Elsevier

Computers & Education

Volume 57, Issue 2, September 2011, Pages 1775-1779
Computers & Education

Effects of identical example–problem and problem–example pairs on learning

https://doi.org/10.1016/j.compedu.2011.03.019Get rights and content

Abstract

Examples are often an integral part of online learning environments directed at the acquisition of problem-solving skills. An unresolved issue, however, is when examples should be provided to learners. Prior research has suggested that example–problem pairs are more effective than problem–example pairs for novice learners. However, in those studies, the problem–example pairs condition may have been hindered by the fact that the examples and problems were not identical within and across pairs. The present experiment therefore employed a between-subjects design with two conditions to compare the effects of learning to solve one particular problem from studying/practicing either in an example–problem–example–problem sequence (EP condition; n = 16) or in a problem–example–problem–example sequence (PE condition; n = 16). Results show that participants in the EP condition outperformed their counterparts in the PE condition during the learning phase, but that this difference had disappeared on the test tasks after participants in the PE group had also studied the example a second time.

Highlights

► Research found example–problem (EP) pairs more effective than problem–example (PE) pairs. ► However, in prior studies, examples and problems were not identical within and across pairs. ► This study shows that when they are identical, EP and PE are equally effective.

Introduction

Worked examples or modeling examples are often an integral part of online learning environments directed at the acquisition of problem-solving skills in domains such as math or science (e.g., Biesinger and Crippen, 2010, McLaren et al., 2008), since a large body of research has demonstrated that providing learners with worked examples or modeling examples in instruction has beneficial effects on learning (Atkinson et al., 2000, Sweller et al., 1998, Van Gog and Rummel, 2010). Worked examples provide students with a written account of how to solve a problem; next to a description of the “givens” and the goal, the solution steps that are to be taken to reach the goal have been worked-out (see e.g., Sweller & Cooper, 1985). In modeling examples on the other hand, a problem-solving procedure is demonstrated to a learner by a (often more competent) model (see Van Gog & Rummel, 2010). Modeling examples are increasingly used in online learning environments, usually taking the form of a digital video of the model him/herself or of the model’s computer screen (i.e., a screen capture; e.g., McLaren et al., 2008, Van Gog et al., 2009). An unresolved issue, however, is when examples should be provided to learners, that is, how example study and problem-solving practice should be sequenced during a learning phase.

Most research on the effects of example study has used example–problem pairs (e.g., Carroll, 1994, Cooper and Sweller, 1987, Kalyuga et al., 2001, Mwangi and Sweller, 1998, Paas, 1992, Rourke and Sweller, 2009, Sweller and Cooper, 1985, Trafton and Reiser, 1993) in which learners first study an example and then attempt to solve an equivalent problem. This allows them to first build a mental model of how to solve a certain type of problem, and then consolidate this model by practicing with a similar but not identical problem of the same type themselves. However, problem–example pairs have also been used (Hausmann et al., 2008, Reisslein et al., 2006, Stark et al., 2000), and it has been argued that when learners first experience deficiencies in their performance during problem solving, they may be more motivated to study the example and may focus their attention particularly on the points where they made errors, thus enhancing learning (Stark, et al. 2000).

To the best of my knowledge, there are only two studies in which the effects of example–problem and problem–example pairs on novices’ learning were directly compared, and these studies have shown example–problem pairs to be more effective for novices than problem–example pairs (Reisslein et al., 2006, Van Gog et al., in press). In the study by Van Gog et al. (in press), the examples did not even seem to contribute to learning in the problem–example pairs condition; participants in this condition showed the same (low) level of test performance as participants in the problem-solving condition, who had not received any examples. Based on these studies, one might conclude that designers of online learning environments should favor example–problem pairs instead of problem–example pairs for novice learners.

