COMPETING DISTRIBUTIONS OF NOMINAL DIMINUTIVE SUFFIXES IN SLOVENIAN: DO NATIVE SPEAKERS’ PREFERENCES REFLECT CORPUS DATA?

Abstract

The purpose of our experiment was to study the relationship between corpus data and native speakers’ preferences regarding word form variation. Specifically, we investigated how well corpus frequency distributions of masculine nominal diminutive suffixes would be reflected in the preferences of native speakers of Slovenian when choosing suffixes to form diminutives. Based on similar research conducted by Makarova (2014), we expected native speakers’ preferences to be in line with corpus distributions. Our stimuli were 33 masculine noun bases, and the diminutive suffixes observed were grouped into four categories -(č)ek, -ič, -iček, and -(e)c. We consulted Gigafida 2.0, the largest corpus of written Slovenian, to obtain frequencies of attestations for each of our target nouns with each of the diminutive suffixes. The most attested diminutive suffix category for each noun was dubbed its expected suffix. With an offline experiment we collected responses from 70 native speakers of Slovenian, who were tasked with transforming nouns in their base form into diminutives within sentence contexts. Our goal was to test whether native speakers would assign the expected suffix to the noun bases more often than some alternative suffix. The results indicated the number of expected responses exceeded the number of non-expected responses. The difference, however, was subtle, and closer examination of individual noun bases revealed a considerable variability in the shares of expected responses. Further analyses investigated potential predictors of suffix choices, including animacy, length, and expected suffix. Linear regression analyses showed the expected suffix category significantly predicted the proportion of expected responses. Notably, the suffixes -(č)ek and -(e)c emerged as significant predictors of the share of expected responses. The reasons for this variation, including the influence of the corpus characteristics, are discussed, and potential avenues for future research are suggested.