Neural substrates of narrative comprehension and memory

Abstract

When reading a narrative, comprehension and retention of information benefit considerably from the use of situation models--coherent representations of the characters, locations, and activities described in the text. Here we used functional magnetic resonance imaging (fMRI) to explore the neural mechanisms supporting situation model processing. Participants read blocks of sentences that were either unrelated to one another or formed coherent narratives. A timecourse-based approach was used to identify regions that differentiated narrative-level comprehension from sentence-level comprehension. Most brain regions that showed modulation of activation during narrative-level comprehension were also modulated to a lesser extent during sentence-level comprehension, suggesting a shared reliance on general coherence-building mechanisms. However, tentative evidence was found for narrative-specific activation in dorsomedial prefrontal cortex. Additional analyses identified spatiotemporally distinct neural contributions to situation model processing, with posterior parietal regions supporting situation model construction and frontotemporal regions supporting situation model maintenance. Finally, a set of subsequent memory analyses demonstrated that the boost in comprehension and memory performance observed for coherent materials was attributable to the use of integrative situation models rather than lower-level differences in sentence-level or word-level encoding. These results clarify the functional contributions of distinct brain systems to situation model processing and their mapping onto existing psychological models of narrative comprehension.

Figure 1

Sample stimuli and memory tests used in the current study.

Figure 2

Regions that showed a significant condition × time interaction. Number labels correspond to timecourse panels in Figure 2 and row IDs in Table 1.

Figure 3

Timecourses of activation in the story and scrambled conditions. Error bars represent 95% confidence intervals. The x-axis in each panel indicates elapsed time (in seconds) since block onset; the y-axis indicates % change in BOLD activation relative to baseline.

Figure 4

A. Regions that showed transient positive activation at block onset in the story condition (blue), scrambled condition (green), or both (turquoise). Clockwise from top left: left medial, left posterior, right posterior, and right medial views. B. Common onset effects in visual cortex (activation reflects average of turquoise voxels). C. Selective onset effect in the story condition in PPC.

Figure 5

A. Regions that showed linear increases in activation as a function of reading time in the story condition (blue), scrambled condition (green), or both (turquoise). Clockwise from left: right ventral, left ventral, left lateral, and right lateral views. Timecourses are presented for representative regions showing either a common linear increase in both conditions (B; activation reflects turquoise voxels) or a selective linear increase in the story condition (C).

Figure 6

A. Regions that showed linear decreases in activation as a function of reading time in the story condition (blue), scrambled condition (green), or both (turquoise). From left to right: left dorsal, right dorsal, left posterior, and right posterior cortical views. Timecourses are presented for representative regions in left postcentral gyrus (B) and left lingual gyrus (C); activation reflects turquoise voxels.

Figure 7

Regions in which block-level (A) or sentence-level (B) activation predicted subsequent recognition memory for sentences in the story condition.