MJMatthew R. Johnson

Matthew R. Johnson

*This is an algorithmically generated author record based on publicly available information.

Current Organizations / Institutes

The organizations/institutes from the scholar's latest article

Previous Organizations / Institutes

The organizations / institutes from the scholar's other articles in the past ten years


Matthew R. Johnson has a total of 12 co-authors. Here are the top co-authors by number of shared publications:


Latest Publications

  • Open Access

    Convolutional neural networks can decode eye movement data: A black box approach to predicting task from eye movements

    by Zachary J. Cole, Karl M. Kuntzelman, Michael D. Dodd, Matthew R. Johnson

    Previous attempts to classify task from eye movement data have relied on model architectures designed to emulate theoretically defined cognitive processes and/or data that have been processed into aggregate (e.g., fixations, saccades) or statistical (e.g., fixation density) features. Black box convolutional neural networks (CNNs) are capable of identifying relevant features in raw and minimally processed data and images, but difficulty interpreting these model architectures has contributed to challenges in generalizing lab-trained CNNs to applied contexts. In the current study, a CNN classifier was used to classify task from two eye movement datasets (Exploratory and Confirmatory) in which participants searched, memorized, or rated indoor and outdoor scene images. The Exploratory dataset was used to tune the hyperparameters of the model, and the resulting model architecture was retrained, validated, and tested on the Confirmatory dataset. The data were formatted into timelines (i.e., x-coordinate, y-coordinate, pupil size) and minimally processed images. To further understand the informational value of each component of the eye movement data, the timeline and image datasets were broken down into subsets with one or more components systematically removed. Classification of the timeline data consistently outperformed the image data. The Memorize condition was most often confused with Search and Rate. Pupil size was the least uniquely informative component when compared with the x- and y-coordinates. The general pattern of results for the Exploratory dataset was replicated in the Confirmatory dataset. Overall, the present study provides a practical and reliable black box solution to classifying task from eye movement data.


  • Open Access

    Reward impacts visual statistical learning

    by Su Hyoun Park, Leeland L. Rogers, Matthew R. Johnson, Timothy J. Vickery

    Humans automatically detect and remember regularities in the visual environment—a type of learning termed visual statistical learning (VSL). Many aspects of learning from reward resemble statistical learning in respects, yet whether and how reward learning impacts VSL is largely unexamined. In two studies, we found that reward contingencies affect VSL, with high-value associated with stronger behavioral and neural signatures of such learning than low-value images. In Experiment 1, participants learned values (high or low) of images through a trial-and-error risky choice task. Unbeknownst to them, images were paired as four types—High-High, High-Low, Low-High, and Low-Low. In subsequent recognition and reward memory tests, participants chose the more familiar of two pairs (a target and a foil) and recalled the value of images. We found better recognition when the first images of pairs have high-values, with High-High pairs showing the highest recognition rate. In Experiment 2, we provided evidence that brain responses were affected by both value and statistical contingencies. When we compared responses between the high-value first image and the low-value first image, greater activation in regions that included inferior frontal gyrus, anterior cingulate gyrus, hippocampus, among other regions were found. These findings were driven by the interaction between statistically structured information and reward—the same value contrast yielded no regions for second-image contrasts and for singletons. Our results suggested that the powerful allocation of attention in response to the high-value first image potentially enables better memory for statistically learned pairs and reward information than low-value first image.

  • Open Access

    Deep learning fMRI classification of temporal codes during naturalistic movie viewing and memory recall

    by Matthew R Johnson, Thomas P O’connell, Marvin M Chun, Marcia K Johnson

    In this fMRI study, participants watched the same short video several times in separate scan runs. We used a novel cognitive task design that afforded direct, time-locked comparisons between perception and imagery-based memory recall for the same information, in contrast to other studies of audiovisual episodic memory recall in which the memory probe took place after the movie, and at a different pace. In some runs of our paradigm, participants saw and heard the full movie (audio + video, A+V); in other runs, they saw only the video and were instructed to imagine the audio from memory (0+V); and in still other runs, they heard only the audio and were instructed to imagine the video from memory (A+0). Using the DeLINEATE toolbox (http://delineate.it), we trained subject-specific deep-learning models based on a sensory cortex (visual + auditory) region of interest to discriminate whether two fMRI volumes from different runs represented the same point in time, or two different points in time. The model performed easily above chance at this task. Although classification was, as expected, best when comparing one full-movie (A+V) run to another, and video-only runs (0+V) tended to classify better than audio-only (A+0) runs, performance was high between all run-type pairs. Critically, this included comparisons between 0+V and A+0 runs, which shared no common sensory information. In fact, classification was higher on data drawn from one 0+V and one A+0 run than on data drawn from two different A+0 runs. We believe this technique provides a powerful new way to assess reinstatement of brain activity patterns in sensory regions during recall of complex, naturalistic information.

  • Monitoring Changes in Gait Adaptation to Identify Construction Workers’ Risk Preparedness after Multiple Exposures to a Hazard

    by Cenfei Sun, Changbum Ryan Ahn, Junseo Bae, Matthew Johnson

    Many construction accidents result from workers’ unsafe behaviors, and evidence demonstrates that individuals who consistently engage in risky behaviors often have a high rate of accidents. These risk-prone workers’ biased evaluation of hazards (risk perception) and high tendency to take risks (risk propensity) leave them insufficiently prepared when encountering hazards that may cause accidents. Therefore, measuring risk perception and propensity and using these measurements to identify risk-prone workers can provide opportunities to effectively interrupt unsafe behaviors. Our previous study indicated that subjects tend to change their gait patterns before encountering a hazard, a result that may occur due to the subject’s perception about the risks of the hazard. This study examines subjects’ gait adaptations when the subject encounters the same hazards repeatedly to determine whether gait adaptations depict changes in the subjects’ perceived risk from the hazards over time. The results from an experiment confirmed the hypothesis that the subject’s gait adaptation upon encountering a hazard decreases as he/she repeatedly encounters the same hazard. This result is in accordance with the theory of “outcome feedback” and represents an opportunity to use gait adaptation as a quantifiable behavioral response indicator for identifying risk-prone workers.