Computers in Human Behavior 27 (2011) 1079–1086

Contents lists available at ScienceDirect

Computers in Human Behavior

journal homepage: www.elsevier .com/locate /comphumbeh

Tacit guidance for collaborative multimedia learning

Daniel Bodemer *

University of Tübingen, Department of Applied Cognitive Psychology and Media Psychology, Germany

a r t i c l e i n f o

Article history:Available online 12 June 2010

Keywords:Computer-supported collaborative learningGroup awarenessMultiple external representationsMultimedia learningCognitive loadRepresentational tool

0747-5632/$ - see front matter � 2010 Elsevier Ltd. Adoi:10.1016/j.chb.2010.05.016

* Address: University of Tübingen, Department of Aand Media Psychology, Konrad-Adenauer-Str. 40, 720+49 7071 979 314; fax: +49 7071 979 300.

E-mail address: d.bodemer@iwm-kmrc.de

a b s t r a c t

Collaborative multimedia learning is a scenario placing various demands on the learners that go beyondunderstanding complex issues and coordinating a learning discourse. On the one hand, individuals haveto mentally interrelate multiple external representations in order to understand the learning materialand the underlying concepts; on the other hand, during collaboration, learners have to use the differentlycoded information in order to exchange conceptual knowledge. In this paper, the development and exper-imental evaluation of a group awareness tool (collaborative integration tool) is presented that is intendedto simultaneously support both individual and collaborative learning processes during dyadic collabora-tive multimedia learning. The tool was experimentally compared with an integration task that alreadyproved to foster meaningful individual learning processes. The results suggest that providing groupawareness can lead to better individual learning gains by reducing demanding processes and by tacitlyguiding learner interactions.

� 2010 Elsevier Ltd. All rights reserved.

1. Introduction

When learners try to understand complex conceptual issues bymeans of multimedia learning material, translating between multi-ple external representations and mentally integrating them is animportant but difficult challenge (Ainsworth, 2006). While inte-grating multiple representations is, in the end, an individual pro-cess, the acquisition and exchange of knowledge about externalrepresentations often occurs during collaboration (Kozma, 2003),which places additional demands on learners (Dillenbourg &Bétrancourt, 2006). However, tools and instructional tasks thatsupport learning with multiple representations usually focus onthe individual. A main requirement in developing a tool that sup-ports collaborative learning with multiple external representationsis the simultaneous consideration of both individual and collabora-tive processes.

In this paper, facilitating group awareness is proposed as a suit-able means for reducing unprofitable collaborative effort, and fortacitly guiding learning-relevant interactions while leaving thescope for individual learning processes and their support. The arti-cle presents the development and experimental evaluation of agroup awareness tool that is intended to support collaborativemultimedia learning in this manner.

ll rights reserved.

pplied Cognitive Psychology72 Tübingen, Germany. Tel.:

1.1. Learning with multiple external representations

Multimedia learning materials commonly comprise differentlycoded external representations, such as texts, formulas, and dia-grams, in order to encourage learning in various ways (Ainsworth,1999). However, learners are frequently unable to utilize thepotentialities of external representations in a meaningful way. Par-ticularly, systematically translating between multiple external rep-resentations has shown to be a challenging task (Ainsworth, 2006;Kozma, 2003). In addition, the simultaneous processing of differ-ently represented information can require a considerable part ofa learner’s working memory capacity (Chandler & Sweller, 1991).As a consequence, learners often concentrate on surface character-istics instead of thematically relevant structures of the externalrepresentations and, therefore, do not recognize the strengthsof particular representations, resulting in disjointed mentalrepresentations.

In order to enable learners to take advantage of the potential ofinformation represented differently, several methods have beensuggested that attempt to support translation processes betweenrepresentations: by reducing visual search processes, such as pre-senting text and pictures in a spatially integrated format (Chandler& Sweller, 1991), or linking multiple representations by varioussymbolic conventions, such as using the same color for corre-sponding entities in different representations (Chandler & Sweller,1991; Kalyuga, 2008; Kozma, 2003). While these instructional sug-gestions have the potential to reduce cognitive workload, they donot directly support learners in active and constructive integrationprocesses. Therefore, Bodemer and colleagues tried to initiatemeaningful mental activity by enabling learners to systematically

1080 D. Bodemer / Computers in Human Behavior 27 (2011) 1079–1086

and interactively integrate differently represented components oflearning material on a computer screen. This task of active integra-tion improved the learners’ understanding in complex conceptualdomains significantly (Bodemer & Faust, 2006; Bodemer, Ploetzner,Bruchmüller, & Häcker, 2005; Bodemer, Ploetzner, Feuerlein, &Spada, 2004).

1.2. Collaborative multimedia learning

External representations are beneficial not only for individuallearning. They can perform important functions during collabora-tive learning that exceed supporting individual knowledge acquisi-tion. For example, research has shown that by means of externalrepresentations, individual contributions can be illustrated andobjectified (Roschelle & Teasley, 1995), but also coordinated andinterrelated (Scardamalia & Bereiter, 1994). Moreover, it wasshown that the type of a shared external representation can influ-ence the focus of the learning partners’ activities (Suthers & Hund-hausen, 2003).

