Mayer, et al (1996)

Embed Size (px)

Citation preview

  • 8/6/2019 Mayer, et al (1996)

    1/10

    Journal of Educational Psychology1996, Vol. 88, No. 1,64-73 Copyright 1996 by the American Psychological Association, Inc.0022-0663/96/$3.00

    When Less Is More: Meaningful Learning From Visual and VerbalSummaries of Science Textbook LessonsRichard E. Mayer, William Bove, Alexandra Bryman, Rebecca Mars, and Lene TapangcoUniversity of California, Santa Barbara

    In a series of 3 experiments, college students who read a summary that contained a sequenceof short captions with simple illustrations depicting the main steps in the process of lightningrecalled these steps and solved transfer problems as well as or better than students whoreceived the full text along with the summary or the full text alone. In Experiment 2, takingaway the illustrations or the captions eliminated the effectiveness of the summary. InExperiment 3, adding text to the summary reduced its effectiveness. Implications for acognitive theory of multimedia learning are discussed; implications for instructional designpertain to the need for conciseness, coherence, and coordination in presenting scientificexplanations.

    Consider the following scenario. A student who is inex-perienced in meteorology reads a textbook lesson explain-ing the cause-and-effect chain of events involved in howlightning storms develop. The explanation is clearly con-tained within the 600 words and five illustrations of thelesson. A few minutes later, we ask the student to writedown the explanation (as a retention test) and to solve someproblems that require using the explanation from the lesson(as a transfer test). For example, as a transfer problem, weask the student to write an explanation for why there can beclouds in the sky but no lightning. Despite exerting consid-erable effort, the student performs poorly on both the reten-tion and the transfer task, indicating a lack of understandingof the process of lightning.

    Unfortunately, this is not a made-up example, but ratherreflects a pattern of results obtained frequently in our re-search laboratory at Santa Barbara (Mayer, Steinhoff,Bower, & Mars, 1995). Students can carefully read a text-book lesson that contains a scientific explanation, and yetnot be able to remember the explanation adequately or touse it to solve problems. Given the importance of textbooksas a commonly used vehicle for prom oting student learning,evidence of students' difficulties in learning from text isparticularly disturbing (Britton, Woodward, & Binkley,1993; Driscoll, Moallem, Dick, & Kirby, 1994; Garner,1992).This predicament raises the question of how to helpstudents understand scientific explanations. By scientific

    explanations, we mean cause-and-effect descriptions of aprocess, such as the step-by-step description of the processof how lightning develops. By understanding, we mean theability to apply what is learned to solving new problems,Richard E. Mayer, William Bove, Alexandra Bryman, RebeccaMars, and Lene Tapangco, Department of Psychology, Universityof California, Santa Barbara.Correspondence concerning this article should be addressed toRichard E. Mayer, Department of Psychology, University of Cal-ifornia, Santa Barbara, California 93106. Electronic mail may besent via Internet to [email protected].

    such as by answering transfer questions. By helping stu-dents, we mean a modification to the textbook lesson, suchas the use of a summary presented in words and illustra-tions.The goal of this article is to examine an instructionaltechnique as a candidate for promoting student understand-ing of scientific explanations, namely, the use of a specialkind of summary that combines visual and verbal informa-tion. During the past decade, researchers have increasinglydemonstrated the role of illustrations in improving the un-derstandability of textbook passages (Houghton & Willows,1987; Mandl & Levin, 1989; Schnotz & Kulhavy, 1994;Willows & Houghton, 1987); in particular, researchers havedemonstrated the value of combining text captions withillustrations to create annotated illustrations (Bernard, 1990;Guri-Rozenblit, 1988). Following Levin's (1982) taxonomyof illustrations, Mayer (1993) showed that the type of illus-tration best suited for summarizing a scientific explanationis an explanative illustration: a sequence of frames depictingthe major steps in a process, such as the stages in theformation of lightning. Furthermore, in a systematic seriesof studies, explanative illustrations were most effective inpromoting retention and transfer when they included con-cise captions describing each frame in words (Mayer, 1989;Mayer & Gallini, 1990; Mayer et al., 1995). In short,previous research has established the value of a series ofannotated illustrations as an adjunct to a text passage; thecurrent study takes this work one step further by exam iningwhether a series of annotated illustrations constitutes auseful summary that, instead of serving as an adjunct to text,can stand alone.

    The main focus of this study is on what can be called amultimedia summary: a sequence of annotated illustrationsdepicting the steps in a process. For example, Figure 1presents a summary of the cause-and-effect explanation forthe process of lightning. The essential elements of thissummary are that (a) the explanation is presented in a shortsequence of simple illustrations depicting the major steps inthe process; (b) the explanation is presented in a shortsequence of brief sentences describing the major steps in the64

  • 8/6/2019 Mayer, et al (1996)

    2/10

    LEARNING FROM SUMMARIES 65

    BrandUpward-moving

    tepped leader

    1. W arm moist air rises , water vaporcondenses and forms cloud. 4. Two leaders meet, negativelycharged particles rush from cloudto ground.iwndrafts

    2. Raindrops and ice crystals dragair downward. 5. Positively charged particlesfrom the ground rush upwardalong the same path.

