Professional Documents
Culture Documents
The Self-Explanation Effect
The Self-Explanation Effect
The Self-Explanation Effect
oneself. It is a self-generated and self-directed constructive activity which requires analysis and
reflection of the underlying principles of one’s action (Roy & Chi, 2005). The self-explanation
effect states that learning is improved when students generate self-explanations about how and
why events or phenomena happen (Chi, Bassok, Lewis, Reimann & Glaser,1989; Segler, 2002).
According to US Next Generation Science Standards (as cited in Villalta-Cerdas & Sandi-
Urena, 2014), the generation of explanations is one of eight practices of science essential for all
Why is the ability to generate self-explanation important? There have been exciting
discoveries regarding why students learn better when they explain to themselves the material
they are learning. First, self-explanation requires students to elaborate on the to-be-learned
information by relating it to prior knowledge and construct new knowledge beyond the learning
materials (Fonseca & Chi, 2011). Simply put, learning is about integrating new information into
addition, the process of generating self-explanation encourages students to engage with the
learning material in a meaningful way (Roy & Chi, 2005; VanLehn, Jones & Chi, 1992). By
actively searching for explanations of the learning materials, students are more likely to go
deeper into the learning process and develop a better understanding of the material they study.
Lastly, the act of self-explaining requires students to be aware of the learning and comprehension
process and thus, influences metacognition (Roy & Chi, 2005; McNamara & Magliano, 2009).
THE SELF-EXPLANATION EFFECT 3
Overall, the findings of these studies suggest that implementing self-explanation in instruction
leads to enhanced learning, more accurate self-assessments, and more effective problem-solving.
What are the theoretical explanations for the self-explanation effect? Roy and Chi (2005)
explain that self-explanation is involved with several cognitive processes which help students
learn better including generating inferences, integrating information, and monitoring and
repairing the learner’s flawed mental model. Similarly, VanLehn and Jones (1993) also view
gap-filling, schema acquisition and enhanced analogical problem solving as the major
contributing factors of the self-explanation effect. Then, Siegler (2002) suggests self-explanation
(2012) claims that self-explaining guides students to look for repeated and unifying patterns, thus
discovering broader generalizations. Across studies, at least the gap-filling and generation of
inferences appear to account for the self-explanation effect. The rationale for gap-filling and
generation of inferences is that self-explanation causes students to discover their gaps in their
knowledge and do something about it; in the process of producing self-explanations, students are
inferring information missing from a text passage or an example’s solution using skills such as
The self-explanation effect is tested in a wide variety of domains from physics problem-
solving (Chi et al.,1989) to geometry (Aleven & Koedinger, 2002) and reading comprehension
(McNamara& Magliano, 2009). Across studies, the general findings demonstrate the benefits of
question how many classroom teachers are aware of the benefits of self-explanation as a learning
strategy and implement it to everyday instruction. Therefore, the purpose of this paper is not only
THE SELF-EXPLANATION EFFECT 4
to identify the advantages of self-explanation, but also to promote the use of self-explaining as an
experimentally supported instructional method that can help teachers create more student-
a wide range of educational settings, I reviewed two studies. These studies investigate the effects
of self-explanation in different domains, contexts and learners. I chose these two articles because
they are idiosyncratic in terms of perspectives, research methods and designs. The first study by
Wylie, Koedinger and Mitamura (2010) tested the self-explanation effect in second language
grammar acquisition in a computer-based learning environment. Most of the studies on the self-
explanation effect are conducted on the math and science domains like geometry (Aleven &
Koedinger, 2002), biology (Chi, De Leeuw, Chiu & LaVancher, 1994), and physics (Chi et al.,
1989). Therefore, this study is unique in evaluating the self-explanation effect within the context
of learning the English article system. The results of this study are informative for me as an
English as a Second Language (ESL) instructor, who wants to help students improve their
grammatical skills.
