There is good empirical evidence to show that the external representations that learners look at and interact with have a strong influence on what they come to understand. However, there is little agreement about why these so called representational effects occur and researchers from different traditions offer radically different interpretations. For example, explanations in empirical studies on learning with educational representations, depending on the theoretical stance, focus on; effects of attention and motivation (e.g. justifying the use of vivid animations), individual differences in learning style and experience (e.g. visualisers/ verbalisers), instructional design and multimedia principles (e.g. cognitive load explanations) and achieving fluency in a disciplinary discourse (stressing the importance of the affordances and constraints of each particular representation). The purpose of this symposium is to bring together researchers from different traditions (including semiotics, cognitive psychology, cognitive ergonomics and discourse analysis) and who apply different methodological paradigms (e.g. experimental studies, usability engineering, qualitative case study research) to explore whether a synthesis between these approaches is possible or even desirable. Each presenter will expose his or her theoretical approach, describe how it explains a variety of empirical results, report on remaining issues within the approach, and suggest new research that should be undertaken. The discussant will look for points of convergence and divergence between the approaches in order to assess progress towards a genuinely integrative approach, which would be of value to researchers from different traditions and whose application could guide educational practice.

Graphic representations accompanying text have a reliable influence on the organization and comprehension of the material students learn. The influence occurs with surprising regularity across a wide variety of subject matter and materials—from the standard textbooks used in schools to the graphical interfaces used for navigation on the web. Much of the research on the influence of graphics has been directed at explaining the higher order cognitive processes involved when students use graphics to learn. However, we believe that other properties are at work—particularly aesthetics—that dictate the degree to which graphics influence what students learn. In the present paper we report on the theory, methodology, and results of three investigations that systematically demonstrate the way aesthetic properties of graphics not only influence what students learn from accompanying text, but also prime the learners to retrieve idiosyncratic and affectively rich schemas from prior knowledge. The results from the three experiments provide strong support for the aesthetic property of graphics in arousing memories of learners during learning, and predicting how this level of arousal dictates what they learn.

Graphic representations accompanying text have a reliable influence on the organization and comprehension of the material students learn.  The influence occurs with surprising regularity across a wide variety of subject matter and materials—from the standard textbooks used in schools to the graphical interfaces used for navigation on the web.  Graphics are powerful visualization tools that influence, direct, and mediate the constructive processes learners deploy to make meaning during learning.

And yet, much of the research on the influence of graphics has been directed at explaining the higher order cognitive processes involved when students use graphics to learn. The influence has been explained in terms of 1) a “conjoint retention effect” in which the propositions of text are linked to specific graphic locations—the organization and comprehension of which are guided by the spatial properties of the graphic during recall  (Kulhavy and Stock, 1996); 2) a mental model derived from a dynamic interaction of inspection and evaluation between multiple levels of graphic-text representation, mediated by a learner’s prior knowledge as to whether inspection and evaluation operate bottom-up or top-down (Schnotz et al., 2002); and 3) an integrated representation comprised from a one-to-one mapping between graphic-text elements, actions, and relations in the visual and verbal representations of working memory (Mayer, 2003). The net effect of all three models is that graphics influence the material students learn because they are rapidly processed at multiple levels of analysis by a sophisticated cognitive system whose primary goal is to derive higher order meaning to ensure deep levels of comprehension.

However, we believe that other properties are at work that dictate the degree to which graphics influence what students learn.  In the present paper, we will build a case based on new research contending that graphics hold properties indigenous to them that operate significantly more primitively and powerfully than the higher order processes the models above reveal. That is, some properties of graphics (e.g. symmetry, color, aesthetics, and form) are processed more rapidly and decidedly because they operate at a much lower level of arousal— a level capable of releasing primitive responses linked to evolutionarily-based approach and avoidance mechanisms and the concomitant learning tendencies that accompany them (Ortony, Norman & Revelle, 2005). That is, objects that are beautiful are attractive and inviting, are evolutionarily speaking, important to be known; objects that are ugly, on the other hand, can be bad, unsafe, or dangerous and are to be avoided.  In short, aesthetically pleasing and displeasing objects stimulate judgments that are immediate, preconscious, and perceptually based—rapid judgments, we contend, that prime individuals for learning.