    Positively[ed particles

    Negatavcharged particles

    3. Negatively charged particles fallto bottom of cloud.Figure 1. A multimedia summary of the process of lightning.

    process; and (c) the visual and verbal explanations arecoordinated so that each visual illustration contains a cor-responding verbal caption.The essential features of this summary are that it is (a)concise, (b) coherent, and (c) coordinated. A summary isconcise if the visual explanation is presented using only asmall number of simple illustrations and the verbal expla-nation is presented using only a small number of words.Furthermore, a summary is concise if these explanationsrefer only to a small number of the parts and actions in thesystem, namely, the parts and actions that are essential forthe explanation. A summary is coherent if the visual andverbal explanations are each given as sequential cause-and-effect chains and a change in the state of one part of thesystem is clearly related to a change in the state of anotherpart of the system. A summary is coordinated if each step inthe explanation is presented in visual form and verbal formso that corresponding illustrations and words are presentedtogether.The theoretical rationale for these three features of sum-maries can be found in a cognitive theory of multimedia

    learning (Mayer & Anderson, 1991, 1992; Mayer et al.,1995), which is based on elements of dual-coding theory(Clark & Paivio, 1991; Paivio, 1990), cognitive-load theory(Mousavi, Low, & Sweller, 1995; Sweller & Chandler,1994; Sweller, Chandler, Tierney, & Cooper, 1990), andgenerative theory (Wittrock, 1974, 1989). As summarizedin Figure 2, a cognitive theory of multimedia learning isbased on the idea that meaningful learning requires that thelearner engage in five active cognitive processes: selectingwords, selecting images, organizing words, organizing im-ages, and integrating words and images.The cognitive process of selecting words involves build-ing a mental representation in verbal working memory ofthe basic propositions in the explanation, such as "negative-ly-charged particles fall to bottom of cloud." The cognitiveprocess of selecting images involves building a mentalrepresentation in visual working memory of the basic im-ages in the explanation such as an image of a cloud withnegative signs toward the bottom. The cognitive process oforganizing words involves building internal connections

    among the propositions such that the statement "negativelycharged particles fall to bottom of cloud" follows "icecrystals drag air downward" and is followed by "negativelycharged particles rush from cloud to ground." The cognitiveprocess of organizing images involves building internalconnections among the images such that an image of neg-ative charges in the bottom of a cloud is followed by animage of negative charges moving from the bottom of thecloud toward the ground, which is followed by an image ofpositive charges moving from the ground toward the bottomof the cloud. Finally, the cognitive process of integratinginvolves building external connections between a proposi-tion such as "negatively charged particles fall to bottom ofcloud" and a corresponding image such as of negativecharges at the bottom of a cloud.

    The cognitive rationale for conciseness is that a concisesummary allows the learner to select the relevant words andimages. By paying attention to the relevant material, thelearner is able to build verbal and visual mental represen-tations of the major states of the system. The cognitiverationale for coherence is that a coherent summary allowsthe learner to organize the relevant words and the relevantimages into respective cause-and-effect chains. By organiz-ing the relevant material, the learner is able to build con-nections among the pieces of verbal information, yielding averbal model, and to build connections among the pieces ofvisual information, yielding a visual model. Finally, thecognitive rationale for coordination is that a coordinatedsummary allows the learner to build connections betweenthe visual and verbal representations of the process.

    In summ ary, this study addresses three kinds of questions:philosophical, instructional, and cognitive. The philosophi-cal question is, What is scientific explanation? For p urposesof this study, a scientific explanation is a cause-and-effectdescription of a sequential process, such as the steps in theformation of lightning as represented in Figure 1. Theinstructional question is, How can instruction help studentsunderstand a scientific explanation? The focus of this studyis a straightforward instructional device that has been shown

  • 8/6/2019 Mayer, et al (1996)

    3/10

    66 MAYER, BOVE, BRYMAN, MARS, AND TAPANGCO

    S T E P l :SELECT WORDSSTEP 2:SELECT IMAGES

    SUMMARY:"negatively chargedparticles fall to bottom ofcloud" _ |

    fall (negatively chargedparticles, bottom of cloud)

    VERBAL WORKING MEMORY VISUAL WORKING MEMORY

    STEP 3:ORGANIZE W ORDSSTEP 4:ORGANIZE IMAGES

    drag (ice cry stals, air(downward))

    fall (negatively chargedparticles, bottom of cloud)V ER BA L WO RK IN G MEM OR Y V IS UA L W OR KIN G ME MO RY

    STEP 5:INTEGRATE WORDSAND IMAGESdrag (ice crystals, air(downward))

    1Tfall (negatively chargedparticles, bottom of cloud)V ER BA L WO RK IN G MEM ORY V IS UA L WO RK IN G ME MO RY

    Figure 2. A cognitive theory of multimedia learning.

    to have some potential in promoting retention and transferof scientific explan ations, namely, a summ ary in the form ofa sequence of annotated illustrations as exemplified in Fig-ure 1. The cognitive question is, Why do annotated illus-trations help students understand scientific explanations?The theoretical framework used to address this question is acognitive theory of multimedia learning that involves theselecting, organizing, and integrating of visual and verbalrepresentations.

