جزئیات اغوا کننده و حواس پرتی توجه - آزمایش ردیاب چشم

Abstract The seductive detail principle asserts that people learn more deeply from a multimedia presentation when interesting but unimportant adjuncts are excluded. The recent eye tracker experiment attempts to expand this principle for learning contexts containing no time limits and examines the moderating effect of attention control. Students (N = 55) received an illustrated introduction to factor analysis and then took a retention and transfer test. Each learner was randomly assigned to one cell of a 2 (with or without seductive text passages) × 2 (with or without seductive illustrations) between subjects factorial design. Students who did not receive seductive text passages or seductive illustrations performed better on transfer, but not on retention than did learners receiving these details. The lower the learner attention control (measured with an anti-saccadic test), the more harmful were seductive text passages on transfer. Eye movements indicate that seductive text passages and illustrations might impede transfer differently.

. Introduction How can a multimedia instructional message be designed to optimize instructional effectiveness? A multimedia instructional message is a communication containing text and pictures intended to foster learning (Mayer, 2005a). For example, consider an illustrated introductory text about factor analysis, a special area of research methods. One conceivable technique for improving instructional effectiveness is to add or remove seductive details (cf. Harp & Mayer, 1997). Seductive details are interesting but unimportant or even irrelevant details that are not necessary to achieve the instructional objective (Mayer, 2005c). In the first studies concerning seductive details, these details refer to written seductive text passages typically added to an instructional text (e.g., Garner, Gillingham, & White, 1989). Latter studies also have included seductive illustrations added to a multimedia message (e.g., Harp & Mayer, 1998). For example, interesting and entertaining text passages about the biography of an important methodologist as well as pictures showing the portrait of a researcher can be added or removed. These text passages and illustrations are only tangentially related to the topic and are unimportant or irrelevant adjuncts not necessary to achieving the instructional objective (i.e., understanding the procedure of the factor analysis). On one hand, many teachers, textbook writers and instructional multimedia designers add these details in an instructional message hoping to compel a learner to pay more attention to the entire instructional material and to foster learning outcome by heightening his or her curiosity, enjoyment, and interest in the topic (cf. Harp & Mayer, 1997). On the other hand, it could be assumed that removing seductive details improves learning outcome, for example, by shifting learners’ attention from irrelevant to relevant details (Harp & Mayer, 1997). In this regard, the seductive detail effect arises when people learn more deeply from an instructional message when extraneous material is excluded rather than included (Mayer, 2005c). The purpose of the present experiment is to investigate, if and under which specific conditions, seductive details in a multimedia instructional message affect learning outcome. In addition, the present experiment examines the causes leading to the seductive detail effect. In the following section, the theoretical explanations and empirical findings concerning the seductive detail effect and its attention distraction explanation are presented. Three predictions are tested in this experiment, and then followed by a discussion about theoretical and practical implications, limitations, and future research directions. 1.1. The seductive detail effect and its theoretical explanations The seductive detail effect, postulated in the context of the cognitive theory of multimedia learning (CTML, see Mayer, 2005a), assumes that students learn more deeply from a multimedia message when extraneous material is excluded rather than included. The CTML (Mayer, 2005a) is based on the assumptions that the human information-processing system contains a visual/pictorial channel and an auditory/verbal channel (dual-channels assumption), that each channel has limited capacity for processing (limited capacity assumption), and that active learning entails carrying out a coordinated set of cognitive processes during learning (active processing assumption). Furthermore, three different memory stores are assumed, which include sensory memory, working memory, and long-term memory. Words and images are processed mainly in the working memory. The major cognitive processes required for learning with words and images are selecting, organizing, and integrating. Design principles derived from the CTML were verified in numerous experiments with different kinds of learning contents. However, it can be criticized (e.g., Rey, 2010) that the CTML does not consider recent