Ullman Michael T, Bulut Talat, Walenski Matthew
Brain and Language Laboratory, Department of Neuroscience, Georgetown University, Washington, DC, USA.
Neurobiology of Language Department, Max Planck Institute for Psycholinguistics, Nijmegen, the Netherlands; Department of Speech and Language Therapy, Istanbul Medipol University, Istanbul, Turkey.
Cognition. 2024 Oct;251:105875. doi: 10.1016/j.cognition.2024.105875. Epub 2024 Jul 16.
Although language depends on storage and composition, just what is stored or (de)composed remains unclear. We leveraged working memory load limitations to test for composition, hypothesizing that decomposed forms should particularly tax working memory. We focused on a well-studied paradigm, English inflectional morphology. We predicted that (compositional) regulars should be harder to maintain in working memory than (non-compositional) irregulars, using a 3-back production task. Frequency, phonology, orthography, and other potentially confounding factors were controlled for. Compared to irregulars, regulars and their accompanying -s/-ing-affixed filler items yielded more errors. Underscoring the decomposition of only regulars, regulars yielded more bare-stem (e.g., walk) and stem affixation errors (walks/walking) than irregulars, whereas irregulars yielded more past-tense-form affixation errors (broughts/tolded). In line with previous evidence that regulars can be stored under certain conditions, the regular-irregular difference held specifically for phonologically consistent (not inconsistent) regulars, in particular for both low and high frequency consistent regulars in males, but only for low frequency consistent regulars in females. Sensitivity analyses suggested the findings were robust. The study further elucidates the computation of inflected forms, and introduces a simple diagnostic for linguistic composition.
虽然语言依赖于存储和组合，但究竟存储了什么或（被）组合了什么仍不清楚。我们利用工作记忆负荷限制来测试组合，假设分解形式会特别加重工作记忆的负担。我们聚焦于一个经过充分研究的范式，即英语屈折形态学。我们预测，在3-back产出任务中，（组合性的）规则形式在工作记忆中比（非组合性的）不规则形式更难维持。对频率、语音、正字法和其他潜在的混杂因素进行了控制。与不规则形式相比，规则形式及其附带的-s/-ing附加填充项产生了更多错误。突出仅规则形式的分解，规则形式比不规则形式产生了更多的裸词干（如walk）和词干词缀错误（walks/walking），而不规则形式产生了更多过去式词缀错误（broughts/tolded）。与之前关于规则形式在某些条件下可以存储的证据一致，规则-不规则差异尤其适用于语音一致（而非不一致）的规则形式，特别是男性中的低频和高频一致规则形式，但仅适用于女性中的低频一致规则形式。敏感性分析表明研究结果具有稳健性。该研究进一步阐明了屈折形式的计算，并引入了一种简单的语言组合诊断方法。