Here, we report on the theory, methodology, and results of three investigations that demonstrate the way aesthetic properties of graphics not only influence what students learn from accompanying text, but also prime the learners to retrieve idiosyncratic and affectively rich schemas from prior knowledge. Parenthetically, aesthetic quality is defined by the nature of the human response to a stimulus pattern whose structural properties give it positive intrinsic hedonic value (Berlyne, 1974).  Hedonic value refers to the pleasurable experience via contact with an object, independent of the access the object affords to other events with beneficial or noxious qualities. 

            Our work is guided by the theory of human response to beauty (Norman, 2004).  Norman proposed a tripartite system of levels to which individuals respond (e.g. visceral, behavioral, and reflective) that explains their initial preconscious, automatic attraction or repulsion toward objects formed on the basis of millennia of natural selection; the way they manipulate and interact with the physical properties of objects (if available) based on the way the objects work; and the way they derive meaning of objects via higher order interpretation and understanding from what they know.  The studies reported here support and elucidate the value of aesthetics for graphics in predicting what is learned relative to the first and third levels of Norman’s theory.

In experiment one, 132 university students read a 650-word passage concerning the nature of romantic relationships with the intention of learning as much from the passage as possible by writing an essay on relationships integrating what they learned with what they know from past experience.  Half of the propositions in the passage conveyed information about the decline and degradation of relationships (negatively valenced); the other half about a relationship’s healthy development and growth (positively valenced).  One third of the students read the passage in the presence of a color pictorial graphic depicting flowers in a vibrant state of growth; for another third the flowers were in a wilted state.  The final third viewed a picture of only the table present in the other two versions (control).  The results revealed that learners remembered no more positive or negative propositions from the passage based on the graphic they viewed.  However, learners added significantly more positively valenced propositions from prior knowledge, when shown both pictures of the flowers, relative to the control.  Focus groups conducted on 20% of participants from each group revealed that the graphics of both flowers were highly aesthetically pleasing relative to the control.

In experiment 2, the same graphics were rated by 49 students for aesthetic value and valence on 30 descriptors. Results revealed no significant differences. Two additional graphics were constructed to control for valence, leaving aesthetic value alone, these were significantly different on all 30 indices of valence but equivalent on aesthetics (growing, dying, control). The results were identical to experiment 1.

In experiment 3, two new graphics were added, the same in valence with their partner graphics in experiment 2 but rated significantly less aesthetically pleasing. Results revealed that learners integrated significantly more positively valenced propositions from the passage when graphics were aesthetically pleasing than when they were not, regardless of their valence.  Additionally, more propositions from prior knowledge were used in the presence of aesthetically appealing graphics than unappealing ones.

Overall, these results provide strong support for the aesthetic property of graphics in arousing memories of learners during learning, and predicting how this level of arousal dictates what they learn.

From a cognitive point of view learning with external representations can be seen as the construction of mental representations of subject topics, which are represented by the external representation. Thus, the question arises how this knowledge construction process works in detail. Currently, there are two prominent models for learning with external representations, which initiated a large amount of empirical work: (1) the cognitive theory of multimedia learning (CTML) and (2) the Cognitive Load Theory (CLT).

The CTML focuses on how different external representational formats like texts and pictures which address different sensory modalities are processed on different levels and on describing the resulting knowledge structures. Hence, it gives a framework for the processes and structures of knowledge acquisition.

The CLT on the other hand focuses on the question of how the limited capacity of working memory can be effectively used for these processes. Thus, it is a capacity theory. Beside these different focuses, both theoretical frameworks inspired empirical research, which aims at defining instructional conditions under which the construction of mental representations is effective without overloading learners’ limited cognitive capacity.

