Formulaicity of affixes in Turkish

This study examines whether suffix sequences in a Turkish corpus distribute as units (formulas). Most research on formulaicity focused on word-level formulas. As for affix-level formulas, most evidence for them comes from psycholinguistic studies, whereas there is less evidence from corpus data. This study examines the pattern of cooccurrence of suffixes on verbs in the Turkish National Corpus. To capture formulaicity between suffixes, this study uses a measurement called risk ratio, which is a novel way to measure collocation. The analysis of the risk ratio data suggests that 1) affix formulaicity likely does exist in the corpus, 2) affix formulaicity is a gradient rather than discrete phenomenon, and 3) formulaicity also holds between affixes and stems. The existence of affix formulas suggests that some polymorphemic sequences are stored as wholes in the mental lexicon, despite their apparent decompositionality. Theoretically, the results support psycholinguistic models of morphological processing with both analytic and holistic processing.