A Comparison of Reasoning Processes in a Collaborative Modelling Environment: Learning about genetics problems using virtual chat

  • Published on

  • View

  • Download


This article was downloaded by: [Laurentian University]On: 11 October 2014, At: 02:55Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UKInternational Journal of ScienceEducationPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/tsed20A Comparison of Reasoning Processes ina Collaborative Modelling Environment:Learning about genetics problems usingvirtual chatKai Pata a & Tago Sarapuu aa University of Tartu , EstoniaPublished online: 23 Feb 2007.To cite this article: Kai Pata & Tago Sarapuu (2006) A Comparison of Reasoning Processes ina Collaborative Modelling Environment: Learning about genetics problems using virtual chat,International Journal of Science Education, 28:11, 1347-1368, DOI: 10.1080/09500690500438670To link to this article: http://dx.doi.org/10.1080/09500690500438670PLEASE SCROLL DOWN FOR ARTICLETaylor & Francis makes every effort to ensure the accuracy of all the information (theContent) contained in the publications on our platform. However, Taylor & Francis,our agents, and our licensors make no representations or warranties whatsoever as tothe accuracy, completeness, or suitability for any purpose of the Content. Any opinionsand views expressed in this publication are the opinions and views of the authors,and are not the views of or endorsed by Taylor & Francis. The accuracy of the Contentshould not be relied upon and should be independently verified with primary sourcesof information. Taylor and Francis shall not be liable for any losses, actions, claims,proceedings, demands, costs, expenses, damages, and other liabilities whatsoeveror howsoever caused arising directly or indirectly in connection with, in relation to orarising out of the use of the Content.This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden. Terms &Conditions of access and use can be found at http://www.tandfonline.com/page/terms-and-conditionshttp://www.tandfonline.com/loi/tsed20http://www.tandfonline.com/action/showCitFormats?doi=10.1080/09500690500438670http://dx.doi.org/10.1080/09500690500438670http://www.tandfonline.com/page/terms-and-conditionshttp://www.tandfonline.com/page/terms-and-conditionsInternational Journal of Science EducationVol. 28, No. 11, 15 September 2006, pp. 13471368ISSN 0950-0693 (print)/ISSN 1464-5289 (online)/06/11134722 2006 Taylor & FrancisDOI: 10.1080/09500690500438670RESEARCH REPORTA Comparison of Reasoning Processes in a Collaborative Modelling Environment: Learning about genetics problems using virtual chatKai Pata* and Tago SarapuuUniversity of Tartu, EstoniaTaylor and Francis LtdTSED_A_143850.sgm10.1080/09500690500438670International Journal of Science Education0950-0693 print/1464-5289 onlineOriginal Article2006Taylor & Francis000002006KaiPatakpata@ut.eeThis study investigated the possible activation of different types of model-based reasoningprocesses in two learning settings, and the influence of various terms of reasoning on the learnersproblem representation development. Changes in 53 students problem representations aboutgenetic issue were analysed while they worked with different modelling tools in a synchronousnetwork-based environment. The discussion log-files were used for the microgenetic analysis ofreasoning types. For studying the stages of students problem representation development, individ-ual pre-essays and post-essays and their utterances during two reasoning phases were used. Anapproach for mapping problem representations was developed. Characterizing the elements ofmental models and their reasoning level enabled the description of five hierarchical categories ofproblem representations. Learning in exploratory and experimental settings was registered as theshift towards more complex stages of problem representations in genetics. The effect of differenttypes of reasoning could be observed as the divergent development of problem representationswithin hierarchical categories.IntroductionTraditionally, reasoning with inductive arguments or reasoning with deductive argu-ments are considered the two modes of creating new knowledge in science (Lawson,2002). Besides these, the complex forms of model-based reasoning, which can berelated to inductive and deductive processes, have been brought to the focus inscience teaching (e.g., Gilbert, Boulter, & Elmer, 2000; Lee, 1999; Mellar & Bliss,1994; Nersessian, 1999; Stewart, Hafner, Johnson, & Finkel, 1992). The current*Corresponding author: Science Didactics Department, University of Tartu, 46 Vanmurse Street,Tartu, 51014 Estonia. Email: kpata@ut.eeDownloaded by [Laurentian University] at 02:55 11 October 2014 1348 K. Pata and T. Sarapuupaper deals with reasoning processes in the network-based synchronous collabora-tive learning environment with modelling aids: hypothetical-predictive inductivereasoning, model-based inductive reasoning, and model-based deductive reasoning.The influence of different model-based reasoning on the structure of problem repre-sentations in the genetics domain is studied.The Nature of Reasoning ProcessesOne theory that explains the general nature of reasoning processes is the theory ofmental models (Johnson-Laird, 1983; Johnson-Laird & Byrne, 1991). The theoryclaims that reasoning is a semantic process of model construction and manipulationin working memory. Johnson-Laird uses the three-stage processing schema forexplaining the reasoning mechanism. In the initial model-construction stage, themental model sets are constructed of separate entities, their properties, andstructural and causal relationships between them. In the conclusion-formulationstage, the inconsistent models are eliminated and consistent ones are joined informing the integrated model. At the final conclusion-validation stage, the personmust look for alternative models that might falsify the conclusion. Accordingly,peoples ability to think and solve problems in a domain depends on the quality ofunderlying mental models for that domain, which they are able to run (Gentner &Gentner, 1983). As several types of reasoning exist, it is also necessary to know whatthe influence of different types of reasoning processes on the development of mentalmodels might be.Several variations of inductive and deductive reasoning have been applied inscience lessons. In inductive reasoning, starting from maximally probable premisesand using correct inductive logic, one should arrive at most probable conclusions(Magnani & Nersessian, 2004). According to Romeyn (2004), the input for induc-tive reasoning includes a data set, usually consisting of observations, and possiblysome further assumptions; and the output may consist of predictions or generalstatements, where predictions concern unobserved singular states of affairs, andgeneral statements, such as empirical generalizations, concern universal states ofaffairs.Two main ways of practising the inductive reasoning in learning situations are thehypothetical-predictive approach (Lavoie, 1999) and the theory-supported approach(Wynne, Stewart, & Passimore, 2001). Hypothetical-predictive reasoning should bepreferable when starting new conceptual topics. According to Lavoie, hypothetical-predictive reasoning is an important thinking skill enhancing problem-solving,stimulating peer-group discussion and logical argumentation, increasing studentsmotivation, revealing prior knowledge, making students aware that a variety of opinionsare held by peers, and facilitating conceptual change. In hypothetical-predictive reason-ing, after observing some patterns, the students will be activating different knowledgefrom long-term memory, constructing separate sets of mental representations thatmight give hypothetical explanations to the phenomena. The final explanations willremain hypothetical and not supported by strong theory. In other cases, students gainDownloaded by [Laurentian