In: Proceedings of IGC2000: First International Workshop on Interactive Graphical Communication. London, August 30-31, 2000 (Working Papers in Computer Science, WP-00-01) Queen Mary, University of London, pp.77-84.
Graphical Interaction and the Emergence of Abstraction Patrick G.T. 2Healey1 (
[email protected]) Nik Swoboda (
[email protected]) Ichiro Umata33 (
[email protected]) Yasu Katagiri (
[email protected]) 1
Department of Computer Science, Queen Mary and Westfield College, University of London, UK 2
Department of Computer Science, Indiana University, Bloomington, US.
3
Media Integration and Communications Laboratories, ATR International, Kyoto Japan Abstract
The relative abstractness of a representation is typically considered to depend on both the nature and availability of underlying domain regularities. A further possible source of influence derives from the medium used in constructing a representation and whether it facilitates the production of fluent, detailed representations or constrains individuals to use more primitive or restricted forms. This paper reports on an experimental investigation of abstraction in graphical representation using a task in which participants were asked to produce drawings of pieces of music. The results indicate that an additional source of constraints, the use of representations in interaction, has a critical influence on abstraction and, for this task at least, is a more significant source of constraints than either domain structure or medium of representation. The findings parallel existing results for linguistic coordination and indicate that explanations of representational structure need to attend directly their communicative use.
Background Early empirical investigations of the structure of graphical representations attended to the influence of both individual and communicative factors as potential influences on form. For example, Bartlett's seminal work investigated the relative influence of both cognitive and social processes in accounting for patterns of schematisation and conventionalisation in graphical representations (Bartlett, 1932). Contemporary research in cognitive science has focused much more on individual, cognitive factors influencing graphical form. Perhaps the most common intuition is that the relative degree of abstraction displayed by representation is a function of the efficiency with which it captures underlying domain structure. It thus depends both on the availability of underlying regularities or structures in the domain to be represented and on the effects of cognitive or perceptual limitations associated with recognising and representing those regularities. Some authors have focussed on the properties of graphical representations as efficient means of reducing memory load or of constraining representations in a way that facilitates reasoning about a domain (e.g., Stenning and Oberlander, 1995). Other authors, such as Tversky (1981, 1989, 1995), have focussed on individual cognitive factors, such as conceptual limitations, schemata, and perceptual processes in normalising and conventionalising the form of graphical representations. A further influence, also sometimes cited as affecting processes of schematisation and abstraction, derives from the properties of different representational media. For example, clay discourages the fluent use of detailed graphical forms, favouring simplified, reduced symbols or scripts instead. The properties of the medium can also interact with frequency of repetition so that simple, more readily executed, patterns will tend to be preserved at the expense of ones that demand more effort, thus also promoting a form abstraction (cf. Tversky, 1995). A focus on cognitive/computational factors that influence representational form and degree of abstraction overlooks what we take to be the primary function of graphical and linguistic representations; their use in communication. The use of external representations as a means of coordinating the activities of different individuals is, we claim, at least as important as their use as external memory or problem solving aids. Evidence is accruing to indicate that significant constraints on representational form derive from distinctively interactional or communicative factors. Garrod and Doherty (1994) showed that, in task-oriented dialogues, choice between a number of distinct spatial description types is strongly influenced by patterns of interaction between individuals performing the task. The description types used in this task vary from figurative, instance-based, descriptions that exploit perceptual salience to relatively abstract,
systematic, coordinate based descriptions. Intuitively, coordinate descriptions encode an efficient, generalisable, abstract representation of the underlying structure of this task domain. However, building on Garrod's and Doherty's findings, Healey (1997, in prep) provides evidence that the sustained use of the more abstract coordinate descriptions depends on interactional constraints that are independent of simple considerations of efficiency or accuracy of the representation. Additionally, although the use of the more abstract descriptions increases with experience, the stability of this choice depends on a shared history of interactions within a subgroup, not individual expertise. If individuals who share a high level of experience, but do not share a common history of sub-group interactions, perform the task together they are much less likely to use the more abstract descriptions. Schwarz (1995) has also provided evidence of communicative constraints on abstraction in graphical representations. He studied the