In the second study by Villalta-Cerdas and Sandi-Urena (2014), the researchers continue
to examine the self-explanation effect but in the context of real college chemistry learning
environments. This study is unique in addressing whether different self-explaining tasks impact
the student self-explaining behavior and assessing how different levels of engagement in self-
explaining influence learning chemistry concepts. Especially, the researchers point out the
disconnect between the theoretical research and the applied research, emphasizing that the
a research laboratory setting. This study suggests that self-explanation can be effectively
instructional strategy for second language grammar learning, specifically, for learning the
English article system. Self-explanation has shown great promise in increasing learning by
encouraging students to focus on key features of the material and has been found to be an
effective domain general strategy for a variety of domains (Roy & Chi, 2005). Yet, most studies
have been done in math and science domains like physics, geometry, and biology and relatively
little work has been done outside of math and science domains. This study moves to the domain
of second language grammar learning and addresses whether the self-explanation effect extends
The researchers chose the English article system as a domain because there are
fundamental differences between math and science learning and second language grammar
acquisition that may affect the effectiveness of self-explanation strategy in the language learning
domain. In addition, the article system is suitable for studying the effects of self-explanation in
second language learning due to its rule-based nature that requires underlying rules to determine
when each article should be used. To examine the effects of prompted self-explanation within the
context of second language grammar learning, the study poses the following hypothesis: students
THE SELF-EXPLANATION EFFECT 6
in the explanation choice condition will show greater learning gains on article choice and
Method
The study took place in the University of Pittsburgh’s English Language Institute during
one 50-minute class period. The sample consisted of 118 adult English language learners (mean
age=27.9, SD=7.2) who represented 13 first languages. There were three class levels:
intermediate (n= 30), intermediate-advanced (n= 61) and advanced (n= 27).
The experimental session started with all participants completing both the article choice
and explanation choice pretests. Next, participants were randomly assigned to two computer-
based tutoring conditions, either the article choice or explanation choice condition. In the article
choice condition, students selected the article (a, an, the, no article) that best completed the
sentence, but were not asked to give an explanation for their choice. Students received immediate
feedback on their answers and had access to hints. In the explanation choice condition, students
were given a sentence with the correct article highlighted and asked to choose the rule of the
sentence that best explained the article use, but do not make any article decisions of their own.
Likewise, students received immediate feedback on their answers and had access to hints.
Finally, all students were assessed on both article choice and explanation choice items as
posttests. The article selection items were the same form as those in the article choice condition
and the explanation selection items were also the identical form of those in the explanation
choice condition.
Since data collection was limited to one 50-minute class period, 86% (101 out of 117) of
the students completed all tasks; however, attrition between conditions was not the same as 95%
(55 out of 58) of students in the article choice condition and 77% (46 out of 60) of students in the
THE SELF-EXPLANATION EFFECT 7
explanation choice condition completed all tasks. Although there was no significant difference in
the pretest scores between conditions after attrition, the researchers used Propensity score
matching (PSM) to adjust for possible differences due to attrition and the results were no
different with or without the PSM variable. The rationale for grouping students using propensity
score is that the propensity score is highly correlated with whether or not a student would
Results
A repeated measures ANOVA indicates that students in both article choice and
explanation choice conditions demonstrate significant pre to posttest improvement on both the
article choice items (F (1, 95) = 29.44, p <0.001) and the explanation choice items (F (1, 91)
=15.09, p < 0.001). In addition, the univariate between-subjects tests show that the conditions
more significantly affected explanation choice posttest scores (p=0.023) than article choice
posttest scores (p=0.153). Cross-task transfer occurs for both conditions: students in the article
choice tutoring group show improvement on the explanation choice posttest, and students in the
explanation choice tutoring group show improvement on the article choice posttest although both
groups had no tutored practice with these tasks. However, the effects of the self-explanation
tutoring condition appear stronger than the article choice tutoring condition.
Figure 4: Normalized gains scores for article choice and explanation choice measures
THE SELF-EXPLANATION EFFECT 8
As normalized gain scores illustrate above, students in the explanation choice condition are
showing equal gains on the article choice posttest compared to students in the article choice
In regard to the processing time, a repeated measures ANOVA indicates that students in
both conditions are significantly faster at completing the posttest than the pretest for both article
choice test (F (1,95) = 76.28, p < 0.001) and explanation choice test (F (1, 91) = 275. 32, p <
0.001). More importantly, a MANCOVA analysis with log-transformed completion times for
article choice and explanation choice posttests as dependent variables shows no main effect for
condition (F (2, 88) = 0.37, p= 0.689). Overall, there is no significant difference between
students in the self-explanation and the article choice conditions in completing the posttests.