Unfortunately, the results of these studies have often been over-generalized and promoted as “thumb-rules”. To avoid these simplifications it is necessary to analyse more detailed the conditions of the learning situation, of the learner him- or herself and to gain deeper insight in the processes of knowledge acquisition on different levels of processing. In the last few years there is increasing interest in all of these three differentiating approaches: (1) by re-analyzing well-known instructional effects under experimentally varied conditions (2) by analyzing interaction effects between learners aptitudes and treatment conditions (ATI-studies) like prior knowledge or spatial abilities and (3) by collecting data within the learning process through thinking aloud protocols or dual task performance data.

From a cognitive point of view learning with external representations can be seen as the construction of mental representations of subject topics, which are represented by the external representation. Thus, the question arises how this knowledge construction process works in detail. Currently, there are two prominent models for learning with external representations, which initiated a large amount of empirical work: (1) the cognitive theory of multimedia learning (CTML) from Mayer (2001) and (2) the Cognitive Load Theory (CLT; Sweller, van Merrienboer & Paas, 1998).

Mayer’s CTML gives a framework for the relevant processes of learning with external representations as well as for the resulting mental representations. By selecting, organizing and integrating information from text and picture the learner constructs mental representations on different levels of processing: first on a base level where learners are able to recall but not to understand the represented subject topic, then – after a semantic analysis – on a deeper level by constructing a verbal or visual mental model of the subject topic. CTML has initiated a lot of empirical research which aimed at specifying the effects of different instructional forms on these different processing levels. One example for these effects is the well-known modality effect, which means that presenting an auditory text in combination with a picture leads to better learning outcomes than presenting a written text.

The CLT on the other hand focuses on the role of memory capacity within the process of learning with external representations. Three different types of cognitive load have been specified which in sum constitute the total amount of cognitive load a learner is experiencing while dealing with a specific learning task: (1) the intrinsic cognitive load, which is due to the complexity of the learning task, (2) the extraneous cognitive load, which derives from the instructional design and (3) the germane cognitive load, which means the amount of cognitive load a learner is investing for schema acquisition. In order to avoid cognitive overload and to enhance learning, most research based on CLT aimed at analyzing conditions for reducing the extraneous cognitive load (e.g. by studying the modality effect; e.g. Mayer, 2001) and for stimulating germane cognitive load (e.g. by fostering the integration process by means of prompts or instructional help, e.g. Seufert & Brünken, 2006).

Beside the different focuses of CTML and CLT, both theoretical frameworks inspired empirical research. This aims at defining instructional conditions, under which the construction of mental representations is effective without overloading learners’ limited cognitive capacity. Unfortunately, the results of these studies have often been over-generalized and promoted as “thumb-rules”. In the last few years there has been increasing interest in more differentiating approaches in order to avoid these simplifications:

- Re-analyzing instructional design effects under experimentally varied conditions.

An example for such a differentiated analysis could be demonstrated by Tabbers (2002). He could show that the modality effect only appears under system paced learning conditions. If the learner is free to choose the learning time of the material by himself (user-paced), the modality effect disappears. This result pattern demonstrates that the effect of instructional conditions can be compensated by the learners’ strategic learning behaviour and highlights the impact of the learners’ active role in knowledge acquisition.

- The impact of learners’ prerequisites

Moreover, in several studies, different researchers could demonstrate the impact of learners’ prerequisites on instructional design effects, particularly on learners’ prior domain specific knowledge (e.g. Seufert, Jänen & Brünken, in press). While learners with low prior knowledge do not profit from instructional support, in case of high prior knowledge, support can be ineffective by producing redundancy (expertise reversal effect; Kalyuga, Ayres, Chandler & Sweller, 2003). Only with a medium level of prior knowledge, learners seem to be sensitive for instructional support.

