Volume 40, Issue 1

Cross-linguistically, it is difficult to tease apart allomorphy from readjustment rules. But regardless, both tend to respect locality and are sensitive to information that is present in the input, not the output. We document a counter-example to these tendencies from Western Armenian, and we discuss how the data falsifies such restrictive models of allomorphy. The Western Armenian theme vowel -i- changes to the theme vowel -e- due to two types of triggers. The first type of trigger is phonological: the change happens when the theme vowel is unstressed in the output. This is a type of allomorphy that is conditioned by output phonology. The second type of trigger is morphological: the change happens when the verb is in the past tense. The +Past morpheme can be either in the verb (adjacent to the theme vowel) or on a separate auxiliary in periphrasis. This amounts to a case of long-distance allomorphy that is conditioned across words, even in suspended affixation. For suspended affixation, I provide semantic and prosodic evidence that suspended affixation is created via base-generation and not via ellipsis. The inability to use ellipsis acts as additional evidence that the allomorphy is long-distance.

The vowel alternations relating the perfectives and imperfectives of Form I verbs in Cairene Arabic are examined in this article. It is argued that such alternations are driven by a coherent system of Ablaut (or apophony), the same system as was proposed by Guerssel and Lowenstamm for Classical Arabic. Mounting evidence from unrelated languages strongly suggests that Ablaut systems are organized in similarly rigid fashion cross-linguistically. The question of the impact of linguistic change on Ablaut thus arises. One possibility is that change may culminate in the abandonment of apophony. Another possibility is that apophony itself drives change. It is argued in this article that several past and ongoing developments in the evolution of the vocalization of verbs in Cairene took place under the control of the apophonic system.

This paper provides new insights into the analysis of vocative structures that co-occur with a sentence by bridging two previously independent domains of linguistic research: wh -interrogatives and vocatives. More specifically, we investigate in which positions Spanish speakers accept vocatives in wh -interrogatives introduced by different wh -phrases. The results of an acceptability judgment task indicate that our participants highly accept initial and final vocatives in all wh- interrogatives. For middle vocatives, the results differ across wh -phrases. While the participants accept middle vocatives in wh- interrogatives introduced by por qué (‘why’) and d-linked wh -phrases, they reject them in bare wh- interrogatives. These findings require a modification of the syntactic analysis of vocatives. Initial vocatives are placed above ForceP, while middle and final vocatives are analyzed in two different positions, in an upper and a lower Voc addr P in the left periphery.

The declarative complementizer has been claimed to have grammaticalized from a relative pronoun in various Indo-European languages. The source construction is assumed to have been the correlative sentence . The initial phase of the hypothesized process, however, has remained unclear; in the explicative clause to which e.g. the Germanic that -type complementizers can be traced back (Mary knows that, that Peter is lying), that is already a complementizer base-generated in C rather than a relative pronoun in Spec,CP. This paper analyzes a similar developmental path, that of hogy ‘that’, the Hungarian general complementizer cognate with the relative proadverb hogy ‘how’, the early stages of which can be reconstructed more completely. It traces hogy back to a canonical correlative construction, and documents the subsequent stages of its evolution from a relative operator binding a variable in a correlative sentence, via a linker introducing an adjunct clause, to a complementizer subordinating a clausal argument to a matrix predicate.

The paper postulates that the propensity of Polish velar consonants to undergo palatalization is the consequence of the activity of a violable constraint which requires autosegmental place nodes to be associated with one and only one element. Since velars are the only consonants which lack place specification, the spreading of the palatality element |I| onto velars leads to the avoidance of the violation of the relevant constraint. As the palatalizing element |I| must entertain the status of the head, the underapplication of Surface Velar Palatalization before the front mid nasal vowel /ɛ̃/, headed by element |L|, and the vowel /ɛ/ found in some borrowings and headed by element |A|, is enforced by the constraint punishing representations in which more than one element plays the role of the head.