Alan Prince

Faithfulness and reduplicative identity

1 Workshop at which this work was first presented (Utrecht, June 22-24, 1994). For comments on this material, we are grateful to them and the other workshop participants, especially Austin have provided valuable feedback; and the comments, questions, and suggestions from the participants in the (eventually joint) UMass and Rutgers Correspondence Theory seminars were particularly important for the development of this work. For useful discussion of numerous points, we would like to thank

The Emergence of the Unmarked: Optimality in Prosodic Morphology

the University of Victoria, UMass, and Rutgers University, particularly to Nora Aion, Akin Akinlabi, Olga Babko-Malaya, Eric Bakovic, Jill Beckman, Jose Benki, Laura Benua, Barry Carlson, Megan Crowhurst, Merce Gonzalez, Sharon Inkelas, Zvi Gilbert, Amalia Gnanadesikan, Rene Kager, Ed Keer, John Kingston, Paul Kiparsky, Marc van Oostendorp, Jaye Padgett, Rossina Petrova, Ivan Sag, Vieri Samek-Lodovici, Lisa Selkirk, Yael Sharvit, Pat Shaw, Paul Smolensky, Rachel Thorburn, Suzanne Urbanczyk, Jennifer Yearley, Laura Walsh, and Arnold Zwicky, for their comments and questions. We owe a special debt to Junko Ito and Armin Mester, who reviewed a draft of the NELS handout and provided much useful feedback. T The Emergence of the Unmarked Optimality in Prosodic Morphology1

Foot and word in prosodic morphology: The Arabic broken plural

This article proposes a theory of prosodic domain circumscription, by means of which rules sensitive to morphological domain may be restricted to a prosodically characterized (sub-)domain in a word or stem. The theory is illustrated primarily by a comprehensive analysis of the Arabic broken plural; it is further supported by analysis of a number of processes from other languages, yielding a formal typology of domain-circumscription effects. The results obtained here depend on, and therefore confirm, two central principles of Prosodic Morphology: (1) the Prosodic Morphology Hypothesis, which requires that templates be expressed in prosodic, not segmental terms; and (2) the Template Satisfaction Condition, which requires that all elements in templates are satisfied obligatorily.

Faithfulness and Identity in Prosodic Morphology

2. Correspondence Theory . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2.1 The Role and Character of Correspondence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 2 2.2 Some Constraints on Correspondent Elements . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 2.3 Correspondence Theory and the P ARSE/FILL Model . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 8

Constraints in Phonological Acquisition: Learning phonotactic distributions

Many essentials of a language’s phonology can be learned from distributional evidence, in the absence of detailed morphological analysis. But distributional learning from positive evidence encounters the subset problem in full force. Here we explore an approach in which the learning algorithm, based on the errordriven variant of Recursive Constraint Demotion (RCD: Tesar 1995, Tesar & Smolensky 1998), is persistently biased to place markedness constraints as high as possible, but aims to place faithfulness constraint as low as possible. The learner seeks only to reproduce the output from identical input, avoiding all concern with nontrivial underlying forms; under the M?F bias this results in significant learning. (Hayes 1999 independently develops a similar approach from the same basic assumptions.) We introduce an explicit measure of the degree to which a hierarchy possesses M?F structure, and we investigate the consequences of trying to maximize this measure by low placement of F in suitably biased versions of RCD. We argue that existing proposals by which M?F structure is carried over from an initial state through a learning procedure blind to the M/F distinction, as in the different conceptions of Smolensky 1996 and Ito & Mester 1999a, cannot accomplish this goal successfully, as they are currently understood. We conclude that Biased Constraint Demotion (BCD) must be used by the learner at each step. The key issue is deciding which F to rank when there is more than one F constraint to choose from. We suggest that the main desideratum is the ’freeing up’ of further M constraints for ranking, though we also show that such decisions have further consequences downstream for the resultant hierarchy that may motivate a certain kind of ‘look ahead’ in the decision-making process. We also consider the issue of the ranking of specialgeneral pairs of faithfulness constraints, arguing that the matter cannot be resolved by examining the structure of constraints in isolation. We show that special/general relations can be derived mid-hierarchy, on the one hand, and on the other, can arise between constraints that appear to be independent. We note that in sharp contrast to the faithfulness situation, special/general relations between markedness constraints are handled automatically by BCD; this provides learning-theoretic motivation for resolving the positional markedness vs. positional faithfulness controversy (Beckman 1998, Zoll 1998) and for deeper scrutiny of faithfulness theory as a whole. Learning Phonotactic Distributions Alan Prince & Bruce Tesar Department of Linguistics Rutgers Center for Cognitive Science Rutgers University, New Brunswick 10/8/1999