Most computer-supported collaborative learning (CSCL) toolsuse shared external representations in some way. However, therehas been little research on collaborative learning with complexmultimedia learning material. Most of the existing studies investi-gate collaborative learning with animations or simulations, focus-ing on dynamic and interactive aspects of multimedia material andnot on the representational code (Rebetez, Bétrancourt, Sangin, &Dillenbourg, in press; Roschelle & Teasley, 1995; Saab, van Joolin-gen, & van Hout-Wolters, 2005; Sangin, Dillenbourg, Rebetez,Bétrancourt, & Molinari, 2008; Schnotz, 1999; Vahey, Enyedy, &Gifford, 2000). Those studies that focus on collaborative learningwith multiple external representations have shown that learnerscan potentially interrelate their knowledge and construct sharedmeaning on the basis of differently coded external representations(Kozma, 2000; Ploetzner, Fehse, Kneser, & Spada, 1999).

Moreover, although spatial distribution of learners is a main po-tential of CSCL, studies on collaborative multimedia learning haveconcentrated on face-to-face scenarios and neglected computer-mediated knowledge communication. One reason for this mightbe the high complexity of such learning situations that combinethe demands of multimedia learning and computer-mediated com-munication. As previously mentioned, learning with multiple rep-resentations is demanding even in individual learning settings.During distributed knowledge communication, learners encounteradditional difficulties, such as (1) establishing references betweenexternal content and collaboration content (Buder, 2007), (2) con-structing a mutual understanding and common ground (Clark &Brennan, 1991) and – associated therewith – constructing a repre-sentation of the learning partner’s knowledge or beliefs (Dillen-bourg, 2006), as well as (3) interacting and discussing with eachother in a structured and goal-oriented way (Bromme, Hesse, &Spada, 2005).

Taking into account both the demands of multimedia learningand those of computer-mediated knowledge communication, itshould be noted that they both involve high cognitive load onthe learners’ working memories. While this is a comprehensivelyinvestigated issue with regard to multimedia learning (e.g., Chan-dler & Sweller, 1991; Mayer, 2001), it is a rather novel perspectivewith regard to collaborative learning (Dillenbourg & Bétrancourt,2006; Kirschner, Paas, & Kirschner, 2009).

When developing an instructional task or a CSCL-tool intendedto support learners during collaborative multimedia learning, it isuseful to distinguish between different types of cognitive load. Fol-lowing current developments of cognitive load theory (Sweller, vanMerriënboer, & Paas, 1998), extraneous cognitive load – load that isimposed by information and activities that do not directly contrib-ute to learning – should be minimized; on the other hand, learn-

ing-relevant germane cognitive load should be encouraged. Theseprinciples are basically applicable to CSCL just as well as to individ-ual multimedia learning.

The aforementioned active integration task already intends toencourage germane load (actively translating between multipleexternal representations) and to reduce extraneous load (con-straining to relevant components; producing an integrated for-mat). With regard to the demands and difficulties ofcollaborative, computer-mediated learning, learners might be sup-ported by (1) reducing extraneous load for establishing referencesbetween content of learning material and communication, (2)reducing their effort for grounding processes and for modelingthe learning partner’s knowledge, and by (3) encouraging germaneeffort with regard to the learners’ communication behavior, such asdiscussing conflicting issues (cf., Doise & Mugny, 1978).

In the following, facilitating group awareness is proposed as asuitable means in order to support learners in using their cognitivecapacities for meaningful individual and collaborative learningactivities.

1.3. A group awareness approach for facilitating and guiding learninginteractions

Group awareness describes the perception and knowledge ofspecific aspects of group members (e.g., Bodemer & Dehler, 2011;Gutwin & Greenberg, 2002), such as where group members are,what they look like, what they are doing, or what they are inter-ested in. As group awareness is more difficult to establish duringcomputer-mediated communication than face-to-face, variousgroup awareness tools have been developed in order to supportusers in accessing and utilizing information about their communi-cation partners. Accordingly, many group awareness tools are in-tended to provide communication conditions that are similar toface-to-face settings. Recently, however, it was emphasized thatgroup awareness tools are suited to provide support that even sur-passes the face-to-face level to some degree by assessing and feed-ing back information that is difficult to yield without technologicalsupport (Bodemer & Buder, 2006; Carroll, Rosson, Farooq, & Xiao,2009).