- Process-oriented data analyses

A third line of experimental research is concerned with the analysis of behavioural data related to processes of knowledge construction. While the analysis of thinking aloud protocols leads to better insights in strategic cognitive and metacognitive behaviour related to learning with external representations (Bannert, 2005), more basic cognitive research is using paradigms from experimental psychology, like dual-task-analysis (Brünken, Plass & Leutner, 2003) and eye-tracking analysis (Folker & Ritter, 2003) to attain empirical evidence for basic aspects of information processing like the allocation of cognitive resources to specific representations.

Based on a brief review of current empirical research, the presentation will focus on the gains and demands of learning with external representations from a cognitive point of view. With a focus on the three approaches for differentiated analyses we will point out the impact of learners’ prerequisites, strategic learning behaviour and working memory capacity on the effective use of external representations.

References

Bannert, M. (2005). Explorationsstudie zum spontanen metakognitiven Strategie-Einsatz in hypermedialen Lernumgebungen. In C. Artelt & B. Moschner (Hrsg.). Lernstrategien und Metakognition: Implikationen für Forschung und Praxis (S. 127-151). Münster: Waxmann.

Brünken, R., Plass, J. L., & Leutner, D. (2003). Direct measurement of cognitive load in multimedia learning. Educational Psychologist, 38, 53-61.

Folker, S. and Ritter, H. (2003). Integration of Multimodal Information - An Eye-Tracking Study. In F. Schmalhofer and R. Young and G. Katz (Hrsg.). Proceedings of the European Cognitive Science Conference 2003 (p. 348). Mahwah/N. J., London: Lawrence Erlbaum Associates.

Kalyuga, S., Ayres, P., Chandler, P., & Sweller, J. (2003). The Expertise Reversal Effect. Educational Psychologist, 38, 23-31.

Mayer, R. E. (2001). Multimedia Learning. Cambridge, UK: Cambridge University Press.

Seufert, T. & Brünken, R. (2006). Cognitive load and the format of instructional aids for coherence formation. Applied Cognitive Psychology, 20, 321-331.

Seufert, T., Jänen, I. & Brünken R. (in press). The Impact of Intrinsic cognitive load on the effectiveness of graphical help for coherence formation. Computers in Human Behavior.

Sweller, J., van Merrienboer, J., & Paas, F. (1998). Cognitive architecture and instructional design. Educational Psychology Review, 10(3), 251-296.

Recently, in addition to the term “cognitive tools”, the term “semiotic tools” has been coined referring to mediating role of the computer in learning. Both terms refer to the affordances of computers for the interpretation, construction, and manipulation of symbolic structures presented on screen. The two terms nevertheless rest on two distinct theoretical perspectives on representations. The prevailing cognitive view can be described as dyadic in two ways: both internal and external representations represent objects and phenomena in the real world. The alternative semiotic view can be described as triadic: upon some sense impression, an idea is evoked in the mind of an individual that corresponds to an entity in the real world. In this contribution, we briefly present both perspectives and examine two implications of introducing the triadic perspective in studies of learning with external representations. The first implication involves taking into account intra-modality variations in external representations, i.e. essentially different ways of representing the same subject matter within a representational category. The second implication involves establishing learners’ prior knowledge of existing and emerging representational formats, in addition to prior domain knowledge, in learning research.

Recently, in addition to the term “cognitive tools”, the term “semiotic tools” has been coined referring to mediating role of the computer in learning. Both terms refer to the affordances of computers for the interpretation, construction, and manipulation of symbolic structures presented on screen. The two terms nevertheless rest on two distinct theoretical perspectives on representations. In this contribution, we briefly present both perspectives and examine two implications of introducing the triadic perspective in studies of learning with external representations.

Cognitive views of representation are fundamentally dyadic (see for instance Palmer 1978). According to such a view, a representation is “something that stands for something else”. Thus, internal mental representations (propositions, mental images and models) stand for objects, states, and events in the world. In interactive learning environments, the term external or educational representations has been coined (Ainsworth, 2006) to designate, roughly speaking, any configuration of inscriptions (or sounds) on a computer screen (or other medium) that has been designed by teachers or programmers to represent the concepts, relations and phenomena in a content domain, and that allows learners to interact with it by interpreting, exploring, constructing and manipulating. Both dyadic views are combined in models of multimedia learning, such as those of Mayer (2001) and Schnotz (2001), which postulate two “routes” for constructing internal representations from external ones, a visual and a verbal track leading up to imagistic and propositional internal representations. Note that the distinction between the two types rests on the relation between the representing world (be it internal or external) and the represented (real) world.

