Is discourse made up of sentences? Focusing on dependent grafted speech in modern standard Japanese

The idea that discourse is made up of sentences has been widespread among linguists. Does this traditional discourse perspective (“sententialism”) apply to casual language in daily communication? This paper examines the validity of sententialism by focusing on a type of speech called “dependent grafted speech” in Japanese conversation. Close examinations of various words, phrases, and sentences reveal that dependent grafted speech is different from sentences on two points: (i) Generally, the lexical accent of the copula at the beginning of dependent grafted speech is a high tone; and (ii) the interaction particle at the end of dependent grafted speech is not uttered with a falling intonation unless it is proceeded by a very abrupt rising intonation (“leaping” intonation). These findings cast doubt on the status of dependent grafted speech as a sentence. Moreover, they demonstrate a new conception of discourse as a mixture of diverse constituents, including sentences, dependent grafted speech, and other utterance types.

Rhythmic Repair of Morphological Accent Assigned Outside of a Metrical Window

This paper studies the interaction of morphologically-assigned accent and a left-edge three syllable metrical window within which primary prominence is restricted. Kager (2012) illustrates that when morphological (or lexical) accent appears outside of a designated metrical window, primary prominence reverts to a default position within the window. I refer to this as 'default repair'. The main claim of this paper is that the Bolivian language Ese’eja demonstrates a novel type of repair called 'rhythmic repair'. In this type, when phonological accent is assigned outside a metrical window, primary prominence does not fall on a default position but rather appears on a position within the window which is rhythmically dependent on the position of accent. I argue that these Ese’eja data cannot be reduced to a type of default repair involving multiple morphosyntactically-conditioned defaults. Further, I formalize the differences between default and rhythmic repair following Kager’s (2012) view of metrical windows as decomposable into common OT constraints. While default repair involves highly ranked word-to-foot alignment constraints (e.g. Align-Wd-L) ranked above lower Faith-Accent and Parse-Syl, in contrast rhythmic repair involves splitting Faith-Accent into Accent-to-Prominence (Acc-Prom) and Accent-to-PrimaryProminence (Acc-PrimProm) constraints, and ranking Parse-Syl, Acc-Prom, and an Edgemost constraint above Acc-PrimProm and word-to-foot alignment constraints.

Against Root Faithfulness in Cupeño Stress

This paper develops a new, optimality-theoretic analysis of word-level stress assignment in Cupeño (Takic, Uto-Aztecan). I argue that primary stress is assigned to the leftmost lexically accented (i.e. stress-preferring) morpheme, else to the word's left edge. I contend that this analysis is simpler and better explains the Cupeño data than previous accounts, which assume that special faithfulness constraints privilege the accentual properties of roots over those of other affixes. The typological implications of this renanalysis of Cupeño stress are then discussed; without empirical support from Cupeño, it is suggested that "root faithfulness" plays no role in determining word stress in lexical accent systems cross-linguistically.

Spanish stress and lexical accent across syntactic categories

In this paper, I provide an analysis of Spanish stress with the following three characteristics: (a) both verbal and non-verbal stress are accounted for in a single, unified, system, (b) the three-syllable window for stress is accounted for in a principled way, and (c) the stress algorithm has no access to the morphosyntactic structure. I do this by extending Roca’s analysis of variable edge parameters for stress in Spanish non-verbs to verbs, and by arguing that morphemes which mark for only person, number, and gender (φ-features) are outside of the domain of stress because they are prosodic adjuncts.

L1 Prosodic Interference: the Case of Slovene Students of Japanese

A bi-directional perception experiment was conducted on Japanese and Slovene subjects to evaluate the result of a full L1 prosodic interference in recognizing (lexical) accent place in declaratives and interrogatives. Perceptual hypercorrection into L1 prosody on the side of the listener was achieved by making the subjects think they were listening to their own language, and results show clear tendencies for errors, which in general agree with predictions. However, mapping from phonetic to phonological representations was found to be asymmetric, suggesting that subjects of the two languages rely on different phonetic cues, as well as that distinctive function of certain phonetic cues, such as duration, has different effects on perception of segmental structure.

In defense of Narten roots

In this paper I argue that at least in Vedic, Narten roots existed. They were distinguished by lexically specified dominant accent, whereas non-Narten roots were underlyingly unaccented. Underlying lexical accent accounts for the static accent in the present tense as well as barytonesis in primary derivatives with recessively accented suffixes. The second characteristic feature of Narten roots, the lengthened grade in the strong stem of the present, is still best explained as an upgrading. This hypothesis is strengthened in a scenario where the analogy is fed by channel bias. In Vedic, the system of lexically accented roots was on the verge of collapse. This explains not only secondary thematic presents like rā́jati, but also sheds new light on forms like sā́hant- and sāhvā́ṃs-.