    Experiment 1Previous research has identified an effective technique forhelping students understand scientific textbook explanationsof how a cause-and-effect system works: adding a briefsummary of the steps in the process depicted by a series ofannotated illustrations (Mayer et al., 1995). In Experiment1, we examined w hether reading a summ ary (summary-onlygroup) is as effective in promoting retention and transferperformance as reading the full passage along with thesummary (passage-and-summary group) or the full textwithout the summary (passage-only group). A no-instruc-

    tion group was included to provide a baseline for compar-ison of retention and transfer scores.One function of annotated illustrations is to guide thelearner's attention toward the verbal explanation given inthe annotation, which fosters the construction of a verbalrepresentation of the explanative material. Consequently,we predicted higher recall scores for groups that receivedthe explanative information in annotated illustrations (i.e.,summary-only and passage-and-summary groups) thangroups that did not receive the explanative informationpresented in annotations (passage-only and no-instructiongroups). A more focused prediction was that students in thesummary-only group would recall more explanative infor-mation than students in the passage-and-summary group,because the summary-only students could focus their atten-tion solely on relevant words, whereas passage-and-sum-mary students had to read many words that were not part ofthe explanation.A second function of annotated illustrations is to encour-age the learner to organize the material into a cause-and-effect system and to integrate the verbal and visual repre-sentations of the system. A cognitive theory of multimedialearning requires that students build representational con-

  • 8/6/2019 Mayer, et al (1996)

    4/10

    LEARNING FROM SUMMARIES 67nections (i.e., verbal and visual representations of the ex-planations) and referential connections between the visualand verbal representations. The groups that are best sup-ported in building connections between words and picturesare those receiving annotated illustrations (summary-onlyand passage-and-summary groups), so we predicted thesegroups would outperform all others on the problem-solvingtransfer test. A more focused prediction was that the sum-mary-only group would perform as well as or better than thepassage-and-summary group on the problem-solving trans-fer test.

    MethodParticipants and design. The participants were 56 college stu-dents who lacked experience in meteorology. They w ere recruitedfrom the Psychology Subject Pool at the University of California,Santa Barbara, and they fulfilled a class requirement by partici-pating in the study. Fourteen participants served in each of fourinstructional groups: passage-and-summary, passage-alone, sum-

    mary-alone, and no-instruction.Materials. The materials consisted of a participant question-naire, three instructional booklets, a recall sheet, and four problem-solving transfer sheets, each typed on 8.5- X 11-in. sheets ofpaper.The participant questionnaire asked students to rate their knowl-edge of meteorology (orweather) on a 5-point scale ranging fromvery little to very much and to place a check mark next to each ofseven items that applied to them, including the following: "Iregularly read the weather maps in a newspaper," "I know what acold front is," "I can distinguish between cumulous and nimbusclouds," "I know what a low pressure system is," "I can explainwhat makes the wind blow," "I know what this symbol means"(symbol for cold front), and "I know what this symbol means"(symbol for warm front).The passage-and-summary booklet consisted of two sheets ar-ranged in a folder as facing pages in an open book. The sheetscontained a 600-word passage (which we refer to as the passage)and five captioned illustrations showing how lightning works(which we refer to as the summary). The passage was created bythe authors on the basis of high school science textbooks andencyclopedia entries, including the World Book Encyclopedia's(1992) entry for lightning. In addition to factual information de-scribing the properties of lightning, the passage included a presen-tation of the chain of causes andeffects that constitutes an expla-nation of how a lightning storm develops. In the summary, fiveillustrations depicted events in the cause-and-effect chain, such aswarm moist air rising to form a cloud. Each illustration was placednext to its corresponding paragraph that described the eventsdepicted in the illustration; each illustration contained a shortcaption that repeated the description of cause-and-effect eventsfrom the corresponding text in the passage (requiring 48 w ords forall illustrations); and each illustration contained labels that re-peated key terms from the passage (requiring 30 words for allillustrations). This booklet is identical to the integrated bookletused by Mayer et al. (1995).The passage-alone booklet was identical to the passage-and-summary booklet except that the summary was deleted. The sum-mary-alone booklet was identical to the passage-and-summarybooklet except that the passage was deleted, as shown in Figure 1.The recall sheet consisted of a statement asking the student towrite down an explanation of how lightning works.Each problem-solving transfer sheet contained one of the fol-

    lowing four problems: (a) "What could you do to decrease theintensity of a lightning storm ?" (b) "Suppose you see clouds in thesky, but no lightning. Why not?" (c) "What does air temperaturehave to do with lightning?" (d) "What causes lightning?"Procedure. Participants were randomly assigned to treatmentgroups and tested in small groups of 1 to 3 per session. Conditionswere throroughly randomized w ith respect to time of day, time ofweek, andweek of the academic quarter. The same experimenteradministered the experiment for all participants in all conditions.All participants in a session received the same treatment. Eachparticipant was seated in an individual booth that contained a deskand partitions on three sides. First, the participant filled out theparticipant questionnaire. Participants who indicated a lack ofknowledge of meteorology on the participant questionnaire (i.e.,those who rated their knowledge as "very little" or next to "verylittle" and who checked fewer than three items on the checklist)were classified as low-experience learners. Participants who werenot classified as low-experience learners (n = 5) were not includedin the experiment. Second, participants in the passage-and-sum-mary, passage-alone, and summary-alone groups were given 5 minto read their corresponding instructional booklet. They were told toread carefully and that afterward they would be asked somequestions about what they had read. The instructional conditionswere contained in the booklets that students read silently whileworking independently. Participants in the no-instruction groupskipped this phase of the study. Next, the booklets were collected.All participants were given the recall sheet and asked to writedown all they could about how lightning works in 6 min. Finally,following instructions, participants were given 2.5 min to generateas many answers as possible for each of the four problem-solvingquestions, presented one at a time in the order previouslyindicated.