Semiotic views on representation are mostly triadic (see for instance Eco, 1988). The classic “semiotic triangle” distinguishes between the signifier, i.e. a material form (a mark or sound) and the signified, i.e. the idea created in someone’s mind. The third pole is the referent, an entity existing in the world. For example, the word /dog/ or the image of a dog (signifiers) evoke the idea “horse” (signified) in the mind of an English speaking individual even in the absence of a real dog (referent). A number of different classifications of signs have been elaborated, of which we name three. First, the modality classification includes the five senses instead of only two: sight (= visual), hearing (=verbal), touch, smell, and taste. Second, a classification based on the type of mapping between signifier and referent has three categories instead of only two: iconic-resemblance (=visual), symbolic-arbitrary (=verbal), and indexical-contiguity relations (Peirce, 1931-1958). A third classification is based on relations between signifier and signified: denotation, connotation and meta-language (Barthes, 1964). For example, a word such as /dog/, depending on the context, may evoke the concept of “dog” (denotation), but also evoke “loyal” or “friend” (connotation).

            One could, in principle, superpose dyadic and triadic views: signifiers are external and signified are internal representations of the referent in the real world. However, this would be an extreme simplification with respect to the philosophical and epistemological traditions underlying them. We identify two main points of interest and implications of the triadic perspective in research on learning with external representations.

            The first point is the dynamic nature of the interpretation process. Whereas research in the cognitive tradition uses the iconic versus symbolic distinction as a basis for designing experimental conditions prior to learning, the semiotic view implies classifying a particular inscription as an icon, a symbol, or an index only as a result of an individual interpretation process. Moreover, semiotics explicitly detaches the two levels of signifiers and signified. One signifier can evoke a number of different signified. Conversely, a number of different signifiers can be used to evoke the same signified. Consider the choice for a rectangle to represent a cylinder in an animation of a bicycle pump (such as those of Mayer, 2001). The cylinder could have been represented using a number of other graphical elements such as domain-specific ones in geometry and technology. Conversely, a rectangle has a number of possible alternative interpretations, e.g. a box, a block, a location, or a label. One direction for research would be to control for such variations and this involves taking into account the fact that there are multiple ways of representing within a category (be it visual or verbal).

            The second point is what we propose to call the “representational paradox”: in order to account for learning by means of interpreting and constructing external representations, we need to postulate learners’ prior knowledge of the content domain, the representational format, or both. For example, it is practically impossible to learn anything from a diagram, such as a structural diagram of a molecule, a flowchart of a manufacturing process, a concept map of the life cycle, or a graphical model of a dynamic system, if one does not already knows either the concepts in those domains or the type of diagram used. The implication for research is the importance of investigating learners’ prior knowledge of a large variety of traditional and modern representational formats (in addition to the standard testing of prior domain knowledge).

Both points, in addition to suggesting future research, also provide potential alternative explanations of findings in multimedia learning research (or account for failures to find significant effects). We will give examples of both in our presentation.

References

Eco, U. (1988). Le signe, histoire et analyse d’un concept. (J.-M. Klinkenberg, Trad.). Bruxelles : Editions Labor. (Edition originale, 1973).

Mayer, R. E. (2001). Multimedia Learning. New York : Cambridge University Press.

Palmer, S. E. (1978). Fundamental aspects of cognitive representation. In E. Rosch & B. B. Lloyd (Eds.), Cognition and categorization (pp. 259-203). Hillsdale, NJ: Lawrence Erlbaum Associates.

