This dataset contains 3 data files, 5 files with R code, and a short read-me file with documentation. The data files contain information about the development of two competing constructions in Russian temporal adverbials. The files with R code give the code for analysis of the databases.

ARTICLE ABSTRACT: What can a corpus do for the historical linguist? How can corpus data shed light on the diachronic development of so-called rival forms, i.e., words or grammatical constructions that appear to be synonyms? This article addresses these questions based on a detailed empirical analysis of two seemingly synonymous constructions in Russian. Corresponding to the English ‘decade construction’ in the twenties, Russian has two rival constructions, viz. v dvadcatye gody [lit. “in the twentieth years”] (with the numeral and noun in the accusative) and v dvadcatyx godax (with the numeral and noun in the locative case). Three hypotheses about rival forms are considered: leveling (whereby one form ousts its rival), sociolinguistic differentiation (whereby the two rivals survive in different varieties of a language) and semantic differentiation (whereby the two rivals develop different meanings over time). Contrary to what has been suggested in the literature, we find little evidence for semantic and sociolinguistic differentiation. Instead, we demonstrate that leveling is taking place, since the accusative construction is in the process of ousting its rival. While our study shows that corpus data facilitate detailed analysis of the interaction between leveling, sociolinguistic differentiation and semantic differentiation, our analysis also points to limitations, especially when it comes to corpus-based analysis of sociolinguistic and semantic factors.