|Title:||Relative Frequency of Patterns and Learnability: The case of phonological harmony||Add Dissertation|
|Author:||Aleksandra Zaba||Update Dissertation|
|Email:||click here to access email|
|Institution:||University of Utah, Linguistics|
|Linguistic Subfield(s):||Linguistic Theories; Phonology; Psycholinguistics;|
|Abstract:||The cross-linguistic frequency of phonological patterns has been commonly linked to their learnability (e.g., Chomsky & Halle (1968); Prince & Smolensky (2003); Steriade (2001)). Experimental work investigating the relation between frequency and learnability has provided contradictory results. For example, Wilson (2003) used an artificial language learning paradigm in a listening experiment, and found that the attested patterns of nasal consonant harmony (triggered by nasal consonants) and nasal consonant disharmony are more learnable by adults than unattested arbitrary patterns. Koo & Cole (2006) used a similar method and found that the more frequently attested pattern of back vowel harmony is not more learnable than the less frequent liquid harmony. My dissertation followed up on studies such as Wilson (2003) and Koo & Cole (2006) to contribute to the question whether frequency and learnability are related. The patterns in focus in my dissertation were types of phonological harmony (different from those in previous experiments) and directionality of harmony.
Experiment 1 of my dissertation tested whether the relative cross-linguistic frequency scale, back vowel harmony>>nasal consonant harmony>>labial consonant harmony (unattested), is related to the corresponding learnability scale. Participants were trained on non-words containing one of the three harmony types and subsequently tested on their learning of the patterns. Results showed that none of the three training conditions learned better than the others. To exclude the possibility that these results were due to individual differences in attention or learning ability, analyses were repeated with participants in each condition whose d-prime scores were within 1 SD from the mean, and with participants whose d-prime scores exceeded 0. Both measures did not provide any support for the relation between frequency and learnability.
Experiment 2 used the same methodology as Experiment 1 to detect whether the relative frequency scale, progressive nasal consonant harmony (P)>>regressive nasal consonant harmony (R), is related to the same learnability scale. Results with all participants included showed that no harmony condition was learned better than the other. When the same cutoff procedures as in Experiment 1 were used, the learning scale was P>>R.
Several possible explanations exist for why only Experiment 2 revealed a relation between frequency and learnability. Among these explanations is that frequency and learnability may be related only in certain patterns. In other patterns, other factors may contribute to frequency. For example, some patterns, such as back vowel harmony, may be more frequent than other patterns since they (frequent patterns) contribute to ease of speech articulation (Oh & Cole 2005).
Among the implications for theory is that the universally preferred ranking of INTEGRITY constraints INTEGRITY (F) Affix>> INTEGRITY (F) Root suggested by Kraemer (2003) is supported. INTEGRITY (F) constraints are constraints against prominence augmentation. For example, INTEGRITY (F) Affix demands that 'no feature of an affix in an input has multiple correspondents in the output' (Kraemer 2003:94). INTEGRITY (F) Root demands that 'no feature of a root in an input has multiple correspondents in the output' (Kraemer 2003:94). The ranking of the constraint INTEGRITY (F) Affix above INTEGRITY (F) Root thus describes a bias toward stem-outward harmony, which was supported in Experiment 2 of my dissertation (where progressive nasal consonant harmony (stem-outward) was more learnable than regressive nasal consonant harmony (suffix-triggered)) for nasal consonant harmony as attested in the Bantu languages.