differences between the problem solving activities of pairs and individuals. He found that although no clear differences in task performance could be established between individuals working alone and individuals working as members of a dyad there were consistent and reliable differences in the forms of representation they produced. Dyads produced abstract problem representations, such as matrices and graphs, significantly more often than individuals working alone. The present paper investigates further the role of interactional influences on representational form and reports two experiments that manipulate the influence of domain structure, media type and interaction on the form of representations. For the present study we had three basic requirements for our task. Firstly that the task domain should be one for which there are a number of basic underlying regularities which could be captured in a representation. Secondly, we were concerned to avoid domains for which there are strong pre-existing representational conventions. Existing experimental studies have focussed almost exclusively on investigations of graphical representations that have a standard interpretation. Experimental tasks and materials have employed representations such as Euler circles, circuit diagrams and program flow charts which have specific intended interpretations and which individuals must learn to read in the intended way (cf. Scaife and Rogers, 1996). Our focus on the development of abstraction motivates the selection of a domain for which participants can, in principle, deploy a variety of possible representations and which does not encourage them to suppose there is a particular ‘correct’ representation. Thirdly, in order to investigate the possible parallels with linguistic dialogue tasks we wanted to develop a communicative task which could be accomplished by exclusively graphical means. The basic requirements were satisfied through the development of a novel graphical communication task. The task involves drawing pictures of pieces of music with the aim of producing a representation which a second person can use, independently, to identify the piece drawn. The only constraint on the drawings produced is that they must not contain any letters of numbers. This satisfies the requirement for a highly structured task domain because of the potentially available elements of musical form e.g., tempo, scale and mode. It also satisfies the requirement that participants must develop and try out new representational strategies rather than relying on pre-existing conventions.
Experiment 1 Experiment one was designed to investigate the effects of domain structure on representational form. On each trial one participant, the Giver, draws a picture of a 30second piece of music. The other member of the pair, the Follower, sees the Giver’s drawing developing on the whiteboard in front of them but is not able to contribute to it themselves. Participants are free to draw anything they like, subject to the restriction than no letters or numbers should be used. The Follower’s task is to determine on the basis of this picture, which of two pieces; the target and a distractor, corresponds to the original drawn by the Giver. Playback and selection of the target and distractor pieces, controlled by buttons at the top of the screen, is self-paced. Followers are asked to make their choice as quickly and as accurately as possible and if a 2minute time limit expired further input is blocked and a dialogue window appears to prompt a final choice. After each selection both participants were given feedback about whether their choice had been correct. This pattern was repeated for 24 trials with the roles of Giver and Follower alternating.
Method: 24 participants were recruited from local universities and divided into 12 pairs. They worked in separate soundproof rooms with a shared virtual whiteboard providing the only connection between them. The whiteboard was displayed on two LCD tablets (combined graphics tablet and screen) connected to two, networked, Macintosh G3’s. The whiteboard, written specifically for this task, consisted of a large white drawing area, a strip palette of eight colours and a set of buttons at the top. All input to the screen was by stylus and lines could be erased by using the reverse end of the stylus. Materials: 36 pieces of music were chosen; 18 jazz pieces and 18 classical pieces. Each piece was a 30second clip of piano instrumental, each by a different composer or artist, avoiding easily recognisable pieces. The distinction between jazz and classical was based on the criteria that the jazz pieces included some improvisation and non-diatonic chord progressions or tones whereas the classical pieces did not.1 Design: The independent variables were; 1. Target type: Jazz vs. Classical, and 2. Discrimination type: Cross Genre (Jazz-Classical) vs. within genre (Classical-Classical and Jazz-Jazz). Two quantitative dependent measures were se1
We note that many contemporary ‘classical’ pieces also use non-diatonic chord progressions or tones. However, to ensure as clear a distinction as possible no pieces of this kind were included in the current study.
lected; response accuracy, measured as proportion correct, and response time, measured as the time elapsed in milliseconds before a choice was made. Music was assigned according to the following constraints: no individual hears the same piece of music twice, (as target or distractor), each piece of music occurs equally often as target and distractor, choice of pieces is counterbalanced across pairs and conditions, and order of presentation is randomised for each pair.
