Two factors have been proposed as the main determinants of phonological typology: channel bias, phonetically systematic errors in transmission, and analytic bias, cognitive predispositions making learners more receptive to some patterns than others. Much of typology can be explained equally well by either factor, making them hard to distinguish empirically. This study presents evidence that analytic bias is strong enough to create typological asymmetries in a case where channel bias is controlled. I show that (i) phonological dependencies between the height of two vowels are typologically more common than dependencies between vowel height and consonant voicing, (ii) the phonetic precursors of the height-height and height-voice patterns are equally robust and (iii) in two experiments, English speakers learned a height-height pattern and a voice-voice pattern better than a height-voice pattern. I conclude that both factors contribute to typology, and discuss hypotheses about their interaction.

Artificial analogues of natural-language phonological patterns can often be learned in the lab from small amounts of training or exposure. The difficulty of a featurallydefined pattern has been hypothesized to be affected by two main factors, its formal structure (the abstract logical relationships between the defining features) and its phonetic substance (the concrete phonetic interpretation of the pattern). This paper, the second of a two-part series, reviews the experimental literature on phonetic substance, which is hypothesized to facilitate the acquisition of phonological patterns that resemble naturally-occurring phonetic patterns. The effects of phonetic substance on pattern learning turn out to be elusive and unreliable in comparison with the robust effects of formal complexity (reviewed in Part I). If natural-language acquisition is guided by the same inductive biases as are found in the lab, these results support a theory in which inductive bias shapes only the form, and channel bias shapes the content, of the sound patterns of the worlds languages.

Artificial analogues of natural-language phonological patterns can often be learned in the lab from small amounts of training or exposure. The difficulty of a featurallydefined pattern has been hypothesized to be affected by two main factors, its formal structure (the abstract logical configuration of the defining features) and its phonetic substance (the concrete phonetic interpretation of the pattern). This paper, the first of a two-part series, reviews the experimental literature on structural effects. The principal finding is a robust complexity effect: Patterns which depend on more features are reliably harder to learn.

Is Optimality Theory a constraining theory? A formal analysis shows that it is, if two auxiliary assumptions are made: (1) that only markedness and faithfulness constraints are allowed, and (2) that input and output representations are made from the same elements. Such OT grammars turn out to be incapable of computing circular or infinite chain shifts. These theoretical predictions are borne out by a wide range of natural phonological processes including augmentation, alternations with zero, metathesis, and exchange rules. The results confirm, extend, and account for the observations of Anderson & Browne (1973) on exchange rules in phonology and morphology.

Linguistic and non-linguistic pattern learning have been studied separately, but we argue for a comparative approach. Analogous inductive problems arise in phonological and visual pattern learning. Evidence from three experiments shows that human learners can solve them in analogous ways, and that human performance in both cases can be captured by the same models. We test GMECCS (Gradual Maximum Entropy with a Conjunctive Constraint Schema), an implementation of the Configural Cue Model (Gluck & Bower, ) in a Maximum Entropy phonotactic-learning framework (Goldwater & Johnson, ; Hayes & Wilson, ) with a single free parameter, against the alternative hypothesis that learners seek featurally simple algebraic rules ("rule-seeking"). We study the full typology of patterns introduced by Shepard, Hovland, and Jenkins () ("SHJ"), instantiated as both phonotactic patterns and visual analogs, using unsupervised training. Unlike SHJ, Experiments 1 and 2 found that both phonotactic and visual patterns that depended on fewer features could be more difficult than those that depended on more features, as predicted by GMECCS but not by rule-seeking. GMECCS also correctly predicted performance differences between stimulus subclasses within each pattern. A third experiment tried supervised training (which can facilitate rule-seeking in visual learning) to elicit simple rule-seeking phonotactic learning, but cue-based behavior persisted. We conclude that similar cue-based cognitive processes are available for phonological and visual concept learning, and hence that studying either kind of learning can lead to significant insights about the other.

Is phonological learning subject to the same inductive biases as learning in other domains? Previous studies of non-linguistic learning found that intradimensional dependencies (between two instances of the same feature) were learned more easily than inter-dimensional ones. This study compares implicit learning of intra-and inter-dimensional phonotactic dependencies. A series of six unsupervised implicit-learning experiments shows that a pattern based on agreement between two instances of the same feature is easier to learn than one based on correlation between instances of two different features. The results are interpreted as evidence for domain-general restrictions on the form of domainspecific learning primitives.

No abstract

No abstract