working memory models (e.g., Baddeley, 2000) and neglect motivational and affective aspects relevant for multimedia learning to a great extent. In addition, the design principles (primarily tested in the field of natural science) derived from the CTML cannot be generalized to other contents like social sciences (De Westelinck, Valcke, De Craene, & Kirschner, 2005). According to the distraction hypothesis, postulated by Harp and Mayer (1998), seductive details reduce learning outcome by drawing the learner’s selective attention away from important information. In this regard, Harp and Mayer (1998) assume that seductive details tend to contain information that requires little attentional effort and are easily understood. Garner, 1992 and Garner et al., 1992 use a light switch metaphor to explain information processing (Anderson, Mason, & Shirey, 1984) where the switch can be turned on (often to seductive details) and off (often to important generalizations). A similar explanation for the distraction hypothesis presumes that seductive details harm learning outcomes only for learner’s with low working memory capacity who are less able to control their attention and focus on relevant information (Sanchez & Wiley, 2006). There are other theoretical explanations for the effects of seductive details (e.g., Harp and Mayer, 1998 and Lehman et al., 2007). They mainly concern working memory overload, schema interference, and coherence disruption (Rey, 2012). 1.2. Empirical findings regarding the seductive detail effect A meta-analysis conducted by Rey (2012) reveals a significant seductive detail effect with small to medium effect sizes (d = 0.30) for retention performance (including 34 effects) and medium effect sizes (d = 0.48) for transfer performance (including 21 effects). Different kinds of seductive details (e.g., seductive text passages or seductive illustrations) might be an important moderating factor for the theoretical explanations of the seductive detail effect as well as for the practical implications for instructional designers. First, different kinds of seductive details might be unevenly difficult to ignore, making them relevant for the attention distraction explanation. Second, a moderating effect of different kinds of seductive details would be valuable information for an instructional designer, because of the importance of knowing which kinds of seductive details should be especially avoided. For seductive text passages, the meta-analysis revealed the significant mean weighted effect sizes of d = 0.27 for retention performance and d = 0.65 for transfer performance. For seductive illustrations, the mean weighted effect size was d = 0.95 for retention performance and d = 0.83 for transfer performance. Overall, the results indicated that the kind of seductive details used might influence the strength of the seductive detail effect. However, other variables like the presence of a time limit might have caused these results ( Rey, 2012). Using a time limit in the learning phase might lead to lower learning performance for learners in the seductive detail condition. These learners might have problems processing the larger amount of instructional material in time or at least have to hurry in order to finish. Note that many studies typically contain 30–40% additional information compared to groups not receiving seductive details (Towler, 2009). A tight time limit in the test phase could also be decreasing the learning outcome of students receiving seductive details, especially if they were advised to recall and write down everything (i.e., the important information and the seductive details) they can remember (Rey, 2012). In addition, a moderating effect of the presence of a time limit would have been valuable information for an instructional designer, because of being able to better adapt instructional materials to specific learning situations and conditions (e.g. time restrictions). The meta-analysis revealed that the presence of a time limit was a significant moderating factor for the seductive detail effect (Rey, 2012). Studies containing a time limit in the learning and test phases showed a significant seductive detail effect with medium to large effect sizes (d = 0.66 each) both for retention and transfer performance. In contrast, studies including no time limit either in the learning or in the test phases indicated no consistent significant seductive detail effect. The effect size for retention performance was very small (d = 0.10) and significant on a 5% level, but not on a 1% level. The effect size for transfer performance was very small (d = 0.06) and not significant, possibly due to the small number of studies (5 experiments). Overall, the results indicate that the seductive detail effect is as yet only confirmed for learning contexts containing a time limit in the learning and test phases. 