With regard to CSCL-related group awareness tools in this lineof research, the assessment and visualization of cognitive variablessuch as knowledge, attitudes, assumptions or rationales of learningpartners is particularly promising and – therefore – has alreadybeen implemented in different ways (e.g., Buder & Bodemer,2008; Dehler, Bodemer, Buder, & Hesse, 2009; Dehler, Bodemer,Buder, & Hesse, 2011; Xiao, 2008). Taking the demands, difficulties,and approaches of support for distributed collaborative learninginto account that are delineated in this paper, group awarenesstools can facilitate learning processes by (1) providing explicit ref-erences between the external content and the communicationpartners (e.g., ratings on contributions in a discussion forum), (2)providing relevant cognitive characteristics of the learning part-ners, and (3) providing information (and providing it in a way) thatdraws the learners’ attention to aspects suitable to induce mean-ingful learning processes (e.g., visualizing diverging conceptions).This visualized offer of information can be described as tacit guid-ance that – contrary to rigid instructive guidance approaches thatexplicitly prescribe meaningful learning behavior – enables largelyself-regulated learning processes.

1.4. A collaborative integration tool for supporting collaborativelearning with multiple external representations

In order to facilitate collaborative multimedia learning, a toolhas been developed that is intended to support learners in meetingthe various demands. A main requirement in developing such type

D. Bodemer / Computers in Human Behavior 27 (2011) 1079–1086 1081

of tool is the simultaneous consideration of both collaborative andindividual learning processes. This tool was based on the previ-ously mentioned active integration task (Bodemer et al., 2004) thathas repeatedly been shown to foster individual integration pro-cesses during multimedia learning. This task of individually assign-ing components of different representations to each other hasseveral characteristics that enable it to be used in a collaborativescenario as well. Regarding the difficulties specified in this paper,(1) it constrains content information in a way that allows for theinterrelation of information gathered from the learning materialand from a learning partner, (2) it provides information about alearner’s knowledge that can be easily compared to representedinformation about a learning partner’s knowledge, (3) it guideslearning processes tacitly and adaptively by externally represent-ing assigned and unassigned information, thereby potentially en-abling learners to switch between individual and collaborativeprocesses in a self-directed way.

Matching the characteristics of the integration task with theaforementioned characteristics of group awareness tools, a collab-orative integration tool was developed that enables two spatiallydistributed learning partners to simultaneously integrate compo-nents of multimedia learning material on computer screens. Learn-ers are provided with a shared visualization that contains thecurrent state of integration of both learning partners (cf. Fig. 1).While interactively integrating different sources of information isintended to support individual elaboration processes by means ofexternal and mental structure mapping, there are other supportingfunctions that address the collaborative scenario.

As learners can assign multiple representations independentlyof each other, the collaborative integration tool visualizes informa-tion about each learner’s knowledge. However, as the tool displayscorresponding assignments of both learners side-by-side, it addi-tionally visualizes information about group knowledge (Buder &Bodemer, 2008), such as which part of the learning material is cov-

Fig. 1. Collaborative integration of multiple external representations during learning stdrop areas adjacent to the visualization (learner A’s assignments on the left side of eachdistribution are highlighted.

ered by at least one of the group members. Furthermore, the spa-tial contiguity of assignments allows for the comparison betweenboth learners for each subset of representations. With regard tothis comparison, four cases of knowledge distribution can be dis-tinguished that are visualized by the tool (cf. Fig. 1): None of thelearning partners has assigned a subset of representations (OO),only one learner has assigned it (XO), both learners have per-formed the same assignment (XX), learners have performed differ-ent assignments (XY).

The visualization of the learning partner’s knowledge in the col-laborative integration tool has the potential to support collabora-tive learning in several ways. (1) It provides a kind of basiccommunication vocabulary consisting of particularly relevant alge-braic and graphical components of the learning material. (2) It mayreduce grounding costs, as each learner is provided with informa-tion about the learning partner’s assumptions. (3) It may structurea learning discourse on the basis of the four cases of knowledgedistribution. If learners have assigned a subset of representationsidentically, they can easily recognize that there is probably no needto deeply discuss the underlying concept. On the other hand, if asubset of representations could not be assigned by any of the learn-ers individually, a joint problem-solving process might help tosolve the integration task. If only one learner has assigned a spe-cific subset of representations, it is apparent that the learner whois not knowledgeable with regard to this subset might benefit fromexplanations by the learning partner. In this case, the visualizedawareness information can also help in the formulation of ques-tions and answers that are adapted to the difference in knowledgebetween the learning partners. A very important case with regardto knowledge construction occurs if both learners have assigneddifferent representation components. Such conflicting issues aresupposed to be especially fruitful for the learning discourse (Doise& Mugny, 1978). It is assumed that learners benefit much morefrom discussing conflicting issues than from repeating issues they

atistics. Two learning partners simultaneously drag algebraic components onto thedrop area, learner B’s assignments on the right side). The four cases of knowledge

1082 D. Bodemer / Computers in Human Behavior 27 (2011) 1079–1086

both agree about. Moreover, as it has shown that conceptual con-troversy produces curiosity (Lowry & Johnson, 1981), the visualiza-tion of conflicting assumptions might tacitly guide learners todiscuss and resolve those conflicts.