University] at 02:55 11 October 2014 Comparison of Reasoning Processes 1349the domain-specific theoretical knowledge before observing the patterns and mustexplain the regularities by selecting from the set of conceptual mental models, whichare based on the previously learned theories. Mental models, not consistent with theobserved phenomena, are left aside or modified until the model that works as a rulefor explaining the pattern was found. The difference with the other inductive approachis that the final explanation will be theoretically sound.The deductive reasoning process is built upon the usage of connectives such asifthen, thus, and so on. Commonly, the method of deductive reasoning isbased upon the accepted theories and controlled data, permitting sound hypothesesto be formulated. True premises plus valid deductive reasoning should yield trueconclusions and the confirming of the theory (Magnani & Nersessian, 2004). Atschool, the deductive approach usually starts from approved theories, which will betested under certain circumstances. The first phase sees the mental processes thatcomprise the formation of the initial mental model, the hypothesis. Next, testingdifferent sets of options follows with some valid data set, which enables the forma-tion of an integrated understanding of the conditions under which the mental modelworks. Several authors (Klahr & Dunbar, 1988; Stewart et al., 1992) have arguedthat the strategies used by the subjects to solve problems could be described in termsof a simultaneous, mutually constraining search through experiment space consist-ing of a set of all possible experiments that might be run, and the hypothesis space,including all possible hypotheses of the solution of the problem. Finally, the alterna-tive hypotheses should be formulated and tested in order to find the contradictionsto the theory.In some cases, the hypotheses may be formed by inductive or hypothetical-predictive means, but tested with deductive methods. These, hypothetical-deductivemethods embody the reasoning with hypothetical entities that interact to produceemergent behaviour (Lehrer & Schauble, 2000). Lawson (2002) describes the hypo-thetical-deductive reasoning on the basis of the inventions of Galileo as the followingif/then/therefore pattern: making a puzzling observation, formulating a causalquestion, formulating one or more hypotheses, using a hypothesis and an imaginedtest to generate expected results/predictions, making actual observations andcomparing them with the expected observations, and drawing conclusions as to theextent to which the initial hypotheses have or have not been supported.Two Model-based Reasoning Processes in ScienceScience education should encourage students to develop insights about science as anintellectual activity (Stewart et al., 1992) that involves the development of sharedmeaning within the community by applying analogy, thought experiments, modelsand visual imagery in the concept, and theory formation by reasoning in order toovercome the working memory limitations (Nersessian, 1999). According to Lee(1999), model-based reasoning is the symbolic processing of an explicit representa-tion of the internal workings of the system (mental models) in order to predict,generate, and explain the resultant behaviour of the system, given details of itsDownloaded by [Laurentian University] at 02:55 11 October 2014 1350 K. Pata and T. Sarapuustructure and the behaviour of its components. Conceptual models serve as externalrepresentations for making interferences, finding characteristics, and structuring theproblem-solving (Reinhard, Hesse, Hron, & Picard, 1997). When composing themodel or when conducting inquiry with the model, different model-based reasoningprocesses may take place that might have an influence on the mental model develop-ment. Mellar and Bliss (1994) distinguish between qualitative, semi-qualitative, andquantitative modelling, which can be related to reasoning with the qualitative, soft,and hard models introduced by Lee (1999). In this study reasoning between qualita-tive and hard models was compared.Qualitative models capture only the fundamental aspects of the phenomena whilesuppressing much of the detail, thus facilitating the basic understanding andmanipulation of the problem. According to Lee (1999), such models are normallybased on strong theories that exploit the principles knowledge, rather than heuristicor empirical data. Qualitative models are often run only internally. An example ofturning inductive reasoning with qualitative models external is the process ofcomposing the schematic figure. Following Lee, hard models are formed on thebasis of reliable and confident data and focus on particularly well-defined problems.They are accurate and precise and well grounded by the strong and accepted domaintheories (Lee, 1999). Many computer-based inquiry models applied at schoolqualify in this category. Mellar and Bliss (1994) have described reasoning with hardmodels as quantitative modelling, which allows constructing models by use of alge-braic relations between variables.In this study, these two modelling methods were applied in computer-basedenvironments that have been differed in terms of their modelling capabilities asexpressive or exploratory (Chalk, 2000; Jimoyiannis & Komis, 2001). In one case,students were supported in the expressive environment (electronic whiteboard) withthe possibility to compose the schematic figure as the problem representation andcould perform inductive reasoning. Students were provided with initial elements ofthe model and had to relate them in order to explain the phenomenon. Roth (2001)assumed that the shared expressive modelling tools, such as argument graphs, mind-maps, and so on, on whiteboards can enhance learning through designarrangingdifferent elements of unique knowledge systems and providing knowledge artefacts.By representing knowledge about the phenomena in the format of visual designstructures, the latter can serve as anchors for grounding unclear matters. Instead ofthe traditional perspective of design, which requires that the learners should haveobtained all the principles of science before actually being engaged in design activity,this type of learning views design as a social process in which rational knowledgeserves as one of the resources, but is secondary in importance to the norms and prac-tices in a particular community (Roth, 2001).The construction of an external model on a whiteboard can enable the students torecognize the faults in their initial interpretation, activating the tunnel effect(Cornujols, Tiberghien, & Collet, 2000)the mechanism of model-based reason-ing with a single situation where the knowledge transfer between the conceptualdomains takes place. A tunnel effect occurs when a model becomes separatedDownloaded by [Laurentian University] at 02:55 11 October 2014 Comparison of Reasoning Processes 1351from its initial justifications in another domain, and upon reinterpretation revealssome inconsistencies of the model leading to the re-conceptualization of the modelin the target domain as well (Cornujols et al., 2000). This type of model-basedreasoning process can be seen as moving from an initial whole mental model to therevised model, while, every time an element or relationship is changed, the revisionsare made considering all the elements of the mental representation as a whole (Yu,2002). As the complete model is visually accessible for learners, it might enhance themental model revision processes in working memory and contribute to the formationof the integrated set of problem representations.In our case, the second approach of modelling enabled the usage of a conceptualinquiry modelthe exploratory learning environmentfor testing certain problemaspects with the valid data. By means of manipulating the model, the students couldobtain the direct feeling for the underlying structure of the represented content thatis often hard to