1.3. Empirical findings for the attention distraction explanation concerning the seductive detail effect One of the first experiments to test the attention distraction explanation for the effects of seductive details was conducted by Harp and Mayer (1998). The two experiments contained seductive illustrations and seductive text passages and used a time limit in the learning phase as well as in the test phase. The selection process toward structurally important ideas was guided by highlighting important information through a bold, italicized font (Experiment 1) and by adding learning objectives preceding the instructional material (Experiment 2). Both techniques did not significantly reduce the seductive detail effect found for recall and transfer performance, putting the attention distraction explanation into question. However, highlighting important information had no significant effect either on recall or on transfer performance suggesting it was insufficient to reduce the seductive detail effect, whereas adding learning objectives had significant effects (d = 0.35 for recall and d = 0.60 for transfer). In contrast to Harp and Mayer (1998), other authors found empirical evidence supporting the attention distraction explanation. The results from Lehman et al. (2007, Exp. 2) confirmed that important information in an instructional message received less attention if seductive details were added. The authors did not use a time limit either in the learning phase or in the test phase. Participants who received seductive text passages spent less time (d = 0.56) with the base text than participants who received the base text without seductive text passages. Additionally, readers who received seductive text passages achieved lower retention scores (d = 0.55) for the important main ideas than readers who did not receive seductive text passages. Peshkam, Mensink, Putnam, and Rapp (2011) used three different types of pre-reading instructions in two experiments concerning space travel and causes of lightning. Time limits were not used. The instructions advised the learners to focus on specific concepts described as relevant, to ignore specific concepts described as irrelevant, or to simply be aware that the text might generally contain irrelevant elements. Results revealed no significant main effect for either pre-reading instructions on reading times or for recall performance concerning the base sentences compared to the condition without these instructions. Only for the seductive detail sentences did the instruction to simply be aware that the text might generally contain irrelevant elements significantly reduce recall performance in the first experiment (d = 0.58) and reading times in the second experiment (approximately d = 1.88) compared to the condition without these instructions. In a study by Sanchez and Wiley (2006) used two different working memory tasks (i.e., the operation span task, OSPAN, and the reading span task, RSPAN) to determine a working memory-capacity score. The authors assumed that individuals possessing lower working memory capacity are generally less able to utilize executive control to ignore irrelevant or interfering information and maintain focus on a specific goal (Engle, Kane, & Tuholski, 1999); that is, controlled attention. Sanchez and Wiley (2006) used a time limit in the learning phase and in the test phase. Adding seductive illustrations (e.g., a photo of a glacier) to an expository text about the causes of ice ages significantly reduced the number of correct causes (0.83 ⩽ d ⩽ 1.03) as well as the performance of an inference verification task (1.15 ⩽ d ⩽ 1.57), but only for learners with low controlled attention. Eye tracking data used in the second experiment revealed that these learners attend to seductive illustrations for a longer time duration (d = 1.92) than learners with high controlled attention. The experiment by Sanchez and Wiley (2006) did not only test the attention distraction explanation, but also the overloading of working memory, both of which could be applied to the seductive detail effect. Instead of measuring learner attention control directly and separately (e.g., with an anti-saccadic task, see below), Sanchez and Wiley (2006) separated the study sample into either low or high working memory capacity. Therefore, this construct was not only relevant for the attention distraction explanation, but also for the working memory capacity explanation (Rey, 2012). In addition, the extreme group comparison and applied dichotomization used in the experiment could be criticized due to methodological reasons (e.g., Cohen, 1983, Humphreys and Fleishman, 1974 and Maxwell and Delaney, 1993). Overall, the attention distraction explanation seems to be one of the most promising explanations for the seductive detail effect (Rey, 2012 and Sanchez and Wiley, 2006). However, due to inconsistent findings or methodological problems the previous findings did not clearly confirm this explanation. 