Discussion
The findings of this study suggest that self-explanation can be an effective instructional
strategy not only for math and science domains, but also for the domain like second language
grammar learning. The self-explanation prompts in the explanation choice condition help
students correctly use articles as well as explicitly identify the underlying rules for their
THE SELF-EXPLANATION EFFECT 9
decisions. In addition, self-explanation does not increase learning time as students in the
explanation choice condition solve the posttest problems as fast as those in the article choice
Especially, students in the explanation choice condition have stronger learning effects
that they show equal gains on the article choice items and substantially higher gains on the
explanation choice items compared to those in the article choice condition. The reasonable
explanation for this result might be that students see the correct article highlighted in the problem
sentence and then combine it with self-explanation prompts, and this process more likely to
encourage students to focus on the features of the sentence that are important for making article
decisions.
In sum, the present findings support the assumption of Roy and Chi (2005) who proposed
that self-explanation prompts encourage students to notice relevant features of the problems and
enable students to become aware of gaps in their own knowledge. This study is one of the first to
investigate the effects of self-explanation on second language learning and suggests positive
results that prompted self-explanation may be beneficial to second language learning students by
Critique
Two weaknesses of this study should be noted. As pointed out by the researchers, there
are limitations related to the generalizability of the results because less than 80% of the students
in the explanation choice group finished all five stages of the study due to the limited data
collection time. The study admitted that the researchers couldn’t make strong claims about those
students who couldn’t finish the posttest and vaguely stated that “perhaps self-explanation is not
appropriate for all students”, which could diminish the credibility of the findings of this study.
THE SELF-EXPLANATION EFFECT 10
Another shortcoming of this study concerns the students’ prior knowledge. Students in all
conditions had received at least some prior instruction on article use. Thus, it is not clear whether
or not the results come from the effects of self-explanation alone. In order to further investigate
these issues, future work would be needed for novice learners with no or little prior knowledge
of article rules.
Despite a few weaknesses, the strength of this study lies in its thorough discussion of the
method and results in a very concise and succinct way. The researchers analyze data with and
without the propensity score matching (PSM) variable to address the possible bias caused by
attrition and report the results that include not only ANOVA and MANCOVA, but also PSM for
a less biased and more accurate description of the findings. In addition, the findings were
reported in four areas that encompass learning gains, processing time, hint usage, and effects of
first language, which are important information for language instructors in determining whether
or not they implement the findings into classroom instructions. As for the contributions, the
additional support that self-explanation is beneficial not only for math and science domains but
also for second language grammar acquisition and can foster both procedural and declarative
knowledge. Further, the findings from this study contribute to the field of second language
learning in terms of the superiority of explicit versus implicit grammar instruction. This study
supports Norris and Ortega’s (2000) view that explicit practice of rules, like the explanation
choice condition, is a more effective instructional strategy than an implicit approach to learning,
This study approaches the self-explanation effect from a different perspective than the
previous article. In this study, the researchers examine whether different self-explaining tasks
modify student self-explaining behavior. Their assumption is that students are more likely to
appropriate instruction in the naturalistic classroom setting can effectively modify students’ self-
explanation behaviors. The researchers identify the need for the applied research of the self-
addition, despite the substantial research literature demonstrating the positive links between self-
explaining and conceptual learning, there are only two articles which have explored the self-
This study of the self-explaining effect is different from other work in several aspects.
First of all, participants in the study are in their normal learning environments as opposed to in
research experimental settings. Data collection also takes place within a naturalistic college level
different levels by modifying the prompts describing the task instead of spontaneous self-
chemistry rather than learning procedural knowledge such as using worked-out examples. In
order to investigate an association between the self-explaining tasks and students’ self-explaining
behavior, the study specifically addresses the following research question: Do tasks that require
Method
THE SELF-EXPLANATION EFFECT 12
The study was conducted in a large, urban, public university in the U.S. and gathered data
from 128 students enrolled in General Chemistry 2. Majors in General Chemistry 2 comprised of
61% pre-professional (pre-Medicine, pre-Pharmacy and Health Sciences), 23% other sciences
(Physics, Biology, Geology, etc.) or Math, 8% Engineering, and 6% Chemistry. The tasks in this
study were embedded within the normal requirements of the course to reflect the complexities of
a live learning environment; therefore, students considered this as part of their normal
assignments.
