The Berkeley Phonetics & Phonology Forum ("Phorum") is a weekly talk and discussion series featuring presentations on all aspects of phonology and phonetics.
The speech categories of a speaker's native language are robust in the face of considerable acoustic variation in natural speech. However, the well-entrenched biases of the perceptual system to these native categories can make it difficult to acquire the speech sounds of a new language in adulthood. In this talk, I discuss two behavioral studies designed to examine the types of information that novice learners find useful when acquiring fledgling phonemic categories. In particular, I'll focus on the interaction of acoustic and articulatory information during training as a means to building more stable novel categories in the early stages of learning.
Note: Special day (Thursday) and time (11:00-12:00)
Umlaut in modern West Germanic languages is understood as a phonologically motivated alternation of a set of vowels in given morphological environments wherein non-front vowels gain the feature [+front]. In Germanic languages, morphological umlaut is frequently distinguished from other types of ablaut alternations as the latter did not develop from a harmony process, therefore there is no expected phonological uniformity between the input-output relations of stem vowels. In this talk, I examine the phonological properties of input-output relations of nominal plural and adjective degree vowel alternation in the West Germanic language Plautdietsch (PDT). While a majority of the alternations ultimately come from Middle Low German umlaut, the system of alternations varies quite considerably from traditional accounts of umlaut in the rest of the family. One of the most striking features of umlaut is that not all outputs are [+front]. PDT vowel alternations are not uniform across the entire system of vowels: some sub-classes of alternations allow feature specifications not found in the input and others do not appear to have a clear featural motivation. I take the position that the vowel alternations in PDT system are best understood as synchronically opaque reflexes of diachronic developments which lie part way between an umlaut system and an ablaut system.
This study examines the production, perception and processing of the Catalan /e/-/ɛ/ and /o/-/ɔ/ vowel contrasts by 60 Spanish-Catalan bilinguals in Majorca (Spain). Results from picture-naming, identification, AX discrimination, and lexical decision tasks show that even though these early and highly proficient bilinguals maintain robust mid-vowel contrasts in their productions and demonstrate a high accuracy in perceptual identification and discrimination tasks, they have difficulties distinguishing between words and non-words in a lexical decision task. These findings provide evidence that making explicit judgments regarding whether a certain sound belongs to a phonemic category (i.e., as accomplished via identification and discrimination tasks) does not entail that listeners have an appropriate representation at the lexical level.
Recent studies suggested that phonological contrasts with low functional load are more likely to undergo phonological mergers than those with high functional load (Wedel et al 2013a, b). It remains unclear, however, how such effects are realized in speech perception and/or production. The current study investigates whether functional load influences the categorical perception of voicing contrasts in English fricatives. I will report results from two experiments (identification and AX discrimination), and provide evidence that, among other factors, phonological contrasts with higher functional load are perceived more categorically. These findings suggest that functional load may influence how categorically a particular contrast is perceived, and thus functional load is a potential factor in sound change.
In Amharic, a Semitic language spoken primarily in Ethiopia, plural agreement on adjectives and iterative marking on verbs involves infixing reduplication. Interestingly, the infixing morpheme is only possible in adjectival and verbal stems containing underlying heavy syllables--those ending in geminate consonants. The infix surfaces immediately preceding the geminate and has the shape CV, where the C shares features with the geminate. In this talk I demonstrate the relationship between the stress and weight systems of Amharic and this infixing reduplication process. Additionally, I provide an Optimality Theory account of the data, demonstrating that our analysis must refer to heavy syllables in order to ensure the correct landing site of the infix. This means that Amharic is the first attested language where infixes target heavy syllables.
In conversation, inter-turn intervals are mostly short (around 200ms), suggesting that listening and speech planning can often overlap in time. However, it is unclear when speakers prefer to begin to plan their utterances. They could either prepare their utterances as soon as they know what to say, or they could wait until right before their turn. I describe a dual-task paradigm involving finger tapping (as an indicator of cognitive load) and turn-taking, which was used to investigate this question. The results show that tapping rate in a turn-taking condition initially did not differ from a listening only condition, but tapping rates dropped shortly before the onset of the participant’s own turn. An eye-tracking experiment confirmed that the participants’ attention turned to their own displays around that time. The combined data suggest that participants adopted a late planning strategy. Additional research shows that such a strategy could be important to allow listeners to process the speech of their interlocutor.
This talk examines a case of parasitic tone harmony from Dioula d’Odienné, in which there is a ganging effect of local and long-distance segmental features that increases the likelihood of tone harmony. I use the Dioula data to revisit the recent claim put forth by e.g., Potts et al. (2010), Pater (to appear) that ganging effects can be captured solely by cumulativity in weighted constraints, and that constraint conjunction—the previous method of obtaining ganging—is no longer necessary given additive constraint weights (cf. Smolensky 2006). In the present analysis, I argue that the simple addition of constraint weights is insufficient to adequately model the probabilistic ganging effects of similarity found in the Dioula corpus. Weighted constraint conjunction, implemented as an interaction term in Maximum Entropy Optimality Theory, provides a significantly improved model of ganging. Differences between weighted constraint conjunction and previous implementations of local constraint conjunction will be discussed, as will methods for model selection and comparison in probabilistic Optimality Theory when addressing quantitative natural language data.