CRC 1412/2: Register in the cross-linguistic diachrony of logical particles (SP B06)

We investigate and model the interaction between language change and register using computational and formal-theoretical models with the goal of integrating register knowledge within the model of language change of H. Kauhanen & Walkden, 2018. We initially focus on the historical semantics of conjunction: Old Indo-European languages had at their disposal two types of conjunction makers: e.g., in Latin, the enclitic particle que and the et marker. Across several different strands of the Indo-European language family, the que-type markers die out and the et-type conjunction takes over. The Latin enclitic conjunction marker que (reconstructable as the Proto-Indo-European particle *kwe) is one of several multifunctional logical particles, which can express both conjunction and quantification depending on the semantic environment they appear in (e.g., the universal quantifier meaning of quis-que in Latin, cf. Bortolussi 2013). Our preliminary work shows that this change correlates with register. As and and its equivalents are among the five most frequent words of most languages, we can efficiently gain a broad view of the entire language family by investigating it in detail. In order to do so, the project builds (on) parsed corpora using text data, where a list of factors including register, year of creation, environment, frequency, meaning, and word order can be determined. Our preliminary work considered formal versus informal registers in the history of Latin, an ab initio low-resolution classification which the project will make more precise using computational methods and an analysis of the textual metadata. Based on initial evidence from Latin, we investigate three main hypotheses: (H1) The first hypothesis is that register effects can interrupt historical change from developing in an S-curved fashion. We test this hypothesis by comparing across different branches of Indo-European the historical rate of change with logistic patterns and investigating the effect of register by classifying the texts. (H2) Our second hypothesis concerns the fact that in many modern Indo- European languages, semantic reflexes of *kwe are retained inside phrases with a quantificational meaning, found, for instance, even in modern Italian qualun-que (‘whatever’) despite these being semantically fully compositional in earlier stages. We hypothesize that retention is sanctioned due to the structural size of the expression in which que appears and onto which register attaches. (H3) Our final hypothesis states that the Constant Rate Effect of historical change holds relative only to a single register, namely that language change is S-curved only register-internally. Therefore, to understand and attain a view of ‘smooth’ language change (i.e., one that is statistically logistic in nature), we must pay careful attention to the various registers of a language, for which old Indo-European languages provide a perfectly documented testing ground.