Peirce, C. S. (1931-1958). Collected Papers (CP) of Charles Sanders Peirce, eds. C. Hartshorne, P. Weiss (Vols. 1-6) and A. Burks (Vols. 7-8). Cambridge : Harvard University Press.

Schnotz, W. (2001). Sign systems, technologies and the acquisition of knowledge. In J. F. Rouet, J. J. Levonen, & A. Biardeau (Eds.), Multimedia learning : Cognitive and instructional issues (pp. 9-30). Oxford : Elsevier.

How do representations mediate learning of disciplinary concepts?

In this paper we explore student learning in terms of becoming fluent in a disciplinary discourse. We define disciplinary discourse as the complex of representations, tools and activities of a discipline, and discuss how these components are related by a disciplinary order of discourse. In our study, physics undergraduates from two Swedish universities were interviewed about their learning experiences in lectures, using a stimulated recall approach. Since we videoed lectures and took field notes, our data best illustrates the representations aspect of disciplinary discourse, i.e. written and oral text, tables, graphs, diagrams, pictures, equations, computer animations, etc.

Naturally, we found instances where students were not ‘fluent’ in the representations of disciplinary discourse. One example of this can be seen in Fig. 1.

Insert figure 1 here

Fig. 1.  Diagram of a transformer drawn by the lecturer on the whiteboard

The majority of students in our study considered this well-known representation of a transformer to be unproblematic. One student, however, claimed not to have ever understood what was being represented. The student knew the various parts (coils, iron core, currents etc.) but not what the diagram represented as a whole.

We also identified situations where students were seemingly ‘fluent’ in a number of representations, but had still not appropriately experienced the associated disciplinary concept. For example, Maxwell’s equations could be used to calculate correct answers but  the meaning of these representations was unclear to students.

Referring to the phenomenological concepts of appresentation and relevance structure, and the observation that different representations have different affordances, our analysis suggests that fluency in a critical constellation of representations may be a necessary condition for meaningful access to disciplinary concepts. The pedagogical implications of this approach are briefly discussed.

Lemke (1998) argues that scientists handle problems that would otherwise be impossible to solve by orchestrating movement between a wide range of discursive resources. Drawing on this, we characterise student learning in terms of achieving fluency in a disciplinary discourse (defining this discourse as the complex of representations, tools and activities of a discipline).

Following Fairclough (1995), the New London Group (2000:20) argue that each semiotic domain has its own ‘grammar’ or order of discourse—that is “a structured set of conventions associated with semiotic activity …”. From here we constituted the notion of discursive fluency—the ability to use a particular representation in a legitimate way with respect to a given concept, i.e. in line with the disciplinary “order of discourse”.

The study examines the feasibility of characterising learning through representations in terms of becoming fluent in a disciplinary discourse with the following research question: How does fluency in the representations of disciplinary discourse relate to learning of disciplinary concepts?

Methodology

Six physics lectures with different lecturers at two Swedish universities were videotaped. Guided by our interest in sampling as many of the representations of disciplinary discourse as possible, the resulting video footage was edited down to eight short segments for each resulting interview. These segments always included whiteboard diagrams and mathematical formula discussions. Twenty-two volunteer students were then interviewed using a semi-structured interview protocol. Students were first asked to talk about learning physics through different representations. The video segments were then used to create a stimulated recall environment (Bloom, 1953; Calderhead, 1981).

Findings

We identified a number of cases where students were not fluent in certain disciplinary representations.

Fig. 1. Diagram of a transformer drawn by the lecturer on the whiteboard

This well-known representation of a transformer was seen as unproblematic by the majority of students in our study. However, during stimulated recall, one student claimed not to understand the diagram:

Interviewer:      What did you think about this part?

Student:            Um. I don’t know what this is. I didn’t know what he was writing…

Interviewer:      Okay, he’s drawing some kind of diagram, but you don’t really know what that is that he’s drawing…?

Student:           No.

Interviewer:      Okay, so…

Student:            —And I think it’s, it’s, quite often like that in the lectures—that he’s drawing something on the whiteboard and he assumes that we know this from before.