achieve with conceptual reasoning alone (Cheng, 1993). Workingwith an exploratory environment, such as simulation, students could study the effectsof changing variable values on the model and could perform deductive reasoning.When students learn with hard models at school they often practise analogicaldeductive reasoning. They must study complex everyday problem situations usingspecialist-developed conceptual models with limited facilities. According to Cornu-jols et al. (2000) analogical reasoning involves the comparison of a well-knownsource case and the target case that must be understood, whereas interpretationtakes place in both the source and the target domains. The relationships between theentities of the mental model of a new problem must be predicted and tested oneafter another with the analogical conceptual model in hand. This type of modellingtends to break down the initial mental representations of the problem during theinvestigation, and presumes the operation with two separate models (Pata &Sarapuu, 2004). The model-based reasoning process during inquiry with modelscan be described as running the mental simulation of a conceptual model andmatching the results with the problem model.Representations during ReasoningIn the case of model-based reasoning in teams, the students must operate with threekinds of representations: internal mental models, external shared verbal and visualrepresentations (progress models), and conceptual visual representations (pedagogi-cal models). The peers can comprehend the structure of each others internal repre-sentations when these are made explicit by writing, talking, or visualizing during thejoint activity. In order to facilitate certain types of reasoning processes at school,educators commonly compose conceptual models with limited facilities. In thegenetics domain, several modelling environmentse.g., EVOLVE (Soderberg &Price, 2003) and BioLogica (Tsui & Treagust, 2003)have recently been developedfor educational purposes.Often, the models that convey theoretical knowledge and represent the complexunderstanding of the phenomena do not match the students internal representationalDownloaded by [Laurentian University] at 02:55 11 October 2014 1352 K. Pata and T. Sarapuuframework. Students must simultaneously operate with their own mental models ofthe phenomena and interpret the expressed progress models of their peers andthe pedagogical models during model-based reasoning (Pata & Sarapuu, 2004). Thenature of models (e.g., expressive, exploratory) applied during learning and theactivity design (e.g., inductive/deductive, theoretically sound/hypothetical-predictive) can be the factors influencing students model-based reasoning patterns,and thereby the development of their internal problem representations.The researchers can capture the structure of mental models at certain phases ofthe problem-solving activity when the students are asked to write down, verbalize, orvisualize their explanations. The discussion recordings enable one to follow thecurrent state of each students mental model at selected time intervals withoutintervention. Besides this, discussion transcripts could be used for re-establishing thereasoning process with mental models. Yet, it cannot be assumed that the wholemental model structure and the nature of reasoning could be traced with this meth-odology because people may not externalize all that is processed in their workingmemory.As the basis for the analysis system for mental models about complex naturalphenomena, three preliminary steps should be considered. First, it is necessary todetermine the common elements of representations for specific problem domains.According to Jonassen (1995), mental models possess representations of objects orevents in systems, and the structural relationships between those objects and events.Second, the representation level of these elements and their properties (see Chi,1997; Wilensky & Resnick, 1999) must be taken into account. Besides understand-ing the observable phenomena, genetics reasoning used in problem-solving has to bebuilt upon understanding of biological sub-cellular processes underlying theconcepts and their relations (see Tsui & Treagust, 2003; Wynne et al., 2001).Depending on students domain knowledge, these two levels of emergence can berepresented in mental models using either concrete or abstract terminology, or both.Students can express themselves on an everyday reasoning level by describing enti-ties with analogous pictures or narratives, or they can refer to them using scientificterminology and symbols (Pata & Sarapuu, 2003).Third, the inter-relations between the elements must be determined not onlyinside one representation level, but also between several levels (Pata, Puusepp, &Sarapuu, 2004). For example, Tsui and Treagust (2003) in their study aboutmodel-based reasoning with the BioLogica model have identified six levels of genet-ics reasoning that are characteristic to novices and experts. These hierarchicallyprogressive levels are mapped at the two-dimensional matrix that considers studentsability of relating genotype to phenotype within or between the generations and theirability of performing cause-to-effect, effect-to-cause, and process reasoning. Inwritten representations, such as short essays about the problems, people usuallydescribe most of the important elements and relationships in a complex and interre-lated way. The persons verbal contributions during discussions, on the other hand,do not have that structured nature. When representations of different origin areintended for comparison, part of the complexity of the mental model structure (e.g.,Downloaded by [Laurentian University] at 02:55 11 October 2014 Comparison of Reasoning Processes 1353inter-relations between elements) must be reduced in order to determine thepersons characteristic representational level.The two-folded relationship of different reasoning processes and the structure ofmental representations were of interest in this study. The following research ques-tions were formulated: 1. What characterizes model-based reasoning in expressive and exploratoryenvironments?2. What influence do different types of model-based reasoning have on thedevelopment of problem representations?3. How does the students initial problem representation level influence theirdevelopment during model-based reasoning?MethodsParticipantsThe participants of the study were 53 secondary school students (aged 1517) fromfour schools in Estonia. They had different knowledge about genetics: at one school,the students had studied the introductory topic of genetics in the previous year; atother schools, they were not familiar with the topic. This sample was chosen becauseit was of interest if the students different knowledge of genetics has influence ontheir development during model-based reasoning. The students of each school wererandomly divided between two different settings of the experiment. Thus, studentswith different knowledge about genetics were distributed into two experimentalgroups, referred to as Group I (who used expressive modelling setting) and Group II(who used exploratory modelling setting). Ten groups were formed with three toseven people working together. Besides this division between two experimentalsettings, an analytical distribution of students according to their problem representa-tion level was performed. In our experimental design, the students problemrepresentation level was measured according to their initial problem-solving essays,which enabled one to divide them into three analytical subgroups AC who hadinitially concrete, semi-abstract and concrete-abstract levels of problem representations.The modelling and discussion process was guided in a chat-room by a human tutorwho had experience in scaffolding cognitive and metacognitive activities.Learning Environment and TasksThe freeware software of the synchronous network-based learning environmentCollaborative Virtual