Variation learning in phonology in line with typological principles

24

2020-06-01

2021-03-17

77000

Variation learning in phonology in line with typological principles

Language learning, Learning simulation, Phonology, Variation

Linguistics and Languages

201910159156

Seed Fund for PI Research – Basic Research

2019

Completed

The acquisition of sound system may often be imperfect, exhibiting discrepancies of sound properties from input. An important question for acquisitionists is to understand why learners have difficulties in learning certain sound patterns more than others. One hypothesis concerns the effect of learning biases. It argues that learners may have prior biases to learn certain sound patterns better than others. Work on learning biases in phonology has focused on discovering underlying mechanisms that make certain phonological patterns harder or easier to learn. Experimental work using artificial language learning paradigms mainly tested two hypotheses: (a) structurally more complex patterns are harder to learn, called complexity bias (Saffran and Thiessen 2003, Cristiá and Seidl 2008, Kuo 2009, Pycha et al. 2003 Chambers et al 2010, Peperkamp 2006, Skoruppa 2009), and (b) patterns lacking phonetic substances are harder to learn, called substantive bias (Toro et al. 2008, Nevin 2010, Koo 2007, Zaba 2008, Campbell 2004, Wilson 2003, Pycha et al. 2003, Peperkamp 201, Wilson 2006, Carpenter 2005, 2006, 2010). Research on learning biases in phonology has produced fruitful results (see Moreton and Pater 2012 for a review). Evidence in the field has consistently supported the complexity bias and the results are mixed for the substantive bias. Notably, there seems to be a critical gap in the literature. Most work which tested phonological learning biases, whether complexity bias or substantive bias, has predominantly assumed categorical phonology. For instance, learners were trained on a language in which [p, t, k] are found exclusively in one phonological environment (e.g., word final positions) and [b, d, g] in other environment (e.g., word medial positions), and the learning outcome was compared to another group where the pattern was restricted [p, d, k] vs. [b, t, g] (Saffran and Thiessen 2003). The argument is that the former classification depends on a single stimulus feature (e.g., voicing), while the latter involves multiple features thus is harder to learn, supporting complexity bias. This experimental evidence leads to an idea that the complexity bias observed in learning is reflected to positional restrictions of phonemes across languages: sounds restricted in one position are likely to share a single feature rather than multiple features cross linguistically (e.g., [p, t, k] in final positions rather than [p, d, k]). Results from such categorical pattern learning provide insights into learning biases in phonology, but natural-language phonology does not purely exhibit categorical patterns. In fact, variable phonological patterns are prevalent in languages. For example, voiceless stops in a word-final position may be pronounced with an unaspirated stop as in [khæt] or with a glottal stop as in [khætʔ] among other variants. Variations also occur beyond word level. For instance, /t, d/ in a word-final consonant cluster in English may remain or be deleted when it is followed by a consonant-initial word, as in last night → [lœst naɪt] or [lœs naɪt]. The shapes of phonological variants are partially predictable to a certain degree when linguistic and social factors are considered, but the predictions are neither deterministic nor absolute. If a goal of a research program of learning biases is to link laboratory evidence to language acquisition, and also potentially language changes and typology, it becomes equally imperative to understand the role of learning biases in the context of learning probabilistic and nondeterministic variable patterns. Artificial language learning work on syntactic variations has been conducted (Singleton and Newport 2004, Hudson Kam and Newport 2005, 2009, Culbertson and Newport 2015), but work focusing on the learning of phonological variations is limited. To the best of our knowledge, no studies in the field of phonology has focused on the learning of sound patterns characterized by inherent variability. The purpose of this project is to investigate the acquisition of unconditioned phonological variation by adults when the two variants represent cross¬linguistic typological differences that are hypothesized to reflect the presence of an inductive bias, whether structural or substantive. The primary objective is to observe whether there are inductive biases of structural complexity and/or phonetic naturalness in adult acquisition of an artificial language. The secondary objective is to observe whether adults are able to match probabilities for unconditioned phonological variation in an artificial input, which can provide insights into the pattern of sound change involving variation.