Children’s Computation of Complex Linguistic Forms: A Study of Frequency and Imageability Effects Cristina D. Dye 1,2 *, Matthew Walenski 3,4 , Elizabeth L. Prado 2,5 , Stewart Mostofsky 6 , Michael T. Ullman 2 * 1 Centre for Research in Linguistics and Language Sciences, Newcastle University, Newcastle Upon Tyne, Tyne and Wear, United Kingdom, 2 Brain and Language Lab, Department of Neuroscience, Georgetown University, Washington, DC, United States of America, 3 School of Speech, Language, and Hearing Sciences, San Diego State University, San Diego, California, United States of America, 4 Center for Research in Language, University of California San Diego, San Diego, California, United States of America, 5 Department of Nutrition, University of California Davis, Davis, California, United States of America, 6 Kennedy Krieger Institute, Johns Hopkins University, Baltimore, Maryland, United States of America Abstract This study investigates the storage vs. composition of inflected forms in typically-developing children. Children aged 8–12 were tested on the production of regular and irregular past-tense forms. Storage (vs. composition) was examined by probing for past-tense frequency effects and imageability effects – both of which are diagnostic tests for storage – while controlling for a number of confounding factors. We also examined sex as a factor. Irregular inflected forms, which must depend on stored representations, always showed evidence of storage (frequency and/or imageability effects), not only across all children, but also separately in both sexes. In contrast, for regular forms, which could be either stored or composed, only girls showed evidence of storage. This pattern is similar to that found in previously-acquired adult data from the same task, with the notable exception that development affects which factors influence the storage of regulars in females: imageability plays a larger role in girls, and frequency in women. Overall, the results suggest that irregular inflected forms are always stored (in children and adults, and in both sexes), whereas regulars can be either composed or stored, with their storage a function of various item- and subject-level factors. Citation: Dye CD, Walenski M, Prado EL, Mostofsky S, Ullman MT (2013) Children’s Computation of Complex Linguistic Forms: A Study of Frequency and Imageability Effects. PLoS ONE 8(9): e74683. doi:10.1371/journal.pone.0074683 Editor: Mark Aronoff, Stony Brook University, United States of America Received February 22, 2013; Accepted August 7, 2013; Published September 9, 2013 Copyright: ß 2013 Dye et al. This is an open-access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Funding: Support for this project was provided to MTU from NIH R01 MH58189, NIH R01 HD049347, the National Alliance for Autism Research, the Mabel Flory Trust, and the Simons Foundation. The funders had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. Competing Interests: The authors have declared that no competing interests exist. * E-mail: cdye930@gmail.com (CD); michael@georgetown.edu (MU) Introduction How is language computed in the mind? Although we now know that language computation requires both storage and composition, it remains unclear which aspects of language are stored and which are (de)composed, and under what circumstanc- es. Moreover, this is even less well understood in children than adults. In adults, a considerable amount of research has investigated this issue by probing for storage effects in the on- line computation of existing inflected forms, in particular in the contrast between regular and irregular inflected variants (e.g., walked vs. dug), which can be equated for phonological, semantic, and other factors. Here, we examine this issue in children. Storage/composition theoretical models A variety of theoretical proposals regarding the computation of existing inflected forms have been put forth. These can be grouped in two broad classes: single-mechanism models and dual-system models. Single-mechanism models propose that all inflected forms are computed with the same basic mechanisms. For example, according to connectionist models, all previously-encountered inflected words are represented in a distributed associative memory [1–6]. Such models suggest that forms that exhibit a common and consistent inflectional pattern (regular forms, as well as some irregulars, such as sing-sang, ring-rang) rely primarily on phonological representations in this distributed memory. Rule- only models constitute another type of single-mechanism model. Here, too, all inflected forms (i.e., both regular and irregular) are handled by the same or similar mechanisms, which in this case, involve rule-based processes [7–10]. Dual-system models, on the other hand, hold that while some inflected forms are stored, the computation of others depends on (de)composition by rule-governed processes in a separate compu- tational and neurocognitive system. Irregularly inflected forms (e.g., dug), which cannot be fully specified by a default rule, are claimed to depend on memorized representations. Regulars hold a different status. Some dual-system models hold that regulars are always composed (e.g., walk +ed) by a default rule [11–13]. Others, however, suggest that even if regulars can be assembled by rule- governed processes, they can also be stored, and indeed often are, with the likelihood of storage depending on various item-level (e.g., frequency) and subject-level (e.g., sex) factors [14–20]. For example, women, who appear to have superior verbal memories as compared to men [21–25], may also be more likely than men to store regularly inflected forms [14,20,25]. These models, as well as similar ones, have been proposed not only for adults, but also for children. Models claiming the dependence of all inflected forms on (an associative) memory suggest that children, like adults, rely on word storage or pattern association for both regulars and irregulars. For example, Bybee PLOS ONE | www.plosone.org 1 September 2013 | Volume 8 | Issue 9 | e74683