Double-gender and double-declension nouns in Croatian: diachronic and synchronic perspective

Recent studies on morphological doublets (overview, Naghzguy-Kohan and Kuteva 2016) have provided many examples that challenge the assumed existence of the blocking effect or synonymy avoidance in language (e.g. Aronoff 1976, Carstairs-McCarthy 2010). Using corpus data, we present this issue in the light of the nominal inflectional morphology of the Croatian language. In principle, the Croatian language displays a clear correspondence between grammatical form and gender. However, some nouns appear with two genders and two inflectional classes. The nouns researched here end in a consonant in the N.sg. and are attested both in the a-declension (masculine gender) and the i-declension (feminine gender). Ten double-gender and double-declension nouns were selected (bol ‘pain’, čar ‘spell, enchantment, charm’, glad ‘hunger, famine’, trulež i gnjilež ‘rot, decay’, živež ‘provisions’, izrast ‘growth’, varoš ‘town’, pelud ‘pollen’, splav ‘raft’). Relevant historical documents from the onset of Croatian literacy were analysed to describe the one-gender stage of the noun (if possible) and to detect the appearance of the second paradigm. Two relevant corpora were used to determine the ratio between two paradigms: CLC, the Croatian Language Corpus (Ćavar and Brozović Rončević 2012) and HrWaC, the Croatian Web Corpus (Ljubešić and Klubička 2014). The study focuses on five questions: 1. Do these nouns appear in two genders in Croatian texts? 2. Will the transition period end in exclusively one declension (and gender)? 3. Is there a constant ratio between the two inflection classes (and genders) through time? 4. How fast is the change, i.e. do rival patterns disappear rapidly? 5. Is this a case of internal or external change? In most current texts, nine out of ten nouns appea r in both genders in their respective declensions, but they differ in the ratio between the two patterns. While some nouns are rarely used in one of the patterns (1%, 4%), for some the ratio is almost equivalent (44%, 56%). Ten double-gender nouns differ in their historical development. Five of them are attested in the oldest documents written in Croatian, while the others entered the Croatian language during the 19th century. Rival forms of some nouns are attested in both genders almost simultaneously, while others obtained their alternative form in different historical periods (from the 16th to the 19th century). The ratio of the two patterns changed over time. For some nouns, there seems to be a developmental shift from one form to another (e.g. varoš ‘town’ m. decreased from 38% to 4%). The areal distribution of the usage of the two forms shows that alternative forms could be the result of language contact phenomena, either through contact between Croatian dialects or through contact with typologically similar languages in the surrounding area. A detailed analysis shows that double forms do not disappear rapidly in most cases. Instead, our results speak in favour of overabundance (Thornton 2011), i.e. a more flexible approach to the blocking phenomenon in inflectional morphology