Why no mere mortal has ever flown out to center field

The English past tense system has recently been used to argue that formal grammatical categories (such as root, rule, and lexical item) may not be necessary to explain the acquisition and knowledge of language. Rumelhart and McClelland (1986) devised a connectionist model relying solely on phonological information; it is often suggested that any deficiencies of such a model can be remedied by supplying it with semantic information. These proposals are incorrect: Grammatical categories and abstract morphological structure are indispensable and cannot be replaced with semantics while preserving the patterns of psychological generalization in the system. Linguists have noted that irregular past tense mappings (e.g., fly/flew; stick/stuck) apply only when a verbs root is marked in the lexicon as having an irregular past. Because nouns are never so marked, verbs with noun roots—denominal verbs—are regular even if they are phonologically identical to irregular verbs, hence: flied out/*flew out to center field; high-sticked/*high-stuck the goalie. Experiment 1 shows that adult subjects are highly sensitive to this principle when rating regular and irregular past tense forms of novel versions of irregular sounding verbs: New verbs formed from nouns were judged as better with a regular past tense (e.g., line-drived was the preferred past of “to hit a line drive”): new verbs formed from verbs were judged as better with an irregular past tense (e.g., line-drove was the preferred past of “to drive along a line”). Experiment 2 replicated the results with noncollege-educated adults, showing that the effect is not due to prescriptive language training. Experiment 3 tested an alternative to the formal grammatical account proposed by Lakoff (1987): When a verb has two meanings, one with an irregular past and one with a regular past, the irregular will belong to the meaning that is more central. Using regression techniques and ratings data, we disconfirm this prediction: In the data from Experiment 1, judgments of regular and irregular forms of a new verb are shown to be affected by whether the verb is derived from a noun or a verb, but not by whether its new sense is near the center or the periphery of the sense of the word it was derived from. Experiments 4 and 5 explain the few apparent counter-examples by gathering independent evidence for a short-circuiting process: When a denominal verb appears to have an irregular past tense form, it is because speakers sometimes interpret such verbs as having been derived directly from a related irregular verb root, bypassing the relevant noun. The experiments serve as a straightforward demonstration that representations of formal grammatical categories and structures are powerful determinants of linguistic behavior, and are not reducible to semantics, phonology, or prescriptive training.

Surgery in Language Learning

The architecture of generative phonology brings with it a difficult challenge for any learner: underlying forms must be acquired at the same time as the phonology – the system of rules or constraint-rankings. Yet, each depends on the other, and neither is known in advance. If the learner had prior knowledge of the underlying forms, then the constraint-ranking could be determined by familiar procedures. If the learner knew the ranking, then the range of viable underlying forms would be greatly limited, simplifying the process of finding them. In addition, the learner must identify a phonology which is maximally restrictive, so that distributional restrictions implicit in the data are enforced, rather than being portrayed as accidental gaps in the lexicon. Since it is impossible to explore all possible lexicon-phonology pairings, an effective learner must use an incremental strategy which goes back and forth between hypotheses about the lexicon and hypotheses about the phonology, testing and improving each until a satisfactory match is found. A principal issue in designing any such procedure is deciding what to do when the mapping fails: should the lexicon be changed or should the phonology be changed?

Rules and connections in human language

Abstract Recently ‘connectionist’ or ‘parallel distributed processing’ (PDP) approaches to brain modelling have attracted an enormous amount of attention. These models are said to be faithful to neurophysiological and to behavioral data in a way that previous approaches based on symbolic computation were not. A PDP simulation by Rumelhart and McClelland of childrens acquisition of the past tense in English has been one of the most famous demonstrations of the advantages of the connectionist approach. In a recent special issue of the journal Cognition devoted to Connectionism and Symbol Systems, Steven Pinker and Alan Prince examine this model and the relevant data in great detail, finding severe limitations in the ability of current PDP models to explain human language and cognition. The key points of their analysis are summarised in the following article.