Despite ample evidence that grammaticalization is accompanied by phonological reduction and ultimately morphological fusion, the latter process is remarkably less common in Turkish – hence its prototypically agglutinating morphology. Since vowel harmony is a means of articulatory reduction, Turkish, as a vowel-harmonic language, therefore shows reduction but (virtually) no fusion. One morphosyntactic consequence of agglutination is that Turkish “suffixes” in many ways continue to behave like free words. To compensate for the resulting lack of clear-cut suffixes, vowel harmony and stress are co-opted to perform affixal functions such as the demarcation of words and encoding of relationships among morphemes. Due to the grammatical function of suffix vowels, however, even grammaticalized items must then remain at least monosyllabic, which constrains the extent of fusion possible. This situation suggests that theories of grammaticalization that do not sufficiently distinguish between reduction and fusion need to be refined. In addition, it highlights the need for language-specific analyses on the diachronic dimension and restores the status of morphological typology as a predictor of certain linguistic variables.

No abstract

This paper investigates what kinds of meanings are expressed by circumfixes cross-linguistically, with an eye toward explaining this rare formal phenomenon. Based on the study of several hundred languages, it shows that circumfixes express negation relatively more often than other functions, followed by a wide array of TAME and derivational meanings and some more specific functions such as adjectival degree or the formation of ordinal numerals. While circumfixes are found in many unrelated languages and in all macro-areas, they nevertheless cluster in specific families, which suggests that it is their diachronic emergence that is rare rather than their synchronic distribution per se. The semantic contributions of circumfixes are explained by different diachronic mechanisms. Negative circumfixes are argued to ultimately follow from the negative cycle, whereas circumfixes encoding other functions are claimed to emerge from the cumulation of functionally compatible morphs, as in the case of certain aspectual and temporal markers. In addition, there is also some evidence that empty morphs may be reinterpreted as parts of circumfixes. All these processes are further constrained by the fact that one half of the developing circumfix has to be a prefix. Since prefixes are relatively rare overall, this necessary condition for circumfixation often is not met. On the other hand, prefixes express negation relatively frequently, and this helps to further explain why circumfixes have strong ties to this particular semantic domain.

Some reference grammars and cross-linguistic works describe all elements that are not clear-cut words as "clitics." As a consequence of this practice, the class of suggested clitics is highly heterogeneous, which reduces the usefulness of the "clitic" label as a whole. In response to this situation, a more nuanced typology of grammatical forms is proposed here. The argument crucially relies on the notion of formal "dependence," which is essentially a synchronic indicator of grammaticalisation status. The resulting system limits the term "clitic" to its prototypical manifestation, which combines a syntactic distribution with some degree of prosodic dependence on a host. Meanwhile, the class of "weak words" subsumes elements that are independent words in every regard except that they do not bear stress and/or tone, whereas "anti-clitics" are affixes except that they share some behaviour with phonological words. Lastly, there are "mobile" and "suspended" affixes, which show types of syntagmatic freedom not found with prototypical affixes. All form classes proposed in this typology are attested across unrelated languages and

No abstract