Results: Although there was some initial hesitation, participants found the task intelligible, engaging and enjoyable. They were able to perform consistently above chance getting, on average, 68% correct (t(44) = 6.38, p (two tailed) = 0.00.). To test whether people became faster with experience the total of 24 trials were divided into four stages consisting of trials 1-6, 7-12, 13-18, 19-24 respectively. The average response time for each pair, for each stage of the experiment, was entered in an analysis of variance with a factor of experience with 4 levels corresponding to the four quarters. There was no reliable overall effect of experience (omnibus F (3,44) = 0.77, p = 0.51) and no evidence of a decline in response times across trials (linear trend t(44) =1.35, p=0.09). To evaluate effects of musical form, the average proportion of correct responses for each pair, in each of the three music conditions, were entered in a 2 factor analysis of variance: Target type (Classical vs. Jazz) crossed with Discrimination type (same or different genre). There was no simple main effect of either target type (F(1,11) = 2.10, p = 0.17) or discrimination type (F(1,11) = 2.10, p = 0.66) on accuracy and no interaction (F(1,11) = 0.58, p = 0.46). The corresponding analysis for average response times also showed no simple main effect of either Target type (F(1,11) =0.84, p = 0.38) or discrimination type (F(1,11) = 2.92, p = 0.12) and no interaction (F(1,11) = 2.87, p = 0.12). Inspection of the drawings suggested that emotional affect of the pieces, represented as, for example, smiling vs. sad faces, was an important aspect of the representations produced. It was hypothesised that this might depend on the mode (Major or Minor of the pieces. To test whether this aspect of musical form was influencing performance the mode of each target piece and distractor was coded. However, an analysis of variance with Target type (Major/Minor) crossed with Discrimination type (Same/Different mode) again showed no evidence of a main effect on proportion correct of either target type (F(1,11) =0.32, p=0.58) or discrimination type (F(1,11) =2.87, p=0.12) and no interaction. The analysis for average response times also provided no evidence of an effect of either target type (F(1,11)=1.312, p=0.28) or discrimination type (F(1,11)=0.87,p=0.37) and no interaction.
Figure 1: Example Abstract Drawing Drawing Types: Informal inspection indicated that the drawings produced in the experiment fell into two general categories. The first, Abstract, involves some representation of musical form, e.g., intensity, pitch, rhythm or tempo typically represented as a contour as illustrated in Figure 1. The second, Figurative, was more heterogeneous involving some depiction of faces, figures, objects or situations as illustrated in Figure 2. A third, smaller category of Mixed drawings was also noted in which some abstract and figurative elements had been combined. Two of the authors independently coded 287 drawings as either abstract, figurative or mixed and showed very high inter-judge reliability of coding (Kappa = 0.9, N = 287, k= 2). It was initially hypothesised that the manipulation of musical form might be reflected in the use of Abstract or Figurative drawings. For example, Jazz is often considered a more abstract form than classical music and, informally, listeners sometimes suggest that Jazz is, in some sense, more difficult to apprehend. On these grounds we might predict that Jazz pieces would be associated with Abstract representations more often than Classical pieces. However, as table 1 indicates, there was no pattern in the distribution of drawing types according to the genre of the
pieces. A one-dimensional Chi2 shows no reliable difference in the distribution of Abstract and Figurative drawings across Jazz and Classical pieces Chi(1)= 0.45, p =0.50. The musical form of the piece being depicted does not appear to influence the form of the drawing produced Table 1: Frequency of Drawing Types According to Musical Form
Classical Jazz
Abstract 37 42
Drawing Type: Figurative 94 89
Mixed 12 12
Figure 2: Example Figurative Drawing To provide an initial test for effects of communicative or interactional context, entrainment scores were calculated for each pair following the method in Garrod and Doherty (1994). This is calculated as the number of drawings produced by an individual that are of the same type (Abstract, Figurative or Mixed) as the immediately preceding drawing produced by their partner. As a result it indexes the degree of coordination or matching of description type between the members of a pair. The average score for pairs was 0.71 whereas the chance level of matching (calculated as the sum of the squared proportions of each graphic type) was 0.49. Members of a pair were thus more likely to be coordinated in their choice of graphic than would be expected by chance (t(11)=4.03, p (2 tailed) = 0.00).