Workplace 4.0 (CVW) (http://cvw.mitre.org) was used in thestudy. The client interface of CVW allowed students to gather in the virtual roomsto talk through chat with one another and share web links. All the interactions thatoccurred in a certain room were displayed to the users in the textual scroll-backwindow. In the expressive modelling activity, it was possible to use an electronicwhiteboard facility. For modelling purposes, a background image was provided onDownloaded by [Laurentian University] at 02:55 11 October 2014 1354 K. Pata and T. Sarapuuthe whiteboard. It contained images of a ladybird, pesticide, and the chromosomes(Figure 1, left). The students could simultaneously model the shared problemrepresentation by adding texts, arrows, and symbols.Figure 1. Example of a whiteboard with the template model about genetics for modelling in the expressive mode (below left). Right-hand side demonstrates the screenshots of the Ghost model for modelling in an exploratory environment about the same topic.For exploratory modelling activity, a web-based inquiry model Ghosts (http://mudelid.5dvision.ee/tondid/) was composed. Students, who worked in the virtualCVW environment, could access the model by clicking the web link deposited in thechat-room. The model (Figure 1, right) enabled one to test the causal effects ofdifferent mutagen levels on the genes and the phenotype of young and elderly ghostparents and their offspring. After selecting the mutagen level (low, medium, high), itwas possible to observe the phenotype of young ghost parents and their offspring, aswell as to examine the genotype of their somatic cells and gametes. Next, it waspossible to observe what the phenotype of ghost parents and their future offspringmight be if the parents had lived in the environmental conditions with a certainmutagen level and had children later. Besides observing the phenotype changes(emerging spots on skin), the students could investigate whether the certain alleles ofsomatic cells and gametes of parents and offspring were dominant or recessive. Theactivities of working with the expressive or exploratory models and the discussionsabout the problem took place at the same time.The genetic problem introduced to students was: Why has the appearance of lady-birds changed after pesticide treatment? The activity aimed to involve the studentsinto the practises of scientific community through using contemporary informationand communication technologies. The students could construct theories about theinfluence of mutagens on organisms as a part of shared meaning-buildingdiscussions in the network-based medium, apply inductive and deductive modellingFigure 1. Example of a whiteboard with the template model about genetics for modelling in the expressive mode (below left). Right-hand side demonstrates the screenshots of the Ghost model for modelling in an exploratory environment about the same topic.Downloaded by [Laurentian University] at 02:55 11 October 2014 Comparison of Reasoning Processes 1355practises in conceptual invention, and develop relationships between everyday andscientific knowledge. A simplified interpretation of mutagens was introducedaccording to which chemical substances and radiation can be possible mutagens.Theoretically the expected solution was related to the influence of certain mutagenlevels to a gene responsible for ladybirds spot colour. The web-based inquiry modelsupported mainly this solution, but in the expressive modelling environment thestudents had more freedom to interpret the elements of the background image andadd components. Besides this solution, some students considered an additionalpossibility that this mutant gene was responsible both for the spot colour and poisonresistance, and these features were inherited to the next generations of ladybirds. Inanother explanation, the students considered the possibility that some ladybirdswere more tolerant to the pesticide treatment due to some previous mutations.Therefore, they survived after pesticide treatment and the mutation in another genecaused changes in the spot colour that was inherited by the offspring. These twoexplanations were interpreted in the analysis as correct ones.The two-group design of the study comprised the following activities: Phase I. Introduction of the Case of ladybirds from situational context and thecomposition of individual Essay 1 for solving the ladybird problem. Phase II. Collaborative hypothetical-predictive reasoning in the chat room forsolving the ladybird problem without conceptual information. Phase III. Introduction with conceptual information about the influence ofmutagens on the genetic variability, and collaborative model-based reasoning inthe chat room for solving the ladybird problem. Phase IV. The composition of individual Essay 2 for solving the ladybird problem1 day after the collaborative activity.The tutors process-related support was applied during the hypothetical-predictivereasoning, and, in addition, the tutor practised conceptual scaffolding promptsduring the modelling phase. The tutors prompts focused on guiding studentstowards representing all the relevant and scientifically correct objects and processesabout the current genetic problem at the concrete and abstract levels of the emer-gence. The tutor encouraged students to discuss until they revised these theoriesthat were not scientifically correct (e.g., related to immunity).Methods of AnalysisThe qualitative investigation of reasoning activities and the combined qualitativeand quantitative analysis of individual students problem representations were usedin the study. The transcripts of group discussions, recorded by the CVW systemduring the activity, and the handwritten individual pre-essay and post-essays aboutthe problem were used for microgenetic content analysis.Frequent observation of rapidly changing competence and the in-depth qualitativeanalysis of this process in the dimensions of path, rate, breadth, source, and variabil-ity of the change are characteristic to the microgenetic method (Siegler & Svetina,Downloaded by [Laurentian University] at 02:55 11 October 2014 1356 K. Pata and T. Sarapuu2002). From these dimensions of microgenetic analysis, the path of change,concerning the sequence of used knowledge states, could be applied for finding outabout the nature of reasoning processes. In this study, the verbal examples of thereasoning processes in the time-line were interpreted for describing the nature ofdifferent reasoning patterns.The microgenetic dimension of the change rate was applied for describing theextent of qualitative shifts in the structure of students problem representations. Forinvestigating the changes, each persons pre-essay and post-essay and their contribu-tions during hypothetical-predictive reasoning (Lavoie, 1999) and model-basedreasoning (Lee, 1999) were extracted. Thus, four different representations could beanalysed for each person. The content of representations was separated into elements.Figure 2 shows the examples how all the elements expressed by the students werecategorized either as concrete or abstract entities. The qualitative map was composedby the researchers on the basis of all the students problem representations in essaysand during the discourse. Thus, both the scientifically correct as well as incorrectconcepts and relationships are present. The representation analysis framework by PataFigure 2. Qualitative map composed by the researchers on the basis of all the students representations about the genetic problem Why has the appearance of ladybirds changed after pesticide treatment? The map visualizes the representation levels of possible elements, and their structural and causal relationships. Note: Immunity was distinguished as not belonging to the correct abstract framework of the current problem model, but represented one of the most frequent misunderstandings of the students.Downloaded by [Laurentian University] at 02:55 11 