The Berkeley Phonetics & Phonology Forum ("Phorum") is a weekly talk and discussion series featuring presentations on all aspects of phonology and phonetics.
* Please note that this meeting is on Thursday, and starts at 3 PM*
download PDF abstract
One of the major assets of Autosegmental Phonology is that it allows to reduce procedural techniques of morphological exponence to a simple generalized concept of concatenation. In particular, the moraic approach to phonological length (Hayes, 1989) gives rise to a maximally simple account of morphologically triggered gemination, vowel lengthening, and coda epenthesis: as affixation of a μ. Although mora affixation is a standard assumption in numerous analyses (e.g. Lombardi and McCarthy (1991); Samek-Lodovici (1992); Davis and Ueda (2002); Grimes (2002); Davis and Ueda (2006); Álvarez (2005); Stonham (2007); Yoon (2008); Haugen and Kennard (2008)), some basic questions about the nature of mora affixation have never been properly addressed, one of them being the question about the linearization of prosodic affixes.
In this talk, we argue that prosodic nodes are assigned to a fixed position on their tier by the morphology and cannot be dislocated by later processes. Prosodic nodes are prefixed or suffixed to specific peripheral or prominent elements of their morphological bases on an affix-specific (i.e. phonologically arbitrary) basis. We therefore extend the assumptions about segmental affixation in Yu (2002, 2007) (cf. also Fitzpatrick (2004)) that affixation targets a specific member of a set of crosslinguistically possible anchor points by lexical subcategorization to prosodic affixation. An important empirical prediction of the subcategorization-based system is that an affix-μ cannot move to different linear positions under the pressure of phonological constraints. Based on a typological survey of quantity-manipulating morphological phenomena, we show that this prediction is true and argue that apparent counterevidence such as Keley-I gemination under the analysis of Samek-Lodovici (1992) is due to a morphological misinterpretation of the data.
Speech units of many sizes--segments, syllables, words or even full clauses--tend to be phonetically reduced when they are more probable. Vowels are more central, segments are more frequently deleted, and duration of the unit is shorter. Yet existing research on Dutch interfixes (Kuperman et al 2007) has shown that more predictable interfixes are phonetically enhanced: they are produced with longer, not shorter, duration. Why does probability have two opposite effects on pronunciation? Is it the case that morphemes simply behave differently from other speech units like words and syllables, or is the difference due to something else?
In my work I investigate two hypotheses respecting the relationship between morphemes and probability. Under the first hypothesis, more probable morphemes are phonetically enhanced, rather than phonetically reduced. Under the second hypothesis, it is necessary to distinguish two sorts of probability: contextual and global. Contextually probable units are probable only in the context of the specific utterance. Globally probable units are those which are more probable compared to related competitors in the language, regardless of utterance context. Under this hypothesis, the frequently observed phonetic reduction patterns are associated with higher contextual probability, whereas the phonetic enhancement observed by Kuperman et al are the result the interfixes' higher global probability.
To test these hypotheses, I am carrying out two experiments, one in English, and one in Russian. In these experiments I investigate the production of subject-verb agreement suffixes in sentences where both singular and plural agreement are possible, but with varying contextual probabilities. In English, these are sentences like "The cleaning staff in the meeting rooms look/looks grumpy." In Russian, these are sentences like "On the table were ([byl-i/o]) four large books." In particular, I measure the duration of the agreement suffixes (singular -s in English, and both singular -o and plural -i in Russian). If it is the case that morphemes simply behave differently from other speech units, then the more probable agreement suffixes should show phonetic enhancement, regardless of whether the probability is measured contextually or globally. On the other hand, if it is the case that contextual probability consistently results in phonetic reduction, then the more contextually probable suffixes should be phonetically reduced, while the more globally probable suffixes (as measured by the ratio of wordform to lemma frequency) should be phonetically enhanced.
Doubly conditioned rounding in Laal: Conditional licensing and correspondence chain
ATR (Advanced Tongue Root) has been recognized as a feature underlying vowel harmony in numerous languages, e.g., Mongolian but especially many West African languages. It has also been attributed to vowel pairs in German, English (and other Germanic languages), and Hindi, e.g., in English, +ATR vs. -ATR for the vowels in beet vs bit and suit vs. soot. But it has also been linked to preceding voiced stops (e.g., Shanghainese, Sundanese) and in such cases the voiced stop has been implicated in modifying the vowel quality, tone, and/or voice quality (i.e., lax voice) on the following vowel. I will report current progress aimed at finding out how ATR may be associated with voiced stops and how ATR on voiced stops might induce lax voice on a following vowel. The evidence will be drawn from acoustic analysis, EGG, fiberoscopy, ultrasound, and MRI.
Phonological theory has long been challenged by the behavior of contour segments and contour tones in harmony patterns. Sometimes these entities participate in phonology as whole units; at other times, their subsegmental parts act independently. This talk builds on insights from Autosegmental Theory, Aperture Theory, and Articulatory Phonology to propose a novel phonological representation for segments: all segments, including contours, are subdivided into a maximum of three ordered, quantized subsegments that host unitary sets of distinctive features and can participate in harmony (and other processes). By incorporating these quantized segmental representations into Agreement by Correspondence, we can offer, for the first time, a united treatment for the behavior of both contour segments and contour tones across observed phonological patterns of harmony.
A question frequently posed in cross-language research on speaking f0 range is whether and in what way bilinguals vary their f0 range according to the language they are speaking. Results in previous studies involving Japanese-(American) English bilinguals have been inconclusive, mainly because analysis is often confounded by sociophonetic influences on f0 range associated with the enactment of strict gender-defining roles in Japanese society. In this study, I report the results of an experiment in which 12 Japanese-AmEng simultaneous bilinguals (6 males, 6 females; all undergrads at UCB) were recorded performing comparable reading tasks in their two languages. The study builds on a relatively new approach to measuring f0 range - proposed by Patterson 2000 and operationalised in Mennen et al. 2012 - that computes its high and low points from actual tonal targets in the intonational phonology. Also, unlike in most previous studies where f0 range is traditionally treated as a one-dimensional measure, f0 range in both languages was measured along two quasi-independent dimensions: level and span. The results reveal that Japanese was realised at a significantly higher level and with a wider range of frequencies (span) than English. This finding provides new insights into the relation between intonational structure and f0 range in two typologically different prosodic systems.