Acoustic Salience and Input Frequency in L2 Lexical Tone Learning: Evidence from a Zapotec Revitalization Program in San Pablo Macuiltianguis
Second language (L2) learners of tone languages do not perceive and produce the different tones of the target language with equal ease. The most common explanation for these asymmetries is that acoustically salient tones are the easiest to learn. An alternative explanation is that tones are easiest to learn when they are highly frequent in the input. This dissertation pits the acoustic salience hypothesis against the input frequency hypothesis by investigating tone acquisition among children learning Macuiltianguis Zapotec, an endangered language traditionally spoken in Oaxaca, Mexico. The learners were seven native Spanish speakers, ages 7-11, studying Zapotec through a language revitalization program. This language is ideal for testing these two hypotheses because tones with low acoustic salience (mid tones) are highly frequent, while the salient (rising and dipping) tones are low-frequency. This begs the question of how learners make use of these competing factors. The study used a video corpus of classroom language and periodic learning measures collected over a ten-month period. Tone token and type frequencies, lexeme frequencies, and relative word frequencies were calculated from the corpus. Vocabulary tests and tone production tests were administered at four time points, and learners also completed a tone perception test and an elicited imitation task. The results show that learners were best at producing mid and rising tones but at perceiving dipping tones. Acoustic salience, tone frequency, and relative word frequency all predicted tone production accuracy. In addition, tone frequency predicted the rate of tone false alarms produced. That is, learners produced high rates of mid tones regardless of the tone target. However, acoustic salience was the best predictor of tone accuracy as measured through d' scores, a statistic that takes response bias into account. That is, learners gave correct productions for the salient rising and dipping tones more often than they produced these tones in error. Acoustic salience and input frequency therefore both play important roles in tone learning, but they influence different aspects of development. The findings emphasize the importance of employing models of L2 phonology that allow for both phonetic substance and statistical properties to drive acquisition.