Which reading method offers the best results for EEG/ERP research?

Reading after Brain Injury

Case Study: How to choose between two research methods?

Portland State University

Roles: Lead Researcher

Skills: A/B Testing, Multivariate Testing, Linear Mixed Effects Modeling, EEG/ERP signal processing, SPSS and R, User flow analysis, Written communication skills, Science communication skills to non-scientific audiences, Teaching and mentoring, Leadership

Project Summary: This discovery project sought to identify and define reading problems for clients after traumatic brain injury. By solving one key challenge, we encountered new design problems. The key 'big picture' problem researching reading after brain injury is that typical behavioral methods don't work. To solve this, we need advanced analytical tools like EEG to look at brainwaves during reading. The problem with EEG/ERP methods is that it's hard to design naturalistic reading materials that can be used with EEG. So, I conducted A/B testing on two reading methods and identified key differences between the methods. This resulted in an improved product for the larger research study and scientific publications.

Design Problem: Advanced analytical tools require changes to the design of a research product that might affect assessment of reading after brain injury

Stakeholders: Research participants, peer-reviewers, brain injury researchers, co-authors

Research Process and Approaches:

This reading project was the core research of the Neurolinguistics Lab. The larger research project was already underway, so comparing the two methodologies had to happen quickly so that data from hard-to-recruit brain injury survivors wasn't lost. We tried a number of approaches to compare the two methodologies, and I ended up consulting with a statistician, Dr. Bodner, in the psychology department to learn how to implement a linear mixed effects model with the data we had.

It turns out that the word-by-word method did change reading, but only for certain sentence types. And, the effects changed based on the memory abilities and vocabulary of the participant. This difference is actually really useful when you want to challenge readers or look at things like predictive text processing, but not so useful when readers are already struggling.

Findings

The primary result of this research method comparison was an improved design for future research products.

We also published a scientific paper about our methods comparison.