Isolating syntax in natural language: MEG evidence for an early contribution of left posterior temporal cortex

2020

Preprint

Self Cite

23In the neurobiology of language, a fundamental challenge is deconfounding syntax from semantics. 24Changes in syntactic structure usually correlate with changes in meaning. We approached this 25 challenge from a new angle. We deployed word lists, which are usually the unstructured control in 26 studies of syntax, as both the test and the control stimulus. Three-noun lists (lamps, dolls, guitars) 27 were embedded in sentences (The eccentric man hoarded lamps, dolls, guitars…) and in longer lists 28 (forks, pen, toilet, rodeo, graves, drums, mulch, lamps, dolls, guitars…). This allowed us to perfectly 29 control both lexical characteristics and local combinatorics: the same words occurred in both 30 conditions and in neither case did the list items locally compose into phrases (e.g. 'lamps' and 'dolls' 31 do not form a phrase). But in one case, the list partakes in a syntactic tree, while in the other, it does 32 not. Being embedded inside a syntactic tree increased source-localized MEG activity at ~250-300ms 33 from word onset in the left inferior frontal cortex, at ~300-350ms in the left anterior temporal lobe 34 and, most reliably, at ~330-400ms in left posterior temporal cortex. In contrast, effects of semantic 35 association strength, which we also varied, localized in left temporo-parietal cortex, with high 36 associations increasing activity at around 400ms. This dissociation offers a novel characterization of 37 the structure vs. meaning contrast in the brain: The fronto-temporal network that is familiar from 38 studies of sentence processing can be driven by the sheer presence of global sentence structure, while 39 associative semantics has a more posterior neural signature. 41Running head: NEURAL PROCESSING OF LISTS WITH AND WITHOUT SYNTAX 3 SIGNIFICANCE STATEMENT 42Human languages all have a syntax, which both enables the infinitude of linguistic creativity and 43 determines what is grammatical in a language. The neurobiology of syntactic processing has, however, 44 been challenging to characterize despite decades of study. One reason is pure manipulations of syntax are 45 difficult to design. The approach here offers a perfect control of two variables that are notoriously hard to 46 keep constant when syntax is manipulated: word meaning and phrasal combinatorics. The same noun lists 47 104 MATERIALS AND METHODS 105 Stimuli & Design106We selected concrete English nouns based on the concreteness rating corpus by Brysbaert, Warriner, 107and Kuperman (2014). From this pool, we then selected nouns that are matched in their log frequency 108 from the SUBTLEX-US corpus (Brysbaert & New, 2009). The critical list nouns were changed from 109 their singular form to plural to block potential noun-noun compounding (e.g. lamp doll could form a 110 Running head: NEURAL PROCESSING OF LISTS WITH AND WITHOUT SYNTAX 7 phrase, but lamps dolls could not). These plural nouns were then used to construct our critical three-111 noun lists such as lamps, dolls, guitars. The lexical characteristics are summarized in Table 2....

mentioning

confidence: 53%

Lists with and without syntax: A new approach to measuring the neural processing of syntax

Law

2020

Preprint

Self Cite

“…On the one hand, as summarized above, there is now a large body of converging evidence behind the hypothesis that sections of the PTL play a role in relational and/or thematic knowledge processing. However, there is also accumulating evidence for the left PTL’s involvement in syntactic processing in language comprehension (Flick & Pylkkänen, 2020; Matchin, Brodbeck, Hammerly, & Lau, 2018; Matchin, Hammerly, & Lau, 2017; Rodd, Longe, Randall, & Tyler, 2010; Rogalsky et al, 2018; Snijders et al, 2008; Tyler, Cheung, Devereux, & Clarke, 2013). The latter could account for the region’s heightened responses to two-word phrases relative to lists and single words, since the latter two conditions do not require the construction of phrasal structure.…”

Section: Discussionmentioning

confidence: 99%

From letters to composed concepts: A magnetoencephalography study of reading

Flick

Abdullah

2020

Preprint

Self Cite

Language comprehension requires the recognition of individual words and the combination of their meanings to yield complex concepts or interpretations. Rather than simple concatenation, this combinatory process often requires the insertion of unstated semantic material between words, based on thematic or feature knowledge of the concepts. For example, the phrase horse barn is not interpreted as a blend of a horse and a barn, but specifically a barn in which horses are kept. Mounting evidence suggests two cortical semantic hubs, in left temporoparietal and anterior temporal cortex, underpin thematic and feature concept knowledge, but much remains unclear about how these putative hubs contribute to combinatory language processing. Using magnetoencephalography, we contrasted source-localized responses to modifier-noun phrases involving thematic relations vs. feature modifications, while also examining how lower-level orthographic processing fed into responses supporting word combination. Twenty-eight participants completed three procedures examining responses to letter-strings, adjective-noun phrases, and noun-noun combinations that varied the semantic relations between words. We found that while color + noun phrases engaged the left anterior temporal lobe (150-300 ms after phrasal head), posterior temporal lobe (150-300 ms), and angular gyrus (300-450 ms), only left posterior temporal lobe responses were sensitive to implicit thematic relations between composing nouns (150-300 ms). We additionally identified a left temporo-occipital progression from orthographic to lexical processing, feeding into ventral anterior areas engaged in the combination of word meanings. Finally, by examining source signal leakage, we characterized the degree to which these responses could be distinguished from one another, using linear source estimation.

“…The language mask (see the light pink region in Figure 4A and 4E) covered regions including the whole temporal lobe, the inferior frontal gyrus (IFG; defined as the combination of BAs 44 and 45), the ventromedial prefrontal cortex (vmPFC; defined as BA11), the angular gyrus (AG; defined as BA39) and the supramarginal gyrus (SMA; defined as BA 40). The left AG and vmPFC have also been implicated in previous literature on conceptual combination (Bemis and Pylkkänen, 2011; Price et al, 2015), and the LIFG and the LMTG have been suggested to underlie syntactic combination (Flick and Pylkkänen, 2020; Haggort, 2005; Lyu et al, 2019; Matchin et al, 2019; Matchin and Hicock, 2020). The medial temporal lobe, on the other hand, has been held critical for associative memory in addition to the anterior temporal lobe (e.g., Day et al, 2003; Ison et al, 2015; Miller et al, 2013; Rey et al, 2018).…”

Section: Methodsmentioning

confidence: 96%

Disentangling semantic composition and semantic association in the left temporal lobe

2020

Preprint

Self Cite

Although composing two words into a complex representation is conceptually different from forming associations between a pair of words, the brain regions supporting semantic composition have also been implicated for associative encoding. Prior work has never directly compared the neural substrates of composition and association. Here, we adopted a two-word magnetoencephalography (MEG) paradigm which varies compositionality ("French/Korean cheese" vs. "France/Korea cheese") and strength of association ("France/French cheese" vs. "Korea/Korean cheese") between the words. We applied both univariate regression analyses and multivariate pattern classification to the source-localized MEG data in a bilateral language network. We show that the left anterior and middle temporal lobe are distinctly modulated by semantic composition and semantic association. Pattern-based directed connectivity analysis further revealed a continuous information flow from the anterior to the middle temporal region. These findings shed light into a functional dissociation within the left temporal lobe for compositional and distributional semantic processing.