In order to evaluate the potential benefits of a collaborative inte-gration tool, an experimental study has been conducted comparingtwo different kinds of support in a collaborative learning scenario:the collaborative integration tool described in this section and theactive integration task investigated by Bodemer et al. (2004, 2005).As the collaborative integration tool is intended to facilitate individ-ual and collaborative learning processes during collaborative multi-media learning, both higher individual learning outcomes as well asmore beneficial discussion processes are hypothesized if learners aresupported by this type of tool. Differences regarding overall cogni-tive load are not expected because the collaborative integration toolintends not only to reduce mental collaborative effort (groundingand model building activities) but also to enhance it (conflict discus-sion and mutual relation processes).

2. Experimental study

2.1. Method

In this experiment, spatially separated dyads were learning intwo consecutive learning phases. (1) In an individual learning phase,they were individually provided with paper-based learning mate-rial about various statistics concepts underlying the one-way anal-ysis of variance. The learning material differed within the dyads inan interdependent way in order to prompt different perspectiveson the learning subject that go along with the representationalcode the material focuses on. While one learner was provided withalgebraic and rather quantitative information, the other learnerwas provided with visual and rather qualitative information (seeFig. 2 for examples). The different learning materials were providedin order to necessitate and thus to initiate representation-baseddiscussions between the learners. (2) In a collaborative learningphase, dyads were provided with a computer-based integrationtool that comprised corresponding algebraic and visual informa-tion. In this phase, learning partners were able to communicateby means of a chat tool.

2.1.1. DesignTwo experimental groups were compared that differed with re-

gard to the visualization of the learning partner’s knowledge. One

Fig. 2. Example section of (a) algebrai

experimental group (individual support only) was provided with anindividual integration tool that enables learners to individually inte-grate algebraic and visual information by dragging and droppingalgebraic components to visual components of the learning mate-rial. This tool corresponds to the aforementioned active integrationtask that proved to be helpful in individual learning settings (Bode-mer et al., 2004). The other experimental group (individual and col-laborative support) was provided with a collaborative integrationtool that also allows individually integrating algebraic and visualcomponents, but additionally visualized the current state of inte-gration of the learning partner and thus supported groupawareness.

2.1.2. ParticipantsForty psychology students (33 females and 7 males) of the Uni-

versity of Tübingen, aged 20–29 years (M = 22.63, SD = 2.43), wererandomly assigned to the two experimental groups. They werepaid for their participation or given course credits. All studentshad attended courses in introductory statistics but were largelyunfamiliar with the specific statistics concepts and visualizationsaddressed in this experiment.

2.1.3. Materials and procedureThe study consisted of five consecutive phases: (1) training

phase (2) individual learning phase, (3) test phase 1, (4) collabora-tive learning phase, and (5) test phase 2.

(1) At the beginning of the experiment, dyad members werespatially separated and received a general introduction intothe experimental environment on the computer where theycould exercise dragging and dropping objects with regard toa neutral domain.

(2) In the individual learning phase dyad members were pro-vided with the interdependent instructional material for20 min in which they learned about the applicationdomain. It consisted of four printed pages explaining var-ious statistics concepts and principles underlying the one-way analysis of variance (ANOVA), such as the conceptsof error and squared error, the principle of least squares,the method of partitioning the sums of squares, and theeffect of outliers. The learning material differed withinthe dyads, giving one learner algebraic and rather quanti-tative information, while the other learner was providedwith visual and rather qualitative information of the con-cepts (cf. Fig. 2).

c and (b) visual learning material.

D. Bodemer / Computers in Human Behavior 27 (2011) 1079–1086 1083

(3) Knowledge test 1 consisted of eight multiple choice ques-tions, which can be classified into four different categories(cf. Fig. 3): (a) basic comprehension questions, referring tothe basic information given in both kinds of learning mate-rial, (b) visual comprehension questions, which requiredthe information of the visual learning material and thusthe interpretation of graphical components, (c) algebraiccomprehension questions, which required the informationof the algebraic learning material and hence the interpreta-tion of the formula components, and (d) transfer questions,which required the integration of both visual and algebraicinformation leading to more interrelated concepts.

(4) In the collaborative learning phase, dyads were providedwith a chat tool and an integration tool that comprised cor-responding algebraic and visual information. Learners had40 min to externally integrate the information and to gainan understanding of the statistics concepts. In this phase,the two types of integration tools were implemented. Onegroup was provided with the individual integration tool thatenables learners to interactively integrate algebraic andvisual components. The other group was provided with thecollaborative integration tool that additionally visualized thecurrent state of integration of the learning partner.

(5) Finally, participants took knowledge test 2, which consistedof 32 multiple choice questions. While the first test com-prised two questions of each type, the second test comprised

Fig. 3. Examples of the four di

eight questions each. In addition, learners reported theamount of mental effort perceived during the collaborationphase on a seven-point Likert scale (1 = very low, 7 = veryhigh mental effort), similar to the scale used by Paas, vanMerriënboer, and Adam (1994). While this measure provedto be reliable, it does not distinguish extraneous from ger-mane cognitive load.

