McNamara, D. (2004). SERT: Self-explanation reading training. Discourse Processes, 38(1), 1–30

RCT with treatment students receiving explicit training in Self-explanation skills associated with more skillful readers. This training helps readers with lower prior domain knowledge as measured by text-based comprehension questions, but doesn’t help higher-domain-knowledge readers, and doesn’t help with questions which require bridging inference (~near Transfer learning).

The main finding here suggests that facilitating self-explanation is only useful for text-level questions, and only for low-knowledge readers. If that’s true, then this seems much less exciting as a direction. That said, looking at the graph, it’s not clear to me that there isn’t an effect on bridging questions for high-knowledge readers; maybe the sample size is just too small (N=42). Searching a bit, it seems that iSTART (2006) finds benefits even for high-knowledge readers.

Q. Main finding?
A. Training helps readers with lower prior domain knowledge as measured by text-based comprehension questions, but doesn’t help higher-domain-knowledge readers, and doesn’t help with questions which require bridging inference

Q. Six strategies?
A. Comprehension monitoring, paraphrase, bridging inference, elaboration, using logic, prediction

Q. Which strategies were included in the “self-explanation score”?
A. elaboration, logic, prediction, comprehension monitoring (i.e. going beyond the text)

Q. What’s meant by “using logic” as a self-explanation strategy?
A. Using domain-general knowledge and common sense to go beyond the sentence; meant to distinguish from elaboration and distant bridging inferences, which are expected to rely on domain-specific knowledge.

Q. Paper’s explanation of how, mechanically, training helped low-domain-knowledge readers?
A. Encouraged them to use domain-general techniques (“logic”) to make more inferences. These techniques are less relevant to higher-knowledge readers.

Q. Papers explanation of why training didn’t help readers with high domain knowledge?
A. The low-cohesion text naturally causes them to make inferences, and they’re able to do it because of their prior knowledge.

Q. Target text?
A. A ~600 word text on cellular mitosis, intentionally modified to have low cohesion.

Q. How does the experimental design avoid simply measuring the effect of cueing self-explanation, like Chi, M. T. H., De Leeuw, N., Chiu, M.-H., & Lavancher, C. (1994). Eliciting self-explanations improves understanding. Cognitive Science, 18(3), 439–477?
A. The control group gets an introduction to self-explanation right before reading the target text, but doesn’t get instruction on strategies. So the results are presumed to measure the effect of teaching specific strategies.

Q. Correlations between reading span / skill and comprehension performance?
A. Null.

Q. How did use of bridging inferences as a strategy relate to comprehension performance?
A. Only distant bridging improved performance, not “near” bridging.

Q. Which strategies were more clearly associated with stronger comprehension scores?
A. elaboration, using logic, and (via post-hoc) distant bridging inferences (“going beyond the text by making connections to prior knowledge or distant ideas is essential to deep comprehension”)