines. The role of lexically-specific tones in languages such as Punjabi are often overlooked in the intonation literature, but are incorporated here as well to introduce further complexity in the system. Overall, this creates a microcosm of what we know of global intonational variation within a region made up of many dialect continua, from heavily stress-marking West Germanic-like prosodic systems in the far north of the Subcontinent to Japanese-like lexical pitch accent systems in the central north to Greenlandic-like systems with little or no effect of tone and stress in the south and east. Furthermore, the current study's approaches to distanced data collection and analysis by non-speakers can also have impacts on how both documentation and typological work can proceed beyond languages most familiar and easily accessible to researchers and their resources