School of Philosophy, Psychology & Language Sciences, University of Edinburgh, UK.
School of Philosophy, Psychology & Language Sciences, University of Edinburgh, UK.
Cortex. 2022 Oct;155:287-306. doi: 10.1016/j.cortex.2022.08.002. Epub 2022 Aug 17.
Language processing requires the integration of diverse sources of information across multiple levels of processing. A range of psycholinguistic properties have been documented in previous studies as having influence on brain activation during language processing. However, most of those studies have used factorial designs to probe the effect of one or two individual properties using highly controlled stimuli and experimental paradigms. Little is known about the neural correlates of psycholinguistic properties in more naturalistic discourse, especially during language production. The aim of our study is to explore the above issues in a rich fMRI dataset in which participants both listened to recorded passages of discourse and produced their own narrative discourse in response to prompts. Specifically, we measured 13 psycholinguistic properties of the discourse comprehended or produced by the participants, and we used principal components analysis (PCA) to address covariation in these properties and extract a smaller set of latent language characteristics. These latent components indexed vocabulary complexity, sensory-motor and emotional language content, discourse coherence and speech quantity. A parametric approach was adopted to study the effects of these psycholinguistic variables on brain activation during comprehension and production. We found that the pattern of effects across the cortex was somewhat convergent across comprehension and production. However, the degree of convergence varied across language properties, being strongest for the component indexing sensory-motor language content. We report the full, unthresholded effect maps for each psycholinguistic variable, as well as mapping how these effects change along a large-scale cortical gradient of brain function. We believe that our findings provide a valuable starting point for future, confirmatory studies of discourse processing.
语言处理需要整合来自多个处理层次的多种信息来源。在之前的研究中，已经记录了一系列心理语言学特性，这些特性对语言处理过程中的大脑激活有影响。然而，大多数研究使用因子设计来使用高度控制的刺激和实验范式来探测一个或两个个体特性的影响。对于更自然的话语中的心理语言学特性的神经相关性，尤其是在语言产生期间，人们知之甚少。我们研究的目的是在一个丰富的 fMRI 数据集中探索上述问题，其中参与者既听取记录的话语段落，又根据提示自己产生叙事话语。具体来说，我们测量了参与者理解或产生的话语的 13 种心理语言学特性，并使用主成分分析（PCA）来解决这些特性中的共变问题，并提取出一组更小的潜在语言特征。这些潜在成分索引词汇复杂性、感觉运动和情感语言内容、话语连贯性和言语数量。我们采用参数方法来研究这些心理语言学变量对理解和产生过程中大脑激活的影响。我们发现，皮层上的效应模式在理解和产生过程中有些收敛。然而，收敛程度因语言特性而异，索引感觉运动语言内容的成分最强。我们报告了每个心理语言学变量的完整、未阈值的效应图，以及这些效应如何沿着大脑功能的大规模皮质梯度变化。我们相信，我们的发现为未来的话语处理确认性研究提供了一个有价值的起点。