Discussion: If the difficulty of the representational task depends on the availability of the underlying domain structure then it would be expected that the manipulation of musical form should affect the difficulty of the task. However, the results of experiment 1 suggest no effect of either musical genre or mode on the effectiveness with which pairs could perform the communication task. The task performance measures of speed and accuracy indicate that both of the difficulty of producing a drawing, as assessed by the effects of target type, and the difficulty of distinguishing between pieces, as assessed by the effects of discrimination type, are unaffected by these aspects of musical form. Of course, it may be the case that some other aspect of the form of the pieces was influencing performance and this requires further investigation. Nonetheless, it is surprising to note that the manipulation of genre did not alter performance as expected. Perhaps more striking are the findings with respect to the form of the drawings produced. Although they can be reliably classified into abstract and figurative drawing types, the distribution of these types also shows no effect of musical form. Jazz and classical pieces are equally likely to be drawn in a Figurative or Abstract style. There was, however, a reliable effect of communicative context, with choice of representation type clearly influenced by partner's choice. Pairs tend to produce matching drawing types much more frequently than would be expected by chance. This parallels Garrod's and Anderson's (1987) finding that choice of description type in task-oriented dialogues depends more on the type of description being produced by an interlocutor than on the structure of what is being described. The findings of experiment one provide initial evidence for the importance of factors relating to communicative or interactive context in influencing the distribution and use of more abstract representational forms. To investigate this further a second experiment was carried out in which the task was altered to allow for greater levels of direct interaction.
Experiment 2 In experiment one participants were unable to interact directly on the whiteboard, they could observe the construction of the drawings but were not in a position to provide feedback to one another during the construction. The only source of feedback they received was at the indication, at the end of a trial, whether the last drawing had been correctly identified. The alternation of roles between Giver and Follower also provided a limited opportunity for modifying or ammending the choice of representation, for example, by adopting the style used by their partner in the last trial. In experiment two the task was modified to make it more genuinely interactive. Both participants could draw and erase freely on the shared whiteboard at any time. Two versions of the interactive experimental task were actually used. In the 'matching' version both members of a pair had one piece of piano music each and the task was for them to determine whether these pieces were the same or different. In the 'discrimination' version each member of the pair had two pieces of music and the task was for them to decide which of the two pieces was the same. Detailed discussion of this task manipulation is deferred for another occasion. The analysis reported below is, however, based on data from both tasks in order to preserve the overall balance of conditions and materials. Experiment two was designed to investigate two specific hypotheses. Firstly, whether increasing the level of interaction relative to experiment one would promote greater representational abstraction. Secondly, to test the prediction that a medium which tended to discourage fluent production of graphics would promote greater use of abstraction. In order to address the second hypothesis, on half the trials participants used the same stylus based interaction as experiment 1 in which they draw directly onto the screen, in the other half they used a mouse, reducing fluency of movement and introducing a spatial separation between input and the screen. It was predicted that the mouse based interaction would favour more reduced, abstract , forms.