The learning event consisted of a textbook passage with a general description of the
Second law of the Thermodynamics. The researchers created four self-explanation tasks (SE-
defining one of the study conditions described in Table 1. The calibration of the conditions is
based on literature reports (Fonseca& Chi, 2010), especially Siegler and Lin’s (2009) multi-
The students were randomly assigned to one of the four SE-Tasks conditions by random
number generation. The number of students in each condition was: SEA (n=29), EADA (n=31),
SEO (n=35), SEIA (n=33). The students read the passage presented to them and wrote a written
response to the given self-explanation task. Think-aloud protocols were not an option for this
study due to the large experimental group and researchers’ desire to gather data in the most
naturalistic environment possible. Materials were printed, administered, and timed in the same
way other in-class assignments were typically given. Written explanations were collected,
The study used two analytical methods for textual analysis of data: a coding scheme and
Latent Profile Analysis (LPA). The final coding scheme for textual analysis and a brief
description of each code type is shown in Table 3. LPA is a model-based statistical technique to
find qualitative differences among participants based on observed variables. In this study, the
researchers analyzed the code-ratios (number of code type divided by total codes in response) in
student’s response as observed variables. LPA helps this study minimize researcher bias in the
Results
The Chi-square test showed that there is a statistically significant association between the
code type and the SE-Task at a 95% confidence level, 2 (9, N = 357) = 22.50, p < 0.05. The
results in Table 4 show evidence for the differences in total number of code types per SE-Task.
The highest percentage of occurrences of BI (bridging inference) came from the SEA
concepts. For DI (deductive inference), the main source was the SEIA (self-explaining-incorrect-
answer) task. The combined count of codes BI and DI was 169 (44% of the total count), which
THE SELF-EXPLANATION EFFECT 15
suggested that the generation of inferences was a considerable component of the responses. In
the case of the code E (elaboration), the highest percentage came from the EADA (self-explain-
agreement/disagreement) task where students brought in information that was not provided in the
materials. Lastly, the code P (paraphrasing) showed the similar high percentage for EADA (self-
associated with less sophisticated explanatory behaviors as they reflect recounting of information
rather than generation of inferences. The code type distribution supported an association between
the code types in the student responses and the SE-Task prompts assigned to them. This
association suggests that the self-explaining prompts the researchers designed for the study
Latent Profile Analysis (LPA) showed five patterns of the student behaviors in terms of
the self-explanation profiles. The Chi-square test showed a significant association between the
SE-Profile and SE-Task at a 95% confidence level, 2 (12, N = 128) = 22.75, p < 0.05. Table 6
The SEA task (self-explain-own-answer) had the highest percentage of students in the SE-
Profiles associated with the more analytic self-explaining behaviors (i.e., bridging inferential and
THE SELF-EXPLANATION EFFECT 16
mixed-behavior). In other words, more students in SEA task engaged in generating inferences
and connecting ideas. On the other hand, SEIA (self-explain-incorrect-answer) and EADA
compared to SEA. One plausible explanation for this is that students in SEA task focus on
repairing what they do not understand without focusing on others (Chi, 2009).
Discussion
Based on the results, the researchers concluded that students’ self-explanatory behaviors
are associated with the tasks of different self-explaining demand. In other words, the way the
tasks are framed for students can positively influence students’ engagement in producing more
self-explaining practices through the thoughtful instructional design in the naturalistic classroom
setting. The findings also showed that some students defaulted to paraphrasing regardless of the
SE-Tasks. This reveals that students may resist explaining when prompted, in addition to not
spontaneously engaging in self-explaining. All in all, the study put emphasis on the fact that a
considerable number of students engaged in self-explaining upon prompting and that the
It is not surprising that the same instructional activity may trigger varied self-explaining
responses in different students. Individual differences in interacting with the prompts might be
because students have different prior knowledge, attitudes, learning strategies, and personal
other instructional devices. For this reason, the researchers suggested that a variety of prompts
may be more effective than a prescriptive, one-size-fits-all type of prompt. Although the
resistance or the lack of ability to construct self-explanation may not be resolved quickly, the
results from this study that the qualities of student self-explanation could be tweaked through the
Critique
The weakness of this study is the analysis of the textual data of students’ written
responses. Although the researchers removed prompts from the responses for the textual
analysis, coders might have inferred the self-explaining task from the structure of the responses.