1.4. The present experiment The present experiment tried to expand the previous findings concerning seductive details. First, the present eye tracking study tried to confirm the seductive detail effect for seductive text passages and seductive illustrations on learning contexts containing no time limit either in the learning or in the test phases. Based on the meta-analysis of Rey (2012) the seductive detail effect is as yet only confirmed for learning contexts containing a time limit in the learning and in the test phases. However, according to the distraction hypothesis as well as to all other postulated explanations for the seductive detail effect, seductive details also should impair both retention and transfer learning performance for learning contexts containing no time limit either in the learning or in the test phases. Second, the experiment tried to test the attention distraction explanation for the seductive detail effect by directly and separately measuring the learner’s attention control as a continuous variable. Interindividual differences in attention control should occur especially in situations in which a specific goal has to been focused on while ignoring irrelevant or interfering information (Engle et al., 1999), e.g. an anti-saccadic task (Everling and Fischer, 1998, Hallet, 1978, Hallet and Adams, 1980 and Hutton and Ettinger, 2006). In an anti-saccadic task an intentional control is required to cope with this task successfully, i.e. the automatic gaze to a suddenly appearing stimulus (pro-saccade) has to be suppressed and an intentional controlled anti-saccade in the opposite direction has to be performed (Bell et al., 2000 and Fischer and Weber, 1997). Therefore, the present experiment tries to expand the results of Sanchez and Wiley (2006) and tries to resolve some of the weaknesses of this study. Moreover, the usage of an eye tracker should enable insights into the on-line processing of seductive details as in other recent studies concerning multimedia learning (e.g., Boucheix and Lowe, 2010, Jarodzka et al., 2010 and van Gog and Scheiter, 2010) as well as insights into the underlying causes of different effect sizes between seductive text passages and seductive illustrations in regard to retention and transfer performance. On the basis of the seductive detail effect, the first hypothesis of the experiment predicts that students who do not receive seductive text passages perform better on retention and transfer than do students who receive seductive text passages. Contrary to the seductive detail effect hypothesis, teachers, textbook writers, instructional designers and some researchers (e.g., Park, Kim, Lee, Son, & Lee, 2005) who advocate adding seductive details in a multimedia instructional message to encourage learners to pay more attention to the entire text by heightening their curiosity, interest and enjoyment of the topic, predict the reverse pattern (i.e. seductive text passages foster both retention and transfer performance). The second hypothesis based on the seductive detail effect predicts that students who do not receive seductive illustrations perform better on retention and transfer than do students who receive seductive illustrations. Again, teachers, textbook writers, instructional designers and some researchers who advocate adding these illustrations predict the reverse pattern (i.e., seductive illustrations improve retention and transfer performance). On the basis of the attention distraction explanation, a learner’s attention control should moderate the impact of seductive details on retention and transfer performance. Therefore, the third hypothesis based on the attention distraction explanation predicts that the lower the learner’s attention control, the more harmful is the seductive detail effect on retention and transfer performance.

3. Results Table 1 shows the mean scores and standard deviations for the four different groups on measures of prior knowledge, retention, transfer, and time spent with the instructional material (without the time spent for the retention and transfer questions and without the time spent with the anti-saccadic task). Prior knowledge scores were considered as covariates in the inferential statistical analysis. Two-way analyses of covariance (ANCOVAs) on each dependent variable were conducted, each with seductive text passages (with or without seductive text passages) and seductive illustrations (with or without seductive illustrations) as between-subjects factors, and retention, transfer, and time spent with the instructional material as dependent measures. Table 1. Mean score on prior knowledge, retention and transfer tests as well as time spent with the instructional material (without the time spent for the retention and transfer questions and without the time spent with the anti-saccadic task) in minutes and their corresponding standard deviations for the four different groups (without seductive text passages and seductive illustrations, without seductive text passages but with seductive illustrations, with seductive text passages but without seductive illustrations or with seductive text passages and seductive illustrations). Group Type of Test Prior knowledge Retention Transfer Time Seductive text passages Seductive illustrations Group size M SD M SD