October 2014 Comparison of Reasoning Processes 1357and Sarapuu (2003) was adopted. Tangible and visible objects and phenomena withtheir properties at the sensory perceptible macroscopic level (e.g., ladybird with blackspots), which were described by everyday explanation, comprised the concrete enti-ties. Objects and complex phenomena observable at a particulate level (e.g., gene) orsupra-macroscopic level over space and time (e.g., adaptation, resistance), which weredescribed by abstract terms or symbolic level formulas and equations, comprised theabstract level. We distinguished the narrative and symbolic forms of abstract levelentities and the scientifically correct and flawed concepts.Figure 2. Qualitative map composed by the researchers on the basis of all the students representations about the genetic problem Why has the appearance of ladybirds changed after pesticide treatment? The map visualizes the representation levels of possible elements, and their structural and causal relationships. Note: Immunity was distinguished as not belonging to the correct abstract framework of the current problem model, but represented one of the most frequentmisunderstandings of the students.Each students overall problem representation type was found by detecting all theused elements and their representation levels. This was done both in essays and forthe two reasoning phases. The following hierarchical categorization scheme fromlower-order to higher-order representations was applied: 1. Concrete: Objects and events are described only at the concrete level.2. Semi-abstract: Objects and events are described mainly at the concrete level; theentities described at the abstract level do not belong to the framework of thecurrent problem situation (are misinterpreted or flawed).3. Concrete-abstract: Objects and events are described and related both at theconcrete and abstract levels.4. Abstract: Objects and events are described only at the abstract level.5. Meta-abstract: Objects and events are described at the concrete and/or abstractnarrative, and abstract symbolic levels (the usage of symbols of alleles).The non-parametric analysis methods of the Wilcoxon Signed Ranks test and theWilcoxon Rank-Sum (MannWhitney) test were used for determining the changesin the problem representation categories due to different types of reasoning. TheSPSS 11.0 and MS Excel 2000 software were used for data analysis.ResultsCharacteristics of Model-based Reasoning in the Expressive and Exploratory ModesTwo divergent reasoning patterns, illustrated by the following example cases, wereobserved during modelling in the expressive and exploratory environments. We usedthe reasoning phases described by Johnson-Laird (1993) for interpreting the activity.The students belonging to Group I, who performed expressive modelling, oper-ated with the general level theoretical model without discussing the concrete details(amount, count, etc.). Initially the group developed the flawed problem representa-tion during hypothetical-predictive reasoning without discussing the abstract genes/chromosomes component and incorrectly using the terms resistance and immunity(Figure 3).Figure 3. Extract I about the teams first hypothesis from the hypothetical-predictive reasoning phase.When these flawed models were discussed and tested in the concrete modellingsettings after students had read the theoretical information, the misunderstandingswere removed in their individual essays. The model construction during the expressivemodelling activity started from the elements expressed verbally by the team in the firstDownloaded by [Laurentian University] at 02:55 11 October 2014 1358 K. Pata and T. Sarapuuhypothesis (Figure 4). After brainstorming about the role of several theoretical prop-erties of the model, the genes that were seen on the template model were included inthe initial shared mental model and the second hypothesis was formulated verbally.During and after the verbal model construction, the pesticide, ladybird, and geneswere linked on the whiteboard in Model I starting from concrete entities and movingtowards abstract ones.Figure 4. Extract II about model-based reasoning in the team in the expressive mode.The visualized Model I initiated the discussion about the discrepancies of thesecond hypothesis and triggered its validation (Figure 5). As a result, the thirdhypothesis was formulated verbally.Figure 5. Extract III about model-based reasoning in the team in the expressive mode.Figure 3. Extract I about the teams first hypothesis from the hypothetical-predictive reasoning phase.Figure 4. Extract II about model-based reasoning in the team in the expressive mode.Figure 5. Extract III about model-based reasoning in the team in the expressive mode.Downloaded by [Laurentian University] at 02:55 11 October 2014 Comparison of Reasoning Processes 1359It was then found that the existing visual Model I does not coincide with the thirdhypothesis (Figure 6). Building the revised Model II now started from abstractelements (mutagen, genes) and ended with visible changes at the concrete level(spots of the ladybirds). All the reasoning during model building took place with thewhole representation of the ladybirds problem.Figure 6. Extract IV about model-based reasoning in the team in the expressive mode.In the decision made in hypothetical-predictive reasoning phase of the exploratorylearning of Group II (Figure 7), a pattern analogical to the behaviour of Group I ofdiscussing the problem without involving the abstract genes/chromosomes compo-nent was found.Figure 7. Extract V about the teams first hypothesis from the hypothetical-predictive reasoning phase.During the exploratory modelling phase it appeared that the students initiallyreasoned only with the analogy model of Ghosts and did not connect it with thesituation model of Ladybirds (Figure 8). This might have been caused by the lackof appropriate guidance by the tutor to build relationships between the Ghostsand the Ladybirds situations. Although the tutors prompt to relate the differentmental models was responded to with the correct answer, the following discussionFigure 6. Extract IV about model-based reasoning in the team in the expressive mode.Downloaded by [Laurentian University] at 02:55 11 October 2014 1360 K. Pata and T. Sarapuuand the continuous tutors prompting showed that the students were operatingprimarily with the Ghosts model.Figure 8. Extract VI about model-based reasoning in the team in the exploratory mode.In the following model-construction phase (Figure 9) two new hypotheses wereformulated, in which the focus was on measurable properties supported by the web-based inquiry model.Figure 9. Extract VII about model-based reasoning in the team in the exploratory mode.It appeared that only after the solution was found from the Ghosts perspectivewas it related with the initial Ladybirds problem model elements on the basis ofanalogy, and then validated (Figure 10).Figure 10. Extract VIII about model-based reasoning in the team in the exploratory mode.The findings indicate that the model-based reasoning processes in two experimen-tal modes were performed differently. The students who worked in an expressivemodelling environment operated constantly with the whole model of the situation.They focused on discussing the theoretical relationships between the modelelements. The verbal and visual progress models were used as the external aids tovalidate the current state of the shared mental model of the problem. The studentswho used an exploratory modelling mode operated mainly with the conceptual peda-gogical model, focusing on the measurable characteristics (age, mutagen level) of theelements of the model one by one. The connections between the pedagogical andprogress models of the situation were discussed only after the problem was solvedfrom the perspective of the analogous pedagogical model.Influence of Different Types of Model-based Reasoning on the