Search

Reduplication is common, but analogous reversal processes are rare, even though reversal, which involves nested rather than crossed dependencies, is less complex on the Chomsky hierarchy. We hypothesize that the explanation is that repetitions can be recognized when they match and reactivate a stored trace in short-term memory, but recognizing a reversal requires rearranging the input in working memory before attempting to match it to the stored trace. Repetitions can thus be recognized, and repetition patterns learned, implicitly, whereas reversals require explicit, conscious awareness. To test these hypotheses, participants were trained to recognize either a reduplication or a syllable-reversal pattern, and then asked to state the rule. In two experiments, above-chance classification performance on the Reversal pattern was confined to Correct Staters, whereas above-chance performance on the Reduplication pattern was found with or without correct rule-stating. Final proportion correct was positively correlated with final response time for the Reversal Correct Staters but no other group. These results support the hypothesis that reversal, unlike reduplication, requires conscious, time-consuming computation.

We present evidence from experiments on novel blend formation showing that adult English speakers have access to constraints that give phonological privilege to heads, nouns, and proper nouns, even though the nonblend phonology provides no evidence that such constraints are generally active in the grammar of English. Our results (i) demonstrate that these positional constraints are universally available; (ii) confirm that the lexical category ‘proper noun’ has the status of a strong position, which has broader implications for the role of lexical categories in positional privilege effects; and (iii) confirm that strong positions based on salience from nonphonetic sources (such as morphosyntactic, semantic, or psycholinguistic salience) participate in position-specific phonological phenomena.

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, 1988a) in a Maximum Entropy phonotactic-learning framework (Goldwater & Johnson, 2003; Hayes & Wilson, 2008) 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 (1961) (“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.

Nowadays natural language user interfaces, such as chatbots and conversational agents, are very common. A desirable trait of such applications is a sense of humor. It is, therefore, important to be able to measure quality of humorous responses. However, humor evaluation is hard since humor is highly subjective. To address this problem, we conducted an online evaluation of 30 dialog jokes from different sources by almost 300 participants -- volunteers and Mechanical Turk workers. We collected joke ratings along with participants» age, gender, and language proficiency. Results show that demographics and joke topics can partly explain variation in humor judgments. We expect that these insights will aid humor evaluation and interpretation. The findings can also be of interest for humor generation methods in conversational systems.

This paper incorporates morphological markedness constraints into a framework in which morphology and phonology directly interact, modeled with interleaving of morphological and phonological constraints in serial OT (Wolf 2008, 2009). Morphological markedness constraints are constraints against realization (or spell-out) of morphologically marked feature sets. The empirical data motivating this proposal mainly come from the case-study of Russian genitive plural allomorphy, which is analyzed as involving tradeoffs between morphological markedness and other constraints in the grammar, including the purely phonological ones. This proposal explains the otherwise apparently arbitrary and unnatural transderivational dependency between the nominative singular and the genitive plural in Russian. Additionally, this account of the genitive plural allomorphy provides a unified explanation for several seemingly exceptional sub-generalizations which upon examining lexical statistics turn out to be regular. Implications and predictions of an interleaved model with morphological markedness constraints are discussed throughout the article.

We present a syntactic account of the derivation of two types of attributivenominal compounds in Spanish, Russian and Greek. These include right-headed “root” compounds, which exhibit more “word”-like properties and single stress domains, and left-headed “semi-phrasal” compounds with more phrasal properties and independent stress domains for the two compound members. We propose that both compound structures are formed on a small clause predicate phrase, with their different properties derived from the merger of the predicate member of the small clause as a root or as a larger nominal unit with additional functional projections. The proposed structures provide an explanation of observed lexical integrity effects, as well as specific predictions of patterns of compound formation crosslinguistically.

It is commonly assumed that patterns of syncretism in inflectional paradigms are restricted in some way. In this article, I show how such restrictions can reflect cognitive constraints on language learning. Namely, I construct a learning algorithm that is biased toward certain types of affix distributions in paradigms, thereby rendering them systematic. In developing this algorithm, I rely on the traditional notions of underspecification and blocking, but recast them in terms of learners' biases toward generalization strategies based on cross-situational intersections and default reasoning. This algorithm allows us to test claims about systematicity of syncretism using typological data and language acquisition studies. In the last part of the article, I present a crosslinguistic survey of verbal agreement paradigms that supports the algorithm's predictions.

This chapter aims to introduce readers not familiar with computational modelling to some approaches and issues in the formal study of learnability, and the relevance of this field to theoretical linguistics and inflectional morphology in particular. After a general overview, the chapter highlights some of the obstacles in learning inflection. Inflection, considered separately from other components of language, is relatively restricted in its expressive power, which should make it easier to learn than syntax. However, inflectional systems are full of irregularities and mismatches between different levels of structure, and such irregularities make learning difficult. Overall, it is concluded that linguistically interesting proposals for machine learning of inflection should provide explanations for the nature and extent of irregularities and for the specific patterns of language acquisition and language change.

In this paper, I propose to extend a model of grammar in which lexical spellout and phonological computation occur in the same component (e.g., serial OT with Optimal Interleaving (Wolf, 2008)) with a family of morphological markedness constraints. These constraints punish realization of morpho-syntactically marked combinations of features. They form a markedness hierarchy in which constraints against more marked feature(s) are ranked higher than constraints against less marked feature(s). Inclusion of such constraints provides a natural way to account for certain cases of non-realization of morphological contrasts in marked contexts (zero marking), as well as cases of syncretism in which a marked paradigm cell in certain contexts becomes occupied by an exponent typically used in less marked cells. Analyses of both types of cases are discussed in this paper. Moreover, this model predicts tradeoffs and interactions between morphological markedness and phonological constraints. Russian Genitive Plural allomorphy, discussed in this paper, provides one example of such interactions.

This paper considers a morphological analogue of a psychological experiment on learning categories with the structure of Boolean connectives AND, OR, and XOR (exclusive OR). This experiment will test the effects of logical structure and the presence of null/overt affixes on paradigm complexity. The results are partially consistent with the previous psychological studies, however, they also differ in several ways from analogous non-linguistic experiments. At this point, it is unclear exactly what factors are responsible for these differences.

In this paper, I show that a problem of learning a morphological paradigm is similar to a problem of learning a partition of the space of Boolean functions. I describe several learners that solve this problem in different ways, and compare their basic properties.