    ResultsScoring. On the recall test, participants received onepoint for each of the major idea units they produced, even if

    their wording differed from the original. The eight idea unitsare major actions in the cause-and-effect chain and werecontained both in the passage and in the summary. The eightidea units are as follows: (1) warm, moist air rises; (2) watervapor condenses (or water vapor forms clouds); (3) rain-drops and ice crystals fall; (4) air is dragged downward; (5)negatively charged particles fall (or move) to the bottom ofthe cloud; (6) two leaders meet; (7) negatively chargedparticles rush (or move) from cloud to ground (or down);and (8) positively charged particles move from groundupward along the same path.

    On the transfer test, participants received one point foreach acceptable answer on each of the four problem-solvingtransfer sheets, so that the total possible score was unlim-ited. A list of acceptable answers was established for each ofthe four transfer questions, although answers did not have tocorrespond verbatim. For example, among the acceptableanswers for the first question were to add positively chargedparticles to the cloud and to warm up the cloud so nofreezing takes place. For the second question, an acceptableanswer was that the cloud was not high enough to be at thefreezing level or that not enough negatively charged parti-cles were in the cloud. For the third question, an acceptableanswer was that there is a difference in temperature between

  • 8/6/2019 Mayer, et al (1996)

    5/10

    68 MAYER, BOVE, BRYMAN, MARS, AND TAPANGCOthe surface and the air or that the cloud's top must be at thefreezing level. For the fourth q uestion, an acceptable answerwas that a difference exists between negative and positiveparticles or that there is a temperature difference within thecloud.Hypothesis 1: Students who receive summaries recallmore explanative information than students who do notreceive summaries. The first hypothesis in Experiment 1was that when students receive verbal summaries (i.e., cap-tions in the illustrations) that concisely state the to-be-recalled explanation, they are more likely to attend to theverbal explanation than when they do not receive verbalsummaries. In our studies, the eight key explanative piecesof information are presented in the summary as well as inthe passage; however, the passage contains much additionalinformation, whereas the verbal summaries contain only theeight explanative idea units. For this reason, we expect thesummaries to guide the learner's attention toward theseeight explanative idea units and to encourage the learner toconstruct a verbal representation of the explanation.

    The left panel of Figure 3 shows the mean number ofexplanative idea units (out of eight) produced by each of thefour groups. Consistent with the hypothesis, a one-wayanalysis of variance (AN OVA) revealed significant differ-ences among the groups, F(3, 52) = 43.05, MSE = 1.66,p < .001. Supplemental Tukey tests (with alpha at .05)indicated that students who received no instruction (no-instruction group) produced fewer explanative idea unitsthan students in each of three groups that received a book letto read, confirming the observation that the instructionalmaterials contributed to student learning and that the pas-sage-alone group did not differ significantly from the pas-sage-and-summary group. More important, students whoreceived the summary-alone booklet outperformed each ofthe other groups, indicating that in this case "less is more";that is, students recalled mo re explanative information w henthey received the annotated illustrations alone than whenthey received the full text passage along with the annotatedillustrations. Similar results have been obtained in otherstudies on text summaries (Britton, Gulgoz, & Glynn, 1993;Reder & Anderson, 1980).These results may be biased because the inclusion of theno-instruction group artificially reduced the MSE. To over-come this problem, we recomputed the foregoing ANOVA

    with just three groups: passage-and-summary, summary-alone, and passage-alone. As in the previous analysis, thethree groups differed significantly in number of idea unitsrecalled, F(2, 39) = 10.97, MSE = 2.16, p < .001, andsupplemental Tukey tests (with alpha at .05) revealed thatthe summary-alone group recalled more idea units thaneither of the other groups, which did not differ from oneanother.Overall, these results are consistent with the idea thatannotated illustrations help the learner to focus on keyexplanative information because the explanations are pre-sented without any distracting verbal information; evenwhen the identical verbal explanative information is pre-sented within the text passage, students are less likely toencode it because they attend to nonexplanative informationin the text passage. In short, we interpret these results asconsistent with one hypothesized function of annotated il-lustrations, namely, that annotated illustrations help to guidethe learner's attention toward explanative information.A possible criticism of using recall as a dependent mea-

    sure is that students in the passage-and-summ ary group mayspend most of their test time writing down nonexplanativeinformation, whereas summary-only students may spend allof their time w riting dow n explanative information. In spiteof the recall results, therefore, students in the passage-and-summary group may actually know more about the expla-nation than do summary-alone students. To examine thispossibility, in the next subsection we focused on problem-solving transfer as the primary dependent measure for as-sessing students' understanding of the explanation of howlightning develops.Hypothesis 2: Students who receive summaries generatemore creative solutions to problem-solving questions than

    students who do not receive summaries. Our second hy-pothesis is based on the idea that annotated illustrations helpthe learner to construct connections between verbal andvisual representations of the explanation, which, in turn,supports problem-solving transfer. The right panel of Figure3 shows the mean num ber of creative solutions generated byeach group on the transfer test. Consistent with the hypoth-esis, a one-way analysis of variance revealed that the groupsdiffered significantly from one another, F(3, 52) = 21.45,MSE = 3.05, p < .001. Supplemental Tukey tests (withalpha at .05) showed that students who did not receive

    Passage-and-summarySummary-alone

    Passage-aloneNo-instruction

    Explanative recall

    1..1 2 3 4 5Mean number of idea units recalled

    Problem-snlving transfer^

    i i i i i

    1 2 3 4 5Mean number of problem solutionsFigure 3. Experiment 1: Mean number of explanative statements produced on the retention testand mean number of acceptable solutions produced on the transfer test for four treatment groups.