Development of Problem RepresentationsIn the previous section it was shown that there were general qualitative differences instudents reasoning patterns in the expressive (Group I) and exploratory (Group II)Figure 7. Extract V about the teams first hypothesis from the hypothetical-predictive reasoning phase.Figure 8. Extract VI about model-based reasoning in the team in the exploratory mode.Downloaded by [Laurentian University] at 02:55 11 October 2014 Comparison of Reasoning Processes 1361modelling settings. The following analysis investigated the changes in the problemrepresentation mode of two experimental groups of students due to model-basedreasoning. The findings indicated that the nature of modelling activities influencedthe development of students problem representations differently. This supportedthe previous findings about the different nature of two model-based reasoningprocesses.No statistically significant difference was found between Group I and Group IIrepresentation types with the MannWhitney test. The Wilcoxon Signed Ranks test(Table 1), which compared the development of individuals during the activity,demonstrated that students in Group I changed their representation level signifi-cantly (p < .01) only in the hypothetical-predictive phase of the activity, whereas inFigure 9. Extract VII about model-based reasoning in the team in the exploratory mode.Figure 10. Extract VIII about model-based reasoning in the team in the exploratory mode.Downloaded by [Laurentian University] at 02:55 11 October 2014 1362 K. Pata and T. SarapuuGroup II such progress was not found. This indicated that the groups might nothave been at an equal level in the beginning.Students from Group I did not make any statistically significant advancement intheir problem representation level during the modelling activity (Phase III), butstudents from Group II appeared to be developing significantly (p = .02) towardsusing higher-order Abstract problem representations. Surprisingly, in Group II thestatistically significant (p = .02) change towards lower Concrete-Abstract representa-tion level was found in final Essay 2 (Phase IV) compared with the students Abstractreasoning level when using the model. In Group I, highly significant (p < .001)development was found between the initial reasoning level in Essay 1 compared withthe final reasoning level in Essay 2. Group II changed their representation categoriesin the final essays at the significance level (p = .02).To conclude, we could observe that the students problem representations devel-oped towards hierarchically higher levels of representations in the modelling phase.The representations of many students who worked in the exploratory modellingsetting became Abstract, suggesting that quantitative conceptual inquiry modelsenhanced reasoning with abstract terminology about the mutagenesis and genotype.In the expressive modelling setting, the students discussions remained at theConcrete-abstract level, indicating that both the observable entities related to pheno-type as well as abstract ones related to the genes and mutagenesis were talked about.The final Essay 2, however, indicated that the students who showed abstract levelreasoning when working with the inquiry model were not able to explain the geneticproblem at the same level after the activity, and used simpler explanations. Thiscould be interpreted as the return to the previous representations due to performingmodel-based reasoning in the context of an analogous problem without relating itwith the situation model.Table 1. Students sequential problem representation development in Groups I and IIDifferences in representation level with the Wilcoxon Signed Ranks testGroup I, expressive modelling (N = 28)Group II, exploratory modelling (N = 25)Compared phases of the activity Z p Z pHypothetical-predictive reasoning (Phase II)Essay 1 (Phase I)2.54 0.01** 1.35 0.17Model-based reasoning (Phase III)Hypothetical-predictive reasoning (Phase II)1.05 0.29 2.32 0.02*Essay 2 (Phase IV)Model-based reasoning (Phase III)0.05 0.95 2.26 0.02*Essay 2 (Phase IV)Essay 1 (Phase I) 3.78 0.001** 2.30 0.02*Note: *p < .05, **p < .01.Downloaded by [Laurentian University] at 02:55 11 October 2014 Comparison of Reasoning Processes 1363Influence of the Students Initial Problem Representation Level on their Representation Development during Model-based ReasoningThe data from the comparison of two experimental groups indicated that thestudents initial problem representation level might have influenced their mentalmodel development during reasoning with models. Certain hierarchical domain-specific reasoning patterns have been described in the study of Tsui and Treagust(2003) when learning with the inquiry BioLogica model, but no comparisons inves-tigating the effectiveness of different modelling activities on learning have beenmade. Therefore, in the further analyses the students data were separated into threeanalytical subgroups on the basis of their problem representations in Essay 1.Subgroup A was initially able to describe the problem only at a Concrete level,Subgroup B had Semi-abstract representations with some flawed understandings,and Subgroup C had Concrete-abstract representations. By this it was possible tofollow the changes in students problem representations from one reasoning level ingenetics to another due to different model-based learning activities.Table 2 reflects the three subgroups mental model development during theactivity. It appeared that the students who had initially Concrete level representa-tions of the problem (Subgroup A) developed their representations further towardsthe higher-order explanation levels only during hypothetical-predictive reasoning(p = .04) and not during model-based reasoning. The comparison of the studentsrepresentations before and after the collaborative activity showed their significantlevel (p = .03) development towards hierarchically higher-order representationcategories. In Subgroup A no statistically significant differences between studentsfrom the expressive (Group I) and exploratory modelling groups (Group II) wereTable 2. The different level of students sequential development of problem representationsDifferences in representation level with the Wilcoxon Signed Ranks testSubgroup A, Concrete level (N = 10)Subgroup B, Semi-abstract level (N = 22)Subgroup C, Concrete-abstract level (N = 21)Compared phases of the activity Z p Z p Z pHypothetical-predictive reasoning (Phase II)Essay 1 (Phase I)2.0 0.04* 2.49 0.01** 0.27 0.78Model-based reasoning (Phase III)Hypothetical-predictive reasoning (Phase II)1.34 0.17 2.18 0.02* 0.86 0.38Essay 2 (Phase IV)Model-based reasoning (Phase III)1.0 0.31 0.52 0.59 1.31 0.18Essay 2 (Phase IV)Essay 1 (Phase I) 2.06 0.03* 4.41 0.001** 0.27 0.78Note: *p < .05, **p < .01.Downloaded by [Laurentian University] at 02:55 11 October 2014 1364 K. Pata and T. Sarapuufound in the development of problem representations during the different phases ofthe activity with the MannWhitney test.Students (Subgroup B), who were initially at the Semi-abstract level of problemrepresentations appeared to develop significantly, using a more advanced explana-tion level in both hypothetical-predictive (p < .01) and model-based reasoningphases (p = .02). Subgroup B also showed highly significant (p < .001) developmentin their Essay 2 compared with Essay 1, representing the problem with higher-orderexplanation levels. With the MannWhitney test no statistically significantdifferences were found between the students who used exploratory and expressivelearning settings.Students who were at the Concrete-abstract representation level in their initialEssay 1 (Subgroup C) did not make significant progress during two reasoningphases and in their initial and final essays. Among the members of Subgroup C astatistically significant (p = .04) divergence caused by a different type of model-based reasoning was found with the MannWhitney test. More students