M SD M SD – – 14 2.21 2.01 28.71 4.21 15.36 2.27 17.47 4.17 – + 14 1.50 1.45 27.43 5.16 13.93 2.17 21.03 5.19 + – 14 2.57 2.50 28.50 4.01 14.50 3.39 21.31 4.25 + + 13 3.15 2.34 29.92 4.99 13.08 2.14 19.79 4.72 Note: “+” means “with”, “–“ means “without”. Potential scores ranged from 0 to 10 for the prior knowledge, from 0 to 40 for the retention and from 0 to 25 for the transfer score. Table options 3.1. Seductive text passages and seductive illustrations Students who did not receive seductive text passages did not score significantly better on the retention test (M = 28.07, SD = 4.67) than did students who received seductive text passages (M = 29.19, SD = 4.48), F(1, 50) = 0.06, p = .81, View the MathML sourceηp2 < .01, d = −0.24. However, students who did not receive seductive text passages scored significantly better on the transfer test (M = 14.64, SD = 2.30) than did students who received seductive text passages (M = 13.81, SD = 2.90), F(1, 50) = 4.14, p < .05, View the MathML sourceηp2 = .08, d = 0.32. In addition, students who did not receive seductive text passages did not differ significantly in the time spent with the instructional material (M = 19.25 min, SD = 4.96) compared to students who received seductive text passages (M = 20.58 min, SD = 4.46), F(1, 50) = 2.84, p = .10, View the MathML sourceηp2 = .05, d = −0.28. Students who did not receive seductive illustrations did not score significantly better on the retention test (M = 28.61, SD = 4.04) than did students who received seductive illustrations (M = 28.63, SD = 5.14), F(1, 50) = 0.01, p = .91, View the MathML sourceηp2 < .01, d < 0.01. However, students who did not receive seductive illustrations scored significantly better on the transfer test (M = 14.93, SD = 2.87) than did students who received seductive illustrations (M = 13.52, SD = 2.16), F(1, 50) = 4.77, p = .03, View the MathML sourceηp2 = .09, d = 0.56. In addition, students who did not receive seductive illustrations did not differ significantly in the time spent with the instructional material (M = 19.39 min, SD = 4.57) compared to students who received seductive illustrations (M = 20.43 min, SD = 4.91), F(1, 50) = 0.68, p = .41, View the MathML sourceηp2 = .01, d = −0.22. In addition, no significant interaction was found between seductive text passages and seductive illustrations on the dependent measures retention, F(1, 50) = 0.47, p = .50, View the MathML sourceηp2 = .01, transfer, F(1, 50) = 0.23, p = .63, View the MathML sourceηp2 < .01 or time spent with the instructional material, F(1, 50) = 3.00, p = .09, View the MathML sourceηp2 = .06. The covariate prior knowledge reached significance on the dependent measures retention, F(1, 50) = 8.91, p < .01, View the MathML sourceηp2 = .15, transfer, F(1, 50) = 9.40, p < .01, View the MathML sourceηp2 = .16 and time spent with the instructional material, F(1, 50) = 6.58, p < .05, View the MathML sourceηp2 = .12. The power (1 − β) for the main effects and for the interaction effects on retention, transfer, and time spent with the instructional material was .44 for an effect size of f = .25 and α = .05. 3.2. Moderating effects of learner attention control in regard to the seductive detail effect Two additional regression analyses were conducted to investigate the moderating effects of learner attention control in regard to the seductive detail effect. Regression analyses were performed instead of median-splits, due to methodological reasons (e.g., Cohen, 1983, Humphreys and Fleishman, 1974 and Maxwell and Delaney, 1993). In the first regression analysis the predictor variables considered were as follows: learner attention control, seductive text passages (with or without seductive text passages), seductive illustrations (with or without seductive illustrations), the interaction between learner attention control and seductive text passages, and the interaction between learner attention control and seductive illustrations. Retention performance was used as a dependent measure. The second regression analysis used the same predictor variables and transfer performance as dependent measure. The results of the first regression analysis revealed no significant correlations between the five predictor variables and the dependent measure retention (−1.41 ⩽ t(49) ⩽ 1.18, .16 ⩽ p ⩽ .66). The results of the second regression analysis also indicated no significant effects for the predictor variables learner attention control, seductive text passages, and the interaction between learner attention control and seductive illustrations (−0.01 ⩽ t(49) ⩽ 0.94, .35 ⩽ p ⩽ .99). However, a significant interaction effect between learner attention control and seductive text passages on the dependent measure transfer was detected, β(standardized) = −.51, t(49) = −2.10, p < .05 (see Fig. 4). The higher the learner’s error rate of the anti-saccadic task, the lower was the mean of transfer performance for learners tested under the condition including seductive text passages. In contrast, for learners tested under the condition not