  • 8/6/2019 Mayer, et al (1996)

    6/10

    LEARNING FROM SUMMARIES 69instruction (i.e., no-instruction group) performed morepoorly than those in each of the other groups; more impor-tant, the two groups who received a summary (i.e., summa-ry-alone and passage-and-summary groups) performed bet-ter than the other groups but did not differ from one ano ther.Again, these results may be biased because the inclusionof the no-instruction group artificially reduced the MSE. Toovercome this problem, we recomputed the foregoingANOVA with just three groups: passage-and-summary,summary-alone, and passage-alone. As in the previous anal-ysis, the three groups differed significantly in number ofcreative solutions produced on the transfer test, F(2, 39) =6.86, MSE = 4.05, p < .001, and supplemental Tukey tests(with alpha at .05) revealed that the passage-alone groupproduced fewer solutions than either of the other groups,which did not differ from one another.Interestingly, adding the text passage to the annotatedillustrations did not increase transfer performance over sim-ply presenting the annotated illustrations alone. Put anotherway, the uninstructed group generated the fewest number ofsolutions, the passage-alone group generated considerablymore than the uninstructed group, and the two groups re-ceiving annotated illustrations each generated approxi-mately 100% more than the passage-alone group. This pat-tern of results is consistent with the prediction that asummary is as effective as a full text along with a summ aryin helping learners to understand a scientific explanation.

    Exper imen t 2Experiment 1 yielded the findings that (a) a visual andverbal summary of a passage is more effective than a fullpassage in helping students to remember and to use an

    explanation of how lightning occurs and (b) a visual andverbal summary is more effective in helping students re-member an explanation and just as effective in helpingstudents use an explanation compared to the combination ofa full text passage with a visual and verbal summary. Thesummary in Experiment 1 was a set of five annotatedillustrations that contained both a visual summary (i.e., fiveframes of illustrations) and a verbal summary (i.e., verbalcaptions and labels for each illustration). In Experiment 2,we attempted to untangle the contributions of the visual andverbal aspects of the summ ary by asking students to read thesame passage and summary as in Experiment 1, the samesummary as in Experiment 1, a summary containing wordsonly, or a summary containing illustrations only. If theeffectiveness of the summary depends on the learner's con-struction of connections between verbal and visual expla-nations, as stipulated by a cognitive theory of multimedialearning, it follows that students receiving the visual-and-verbal summary will perform better on transfer tests thanstudents receiving the visual summary alone or the verbalsummary alone.Method

    Participants and design. The participants were 68 college stu-dents who lacked knowledge of meteorology and who were re-

    cruited from the Psychology Subject Pool at the University ofCalifornia, Santa Barbara. Seventeen students served in each offour treatment groups: passage-and-summary, summary-alone,verbal-summary-alone, and visual-summary-alone.Materials. The participant questionnaire, retention sheet, andfour problem-solving transfer sheets were identical to those used inExperiment 1. The passage-and-summary booklet and the summ a-ry-alone booklet were identical to those used in Experiment 1. Theverbal-summ ary-alone booklet was identical to the sum mary-alonebooklet except that all illustrations were deleted; the visual-sum-mary-alone booklet was identical to the summary-alone bookletexcept that all words (i.e., captions and labels) were deleted.

    Procedure. The procedure w as identical to that used in Exper-iment 1 except that during the study period, students receivedeither the passage-and-summary, summary-alone, verbal-summa-ry-alone, or visual-summary-alone booklet.

    Results and DiscussionScoring. The retention and transfer tests were scored asin Experiment 1.Hypothesis 3: Students who receive summaries performas well as or better than students who receive passages withsumm aries on tests of retention and transfer. A secondarygoal of Experiment 2 was to determine whether the resultsof Experiment 1 could be replicated. As in Experiment 1,the summary-alone group recalled more explanative ideaunits than the passage-and-summary group (Ms = 5.94 and3.45, respectively), f(32) = 4.43, p < .001. Also as inExperiment 1, the summary-alone group performed as wellon the transfer test as the passage-and-summary group; infact, in contrast to Experiment 1, the summary-alone groupproduced significantly more creative solutions than the pas-sage-and-summary group (M s = 5.00 and 3.85, respective-

    ly), f(32) = 2.06, p < .05. Taken together, the results ofExperiments 1 and 2 show that a summary, in the form ofannotated illustrations, can be either as effective as or moreeffective than a full passage combined with a summaryin helping students to remember and to use a scientificexplanation.Hypothesis 4: Students who receive summaries contain-ing verbal information recall more explanative informationthan students who receive summaries that do not containverbal information. The main goal of Experiment 2 was todetermine whether a visual summary (i.e., a series of fiveillustrations depicting the major events in the process oflightning), a verbal summary (i.e., a series of five captionsdescribing the major events in the process of lightning), orboth, are needed to produce an effective summary.The left panel of Figure 4 shows the mean number ofexplanative idea units produced on the retention test. Ananalysis of variance performed on the recall data revealedthat the four treatment groups differed significantly in themean number of idea units recalled, F(3, 64) = 50.25,MSE = 2.27, p < .001 . Supplemental Tu key tests (withalpha at .05) revealed that the visual-summary-alone groupproduced fewer explanative idea units than each of the othergroups, presumably reflecting th e lack of verbally presentedmaterial. More interestingly, the summary-alone and verbal-summary-alone groups produced significantly more idea