Method: Excepting the changes in interactivity and media, essentially the same method as experiment 1 was followed. 24 participants, (16 male and 8 female, average age 19) were recruited from a variety of disciplines at local colleges and universities. They were paid an honorarium for taking part. Design: Experiment 2 employed a within-subjects, factorial design with task (Matching vs. Discrimination) crossed with Media (Mouse vs. Stylus). Selection of music was constrained so that the combinations of form (Jazz, Classical) and Mode (Major vs. Minor) were counterbalanced across conditions. Each piece was also classified according to its tempo with selection of tempo randomised across conditions. As before, no one heard the same piece of music twice. To provide enough combinations of materials a total of 112, 20 second, piano solo pieces were used. This design resulted in a total of 68 trials per pair with order of conditions and materials counterbalanced.
Results: The effects of media and experience on task performance were assessed in two analyses of variance. To index experience, the 68 trials were divided into four blocks. For each quarter of the experiment, and each pair, the proportion of correct responses and the average time to respond were calculated. Analysis of variance on the proportion of correct responses, with media and experience as within subjects factors, showed a reliable main effect of experience (F(3,24) = 4.84, p=0.01) but no effect of media (F(1,24) = 0.04, p=0.84) and no interactions. Linear trend analysis confirmed that participants became more accurate with experience (Omnibus F(1,33) = 12.12, t(33) = 3.48, p(one tailed) = 0.00) . The parallel analysis for time to respond also showed a main effect of experience (F(3,47) = 4.07, p=0.18) and again, no effect of media (F(1,47) = 0.01, p=0.92). Linear trend analysis confirmed that participants were becoming faster at the task with experience (Omnibus F(1,33) = 5.69, t(33) = 2.38, p(one tailed) = 0.01). Overall, the results suggest that the media used in the production of the representation has no effect of participants' ability to carry out the task. In contrast to experiment one, however, they show a reliable improvement in performance across trials although this may be attributable to both the larger number of trials in experiment two as well as the possibility of direct interaction. Drawing Types: To allow for independent classification of the drawings produced by each member of a pair, their whiteboard input was separated into two files consisting only of what each individual participant had drawn. As before, the pictures were classified as Abstract, Mixed, or Figurative. An additional category of 'None' was introduced to deal with a small number of cases (3%) where one or both of the partners had not drawn a picture of a piece on a given trial. The distribution of each drawing type across all trials is given in table 2. A strong test of the prediction that the manipulation of media should affect distribution of drawing types was assessed by scoring, for each pair, the proportion of drawings that were classified as Abstract. This was analysed in an analysis of variance with medium, task and experience as within subjects factors. There was no simple main effect of media (F(1,24) = 2.22, p= 0.15) and no reliable interactions. There was a marginal overall effect of experience (Omnibus F(3,24) = 2.84, p= 0.06) and a Linear trend analysis showed that there was a reliable increase in the proportion of Abstract drawings produced with experience (Omnibus F (1,33) = 5.69, t(33) = 2.38, p(one tailed) = 0.01). Although choice of representation type is affected by task experience, the results indicate that the manipulation of media did not have the predicted effect on participants’ production of abstract representations.
Table 2: Distribution of Drawing types, Experiment 2.
Frequency Proportion
Abstract 970 59%
Drawing Type: Figurative 345 21%
Mixed 257 16%
Entrainment scores were again calculated to provide an indication of the extent to which the members of a pair were coordinating their choice of representation type. In this case scores were calculated as the proportion of trials in which drawings of the same type were produced, excluding trials in which one or both participants produced no drawing. The average entrainment score was 0.79, reliably above the chance level of 0.42 (t(11) = 6.75, p = 0.00).