As a result, coders may be prone to categorize student’s self-explaining behaviors based on the
condition and not on the analysis of the responses. This might lead to a potential coder bias. One
thing I wish the researchers investigated further is that how different kinds of self-explaining
A key strength of this study lies in its attempt to minimizing “observer’s paradox”, which
is the notion that participants tend to adjust their behavior some other way when they are under
the impression of being observed. Unlike other tightly controlled experimental studies, the
participants in this study had no cues that this learning event was part of educational research as
the procedures were consistent with their normal in-class assignments. Therefore, evidence
gathered from this study would reflect more real-world situation. By conducting research in the
context of real college science learning environments, this study contributed to improving the
college instructors about the positive links between self-explaining and learning, thus influencing
THE SELF-EXPLANATION EFFECT 18
Connections
The findings from the two studies support the assumption that self-explanation is a
constructive cognitive activity that helps students engage in active learning and knowledge-
material in a more meaningful way by integrating concepts and generating inferences while
effectively monitoring their comprehension. Results observed in these studies have implications
for teachers and students. The study by Wylie et al. (2010) demonstrateed that self-explanation is
beneficial for second language grammar acquisition and increases students’ procedural
knowledge (article choice) and declarative knowledge (rationale behind the article choice). This
suggests a clear reason for ESL teachers to integrate self-explanation prompts into their grammar
The second study by Villalta-Cerdas and Sandi-Urena (2014) similarly suggests that the
ability to generate self-explanation become an essential learning outcome for students because it
helps students focus on connections between parts and cause-effect relationship. Therefore,
chemistry instructors should take advantage of the benefits of self-explaining and integrate this
evidence-based learning strategy into their instructional design to improve science education.
Results of this study also have implications for students. Self-explaining is a learnable strategy
and may become habitual upon practice through effective implementation. Students should take
strategy.
THE SELF-EXPLANATION EFFECT 19
The two studies proposed some interesting possibilities for future research. As for the
first study by Wylie et al. (2010), future work is needed to investigate how being able to explain
article rules could enhance students’ article use in natural speech. As the ultimate goal of
learning article rules is a spontaneous article production in speaking without reflection on rules,
the explicit declarative knowledge of the rules might not be required for successful
communication. Hence, it may be a good opportunity for a future study to investigate further
how attained declarative knowledge may or may not improve student’s communicative
competence in terms of accurate article use. The second study by Villalta-Cerdas and Sandi-
Urena (2014) suggested future investigation of what individual characteristics may be associated
with differential self-explaining behavior. Examining what triggers the individual differences in
self-explaining responses to the same self-explaining prompts would be an important next step.
elicit self-explaining can generate many learning outcomes including problem-solving skills,
conceptual understanding and monitoring the comprehension process (Siegler & Lin, 2009).
However, as mentioned in the introduction section, self-explaining has not gained sufficient
recognition by many educators due to the lack of awareness of its potential to promote learning.
disconnect between what researchers have figured out and the practice of everyday instruction. It
is difficult to modify an instructional strategy. This may be in part because students become
accustomed to learning environments where they are not expected to actively engage with the
learning materials (this is the case for most Asian students). But for the most part, I believe it is
probably because instructors tend to teach the way they were taught and are more comfortable
with the use of traditional teacher-centered and lecture-based methods. It might take some time
THE SELF-EXPLANATION EFFECT 20
and effort to persuade students and teachers to act differently. I hope this paper is convincing and
able to contribute in promoting the use of self-explaining as a learning and instructional strategy
References
doing and explaining with a computer‐based cognitive tutor. Cognitive science, 26(2),
147-179.
Chi, M. T., Bassok, M., Lewis, M. W., Reimann, P., & Glaser, R. (1989). Self-explanations:
How students study and use examples in learning to solve problems. Cognitive
Chi, M. T., De Leeuw, N., Chiu, M. H., & LaVancher, C. (1994). Eliciting self-explanations
Denancé, V., & Somat, A. (2015). Learning by explaining: Impacts of explanations on the
Deutsch, D. (2011). The beginning of infinity: Explanations that transform the world. Penguin
UK.
Lombrozo, T. (2012). Explanation and abductive inference. Oxford handbook of thinking and
reasoning, 260-276.
McNamara, D. S., & Magliano, J. P. (2009). Self-explanation and metacognition: The dynamics
Roy, M., & Chi, M. T. (2005). The self-explanation principle in multimedia learning. The
VanLehn, K., Jones, R. M., & Chi, M. T. (1992). A model of the self-explanation effect. The
VanLehn, K., & Jones, R. M. (1993). What mediates the self-explanation effect? Knowledge
Villalta-Cerdas, A., & Sandi-Urena, S. (2013). Self-explaining and its use in college chemistry
Wylie, R., Koedinger, K., & Mitamura, T. (2010). Extending the self-explanation effect to
Sciences.