  • 8/6/2019 Mayer, et al (1996)

    7/10

    70 MAYER, BOVE, BRYMAN, MARS, AND TAPANGCOProblefn-RQlYJnP transfer

    Passage-and-summarySummary-alone

    Verbal-summary-aloneVisual-summary-alone

    1 2 3 4 5Mean number of idea units recalled 1 2 3 4 5Mean number of problem solutions

    Figure 4. Experiment 2: Mean number of explanative statements produced on the retention testand mean number of acceptable solutions produced on the transfer test for four treatment groups.

    units than either of the other two groups but did not differfrom one another. If the goal of instruction is solely toproduce good retention of a verbal explanation, the bestinstructional method seems to be to provide the learner witha verbal explanation. However, as noted in the previoussection, we focused on problem-solving transfer as ourprimary measure of understanding.

    Hypothesis 5: Students who receive summaries contain-ing both visual and verbal information produce more cre-ative solutions than students who receive summaries con-taining only visual or only verbal information. The rightpanel of Figure 4 shows the mean number of creativesolutions produced on the problem-solving transfer test foreach of the four treatment groups. An analysis of variancerevealed that the groups differed significantly from oneanother in the mean number of problem solutions, F(3,64) = 14.53, USE = 2.50, p < .001. Supplemental Tukeytests (with alpha at .05) revealed that the summary-onlygroup did not differ significantly from the passage-and-summary group, replicating the results of Experiment 1.More imp ortant, a major new finding in Experim ent 2 is thatthe summary-alone group outperformed both the visual-summary-alone and the verbal-summary-alone groups. Weinterpret these results as indicating that both verbal andvisual aspects of the summary are useful in helping inex-perienced learners build a coherent understanding of a sci-entific explanation. It is interesting that, although studentsin the verbal-summary-alone group remembered as much ofthe verbal explanation as did students in the summary-alonegroup, they were not as effective as the summary-alonestudents in applying their verbal knowledge to solve prob-lems. If the goal of instruction is to produce good retentionand transfer of scientific explanation, then the best instruc-tional method seems to be a coordinated verbal and visualsummary of the major steps in the to-be-explained process.

    In conclusion, the most effective summary, according toour research, is one that coordinates a verbal and visualsummary of the explanation, rather than one that summa-rizes the explanation only in words or only in pictures.Exper imen t 3

    Experiment 2 provided a replication of one of the majorfindings in Experiment 1, in which students who read a

    summary performed as well as or better than students whoread a full text along with a summ ary. In addition, the majornew finding in Exp eriment 2 was that the positive effects ofthe summary occurred most strongly when both visual andverbal information were presented and not when the sum-mary w as presented solely in verbal form or solely in visualform. Thus, the results of Experiment 2 showed that takingaway the illustrations or the captions from the summaryreduced its effectiveness.

    This pattern is consistent with a cognitive theory of mul-timedia learning, which states that both visual and verbalrepresentations are needed to build referential connections.When only a verbal summary is presented, an inexperiencedlearner may be able to generate a verbal representation butmay have difficulty in constructing a visual representationon the basis of the verbal information; when only a visualsummary is presented, an inexperienced learner may be ableto generate a visual representation but may have difficultyin constructing a verbal representation on the basis of thevisual information.In Experiment 3, we examined the effects of adding textto the captions in the summary. According to a cognitivetheory of multimedia learning, adding text may disrupt theprocess of selecting relevant wo rds. When extraneous text isadded, the learner may have difficulty in identifying thecore steps in the causal chain; this task is easier when thetext consists only of the core steps in the causal chain. Itfollows that students who read a summary will performbetter on tests of retention and transfer than will studentswho read a summary containing additional words.

    MethodParticipants and design. The participants were 39 college stu-dents who lacked experience in meteorology. They w ere recruitedfrom the Psychology Subject Pool at the University of California,Santa Barbara, and they fulfilled a class requirement by partici-pating in the study. Thirteen students served in each of threeinstructional groups: summary, summary-plus-50-words, summa-ry-plus-550 words.Materials. The materials consisted of the same participantquestionnaire, recall sheet, and four problem-solving transfersheets as in Experiments 1 and 2; in addition, the materialsincluded three instructional books: the summary booklet, a sum-mary-plus-50-words booklet, and a summary-plus-550-words

  • 8/6/2019 Mayer, et al (1996)

    8/10

    LEARNING FROM SUMMARIES 71booklet. The summary booklet was identical to the one used inExperim ents 1 and 2; it contained five captioned illustrations w ithapproximately 50 words in the captions. The summary-plus-50-words booklet was the same as the summary booklet except that amore in-depth description of each illustration was given in eachcaption, yielding a total of approximately 100 words in the cap-tions. Like the summary captions, the expanded captions weretaken verbatim from the full text passage used in Experiments 1and 2. The expanded captions were as follows:

    1. Warm moist air near the earth's surface rises. As the airin this updraft cools, water vapor condenses into water drop-lets and forms a cloud. The cloud's top extends beyond thefreezing level, so tiny ice crystals form in the upper portion ofthe cloud.2. Eventually, the ice crystals become too large to besuspended by updrafts, so they fall through the cloud. Theydrag air downward, producing downdrafts. When downdraftsstrike the ground they produce gusts of cool wind.3. The rising water droplets collide with the falling ice,producing electrical charges. Negatively charged particles fallto the bottom of the cloud, and positively charged particlesrise to the top.4. A negatively charged stepped leader moves downwardfor the cloud in a series of steps. A positively charged upward-moving leader travels up from the trees and buildings to meetthe negative charges. When the two leaders meet, negativelycharged particles rush from the cloud to the ground.5. As the leader stroke nears the ground, it induces anopposite charge. Positively charged particles from the groundrush upward along the sam e path. This return stroke producesthe bright light that people notice in a flash of lightning.