2.2. Results

2.2.1. Learning outcome and cognitive loadWith regard to knowledge test 1 (cf. Table 1), a two-tailed t-test

revealed a significant difference between the two experimentalgroups (t(38) = �2.11, p < .05). In the condition individual supportonly, a higher level of knowledge was measured. This differencewas unexpected, as no experimental variation had taken place be-fore. However, performing correlations and analyses of covarianceshowed no significant influence of this difference on any otherdependent variable (except for learning gain, which comprisesknowledge test 1).

Regarding knowledge test 2 (cf. Table 1), a one-tailed t-test re-vealed no significant effect (t(38) = 0.28, p = .389): Learners per-formed nearly equal in knowledge test 2 no matter if they wereprovided with group awareness support or not.

In order to take the unexpected (pre-) knowledge differenceinto account, another t-test regarding the learning gain (difference

fferent types of questions.

Table 1Knowledge test performance.

Individual and groupsupport

Individual supportonly

Overall

M SD M SD M SD

Knowledge test 1 .35 .12 .44 .16 .40 .15Knowledge test 2 .61 .17 .61 .18 .61 .17

Table 3Test performances as a function of the representational code of the learning material.

Visual learning material Algebraic learning material Overall

M SD M SD M SD

Knowledge test 1Visual .45 .22 .20 .25 .33 .27Algebraic .60 .21 .60 .21 .60 .20

Learning gainVisual .17 .36 .35 .22 .26 .31Algebraic .03 .32 .03 .31 .03 .31

1084 D. Bodemer / Computers in Human Behavior 27 (2011) 1079–1086

between the relative learning outcomes of both knowledge tests)was performed. It revealed a marginally significant effect(t(38) = 1.32, p = .097, one-tailed). As expected, the individuallearning gain was, on average, higher with the collaborative inte-gration tool than with individual support only.

Further analyses regarding the types of questions showed thatthe benefit of the collaborative integration tool appeared particu-larly in the most demanding category of the knowledge tests thatrequired the integration of visual and algebraic information:transfer knowledge (t(38) = 1.89, p < .05, one-tailed). Regardingthe other three types of questions, one-tailed t-tests revealed nosignificant difference: basic knowledge (t(38) = 0.54, p = .291),visual knowledge (t(38) = 0.54, p = .291), algebraic knowledge(t(38) = 0.19, p = .426). Table 2 summarizes the means and stan-dard deviations (relative frequencies) for learning gains.

In this context, another aspect to look at is the influence of pro-viding learners with differently coded learning materials on thelearners’ performances. It was hypothesized that learners providedwith visual material (learning phase 1) outperform learners pro-vided with algebraic material in answering visual test questions(knowledge test 1) and gain more algebraic knowledge during col-laboration (learning phase 2). Vice versa, it was hypothesized thatstudying algebraic material lead to better algebraic performancesin knowledge test 1 and better visual learning gains during collab-oration. One-tailed t-tests revealed the hypothesized effects for vi-sual test questions (knowledge test 1: t(38) = 3.32; p < .01; learninggain: t(38) = 1.91; p < .05) but not for algebraic test questions(knowledge test 1: t(38) = 0.00; p = .500; learning gain: t(38) = -.063; p = .525). Algebraic test questions were solved on a very highlevel even before learners were able to collaborate (see Table 3 formeans and standard deviations).

With regard to cognitive load imposed by the two conditions(regardless of whether the load was beneficial or detrimental tolearning), as expected, no significant difference occurred(t(38) = -1.16, p = .126). Learners perceived the same mental effortwith (M = 2.35, SD = 1.60) and without (M = 2.95, SD = 1.67) groupawareness support. Likewise, the representational format of thelearning material did not affect the perceived mental effort(t(38) = 0.57, p = .286; algebraic: M = 2.80, SD = 1.77; visual:M = 2.50, SD = 1.54).

Table 4Number of assignments.

2.2.2. Group outcome and processes of collaborationIn order to gain information regarding the dyads’ performance

and first insights into the learners’ collaboration processes, log filesof the learners’ external integration behavior during the collabora-

Table 2Learning gain regarding the different types of questions (relative frequencies).

Individual and group support Individual support only Overall

M SD M SD M SD

Basic .36 .33 .30 .39 .33 .36Visual .29 .35 .23 .27 .26 .31Algebraic .04 .28 .02 .35 .03 .31Transfer .36 .42 .12 .42 .24 .44Overall .27 .22 .17 .24 .21 .23

tive learning phase were analyzed. As previously mentioned thecase of unequal assignments is particularly relevant for learningand can be differentiated into two sub cases: conflicting assign-ments (XY) and partial assignments (XO). Regarding the numberof unequal assignments appearing during the collaborative learn-ing phase (that persisted for at least one minute), a two-tailed t-test showed no significant difference between the two experimen-tal groups for neither the number of conflicting assignments(t(18) = �0.45, p = .331) nor the number of partial assignments(t(18) = 0.28, p = .390). Hence, as expected, in both conditions asimilar amount of unequal assignments appeared.