Comparison of Experiments 1 and 2 To test for effects of the manipulation of interactivity additional analyses were performed directly comparing the data from experiments 1 and 2. For these contrasts only data from the first 12 trials of experiment 2 were used. This was in order to balance levels of task expertise; each participant having completed approximately 12 drawings in each experiment. To compare level of coordination in choice of drawing type a t-test was performed on the average entrainment scores for each pair with experiment (1 vs. 2) as a between subjects factor. This indicated that levels of matching were not reliably different (t(22) = 0.74, p (2 tailed) = 0.46). The ability to interact directly did not appear to have any effect on the extent to which pairs tended to match their choice of representation type.
70 60 Percentage of Drawings
Experiment 1
Experiment 2
50 40 30 20 10 0 Abstract
Figurative
Mixed
Drawing Type Figure 3: Choice of Drawing Types in Experiments 1 and 2 Although degree of matching did not differ between experiments 1 and 2, markedly different patterns of choice in drawing types were observed. As figure 3 illustrates, during the first 12 trials of experiment 2 far more abstract and far fewer figurative drawings were produced than in experiment 1. The contrast in relative frequency of Abstract and Figurative drawings confirmed the reliability of this pattern (Chi2(1) = 50.7, p = 0.00). The results show that the ability to interact directly has a substantial effect on the distribution of drawing types leading to a much greater use of abstract drawings.
Discussion Considered together, the results of experiments one and two provide clear evidence of the influence of interactional or communicative constraints on representational form. Support for this claim derives firstly from the observation that partner's choice of drawing is a better predictor of distribution of drawing type than perhaps the most intuitively salient domain characteristics; genre and mode. Although, in order to be effective, a representation must address itself in some non-arbitrary way to domain structure, the specific aspects of domain structure tested here underdetermine choice of drawing type. As noted, this parallels findings for linguistic representations. Garrod and Anderson (1987) argue that, while domain properties may favour some types of verbal description over others, the main factor is the pressure
to coordinate with an interlocutor as a means of sustaining the coherence of an interaction. Of course, it may be that other domain characteristics would have an effect on representation type and this issue would require further empirical work to resolve. Partners choice was also a better predictor of drawing type than drawing medium. Intuitively, we might expect the simpler contours of Abstract drawings to be favoured where drawings have to be completed using a mouse. The reduced dexterity of mouse movements compared to a stylus and the separation of mouse movements and screen activity could be expected to increase the difficulty of executing a drawing. Nonetheless, no evidence was found for an influence of media on drawing type. It may be that the difference in ease of execution between a mouse and a stylus, while marked, is in itself insufficient to affect representational style. However, on the basis of the present study it seems that other pressures to maintain a particular choice outweigh the effects of media. The most compelling evidence for the importance of constraints deriving from interaction comes from the contrast between experiments one and two. The comparison of the first twelve drawings produced in each case indicates that degree of abstraction deployed in the representations is highly sensitive to the level of interaction that is possible. In experiment two participants made much greater use of the Abstract drawing types than in experiment one. It is notable that this occurs even though the level of entrainment or matching is not reliably different in the two experiments. Although partner's choice is a better predictor of representation type than either media or musical form it doesn’t account for this shift in the distribution of drawing types between the two experiments. The interpretation of these results raises two questions; what does the difference between Abstract and Figurative drawings consist in and what are the specific interactive constraints that have such a strong influence on their distribution? Detailed discussion of the notion of abstraction merits a paper in itself (see e.g., Swoboda, in prep). However, it is useful for current purposes to distinguish between a 'perceptual perplexity' notion of abstraction and an 'underlying structure' notion. The perceptual perplexity notion reflects the 'iconicity' or 'resemblance' of representations. The 'underlying structure' notion of abstraction refers to the isolation of invariant or generalisable structure that covers a number