You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster B7, Wednesday, November 8, 3:00 – 4:15 pm, Harborview and Loch Raven Ballrooms

Prediction-related activity in the medial prefrontal cortex reflects processing of cataphor cues

Andrew Jahn1, Dave Kush2, Ashley Lewis1, Julie Van Dyke1;1Haskins Laboratories, 2Norwegian University of Science and Technology

Introduction: The role of executive function in language - in particular, prediction and prediction error (PE) – is an area that has received increasing attention in linguistic research. To date, most linguistic studies using fMRI have focused on the left inferior frontal gyrus (LIFG) as a key hub in both syntax-processing and cognitive control of interference (Novick et al., 2005; Rogalsky et al., 2008). However, recent computational models of executive function have shown the medial prefrontal cortex (mPFC) to be involved in prediction and PE across a wide range of experimental contexts. We sought to test the predictions of one of these computational models (Alexander & Brown, 2011) in a linguistic context. According to this model, prediction-related activity should linearly increase over time until the predicted event occurs. Methods: We scanned 23 young adults (ages 18-27) reading sentences containing a cataphor construction, wherein a referring pronoun ("he") is placed before its antecedent ("John"), as below: When he returned the car that had been rented two days ago, John went home. The cataphor creates an expectation for an antecedent to appear later in the sentence. According to model predictions, there should be a ramping up of activity in the mPFC starting at the cataphor and ending at the antecedent. To assess this, we ran a finite impulse reponse (FIR) model to estimate activity at each second up to 5 seconds before the onset of the antecedent. We extracted this data from a 5mm sphere placed on the peak activity for PE found in a study by Jahn et al. (2016). To validate our paradigm, we also examined gender mismatch effects; conditions similar to the example sentence above, but with “John” replaced by “Sarah." Results were voxel-wise corrected at p<0.001 and cluster corrected with an updated version of AFNI’s 3dClustSim (cf. Eklund et al., 2016). Results: We found significant mismatch effects in the dorsal anterior cingulate cortex (dACC) and left inferior frontal gyrus (LIFG), corroborating previous studies of gender mismatch. Our ROI analysis based on the Jahn et al. 2016 study showed a significant linear trend of increasing activity in the 5 seconds before the onset of the antecedent. Conclusion: This study is one of the first to apply computational modeling from the decision-making literature to fMRI studies of language. These results suggest that the mPFC is associated with increased prediction over time, and that this is a domain-generalizable function. This represents a step towards combining modeling across different fields, and may yield novel hypotheses about which brain regions may contribute to reading disability.

Topic Area: Control, Selection, and Executive Processes

Back to Poster Schedule