With regard to the final assignments at the end of the collabo-rative learning phase, a one-tailed t-test revealed more conflictingfinal assignments (XY) in the group with individual support onlythan in the group with individual and collaborative support(t(18) = �2.21, p < .05). Regarding the frequency of partial assign-ments (XO) at the end of the collaborative learning phase, therewas a marginally significant effect in the same direction(t(18)=�1.55, p = .068). Hence, although the average number of un-equal (conflicting and partial) assignments was similar in bothconditions, learners in the collaborative integration conditionrather agreed at the end of the collaboration phase. In general, thissupports the expectations of a beneficial effect of the collaborativeintegration tool on learning-relevant collaboration processes.

This assumption is also supported by the results regarding thecorrectness of assignments at the end of the collaborative learningphase, revealing a marginally significant difference between thetwo groups (t(18) = 1.40, p = .089). Thus, learners provided withthe collaborative integration tool to some extent achieved morecorrect final assignments than those provided with individual sup-port only. Means and standard deviations for the different types ofassignments are displayed in Table 4.

In order to shed some light on the aforementioned results, par-ticularly, on the question why more unequal assignments are re-solved if learners were provided with group awareness support, apreliminary analysis of the chat data was performed. It identifiedwhether unequal assignments were mentioned in conversationor not and calculated the ratio of the number of discussed to theoverall number of unequal assignments. A descriptive comparison

Individual andgroup support

Individualsupport only

Overall

M SD M SD M SD

During collaborationConflicting assignments (XY) 3.40 2.99 4.00 2.87 3.68 2.87Partial assignments (XO) 13.20 5.79 12.56 3.78 12.89 4.82

After collaborationConflicting final assignments (XY) .20 .63 2.60 3.37 1.40 2.66Partial final assignments (XO) .20 .42 1.20 1.99 .70 1.49Correct final assignments 11.05 2.49 8.75 4.55 10.22 3.59

D. Bodemer / Computers in Human Behavior 27 (2011) 1079–1086 1085

of the experimental groups showed that learners supported by thecollaborative integration tool discussed more conflicting (M = 0.51,SD = 0.44) and partial (M = 0.44, SD = 0.24) assignments than learn-ers with individual support only (conflicting: M = 0.30, SD = 0.37;partial: M = 0.34, SD = 0.26). However, this difference failed toreach statistical significance in one-tailed t-tests (conflicting:t(15) = 1.08; p = .150; partial: t(17) = .861; p = .201).

3. Discussion

This paper reports on the development and experimental eval-uation of a group awareness tool that is intended to support collab-orative multimedia learning. The tool is based on an externalintegration task that proved in earlier studies to facilitate the indi-vidual mental interrelation and integration of multiple representa-tions. The newly-developed collaborative integration toolenhanced the original version by visualizing the learning partner’sassignments of representations as indicators for the partner’s con-ceptual assumptions or knowledge.

An experimental study compared the collaborative integrationtool to the individual version. It was hypothesized that the collab-orative integration tool supports both individual and collaborativelearning processes to a greater extent during collaborative learningwith multiple external representations. Thus, higher individuallearning outcomes as well as more beneficial collaborative learningprocesses were expected.

While the results of a post knowledge test could not supportthis assumption, the analysis of the learning gains – especiallyregarding transfer knowledge, which required the integration ofvisual and algebraic information – revealed better performanceswhen learners were provided with the visualization of the part-ner’s knowledge. The benefits of group awareness support on thelearning process were analyzed by the number of unequal assign-ments during and at the end of the collaborative learning phase.Different assignments by the learners, as well as assignments thathad been performed by only one of the learning partners, wereboth resolved to a much greater extent during the collaborationprocess if learners were aware of the inequality of assignments.Moreover, more correct final assignments occurred in the collabo-rative integration group.

With regard to the assumptions about the support of collabora-tive discussion processes, the analysis of unequal assignments asindicators for the emergence, discussion, and resolving of concep-tual conflicts can give some insight into the way a collaborativeintegration tool might tacitly structure a learning discourse. For in-stance, the results suggest that learners talked in a more meaning-ful way about conflicting issues if they were externally providedwith information on conflicting knowledge distribution. However,in order to gain a deeper understanding in the functionality of thetool, the reported variables have to be complemented by furtherquantitative and qualitative analyses.

Therefore, currently, a set of categories and rules are defined inorder to systematically analyze the learning discourse. The analyt-ical categories are based on the Rainbow framework (Baker,Andriessen, Lund, van Amelsvoort, & Quignard, 2007) and are par-ticularly intended to identify meaningful processes of interactiveknowledge elaboration. Applying the preliminary categories to se-lected contrasting cases of the sample, it suggests that learnerswith group awareness support were more involved in meaningfuldiscussions and spent less time for grounding and modeling pro-cesses. Moreover, it seems that learners adapted their discussionbehavior to their awareness of knowledge distribution (i.e., talkingabout perceived conflicting perspectives in a more interactiveway), which might emphasize the benefit of visualizing cognitiveconflicts.