Interviewer:      So er, you’ve got no idea what this transformer thing is?

Student:           [laughing] No.

Interestingly, this student identified the various parts (coils, iron core, currents etc.) but not what the diagram represented as a whole, suggesting that fluency in a particular representation may be a necessary condition for gaining access to a disciplinary concept.

We also documented instances where students were fluent in representations of disciplinary discourse, but still without an experience of the corresponding concept.

Interviewer:      You’ve seen these equations before..?

Student:            Yeah I’ve seen them before er… but I really don’t know exactly what they mean [laughs].

Interviewer:      Can you tell me what this means to you?

[curl E=0]

Student:            Um, I think the E is the intensity of an electric field. And then the curl of E…[quietly to herself] equals zero…

                        Er, I think this is a conservative vector field—and I know how to calculate it but I don’t know what it means.

We suggest that this student displays fluency in two representations, the equation and the corresponding oral text “conservative vector field”. In fact without the expression of incomprehension we may have assumed that the student had experienced the disciplinary concept. We suggest that fluency in a number disciplinary representations may still prove insufficient for experiencing a disciplinary concept.

Theoretical and educational significance

So, how does fluency in the representations of disciplinary discourse relate to learning of disciplinary concepts?

Our analysis suggests that in order to experience a concept appropriately through a given disciplinary representation, we need to spontaneously generate other relevant representations, which we are fluent in (c.f. Marton and Booth’s (1997) discussion of the phenomenological ideas of appresentation, and relevance structure). At the same time our analysis also shows that fluency in a number of representations is necessary but not sufficient. Why is this the case? Kress et al. (2001) and Duval (2006) discuss the way in which representations have different affordances or, in more tangible terms, different possibilities for representing disciplinary concepts. Constituting all the above leads us to the suggestion that coming to appropriately experience disciplinary concepts depends on fluency in a critical constellation of representations.

Based on this depiction, we argue that multi-representational science teaching has the distinct potential to lead to improved learning outcomes compared with teaching with a reduced number of representations. Hence there is a need for teachers’ practice to be informed about which constellation of representations best opens up the possibility for learning. Moreover, students need opportunities to become fluent in these representations. Without this, we argue that there is a risk that teaching may not develop the necessary fluency in the constellation of representations needed to make disciplinary knowledge accessible to students.

Since the beginning of information and communication technology, many computer programs for learning were designed and distributed to schools and universities. But sadly, many of them are not used. Consequently, it is both interesting and useful to evaluate and to improve these computer programs before trying to distribute them. Ergonomic approaches attempt to fulfil this type of goal: evaluation in order to improve. But these approaches are very often confused with usability evaluation. Confusing usability approaches and ergonomic approaches can lead to a deadlock. We conducted different studies in the evaluation of computer programs, containing a variety of external representations for learning. These studies show that a designed artefact can be usable but not used, and vice versa. That is the reason why we propose an ergonomic framework where the quality of an artefact is defined by its utility, usability and acceptability. We present data, from different empirical studies, which allow us to say that these three dimensions are equally important and interrelate. We also found that these relationships are not always similar. Then, we defend an inductive approach to the analysis of the relationships between utility, usability and acceptability.

Since the beginning of information and communication technology, many computer programs were designed and distributed to schools and universities. But many of them are not used. Consequently, it is both interesting and useful to evaluate and to improve these computer programs before trying to distribute them. Ergonomic approaches attempt to fulfil this type of goal: evaluation in order to improve. But these approaches are very often confused with usability evaluation. Confusing usability approaches with ergonomics approaches can lead to a deadlock.

Amiel et al. (2002) analyzed an interactive learning environment (ILE) designed for the professional training of the maintenance men in the car industry, but which was rarely used by the 400 trainers. The ILEs did not have overwhelming defects in terms of utility and usability. The problem was linked to the organization of time, space and work, which were not compatible with the use of the environment. Galy et al. (2006) analyzed an ILE, based on external representations, used in pre-school by 72 teachers. They observed that the teachers do not link usability with ILE’s acceptance. Acceptance was influenced by perceived utility and motivation.