    In the summary-plus-550 words booklet, the captions for the fiveillustrations consisted of the entire 600-word passage as used inExperiments 1 and 2. This is just like the passage-and-summarycondition of the previous experiment but presented in a differentformat, that is, each portion of text is placed with its correspondingillustration.

    Procedure. The procedure was identical to that used in Exper-iments 1 and 2, except that during the study period, studentsreceived either the summary, summary-plus-50-words, or summa-ry-plus-550-words booklet.

    Results and DiscussionScoring. The retention and transfer tests were scored asin Experiment 1.

    Hypothesis 6: Students who receive concise summariesrecall more explanative information than students who re-ceive expanded summaries. A major hypothesis in Exper-iment 3 is that students who receive simple illustrations withbrief captions depicting the steps in the process of lightningwill recall those steps better than will students who receivesimple illustrations with longer captions. A rationale for thisprediction is that lengthy captions may overload verbalworking memory so that some of the explanative informa-tion is lost.The left panel of Figure 5 shows the mean number ofexplanative idea units (out of eight) produced by each of thethree groups on the retention test. Consistent with the hy-pothesis, a one-way ana lysis of variance revealed significantdifferences among the groups in mean number of explana-tive ideas recalled, F(2, 36) = 20.94, MSE = 2.15, p

  • 8/6/2019 Mayer, et al (1996)

    9/10

    72 MAYER, BOVE, BRYMAN, MARS, AND TAPANGCOcreative solutions to problem-solving questions than willstudents who receive simple illustrations with longer cap-tions. A rationale for this hypothesis is that lengthy ca ptionsmay overload verbal working memory, disrupting the pro-cess of building a verbal representation and connecting itwith a visual representation of the lightning system.

    The right panel of Figure 5 shows the mean number ofcreative solutions generated by each group on the problem-solving transfer test. An analysis of variance revealed sig-nificant differences among the groups in their problem-solving performance, F(2, 36) = 3.19, MSE = 2.34, p =.05. Supplemental Tukey tests (with alpha at .05) showedthat the summary group performed significantly better thanthe summary-plus-550-words group on the problem-solvingtransfer test but not better than the summary-plus-50-wordsgroup. These results are consistent with the idea that addinga lot more text to the captions can reduce students' effi-ciency in abstracting the core verbal explanation and inconnecting it with the visual explanation.

    ConclusionWhen the goal of instruction is to help students be able toexplain a scientific system in words (retention) and to usethis explanation to solve problems (transfer), a commoninstructional practice is to provide a lengthy verbal expla-nation, such as a textbook passage or a classroom lecture.Indeed, instructors may believe that providing a lengthyverbal explanation fulfills their responsibility to provideinformation to the learner. Unfortunately, this practice is notvery efficient for many students, presumably because stu-dents do not process the information effectively. In the

    present study, for example, there was no instructional treat-ment, including a 600-word passage with summary, thatproved to be more effective in promoting retention andtransfer than a summary. By reducing the load on thecognitive system, summaries may enable students to carryout the cognitive processes necessary for meaningful learn-ing as summarized in Figure 2.Our research extends earlier research on text-based sum-maries (Reder & Anderson, 1980) by examining the natureof summaries that are based on both text and illustrationswhat can be called multimedia summaries. In particular, ourresearch suggests that a verbal summary is not as effectiveas a multimedia summary that combines both visual andverbal formats and that a multimedia summary is moreeffective when it contains a small am ount of text rather thana large amount.What constitutes an effective multimedia summary? Ourmultimedia summary was constructed on the basis of threecriteria: conciseness, in that only a few illustrations andsentences were presented; coherence, in that the images andsentences were presented in cause-and-effect sequence; andcoordination, in that the images were presented contigu-ously with their corresponding sentences (i.e., each illustra-tion had a verbal caption). In Experiment 3, we varied theconciseness of the verbal explanation and found evidencethat students learn more effectively from a more concise

    summary. In Experiment 2, we varied the coordination andfound that students learn more effectively when words andillustrations are presented together rather than separately (asa words-only or illustrations-only treatment). However, wedid not directly compare presenting a coordinated multime-dia summary (i.e., verbal captions with each illustration)with presenting a verbal summary on a different page froma visual summary. Furthermore, we did not vary the coher-ence of the summaries by presenting the illustrations andsentences in random order versus sequential order.The quality of the verbal explanation is another issue thatdeserves additional research attention, because well-writtentext may not have the same effects as poorly written text.Although we attempted to use well-written text in the 600-word passage, we may not have succeeded. Therefore, ad-ditional research is needed to determine whether or not thepassage-and-summary treatment would foster better resultsif the quality of the passage was improved.Additional limitations concern the n ature of the materials,the learners, and the dependent measures. The materials

