Cognitive control during sentence generation

ABSTRACT

Sentence generation involves selecting the right words out of multiple co-activated representations for the right utterance positions. This selection process is hypothesized to involve cognitive control. Previous research has linked cognitive control during language processing to regions within the frontal lobe, including the anterior cingulate cortex (ACC) and left ventrolateral prefrontal cortex (VLPFC). We examined the neural correlates of word and sentence generation to determine whether these cognitive control regions were activated more for sentences than words. We also manipulated the phonological overlap between nouns in the sentence to determine whether cognitive control recruitment increased under phonological competition. Whole-brain analysis revealed increased activation in the ACC and bilateral frontal cortex for sentences over words. Region of interest analysis revealed an effect of phonological competition within the ACC. Psychophysiological interaction (PPI) analysis further showed that functional connectivity within a task-maintenance network increased under phonological competition. These results suggest a role for conflict detection and monitoring during sentence generation, which is consistent with a theoretical framework incorporating domain-general cognitive control processes into language processing.

KEYWORDS:

• Sentence generation
• cognitive control
• phonology
• anterior cingulate cortex
• prefrontal cortex
• fmri

Disclosure statement

No potential conflict of interest was reported by the authors.

We thank Katherine O’Connor for assistance with stimuli, and John VanMeter and Manish Khatiwada for scanning help. This research was partly funded via a university facilitating fund from The George Washington University to the first author.

Notes

¹ The role of the ACC has not been investigated much in this population because lesions to the ACC are not common in left MCA strokes.

² Consistent with a cognitive control account, the ACC ROI also showed a significant effect (Estimate = 174.82, p < .01) while a control low-level visual ROI (suprathreshold occipital pole, probability > 40) did not (Estimate = 42.21, p > .1).