These studies and others conducted in our lab have lead us to propose an ergonomic framework, where the quality of an artefact is defined by its utility, usability and acceptability. We propose to apply this framework to external representations designed for learning.

-         Utility is synonymous with relevance or efficacy. It answers the following question: Does the external representation allow the learner to reach his/her learning goal? It can be evaluated by learning measures, with tasks like recall, problem solving, and so on.

-         Usability refers to ease of use. It can be evaluated with criteria like learnability, efficiency, memorability, less errors, satisfaction (Nielsen, 1994).

-         Acceptance addresses the learners’ desire to use the external representation. It answers the following question: is the external representation compatible with a learner’s motivation, affect, culture, values, and also does it fit within the constraints of the context.

We also found that the relationships between those three dimensions are variable.

These data contradict other authors’ point of view. For Nielsen (1994) acceptability has two dimensions: practical and social. Usability and utility (i.e. the goal to which the device is designed to reach) are two dimensions of the usefulness (i.e. the goal which the device really allows to reach). Usefulness is itself an underused dimension of practical acceptability. The Nielsen’s model has been criticized. In this model, acceptability cannot have effect on usability or utility. Dillon and Morris (1996) note that in this model, usability, which is however in the main concern of Nielsen and the majority of appraisers, (1) is necessary but not sufficient (2) makes it possible nothing to say about the use. According to Dillon and Morris’ model  (inspired by Davis' Technology Acceptance Model), it is necessary to introduce the concept of attitude of the user (perceptions, affects) to be able to include/understand the relations between utility, usability and acceptability. In particular, these authors insist on the utility and usability perception. These perceptions can be very different between them and very different from effective usability or usefulness: I can believe that this external representation will be easy to use but little useful for me, whereas it is actually difficult to use but useful for the development of my competences.

Then, we defend an inductive approach of the analysis of the relationships between utility, usability and acceptability: how the relationships between three series of tests measuring the utility, acceptability and usability should be interpreted? Let us imagine the case of an ILE that, following a set of tests, appears acceptable but unusable and useless. How do we know what it is necessary to improve? Can we believe that the improvement of usability will involve the improvement of utility? Should the utility also be improved? Or can we expect that the improvement of the utility will involve usability improvement? These questions concern the rationality problem, i.e. the problem of the relations between means and goals (Searle, 2001). Working out a rationality model is working out a model of the relationships between means and goals. But working out such a model does not guarantee the value of this model. A rational criterion in a field can be irrelevant in another field (Tricot et al., 1999). That is why we prefer an empirical approach to rationality. An observed relation between attained goal and invested means can be said to be rational. We consider moreover that for each ILE, there is a mental representation (individual or collective) of this device (e.g. of its physical attributes) and of its rationality. This representation is constrained by the context (i.e. by the organization of time, space, work, knowledge in the school). We call acceptability the value of this representation. We defend the following position whatever the ILE to evaluate, and with whatever relation between usability, utility and acceptability.

What we propose to interpret is thus a set of results corresponding to an analysis with three dimensions: usability, utility, acceptability. Each dimension is a variable, on which one can realize measurements independent of the two other variables. Each user has a value on each of the variables and can be represented as a dot in a three-dimensional space. The form of a set of points corresponding to N learners allows to describe the relationships between the three dimensions.

Amiel A., et al. (2002). Acceptabilité de Form@lion : Évaluation et recommandations, rapport CERFI, IUFM de Midi-Pyrénées.

Dillon A., Morris M., (1996). User acceptance of information technology: theories and models, Annual Review of Information Science and Technology, 3-32.

Galy, E. et al. (2006). An investigation of factors impacting teachers’ acceptance of technological tool in preschool. ERT “Hypermedia and Learning” report, Toulouse.

Nielsen J. (1994). Usability engineering, Boston, Academic Press.

Searle J. (2001). Rationality in action, Cambridge, MIT Press.