The Strength of Segmental Contrasts: A Study on Laurentian French

Title: The Strength of Segmental Contrasts: A Study on Laurentian French
Authors: Stevenson, Sophia Diana
Date: 2015
Abstract: The dichotomy of contrastive and allophonic phonological relationships has a long-standing tradition in phonology, but there is a growing body of research (see Hall, 2013, for a review) that points to phonological relationships that fall between contrastive and allophonic. The criteria most commonly used to define phonological relationships or resolve cases of ambiguous phonological relationships – namely (a) predictability of distribution, and (b) lexical distinction – are not always able to account for observed sound patterns. The main goal of this dissertation is to identify and apply quantitative measures (relative frequency and minimal pair counts) to the traditional criteria in order to better account for cases of intermediate phonological relationships or, in other words, to account for different strengths and degrees of contrast. Twenty native speakers of Laurentian French (LF) participated in Experiment 1, an AX discrimination task, and Experiment 2, a four-interval AX (4IAX) task, which tested the broader relationships of allophony and contrast. It was hypothesized, based on previous experiments (Boomershine et al., 2008; Dupoux et al., 1997; Ettlinger & Johnson, 2009; Johnson & Babel, 2010; Kazanina et al., 2006; Peperkamp et al., 2003; Pruitt et al., 2006), that phones in an allophonic relationship would be more difficult to perceive than phones in a contrastive relationship. Results confirmed previous findings, with longer reaction times for allophonic pairs as compared to contrastive pairs in the AX task (p<.001), as well as in the 4IAX task (p = .004). For Experiments 3, 4 and 5, thirty native speakers of LF participated in an AX, a 4IAX and a similarity rating task. Measures of functional load, frequency and acoustic similarity were applied to pairs of phones in allophonic and phonemic relationships in order to quantify the degree of contrast between pairs. If a gradient view of contrast was supported, it was hypothesized that High Contrast vowels [a-ɔ] would yield higher accuracy, faster reaction times and lower similarity ratings; Low Contrast vowels [y-ʏ] would yield lower accuracy, slower reaction times and higher similarity ratings; and Mid Contrast vowels [o-ʊ] would yield results that fell between the two extremes. If, on the other hand, a strict binary interpretation of contrast was supported, High Contrast vowels and Mid Contrast vowels should yield similar results since these vowels are considered to be in a phonemic relationship, with higher accuracy, faster reaction times and lower similarity ratings, while Low Contrast vowels [y-ʏ], in an allophonic relationship, should yield lower accuracy, slower reaction times and higher similarity ratings. The results from Experiments 3 (AX) and 4 (4IAX) show that the High Contrast pairs yielded significantly higher accuracy scores and faster reaction times than both Mid and Low Contrast pairs (Experiment 3: p<.001 for both High vs. Mid and High vs. Low comparisons; Experiment 4: p = .039 for High vs. Mid, p = .055 for High vs. Low comparisons). However, no significant differences were found between Mid and Low Contrast pairs in these two experiments. The results from Experiment 5 matched gradient predictions, showing significant differences between High, Mid and Low conditions, with similarity being judged highest for Low pairs, lowest for High pairs, and ratings for Mid pairs falling exactly between the other two levels (p<.001 for all comparisons). While results do not perfectly match gradient predictions, the findings provide evidence counter to a strict binary interpretation of contrast since traditionally phonemic pairs (High [a-ɔ] and Mid [o-ʊ]) were significantly different from one another in all experiments. The lack of difference between Mid and Low Contrast pairs could be due to the measures of functional load and frequency for Mid pairs being closer to those of Low pairs, and thus did not reflect a level of contrast that was equidistant between High and Low Contrast. Nevertheless, taken together with the results from Experiment 5, the results appear to support a gradient view of phonological relationships rather than a strictly dichotomous view. Quantitative measures therefore show promise in accounting for cases of intermediate phonological relationships.
CollectionThèses, 2011 - // Theses, 2011 -