Measuring Neural Correlates of Attention in Learning Activities using Mobile EEG

• In Event: Measuring Attention and Engagement in Real-world Settings

Abstract

Attention skills are crucial for children’s behavioral regulation and academic success in schools (Duncan et al., 2007). Traditional measures of children’s attention abilities, such as teacher/parent report and observational coding, can be biased and blind to unobservable cognitive phenomena. With the development of mobile EEG technology, researchers have begun to investigate neural underpinnings of attention in activities mimicking real-life experience, a key piece to understanding how these skills influence learning opportunities at the moment and, conversely, how they can be supported by making changes within the child’s environment. However, concerns regarding design rigidity and data quality have risen for real-world EEG studies given their limited control over environmental stimuli. Thus, the current presentation explored the feasibility and trade-offs of using mobile EEG to measure neural correlates of student attention in different learning activities, by gathering data from two studies using the same semi-naturalist paradigm in group and individual settings.

In both studies, kindergartener to 6th-grade students (Group: N = 10, Mage= 7.36 yrs; Individual: N = 41, Mage = 8.08 yrs) participated in a 60-min lesson implemented by a trained researcher during which continuous EEG recordings were conducted. In the group study, two or three students in each session were outfitted with EEG caps and joined a group of students in a classroom to participate in the lesson, while students in the individual study join the lesson alone in a quiet conference room. Learning activities for both studies included a baseline activity (e.g., mindfulness session), teacher-led instruction (e.g., lecture and zoom class), and student-led activities (e.g., seated work). To access data quality, we compared the data loss, non-brain artifacts identified, and EEG epoch deletion from both studies. Additionally, alpha frequency bands (7–11.5 Hz) were used as the indicator of student attention and were calculated using the mean power value for each spectrum range.

After EEG preprocessing, 55% of EEG epochs were retained for all but one child in the group study, while 6 were deleted from the individual study. The average number of deleted electrodes was 1.33 (SD = 1.41). About 0.22% (SD = 0.39%) of continuous EEG data was marked as containing nonbrain artifacts and was deleted from the power spectra analysis. On average, 18.21% (SD = 18.78) of epochs were excluded. Preliminary results of the individual study showed that an average of 2 (SD = 1.18) electrodes were deleted and about 25.42% (SD = 15.00%) of epochs were excluded. The percentage of removed epochs did not differ for activities in the group study, χ 2(2) = 0.71, p = .70, however, more epochs were removed during in-person and zoom lectures than during video watching, F (3, 42) = 8.35, p < .001.

These findings demonstrate the feasibility of collecting quality data from young students using mobile EEG technology while engaging in naturalistic learning activities. Overall, findings also highlight necessary trade-offs between ecological validity and data quality for researchers to consider when designing real-world EEG studies.

Authors

• Keye Xu, University of California - Los Angeles

• Fang Yu Chang, University of California, Los Angeles

• Agatha Lenartowicz, University of California, Los Angeles

• Jennie K Grammer, University of Virginia - Charlottesville
  Presenting Author