performedreasoning at the Meta-abstract level in the exploratory modelling environment.These findings indicate that, in genetics reasoning, the students representationaldevelopment during discussions and modelling activity depends on their initial levelof problem representations. A low level of understanding the problems (observingmainly the phenotype, not considering the inheritance to the next generations) limitsthe students capability of reasoning with conceptual genetics models. Misunder-standings in the mutagenesis process (relating it with immunity) can be removedboth by expressive and exploratory model-based reasoning in teams. The externalmodels direct the students to focus on certain aspects of the problem and thestudents make changes in their problem representations by criticising each othersproblem representations. Only the students who had correct conceptual knowledgeabout the mutagenesis could use scientific terminology and abstract symbols whenreasoning with the exploratory inquiry model. This indicates that, even though allthe students could work with the inquiry model, they did not understand all theconceptions the model expressed.DiscussionDue to the increasing availability of computer-based modelling facilities in schools,the in-depth understanding of the influence of these learning methods has becomenecessary. Several comparative studies have been carried out, conducting expressiveand exploratory learning activities with qualitative, semi-quantitative, and quantita-tive models (e.g., Mellar, Bliss, Boohan, Ogborn, & Tompsett, 1994). These studiesinvestigated the reasoning differences with certain type of models when practisingexpressive or exploratory tasks. The focus of these studies was on the aspects ofqualitative, semi-quantitative, and quantitative reasoning. In this paper, a differentapproach was taken. It was assumed that the certain type of models favour learningin expressive or exploratory modes, which in turn affects reasoning and the develop-ment of mental models. This idea was theoretically grounded by the studies thatDownloaded by [Laurentian University] at 02:55 11 October 2014 Comparison of Reasoning Processes 1365indicate that model construction as an inductive reasoning activity might activatestudents mental representations about the problem differently to explorations withthe model, which presumes deductive reasoning patters (Cornujols et al., 2000;Pata & Sarapuu, 2004; Yu, 2002). Two types of activitiesthe model constructionin an expressive team-learning environment and collaborative inquiry with theexploratory web-based modelwere compared.The findings demonstrated the clear qualitative difference of these two types ofmodel-based reasoning. The model construction activated the representation of theproblem from the situation framework and the changes in it were made consideringthe model as a whole. The students verbal and visual progress models served as aidsfor validating the current state of the shared mental model of the problem in a team.Thus, it became possible that during the expressive modelling activity, when thesame model was constantly revised and validated, the students were capable ofmaking changes in the structure of their mental model of the problem by developingit towards more abstract and theoretically sound explanations.Inquiry with the model activated two different mental representations: the peda-gogical model and the situation model. Nevertheless, the explorations were madeonly in the context of an analogous pedagogical model. The hypotheses could betested with an exploratory model only if separate pierces of the mental model wereactivated through experiments. Thereby, the mental model of the pedagogical prob-lem was broken to smaller units as it was complicated to make revisions in themental model of the problem situation at the same time. This interpretation is inaccordance with the studies of Mellar and Bliss (1994), which emphasized that inthe expressive mode of learning the students are modelling their own assumptions,while in the exploratory mode learners are led to confront anothers view of a prob-lem and must bring their own ideas to interact with those provided in the model.Our findings also support the results of Mellar et al. (1994) that the expressive typeof learning enhanced the students to focus in reasoning on the theoretical qualitativeelements and properties of the model (qualitative reasoning); in the exploratorylearning activity, the students centred in reasoning on the measurable quantitativefeatures (quantitative reasoning).Two tendencies were observed about the influence of different types of model-based reasoning on the development of students mental models of the problem.Students who performed reasoning in an exploratory learning environment appliedhigher-order Abstract or Meta-Abstract reasoning levels when working with theanalogy model but were not able to bring this conceptual knowledge into theirexplanations of the initial problem situation. As no similar negative development wasfound in the expressive modelling environment, it was possible to assume that theexploratory modelling with an analogous situation was more demanding from theperspective of developing students in-depth understanding about natural phenom-ena than the expressive modelling activity. As the exploratory modelling task withthe analogy model turned out to be more complex for explaining similar situations, itis recommended that this type of modelling activity should be designed as a backand forth movement between the analogy model and the situation model.Downloaded by [Laurentian University] at 02:55 11 October 2014 1366 K. Pata and T. SarapuuAccording to Ploetzner, Spada, Stumpf, and Opwis (1990), effective teachingpresumes that each different level of mental representation has to be addressed byspecific levels of instructional presentation. The results of current researchsupported this general claim with characteristics that must be considered in planningthe application of modelling activities. The students conceptual progress, due tomodel-based reasoning activity, appeared to be influenced by their initial ability ofrepresenting the problem. The students who were initially able to focus only at theconcrete objects and properties did not benefit from the model-based reasoningactivity. The modelling was most effective for the students who had some under-standing of the abstract level objects, properties, and relationships of the problembut whose interpretation was flawed. Only those students who already had theConcrete-abstract type of problem representations appeared to be influenced by thetype of model-based reasoning activity. The expressive modelling activity did nottrigger the students to use significantly more higher-order problem representationsduring discussions and on their visual model. The exploratory modelling activityfavoured the students to use Meta-abstract representations during modelling. It canbe concluded that model-based reasoning, and especially exploratory modelling,should be practised with students who already have some theoretical understandingof the phenomenon under investigation.ConclusionsThis study aimed to investigate the relationship between the two types of modellingactivities (expressive, exploratory), possible different reasoning processes appliedduring model-based learning, and the influence of two modes of reasoning on thedivergent development of mental models about genetics.Earlier studies (Mellar et al., 1994) relate the effects of model-based learning withthe development of qualitative, semi-quantitative, and quantitative reasoning. Thisstudy provided some evidence that the nature of modelling activity could favour theusage of inductive and deductive reasoning. Further on, it was found that modellingin an expressive mode (constructing the model) triggered the inductive reasoningmechanisms and enabled the learners to develop and revise the whole situationmodel