However, in both of the abovementioned studies (Reisslein et al., 2006, Van Gog et al., in press), the example and problem in each pair had the same structural features, but different surface features. For example, in the study by Van Gog et al. (in press), electrical circuits troubleshooting tasks were used, in which students first had to calculate (problem) or study how to calculate (example) what current they should measure at each ammeter based on the voltage and resistance indicated in the circuit diagram; then they had to compare this calculation to given measurement values at each ammeter; and subsequently, they had to determine what the fault in the circuit was based on this comparison (problem) or study how to determine what the fault was (example). In the problem and example in the first pair, the fault was the same: the current in one of the branches was too high, which was caused by too low resistance. The surface features of the example and problem within this first pair (i.e., the values of voltage, resistance, and current) were different though. Because the fault was the same, the example could provide feedback to learners in the problem-example pairs condition about their errors during problem solving, but to recognize this, they would need to have the ability to look past the surface features (i.e., the values) to the structural features that really matter (i.e., how to determine what the fault is).

Moreover, in the Van Gog et al. (in press) study, the structural features between the two pairs of problems in the learning phase differed. While the fault in the first pair was that the current was too high in one of the branches, caused by too low resistance, the fault in the second pair was that the current in one of the branches was too low, which was caused by too high resistance. In other words, the structural features of the example and the problem within each pair were the same, but they differed between pairs. As a consequence, students in the problem–example pairs condition could not use the first example they studied directly for solving the second problem they were given, because this problem had different structural features, and so the use of the example for solving this problem would require transfer.

Thus, in those previous studies, the beneficial effects on learning of example–problem pairs compared to problem–example pairs, may have been a result of students in the problem–example pairs condition being hindered by the fact that the examples and problems were not identical within and between pairs. If that was indeed the case, then it should not matter how examples and problems are alternated when the examples and problems within and between pairs are identical, that is, focus on attempting to solve or studying how to solve the exact same problem.

Therefore, the question addressed in this study is, does it matter for learning outcomes how problem solving and example study is alternated (i.e., problem–example–problem–example vs. example–problem–example–problem) when there is only a single problem-solving procedure to be learned? This question is addressed using a computer-based problem and modeling examples consisting of a digital video (screen capture) of an expert model solving that problem.

Section snippets

Participants and design

Participants were 32 students from a Dutch University (8 male; age M = 20.22, SD = 2.24). The majority of students (ca. 90%) were Psychology students, the others were enrolled in other programs of the Faculty of Social Sciences. Participants were randomly assigned to either the example–problem (EP; n = 16) or the problem–example (PE; n = 16) condition.

Problem

The problem that participants had to learn to solve is known as Frog Leap (http://www.addictinggames.com/frogleap.html; see also Van Gog et al.,

Results

A t-test on the mean performance on the two problems in the learning phase showed that participants in the EP condition outperformed participants in the PE condition (EP: M = 7.59, SD = 4.40; PE: M = 4.00, SD = 3.07), t(30) = 2.68, p = 0.12, Cohen’s d = .95 (large effect). Analysis of the mean performance on the two test problems no longer showed a significant performance difference between the conditions (EP: M = 8.41, SD = 3.82; PE: M = 8.47, SD = 5.07), t(30) = .039, ns. Half of the

Discussion

The finding that participants in the EP condition outperformed participants in the PE condition in the learning phase is not surprising considering that participants in the PE condition had not seen the example prior to attempting to solve the problem for the first time, and had only seen the example once prior to attempting to solve the problem the second time. Indeed, after participants in the PE condition had also studied the example a second time, that is, on the test, performance did no

Acknowledgment

This work is part of a research project funded by the Netherlands Organization for Scientific Research (Veni Grant 451-08-003). The author would like to thank Jeroen Mikkers for his assistance with this study.

References (20)

There are more references available in the full text version of this article.

Cited by (45)

  • Effects of a review video and practice in video-based statistics training

    2020, Computers and Education
    Citation Excerpt :

    One argument is that practice can help consolidate learning. The worked example provides users with a mental model of the solution process, which can then subsequently be fortified by practicing with a similar problem (e.g., Hodges & Coppola, 2015; van Gog, 2011). Similarly, a testing effect may occur, with practice strengthening the memory trace (e.g., Brewer, Marsh, Meeks, Clark-Foos, & Hicks, 2010).

View all citing articles on Scopus
View full text