    consisted of a scientific explanation of a cause-and-effectprocess; if we had used a narrative or purely descriptivepassage rather than an explanative passage, the results mayhave been quite different. The learners were inexperiencedin the subject domain; if we had used experienced learners,we would not expect to find differences among the treat-ment groups. Finally, we focused on retention and transferof the explanative information as our dependent measures;if we had focused on overall amount remem bered, we wouldnot expect to find the same kinds of differences among thegroups. In short, these experiments p rovide new informationconcerning the circumstances in which a summary is effec-tive, namely, when the material is potentially structured,when the learners do not norm ally identify the structure, andwhen the summary conveys the structure visually andverbally.Last, it would be incorrect to conclude that in all casesstudents can learn as much from studying a summary asfrom studying a full lesson along with a summary. Thisstudy provides consistent evidence of a situation in which asummary can be effective in promoting student understand-ing of a scientific explanation. Further research is needed todetermine the role of carefully constructed multim edia sum-maries that are sensitive to the cognitive loads on visual andverbal working memories.

    ReferencesBernard, R. M. (1990). Using extended captions to improve learn-ing from instructional illustrations. British Journal of Educa-tional Technology, 21, 215-225.Britton, B. K., Gulgoz, S., & Glynn, S. (1993). Impact of good andpoor writing on learners: Research and theory. In B. K. Britton,A. Woodward, & M. Binkley (Eds.), Learning from textbooks:Theory and practice (pp. 1-46). Hillsdale, NJ: Erlbaum.Britton, B . K., Woodw ard, A., & Binkley, M. (Eds.). (1993).Learning from textbooks: Theory and practice. Hillsdale, NJ:Erlbaum.Clark, J. M., & Paivio, A. (1991). Dual coding theory and educa-tion. Educational Psychology Review, 3, 149-210.

  • 8/6/2019 Mayer, et al (1996)

    10/10

    LEARNING FROM SUMMARIES 73Driscoll, M., Moallem, M., Dick, W., & Kirby, E. (1994). Howdoes the textbook contribute to learning in a middle schoolscience class? Contemporary Educational Psychology, 19, 7 9 -100.Gamer, R. (1992). Learning from school texts. Educational Psy-chologist, 27 , 53-63.Guri-Rozenblit, S. (1988). The interrelationship between diagram-

    matic representations and verbal explanations in learning fromsocial science text. Instructional Science, 17, 219-234 .Houghton, H. A., & Willows, D. M. (Eds.). (1987). The psy-chology of illustrations: Vol. 2: Instructional issues. Berlin:Springer-Verlag.Levin, J. R. (19 82). Pictures as prose-learning de vices. In A.Flammer & W. Kintsch (Eds.), Discourse processing (pp. 412 -444). Amsterdam: North-Holland.Mandl, H., & Levin, J. R. (Eds.). (1989). Knowledge acquisitionfrom text and pictures. Amsterdam: North-Holland.Mayer, R. E. (1989). Systematic thinking fostered by illustrationsin scientific text. Journal of Educational Psychology, 81, 2 4 0 -246.Mayer, R. E. (1993). Illustrations that instruct. In R. Glaser (Ed.),Advances in instructional psychology (Vol. 4, pp. 253-284).Hillsdale, NJ: Erlbaum.Mayer, R. E., & A nderson, R. (1991 ). Animations need narrations:An experimental test of a dual-coding hypothesis. Journal ofEducational Psychology, 83, 4 8 4 - 4 9 0 .Mayer, R. E., & Anderson, R. (1992). The instructive animation:Helping students build connections between w ords and picturesin multimedia learning. Journal of Educational Psychology, 84,4 4 4 - 4 5 2 .Mayer, R. E., & Gallini, J. (1990). When is a picture worth tenthousand words? Journal of Educational Psychology, 82, 7 1 5 -727.

    Mayer, R. E., Steinhoff, K., Bower, G., & Mars, R. (1995). Agenerative theory of textbook design: Using annotated illustra-tions to foster meaningful learning of science text. EducationalTechnology Research and Development, 43, 3 1 - 4 3 .Mousavi, S. Y., Low , R., & Sw eller, J. (1995). Reducing cognitiveload by mixing auditory and visual presentation modes. Journalof Educational Psychology, 87, 319-334 .Paivio, A. (1990). Mental representations: A dual-coding ap-proach (2nd ed). New York: Oxford University Press.Reder, L. M., & A nderson, J. R. (1980). A comparison of texts andtheir summaries: Memorial consequences. Journal of VerbalLearning and Verbal Behavior, 19, 121-134.Schnotz, W., & Kulhavy, R. W. (E ds.). (1994). Comprehension ofgraphics. Amsterdam: North-Holland.Sweller, J., & Chandler, P. (1994). Why some material is difficultto learn. Cognition and Instruction, 12, 185-233.Sweller, J., Chandler, P., Tierney, P., & Cooper, M. (1990).Cognitive load and selective attention as factors in the structur-ing of technical material. Journal of Experimental Psychology:General, 119, 176-192.Willows, D. M., & Houghton, H. A. (Eds.). (1987). The psychol-

    ogy of illustration: Vol. 1, Basic research. Berlin: Springer-Verlag.Wittrock, M. C. (1974). Learning as a generative activity. Educa-tional Psychologist, 11, 87-95.Wittrock, M. C. (1989). Generative processes of comprehension.Educational Psychologist, 24, 345-376.Received June 23, 1995Revision received August 24, 1995Accepted August 29, 1995