Neural underpinnings of processing combinatorial unstated meaning and the influence of individual cognitive style

Abstract

We investigated the neurocognitive mechanisms underlying the processing of combinatorial unstated meaning. Sentences like "Charles jumped for 5 minutes." engender an iterative meaning that is not explicitly stated but enriched by comprehenders beyond simple composition. Comprehending unstated meaning involves meaning contextualization-integrative meaning search in sentential-discourse context. Meanwhile, people differ in how they process information with varying context sensitivity. We hypothesized that unstated meaning processing would vary with individual socio-cognitive propensity indexed by the Autism-Spectrum Quotient (AQ), accompanied by differential cortical engagements. Using functional magnetic resonance imaging, we examined the processing of sentences with unstated iterative meaning in typically-developed individuals and found an engagement of the fronto-parietal network, including the left pars triangularis (L.PT), right intraparietal (R.IPS), and parieto-occipital sulcus (R.POS). We suggest that the L.PT subserves a contextual meaning search, while the R.IPS/POS supports enriching unstated iteration in consideration of event durations and interval lengths. Moreover, the activation level of these regions negatively correlated with AQ. Higher AQ ties to lower L.PT activation, likely reflecting weaker context sensitivity, along with lower IPS activation, likely reflecting weaker computation of events' numerical-temporal specifications. These suggest that the L.PT and R.IPS/POS support the processing of combinatorial unstated meaning, with the activation level modulated by individual cognitive styles.

Keywords: combinatorial unstated meaning; dual-process accounts; individual autistic tendency; language processing; semantic composition.

Fig. 1

Left: Results of naturalness rating in the norming questionnaire (mean ± 1 standard errors). Right: Results of Iterativity judgments in the norming questionnaire (responses of “more than twice”, “more than 10 times”, and “more than 100 times” are grouped as the [iterative] response type).

Fig. 2

Trial procedure in the fMRI experiment. The translation texts at the left side of the frames were provided for illustration only and were not shown to the participants.

Fig. 3

Iterativity judgments in the fMRI scanning session (n = 18). The participants determined how many times the event denoted by the verbal predicate occurred in the sentential context after reading, among four options: (i) once, (ii) 2–10 times, (iii) 11–100 times, or (iv) >101 times. Left: Averaged responses using the options (i)–(iv) across participants. Right: The proportions of each response option chosen per condition.

Fig. 4

Correlations between individual AQ scores and iterativity judgments in the fMRI session (n = 18). The PuncIter, DurIter and GapIter conditions were grouped as ITER. Pearson correlation coefficients were computed between each participant’s AQ scores and iterativity judgments (once, 2–10 times, 11–100 times, >100 times) per condition.

Fig. 5

Imaging results of the whole-brain activation in the (A) “ITER > Multiple” and (B) “Multiple > Once” contrasts. (C) Correlations between the individual brain activation (beta-value) and AQ in the “ITER > Multiple” and “Multiple > Once” contrasts across all ROIs. The ROIs were chosen by the maxima of the activated regions in the contrasts identified in the whole-brain analysis (Table 2). (D) The distributions of the ROI-AQ correlations in the ROIs reported in (C), for the “ITER > Multiple” (black) and “Multiple > Once” (white) contrasts.

Fig. 6

Illustration of the iterative meaning in the PuncIter, DurIter, and GapIter conditions.