Abstract
We describe the principal features of the Database of Loanword Pronunciation Variants (http://dvucs.ff.cuni.cz), realised with the aim of bridging a long gap in the empirical study of loanword phonology in Czech. The Database contains pronunciation forms obtained for 300 phonetically problematic loans and proper names from a balanced sample of 300 Czech speakers.
After considering the general principles that underlie the phonological adaptation of loanwords, we describe the design of the Database and the methodology used. Although the validity of the data for the study of loanword phonology is undoubtedly high, we note some factors that might cause bias, such as the influence of spelling or the item selection procedure.
The average number of different pronunciation variants per item is 8.02 (1.97 after variants with less than 5% frequency have been excluded). 34 different paradigms of phonological variability were identified in the database, some of which are typical of particular source languages. A number of criteria can be used when analysing the data: the phonological phenomenon, word type (e.g. common vs. proper noun), source language, as well as informant features (region, sex, age, education).
Examples of analysis are given for each of these criteria. The Database represents a large sociolinguistic survey which can be used both as an empirical platform for the study of Czech loanword phonology and as a valuable source for lexicographic work.