While the reported experimental study revealed some promis-ing findings regarding simultaneously supporting individual andcollaborative learning processes during collaborative multimedialearning, there are still open questions that should be addressedin further studies. One question concerns the assumptions of col-laborative aspects of cognitive load that need to be verified by sys-tematically investigating components of the tool that reduce orenhance mental effort. Another issue addresses individual differ-ences that could be adaptively considered by the tool. For example,visualizing cognitive conflicts might only guide learners who aretolerant for complexity and ambiguity (e.g., McLain, 1993).

An experimental aspect that is partly considered in currentlyconducted follow-up studies is the comparison of the collaborativeintegration scenario presented in this paper with other theoreti-cally and practically relevant scenarios of collaborative informa-tion integration. For instance, a common CSCL-scenariocomprises the use of one shared representation both learners canmodify. With regard to collaborative integration, this implies thatboth learners have to agree on a single assignment. Such type oftool might look more easy to use but does not consider individuallearning processes during collaboration and does not providelearners with group awareness information on an individual level.Another common scenario is using multimedia learning materialface-to-face instead of spatially distributed. However, until re-cently, a collaborative integration was difficult to implement forone shared computer as it requires the simultaneous identificationof different learner interactions. Latest technological developmentsallow for simultaneously touching a shared screen and for differen-tiating user interactions. On this technological basis, it is possibleto implement collaborative integration in face-to-face settingsand thus to better compare it to other mostly face-to-face-basedresearch on collaborative multimedia learning.

General insight on group awareness tools could be gained byexperimentally disentangling the guidance mechanisms underly-ing the collaborative integration tool. Some representational guid-ance (Suthers and Hundhausen, 2003) might be given by the emptydrop areas that provide a perceptually salient cue for missingknowledge. Further guidance might be initiated as a result of thevisualized assumptions or knowledge of both learners that canbe interpreted by the learners beyond representational salienceon an informational level. In many group awareness tools bothguiding principles structure computer-supported collaborativelearning processes implicitly. Frequently, both guiding principlesinteract inasmuch as salient representational features direct thelearners’ attention to relevant informational aspects.

The reported tool used these principles in order to tacitly guidecollaborative learning activities, thereby saving cognitive capaci-ties for meaningful individual and collaborative processes. This isespecially important during collaborative learning of complex con-cepts with multimedia learning material but can also be helpful inmany other areas of computer-supported collaborative learning.

References

Ainsworth, S. (1999). The functions of multiple representations. Computers andEducation, 33, 131–152.

Ainsworth, S. (2006). DeFT: A conceptual framework for considering learning withmultiple representations. Learning and Instruction, 16, 183–198.

Baker, M. J., Andriessen, J., Lund, K., van Amelsvoort, M., & Quignard, M. (2007).Rainbow: a framework for analysing computer-mediated pedagogical debates.International Journal of Computer-Supported Collaborative Learning, 2, 315–357.

Bodemer, D., & Buder, J. (2006). Supporting collaborative learning with augmentedgroup awareness tools. In R. Sun & N. Miyake (Eds.), Proceedings of the Twenty-Eighth Annual Conference of the Cognitive Science Society (pp. 77–82). Mahwah,NJ: Lawrence Erlbaum.

Bodemer, D., & Dehler, J. (2011). Group awareness in CSCL environments. Computersin Human Behavior, 27, 1043–1045.

Bodemer, D., & Faust, U. (2006). External and mental referencing of multiplerepresentations. Computers in Human Behavior, 22, 27–42.

1086 D. Bodemer / Computers in Human Behavior 27 (2011) 1079–1086

Bodemer, D., Ploetzner, R., Bruchmüller, K., & Häcker, S. (2005). Supporting learningwith interactive multimedia through active integration of representations.Instructional Science, 33, 73–95.

Bodemer, D., Ploetzner, R., Feuerlein, I., & Spada, H. (2004). The active integration ofinformation during learning with dynamic and interactive visualizations.Learning and Instruction, 14, 325–341.

Bromme, R., Hesse, F. W., & Spada, H. (2005). Barriers, biases and opportunities ofcommunication and cooperation with computers: Introduction and overview.In R. Bromme, F. W. Hesse, & H. Spada (Eds.), Barriers and biases in computer-mediated knowledge communication – and how they may be overcome (pp. 1–14).New York: Springer.

Buder, J. (2007). Net-based knowledge communication in groups. Zeitschrift fürPsychologie/Journal of Psychology, 21, 209–217.

Buder, J., & Bodemer, D. (2008). Supporting controversial CSCL discussions withaugmented group awareness tools. International Journal of Computer-SupportedCollaborative Learning, 3, 123–139.

Carroll, J. M., Rosson, M. B., Farooq, U., & Xiao, L. (2009). Beyond being aware.Information and Organization, 19, 162–185.

Chandler, P., & Sweller, J. (1991). Cognitive load theory and the format ofinstruction. Cognition and Instruction, 8, 293–332.

