Typological Intersections: Prosody, Morphology, and Syntax

Marianne Mithun

University of California, Santa Barbara

We know that intonation can play a major role in conveying information structure, distinguishing such elements of connected speech as focus and topic, and such features of referents as givenness and identifiability. We also know that languages and even dialects can vary substantially in their intonation patterns. The variability raises questions about the universality of the relationship between prosody and information structure. If there are differences among languages, we can ask whether any other features correlate with such differences and if so, whether we can discern any cause-effect relationships among them.

One of the major typological parameters proposed for prosodic structure is [± rightmost] (Ladd 1996:160). In various Romance languages, sentence accent always appears on the rightmost element of the nuclear clause (Vallduví 1992, Lambrecht 1994). In various other languages, such as English and Hungarian, sentence accent can appear in different positions to signal focus on different elements. There are, however, languages which show neither consistent rightmost sentence accent nor movable accent. Such a language is Mohawk, where sentence accent is consistently leftmost. This seeming typological anomaly may not be an isolated idiosyncrasy. Mohawk is at another typological extreme: degree of synthesis. Much of what is expressed syntactically in many other languages is expressed morphologically in polysynthetic languages like Mohawk. By considering the chains of events behind the development of these prosodic and grammatical structures, we can see how such typological dimensions can be related.

'Fish', 'crab', and 'fig': Can we make sense of the Indus pictograms?

The Harappan or Indus Civilization was the most extensive culture of its time, c. 2600-1900 BC. It is famed for its town planning and water engineering, but in comparison to the other early civilizations, we know very little of it. Yet it was a literate civilization -- about 5000 inscriptions have been excavated, but they are all short and written in a forgotten script and in an unidentified language. There are no translations into known scripts and languages, nor any historical information of the kind that has been instrumental in the decipherment of other early scripts.

The Indus script thus presents a tantalizing puzzle. Many solutions have been published, but none has won general acceptance. One recent attempt to solve the problem even claims that the Indus pictograms do not represent language-bound writing at all, but are non-linguistic symbols. The hypothesis is discussed in a separate workshop of the Institute, but not in this lecture beyond its general thrust. The aim is to show by means of a few select pictograms that the Indus script is a phoneticized logo-syllabic writing system based on a Dravidian language, which made good use of the rebus principle, like all other early scripts. The emphasis is on methodology and on checking the hypotheses.

Divergence

William Labov

When speech communities become widely separated by migration, dialect divergence over time is to be expected and no special accounting is needed. The major problems for explanation arise when neighboring dialects diverge even when they appear to be in continuing communication. Recent studies have found such increasing divergence between adjacent regional dialects of North America and between communal dialects of the same community.

Origins of diversity are found at linguistic forks in the road, where linguistic instability can be resolved in several equally probable ways. Continued divergence is found when one or both of the neighboring dialects become engaged in unidirectional change, typically driven by internal factors: chain shifting, merger or unidirectional patterns of grammaticalization. Such divergence is the product of continued incrementaton in transmission by child language learners.

A second source of diversity is found in the diffusion of dialect features across geographic or social boundaries. This is primarily the result of contact among adult language learners, which produces less faithful copies of the original, particularly when structural conditions of any degree of abstraction are involved.

These sources of divergence will be illustrated by an examination of American dialects in contact across the North/Midland line, differing chain shifts initiated by the low back merger, the differentiation of short-a systems, and the development of tense, mood and aspect markers in African American Vernacular English.

Statistical Language Learning: Computational and Maturational Constraints

In recent years a wide variety of studies have shown that infants, young children, and adults can successfully utilize the statistics of distributional linguistic information to find candidate words in a speech stream, form and alter phonetic categories, discover grammatical categories, and acquire simple syntactic structure in miniature languages in the laboratory. A major question we face, then, is how to think about the broader picture of statistical learning: How many kinds of statistical computations can learners perform? How are these computations organized, and how are they constrained? Must such mechanisms be combined with qualitatively different, more traditional mechanisms that form symbolic rules or set linguistic parameters?

I will address these questions by presenting findings from recent studies of statistical learning of syntax, examining the effects of complex multiple cues and also comparing child and adult learners given inconsistent input and (sometimes) forming rule-like generalizations. The results of these studies suggest the outlines of several distinct but suitably sophisticated candidate statistical learning mechanisms and raise questions for future research regarding how to develop a theory of statistical language learning.

Probing Lexical Processing in Modern Dutch Poetry

Most studies investigating lexical processing have made use of experimental techniques presenting words in isolation. These studies have revealed a wide range of factors co-determining lexical processing, but leave us with questions concerning the ecological validity of the results obtained. For instance, Pinker & Ullman (2002) admit that frequency effects for regular inflected words are attested, contrary to the original assumptions of their dual mechanism model, but they maintain that such frequency effects are task artifacts that are not used in normal listening, reading, or speaking.

In this presentation, I report the results of a large self-paced reading study addressing lexical processing in modern Dutch poetry (Breukers, 2006). A linear mixed-effects analysis of some 300,000 logarithmically transformed reading latencies, with Subject, Word and Poem as random effects, confirmed that the specific form frequency of an inflected word is the key frequency measure predicting performance not only in isolated word recognition but also in reading poetry. This frequency effect was attenuated (but still significant) for males, in line with the hypothesis of Ullman et al. (2002) that females have slightly superior declarative memory. Conversely, males emerged with an advantage reading multiply complex words, in line with Ullman's hypothesis that males have slightly superior procedural memory. Measures gauging paradigmatic morphological connectivity in the mental lexicon turned out to be predictive mainly for words in the initial position of a line of poetry, a position generally eliciting relatively long reading latencies that allow these semantic effects to emerge. Assonance and rhyme led to faster reading latencies, whereas repetition (measured by the number of times a word had occurred) slowed reading, especially for older subjects.

This study shows that it is now becoming possible to model lexical processing of individual words read by individual subjects, taking into account a wide range of factors ranging from characteristics of the reader to the position of a word in a verse of poetry, and from structural factors relating to morphological complexity to poetic devices such as repetition, assonance and rhyme.

Breukers, C. (2006). 25 jaar Nederlandstalige poezie 1980-2005 in 666 en een stuk of wat gedichten. De Contrabas Bloemlezing, BnM Publishers, Nijmegen.

Pinker, S. and Ullman, M. (2002). The past and future of the past tense, Trends in the Cognitive Sciences, 6, 456-462.

Ullman, M. T., Estabrooke, I. V., Steinhauer, K., Brovetto, C., Pancheva, R., Ozawa, K., Mordecai, K. & Maki, P., (2002). Sex differences in the neurocognition of language. Brain and Language 83, 141-143.