Dialect geography and dialectometry of Ladakh (UT Ladakh, India)

01/01/2026 to 31/12/2028

The Ladakhi dialects form three greater dialect groups or even languages: 1. Shamskat ‘the lower language’: the dialects of Lower Ladakh, Ldumra (a.k.a. Nubra), and East-Purik. Most distinctive for this group is that transitive agent and possessor are distinguished morphologically. 2. Kenhat ‘the upper language’: the dialects of Upper Ladakh and Zanskar, displaying ergative-genitive syncretism. 3. The Baltipa-Purikpa group as a Shamskat subgroup, lacking a special marker for non-visual perceptions with signifcant differences in the lexicon. Alternative binary classifications are a differentiation between ‘archaic’ dialects that show initial consonant clusters and ‘innovative’ dialects that lack these clusters or a differentiation between a ‘northwestern’ group of dialects without suprasegmental features or tone and a ‘western’ group of tonal dialects, albeit the ±tonal divide crosscuts both groups. The classificatory boundaries vary thus significantly, but none of them is based on statistics. The Project aims to close this lack of statistical data by using a set of more than 100 morphological items and a set of more than 600 everyday vocabulary items.

Computational dialectometry focuses on sound correspondences in cognate word forms. These features are aggregated to compute pairwise distances between dialects. In a subsequent step, these distances are analysed through clustering, multidimensional scaling, and spatial regression via generalised additive models. The results often coincide with traditional isogloss boundaries but are also capable of revealing patterns not readily accessible through qualitative methods. Phylogenetic linguistics focuses on cognate sets, and thus quantifies lexical replacement. Moreover, phylogenetic approaches assume that linguistic diversity arises through repeated splits of language communities, resulting in a branching genealogical structure. The primary goal is to infer the tree—or set of trees—that best explains the observed distribution of features under probabilistic models of evolution. While these traditions are often pursued separately, they can be fruitfully integrated. Features based on automatically inferred sound correspondences are a viable alternative to manually annotated cognate sets in phylogenetic inference. Newer statistical methods factor out the effects of contact when inferring phylogenetic relationships, thus enabling a more nuanced view of linguistic diversification in contact zones such as Ladakh.

With respect to such computerised language measurements and tree modelling, many linguists hold that differences in phonology, lexicon, and morphology are equally important or that the differences in the lexicon, which are naturally higher in number than the differences in morphology, weigh more than the latter. Nevertheless, it is also generally known that innovations in phonology and in the lexicon develop and spread faster than grammatical features. The lexicon is particularly susceptible to influences from prestige or dominant languages or dialects. We thus also want to show how different approaches, e.g., using only the lexical data, only the morphological data, both sets together, as well as different weighing might lead to different results. All results shall be discussed in view of the speakers’ own assessments, geography, climate, and ethno-historical facts to figure out the more convincing ones.