How hierarchical is language use?

Frank, Stefan L.; Bod, Rens; Christiansen, Morten H.

doi:10.1098/rspb.2012.1741

Cited by 154 publications

(111 citation statements)

References 92 publications

(111 reference statements)

Supporting

Mentioning

103

Contrasting

Order By: Relevance

“…This is consistent with recent claims that language acquisition, production, and comprehension rely primarily on sequential information rather than hierarchical structure (19,20). This perspective also offers an explanation for cross-linguistic variation; aside from general constraints on the cognitive mechanisms involved (18, 21), language-specific patterns in the input should be the major determinant of what is learned.…”

supporting

confidence: 72%

Language learners privilege structured meaning over surface frequency

Culbertson

Adger

2014

Proc. Natl. Acad. Sci. U.S.A.

111

View full text Add to dashboard Cite

Although it is widely agreed that learning the syntax of natural languages involves acquiring structure-dependent rules, recent work on acquisition has nevertheless attempted to characterize the outcome of learning primarily in terms of statistical generalizations about surface distributional information. In this paper we investigate whether surface statistical knowledge or structural knowledge of English is used to infer properties of a novel language under conditions of impoverished input. We expose learners to artificial-language patterns that are equally consistent with two possible underlying grammars-one more similar to English in terms of the linear ordering of words, the other more similar on abstract structural grounds. We show that learners' grammatical inferences overwhelmingly favor structural similarity over preservation of superficial order. Importantly, the relevant shared structure can be characterized in terms of a universal preference for isomorphism in the mapping from meanings to utterances. Whereas previous empirical support for this universal has been based entirely on data from cross-linguistic language samples, our results suggest it may reflect a deep property of the human cognitive system-a property that, together with other structure-sensitive principles, constrains the acquisition of linguistic knowledge. A central goal of linguistics is to provide a formal characterization of human knowledge of language. It has long been argued that this knowledge crucially involves rules that refer to abstract structure rather than surface word order (1-3). One classic example is the relationship between English declarative and interrogative sentences. Although in many cases a rule forming the interrogative from the declarative could simply change the surface position of the auxiliary verb, the full range of English facts can only be captured by a rule making reference to the structural position of the auxiliary. For example, to generate the correct interrogative for complex declaratives such as "The man who is a fool is amusing" and "The man is a fool who is amusing," a rule referring to the superficial surface position of the auxiliary-for instance, leftmost or rightmost-will not do. Rather, the rule must pick out the auxiliary in the main clause. Because these complex cases are relatively rare, language learners could initially entertain a surface-based rule before converging on the structure-based alternative. Interestingly, though, children acquiring English do not seem to do this, suggesting that structurebased generalizations are preferred from the very start (4).The idea that explaining the syntax of natural languages requires abstract structure-and that learners posit structure-based generalizations early in the acquisition process-has been challenged from multiple angles. For example, simple recurrent networks trained on English input can replicate correct interrogative ordering patterns, suggesting that surface-based generalizations may be sufficient to characterize linguistic knowledge (...

show abstract

supporting

confidence: 72%

Language learners privilege structured meaning over surface frequency

Culbertson

Adger

2014

Proc. Natl. Acad. Sci. U.S.A.

111

View full text Add to dashboard Cite

show abstract

“…Here I propose that this can be modeled as selection from parallel streams. This proposal is an extension of the one made by Frank et al (2012), showing how syntactic relationships that have previously been explained in terms of hierarchical structure can instead be accounted for in a model using sequential, non-hierarchical structure. In their model, items are combined from parallel sequential streams, eliminating any need for hierarchical processing.…”

Section: Theories Messy Realities and By A Grant From The Centre Formentioning

confidence: 99%

“…The parallel sequential streams model proposed by Frank et al (2012) is primarily aimed at comprehension and at syntax, but there is in principle no reason why it cannot be extended to production data (such as what we find in a corpus) and to other kinds of constructions (such as morphemes, preposition and case constructions, and lexical items). Furthermore, there is no reason why parallel streams should not include alternative rival forms in addition to the forms that are actually selected.…”

Section: Theories Messy Realities and By A Grant From The Centre Formentioning

confidence: 99%

“…Rejecting traditions that have assumed elaborate hierarchical structures in language, a recent proposal (Frank, Bod and Christiansen 2012) offers a model based instead on sequential structure. In this model, the constructions of a language appear in parallel, and complex sequences are aggregated by navigating among parallel streams.…”

Section: Theories Messy Realities and By A Grant From The Centre Formentioning

confidence: 99%

“…put X down: put down knife & fork: knife & fork your X: your Fig. 6 The parallel sequential streams model for combining elements of a construction (after Frank et al 2012) This model can be compared to a musical score for a symphony in which different instruments represent different parallel streams. Each instrument is silent for a while and then plays for a certain time, and then stops again.…”

Section: Theories Messy Realities and By A Grant From The Centre Formentioning

confidence: 99%

See 2 more Smart Citations