of instances. Although highly correlated in practice, these two notions are logically independent. The prediction that media type should affect degree of abstraction only clearly applies to abstraction in the former sense. Where more effort is required to produce a picture the overall quality and recognisability of the representation may suffer, particularly with repetition. It is much less clear why the manipulation of media type should affect abstraction in the latter sense; the accuracy or effectiveness with which domain structures can be captured. In the current study, the Abstract pictures depend primarily on abstraction of the latter kind. The contours of melody or intensity or tempo represent underlying musical structures and, unlike the Figurative drawings, each type of Abstract drawing can generalise to any piece of music. There is always an intensity contour for every piece of music but there is not always a street scene or landscape. It does not seem plausible to claim that participants in experiment one were more or less sensitive to these possible structures in the music than those in experiment two. We suggest that it is the difficulty in coordinating their interpretations of which domain regularities are being represented and at what level of granularity that accounts for the importance of direct interaction. The successful use of a contour depends, amongst other things, on coordinating on the ontology it implies. For example, the dimensions it aims to capture (e.g, intensity, melody, tempo) and the granularity of the representation (e.g., a few notes, a few bars, the whole piece). This problem is much less acute for Figurative drawings which do not invoke an underlying domain structure. They take advantage of ad hoc associations (such as a café bar for a Jazz piece) more or less specific to each instance rather than systematic generalisations. The present results indicate that it is direct interaction that is critical to the use of abstract drawing types (cf. Healey, 1997, in prep). In experiment one, participants have only the interchange of roles and the feedback from each trial to make inferences about what possible system of representation is being used. In experiment 2 they can promote much greater coordination by, for example, circling, underlining or modifying parts each other's representations, using lines, between two contours to indicate possible alternative alignments. These phenomena were frequently observed with some interactions involving such rapid and fluid exchanges that they appeared almost conversational in character. It appears that this ability to interact during the production of a representation is critical to sustaining the use of representations that abstract, in the sense of generalise, across instances. The implication of this is that an effective representation is one which, first and foremost reflects the structure of interaction, and the kinds of coordination that it makes possible, and only secondarily reflects the structure of the domain.
Acknowledgments This work was generously supported by ATR Media Integration and Communications Laboratories. We would also like to thank Professor Rosemary Bailey for patient help with the statistical analyses.
References Bartlett, F. C. (1932) Remembering: A Study in Experimental and Social Psychology. CUP: Cambridge Garrod, S.C. and Anderson, A. (1987) Saying what you mean in dialogue: A study in conceptual and semantic coordination. Cognition, 27, 181-218. Garrod, S.C. and Doherty, G. (1994) Conversation, coordination and convention: An emprirical investigation of how groups establish linguistic conventions. Cognition, 53 181--215. Healey (in preparation). Semantic Coordination in Dialogue. Manuscript in preparation.
Healey, P. (1997) Expertise or expert-ese: The emergence of task-oriented sub-languages. Proceedings of the Ninteenth Annual Conference of The Cognitive Science Society. M.G. Shafto and Langley, P. (Eds.) August 7th10th, Stanford University, CA. pp. 301-306. Scaife, M. and Rogers, Y. (1996) External Cognition: How do graphical representations work? International Journal of Human-Computer Studies, 45, 185-213. Schwarz, D. L. (1995) The emergence of abstract representations in dyad problem solving. The Journal of the Learning Sciences. 4(3). 321-354. Stenning, K and Oberlander, J. (1995) A cognitive theory of graphical and linguistic reasoning: logic and implementation. Cognitive Science, vol. 19, 97-140. Swoboda, N. (in prep) Exploring the Notion of Abstraction for Diagrammatic Representations. Manuscript in preparation. Tversky, B. (1995) Cognitive Origins of Graphic Conventions. In F.T. Marchese (ed.) Understanding Images. New York: Springer-Verlag. pp.29-53. Tversky, B. (1989). Perceptual and conceptual factors in distortions in memory for graphs and maps. Journal of Experimental Psychology: General. vol.118, No.4, 387-398 Tversky, B. (1981). Distortions in memory for maps. Cognitive Psychology, 13, 407-433.
