题目：Sifting microvariation data through elementary statistics

主讲人：Diego Pescarini（法国国家科学研究中心教授）

时间：2023-10-27 星期五 20:50-21:35

地点：腾讯会议 579-660-070

主办单位：北京语言大学语言学系、乔姆斯基研究所

摘要：

In recent decades, a considerable effort has been put into reviving existing collections of linguistic data (e.g. atlases such as AIS/ALF or dialectological databases such as ASIt/Thesoc). Such collections are now freely available thanks to the recent open data revolution.

These digital resources, however, collect single manifestations of grammars, whereas, to study grammatical systems, sentences need to be translated (manually) into sets of variables. A variable is a discrete numerical index (e.g. 0/1) that corresponds to a Labovian variable, i.e. a linguistic trait that is “high in frequency”, has “a certain immunity from conscious suppression”, and is “easily quantified” (Labov 1966/1982: 49).

Through elementary statistical analysis (e.g. chi-square tests) we can find out whether/how variables are clustered. The preliminary findings of my pilot studies suggest that clusters of variables are not organized into chains of entailments, as suggested in hierarchical models of Parametric Theory. Instead, clusters tend to revolve around a single primary variable, which in  turn is associated with multiple secondary variables.