- Should only include sentences that prompt inference and discourse-level processes in think aloud techniques
- Study the contributions of other factors, including verbal fluency, vocabulary, memory, world knowledge, metacognition
- Could ask participants questions following each sentence rather than have them speak or write down their thoughts 

University of Michigan - Ann Arbor

- Prediction of reading comprehension
- Further research

Measuring discourse-level processes with verbal protocols and latent semantic analysis - discussion

Millis, K., Magliano, J., & Todaro, S. (2006). Measuring discourse-level processes with verbal protocols and latent semantic analysis. Scientific Studies of Reading, 10(3), 225-240. https://www.tandfonline.com/doi/abs/10.1207/s1532799xssr1003_2?casa_token=t1DCNTJm4G0AAAAA:4Zmsq1hksZB2rZsEYOs0VuIGfn16rDbUviE39Om4DyMxLz0HWOC7RzN9yyymRidOj9nS1VQqXTtWZA

Measuring discourse-level processes with verbal protocols and latent semantic analysis

- R^2 between LSA cosines and reading comprehension results ranged from 0.17 to 0.22 much lower than the R^2 values of 0.38 to 0.45 obtained from a similar study done by Magliano and Millis (2003) on fairy tales 
- Low R^2 value could be attributed to presence of prior sentences, fatigue from responding to every sentence, unfamiliarity with the content presented, separately studying local from distal causes
- Distal causes predicted short answer tests – focus on information retrieval and long-term memory 
- Local causes predicted Nelson-Denny test – not as much emphasis on inference and the use of many sentences
- Cosines affected by irrelevant words spoken/written (i.e. I think, I didn’t know…)

Learn Before

Related