including seductive text passages, the interrelation was weaker and reversed (see Fig. 4). However, both regression lines failed to reach significance (p > .05). Furthermore, a marginally significant effect for the predictor variable seductive illustrations was found, β(standardized) = −.45, t(49) = −1.93, p = .06. Interaction between learner attention control (i.e., the learner’s error rate of ... Fig. 4. Interaction between learner attention control (i.e., the learner’s error rate of the anti-saccadic task shown on the x-axis) and seductive text passages (with or without seductive text passages) on the dependent measure transfer (shown on the y-axis). The figure shows regression lines, not observed data points. Figure options 3.3. Eye tracking data concerning the learning materials Table 2 shows the mean scores and standard deviations for the four different groups on the mean total fixation time of seductive text passages, and seductive illustrations as well as the base text and the base illustrations per seductive detail or per page in seconds. The mean total fixation time of a seductive text passage (M = 20.14, SD = 5.25) was more than nine-fold longer than the mean total fixation time of a seductive illustration (M = 2.20, SD = 1.10). Table 2. Mean total fixation time per page or per seductive detail in seconds on the nine seductive text passages, the eight seductive illustrations as well as the base text and the base illustrations and their corresponding standard deviations for the four different groups (without seductive text passages and seductive illustrations, without seductive text passages but with seductive illustrations, with seductive text passages but without seductive illustrations, or with seductive text passages and seductive illustrations). The mean total fixation time represents the mean total time in seconds for every time a person looked within an area of interest (AOI), starting with a fixation inside the AOI and ending with a fixation outside the AOI. Group Mean total fixation time per page or per seductive detail in seconds of Seductive text passages Seductive illustrations Base text and base illustrations Seductive text passages Seductive illustrations Group size M SD M SD M SD – – 12 62.80 15.77 – + 12 2.27 1.00 75.14 17.72 + – 12 20.74 4.22 66.49 16.06 + + 12 19.55 6.24 2.12 1.24 56.18 15.43 Note: “+” means “with”, “–“ means “without”. Due to data recovery problems, only the first 12 participants of each group could be analyzed. Table options Furthermore, a two-way analysis of variance (ANOVA) on the dependent variable mean total fixation time of the base text and the base illustrations was conducted, with seductive text passages and seductive illustrations as between-subjects factors. Neither of the main effects for seductive text passages, F(1, 44) = 2.64, p = .11, View the MathML sourceηp2 = .06, nor those for seductive illustrations, F(1, 44) = 0.05, p = .83, View the MathML sourceηp2 < .01 were found for the dependent variable. However, a significant interaction effect between seductive text passages and seductive illustrations on the dependent measure was detected, F(1, 44) = 5.82, p = .02, View the MathML sourceηp2 = .12. Students who received no seductive details at all had a significantly shorter total fixation time, t(22) = −1.80, p = .04 (one-sided), d = −0.74, on the base text and the base illustrations (M = 62.80, SD = 15.77) than students who received seductive illustrations only (M = 75.14, SD = 17.72). In contrast, students who received seductive text passages and no seductive illustrations had a marginally significant higher total fixation time, (t(22) = 1.60, p = .06 (one-sided), d = 0.65) on the base text and the base illustrations (M = 66.49, SD = 16.06) than students who received seductive text passages and seductive illustrations (M = 56.18, SD = 15.43). Moreover, students who received seductive illustrations only had a significant higher total fixation time (t(22) = 2.80, p = .01 (two-sided), d = 1.14) on the base text and the base illustrations (M = 75.14, SD = 17.72) than students who received seductive text passages and seductive illustrations (M = 56.18, SD = 15.43). All other post-hoc tests failed to reach significance (-0.57 ⩽ t(22) ⩽ 1.25, .22 ⩽ p ⩽ .58). Furthermore, in the correlation analyses with learner attention control, the total fixation time of seductive text passages and the total fixation time of seductive illustrations were performed. While learner attention control did not correlate significantly with the total fixation time of seductive text passages (r < .01, p > .99), a significant correlation between learner attention control and the total fixation time of seductive illustrations was found (r = .41, p < .05). Significant correlations were detected neither between the total fixation time of seductive text passages and the total fixation time of seductive illustrations (r = .17, p = .61) nor between learner attention control and the total fixation time on the base text and the base illustrations (r = .04, p = .78).