Clark, H. H. & Brennan, S. E. (1991). Grounding in communication. In L. B. Resnick, J.Levine, & S. D. Teasley (Eds.), Perspectives on socially shared cognition (pp. 127–149). Washington: APA.

Dehler, J., Bodemer, D., Buder, J., & Hesse, F. W. (2011). Guiding knowledgecommunication in CSCL via group knowledge awareness. Computers in HumanBehavior, 27, 1068–1078.

Dehler, J., Bodemer, D., Buder, J., & Hesse, F. W. (2009). Providing group knowledgeawareness in computer-supported collaborative learning: Insights into learningmechanisms. Research and Practice in Technology Enhanced Learning, 4, 111–132.

Dillenbourg, P. (2006). The solo/duo gap. Computers in Human Behavior, 22,155–159.

Dillenbourg, P., & Bétrancourt, M. (2006). Collaboration Load. In J. Elen & R. E. Clark(Eds.), Handling Complexity in Learning Environments: Theory and Research(pp. 142–163). Amsterdam: Elsevier.

Doise, W., & Mugny, G. (1978). Socio-cognitive conflict and structure of individualand collective performances. European Journal of Social Psychology, 8, 181–192.

Gutwin, C., & Greenberg, S. (2002). A descriptive framework of workspace awarenessfor real-time groupware. Computer Supported Cooperative Work, 11, 411–446.

Kalyuga, S. (2008). Relative effectiveness of animated and static diagrams: An effectof learner prior knowledge. Computers in Human Behavior, 24, 852–861.

Kirschner, F., Paas, F., & Kirschner, P. A. (2009). A cognitive-load approach tocollaborative learning: United brains for complex tasks. Educational PsychologyReview, 21, 31–42.

Kozma, R. B. (2000). The use of multiple representations and the social constructionof understanding in chemistry. In M. Jacobson & R. B. Kozma (Eds.), Innovations

in science and mathematics education: Advanced designs for technologies oflearning (pp. 11–46). Mahwah, NJ: Erlbaum.

Kozma, R. B. (2003). The material features of multiple representations and theircognitive and social affordances for science understanding. Learning andInstruction, 13, 205–226.

Lowry, N., & Johnson, D. W. (1981). Effects of controversy on epistemic curiosity,aehievement, and attitudes. Journal of Social Psychology, 115, 31–43.

Mayer, R. E. (2001). Multimedia Learning. NY: Cambridge University Press.McLain, D. L. (1993). The MSTAT-I: A new measure of an individual’s tolerance for

ambiguity. Educational and Psychological Measurement, 53, 183–189.Paas, F., van Merriënboer, J. J. G., & Adam, J. J. (1994). Measurement of cognitive load

in instructional research. Perceptual and Motor Skills, 79, 419–430.Ploetzner, R., Fehse, E., Kneser, C., & Spada, H. (1999). Learning to relate

qualitative and quantitative problem representations in a model-basedsetting for collaborative problem solving. The Journal of the LearningSciences, 8, 177–214.

Rebetez, C., Bétrancourt, M., Sangin, M., & Dillenbourg, P. (in press). Learning fromanimation enabled by collaboration. Instructional Science, doi:10.1007/s11251-009-9117-6.

Roschelle, J., & Teasley, S. D. (1995). The construction of shared knowledge incollaborative problem solving. In C. O’Malley (Ed.), Computer SupportedCollaborative Learning (pp. 69–97). Berlin: Springer.

Saab, N., van Joolingen, W. R., & van Hout-Wolters, B. H. A. M. (2005).Communication in collaborative discovery learning. British Journal ofEducational Psychology, 75, 603–621.

Sangin, M., Dillenbourg, P., Rebetez, C., Bétrancourt, M., & Molinari, G. (2008). Theeffects of animations on verbal interaction in computer supported collaborativelearning. Journal of Computer Assisted Learning, 24, 394–406.

Scardamalia, M., & Bereiter, C. (1994). Computer support for knowledge-buildingcommunities. Journal of the Learning Sciences, 3, 265–283.

Schnotz, W. (1999). Visual learning with new technologies: Introduction. EuropeanJournal of Psychology of Education, 14, 163–165.

Suthers, D., & Hundhausen, C. (2003). An experimental study of the effects ofrepresentational guidance on collaborative learning processes. Journal of theLearning Sciences, 12, 183–219.

Sweller, J., van Merriënboer, J. J. G., & Paas, F. (1998). Cognitive architecture andinstructional design. Educational Psychology Review, 10, 251–296.

Vahey, P., Enyedy, N., & Gifford, B. (2000). The probability inquiry environment:Learning probability through the use of a collaborative, inquiry-basedsimulation environment. Journal of Interactive Learning Research, 11, 51–84.

Xiao, L. (2008). The effects of awareness of group members’ rationale oncollaborative learning activities. In C. Chinn, G. Erkens, & S. Puntambekar(Eds.), Proceedings of the 7th Computer Supported Collaborative LearningConference (pp. 874–876). New Brunswick: International Society of theLearning Sciences, Inc.