A demonstration of the uncomputability of parametric models of language acquisition and a biologically plausible alternative
Leivada, Evelina (Universitat Rovira i Virgili)
Murphy, Elliot (University of Texas Health Science Center at Houston)

Data:	2022
Resum:	The logical problem of language acquisition has been at the forefront of psycholinguistics and behavioral neuroscience for decades. One of the most influential answers to the problem of how successful acquisition occurs on the basis of noisy input suggests that the child is aided by innate principles and parameters (P&P). These are conceived aspart of our biological endowment for language. Previous work on the computability of parametric models has focused on the process of parameter-setting,leaving settability unaddressed. Settability is a key notion in parametric models since it provides an answer to the logical problem of language acquisition: the setting of one parameter carries implications for the settability of others, minimizing the child's task. However, a mathematical analysis of the expected probability of successful computation of settability relations has not been carried out. We report results from a novel program developed to calculate the probability of successful computation of a network of 62 linguistic parameters as attested in 28 languages, spanning across 5 language families. The results reveal that some parameters have an extremely low probability of successful computation, such that trillions of unsuccessful computations are expected before a successful setting occurs. Using the same program, we performed an additional second analysis to a different network, covering 94 parameters from58 languages and 15 language families. In this case, the estimated number of expected unsuccessful computations rose from trillions to quadrillions. These results raise concerns about the computational feasibility of the highly influential P&P approach to language development. Merging insights from various acquisition models, including some developed within P&P, a biologically plausible alternative is offered for the process of deciphering a target grammar in the acquisition of both spoken and signed languages. Overall, our analysis of the P&P approach to language acquisition centers learnability and computability constraints as the major factors for determining the psychological plausibility of grammar development.
Nota:	This work received support from the European Union's Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement n° 746652 and from the Spanish Ministry of Science, Innovation and Universities under the Ramón y Cajal grant agreement n. RYC2018-025456-I (to EL).
Drets:	Aquest document està subjecte a una llicència d'ús Creative Commons. Es permet la reproducció total o parcial, la distribució, la comunicació pública de l'obra i la creació d'obres derivades, sempre que no sigui amb finalitats comercials, i sempre que es reconegui l'autoria de l'obra original.
Llengua:	Anglès
Document:	Article ; recerca ; Versió publicada
Matèria:	Acquisition ; Computation ; Learnability ; Parameter ; Universal Grammar
Publicat a:	Language Development Research, Vol. 2 Issue 1 (2022), p. 105-138, ISSN 2771-7976

DOI: 10.34842/2022-585

34 p, 3.7 MB

El registre apareix a les col·leccions:
Articles > Articles de recerca
Articles > Articles publicats