until the more scientific understanding of the phenomenon had been devel-oped. The modelling in an exploratory environment demanded the activation ofdeductive reasoning and the construction of mental models for both the situationmodel of the problem and the pedagogic model that was offered for explorations.The process of making revisions in the mental model of the situation appeared to beinhibited in the exploratory environment due to the need for operating with twodifferent mental models, which was a cognitively more demanding process thanrevising only the situation model. Working with the exploratory model presupposedtesting of certain relationships between the model elements, which hindered theprocess of forming the overall mental model of the problem.In future studies the possible relationships between the nature of model-basedlearning on the reasoning processes and the mental model development should beDownloaded by [Laurentian University] at 02:55 11 October 2014 Comparison of Reasoning Processes 1367studied in depth. The findings of the study indicate that it is not sufficient to analyseonly the students problem representations before and after the activity, but theinformation of the microgenetic changes in mental model during learning andreasoning are necessary to explain the influence of modelling. The application ofcollaborative hypothetical-predictive reasoning activities before modelling might bean effective means for predicting which type of modelling activity might favour thestudents conceptual change the most.AcknowledgementsThis research was supported by the ESF grant 5996 and MER funding 0182542s05.The authors express their appreciation for the cooperation and assistance of KristjanAdojaan in programming the model, of Arle Puusepp for assistance, and of the Esto-nian science teachers and the students of the in-service course 20022003.ReferencesChalk, P. (2000). Assessing students learning when using software engineering webworlds,Proceedings of the 1st LTSN-ICS Annual Conference (pp. 6773). Edinburgh: Heriot-WattUniversity.Cheng, P. C.-H. (1993). Multiple interactive structural representations systems for education. InR. Cox, M. Petre, P. Brna, & J. Lee (Eds.), Proceedings of the Workshop on Graphical Represen-tations, Reasoning and Communication (pp. 2528). World Conference on Artificial Intelligencein Education, Edinburgh, Scotland.Chi, M. T. H. (1997). Creativity: Shifting across ontological categories flexibly. In T. B. Ward, S.M. Smith, & J. Vaid (Eds.), Creative thought: An investigation of conceptual structures andprocesses (pp. 209234). Washington, DC: American Psychological Association.Cornujols, A., Tiberghien, A., & Collet, G. (2000). A new mechanism for transfer betweenconceptual domains in scientific discovery and education. Foundations of Science, 5, 129155.Gentner, D., & Gentner, D. R. (1983). Flowing waters or teeming crowds: Mental models of elec-tricity. In D. Gentner & A. Stevens (Eds.), Mental models. Hillsdale, NJ: Lawrence ErlbaumAssociates.Gilbert, J. K., Boulter, C. J., & Elmer, R. (2000). Positioning models in science education and indesign and technology education. In J. K. Gilbert & C. J. Boulter (Eds.), Developing models inscience education (pp. 317). Dordrecht, The Netherlands: Kluwer Academic Publishers.Jimoyiannis, A., & Komis, V. (2001). Computer simulations in physics teaching and learning: acase study on students understanding of trajectory motion. Computers and Education, 36(2),183204.Johnson-Laird, P. N. (1983). Mental models: Towards a cognitive science of language, inference, andconsciousness. Cambridge, MA: Harvard University Press.Johnson-Laird, P. N., & Byrne, R. M. J. (1991). Deduction. Hove, UK: Lawrence Erlbaum Associates.Jonassen, D. (1995). Operationalizing mental models: Strategies for assessing mental models tosupport meaningful learning and design-supportive learning environments. In J. L. Schnase &E. L. Cunnius (Eds.), Proceedings of CSCL95 (pp. 182186). Mahwah, NJ: LawrenceErlbaum Associates.Klahr, D., & Dunbar, K. (1988). Dual space search during scientific reasoning. Cognitive Science,12(1), 148.Lavoie, D. R. (1999). Effects of emphasizing hypothetico-predictive reasoning within the sciencelearning cycle on high school students process skills and conceptual understandings in biol-ogy. Journal of Research in Science Teaching, 36(10), 11271147.Downloaded by [Laurentian University] at 02:55 11 October 2014 1368 K. Pata and T. SarapuuLawson, A. E. (2002). What does Galileos discovery of Jupiters Moons tell us about the processof Scientific Discovery? Science and Education, 11, 124.Lee, M. H. (1999). On models, modelling and distinctive nature of model-based reasoning. AICommunications, 12, 127137.Lehrer, R., & Schauble, L. (2000). Developing model-based reasoning in mathematics andscience. Journal of Applied Developmental Psychology, 21(1), 3948.Magnani, L., & Nersessian, N. (2004). Preface. Foundation of Science, 9, 213218.Mellar, J., & Bliss, J. (1994). Introduction: Modelling and education. In H. Mellar, J. Bliss, R.Boohan, J. Ogborn, & C. Tompsett (Eds.), Learning with artificial worlds. Computer-basedmodelling in the curriculum (pp. 18). London: Falmer Press.Mellar, J., Bliss, J., Boohan, R., Ogborn, J., & Tompsett, C. (Eds.). (1994). Learning with artificialworlds. Computer-based modelling in the curriculum (pp. 18). London: Falmer Press.Nersessian, N. J. (1999). Model-based reasoning in conceptual change. In L. Magnani, N. J.Nersessian, & P. Thagard (Eds.), Model-based reasoning in scientific discovery (pp. 522.) NewYork: Kluwer Academic/Plenum Publishers.Pata, K., & Sarapuu, T. (2003). Framework for scaffolding the development of problemrepresentations by collaborative design. In B. Wasson, S. Ludvigsen, & U. Hoppe (Eds.),Designing for Change in Networked Learning Environments. Proceedings of CSCL 2003 Confer-ence (pp. 189198). Dordrecht, The Netherlands: Kluwer Academic Publishers.Pata, K., & Sarapuu, T. (2004). Information-processing and conceptual change in network-basedsynchronous modelling environments. World Conference on Educational Multimedia, Hypermediaand Telecommunications, 2004(1), 39934000.Pata, K., Puusepp A., & Sarapuu T. (2004). Developing students problem solving skills in physicsusing exploratory and expressive synchronous learning environments. In A. Laine, J. Lavonen,& V. Meisalo (Eds.), Current research on mathematics and science education. Proceedings of theXXI Annual Symposium of The Finnish Association of Mathematics and Science EducationResearch (pp. 550569). University of Helsinki, Finland.Ploetzner, R., Spada, H., Stumpf, M., & Opwis, K. (1990). Learning qualitative and quantitativereasoning in a microworld for elastic impacts. European Journal of Psychology of Education,V(4), 501516.Reinhard, P., Hesse, F. W., Hron, A., & Picard, E. (1997). Manipulable graphics for computer-supported problem solving. Journal of Computer Assisted Learning, 13, 148162.Romeyn, J. W. (2004). Hypotheses and inductive predictions. Synthese, 141, 333364.Roth, W. M. (2001). Modelling design as situated and distributed process. Learning andInstruction, 11, 211239.Siegler, R. S., & Svetina, M. (2002). A microgenetic cross-sectional study of matrix completion:Comparing short-term and long-term change. Child Development, 73(3), 793809.Soderberg, P., & Price, F. (2003). An examination of problem-based teaching and learning inpopulation genetics and evolution using EVOLVE, a computer simulation. InternationalJournal of Science Education, 25(1), 3355.Stewart, J., Hafner, R., Johnson, S., & Finkel, E. (1992). Science as models building: Computersand high-school genetics. Educational Psychologist, 27(3), 317336.Tsui, C.-Y., & Treagust, D. F. (2003). Genetics reasoning with multiple external representations.Research in Science Education, 33, 111135.Wilensky, U., & Resnick, M. (1999). Thinking in levels: A Dynamic Systems Approach to makingsense of the world. Journal of Science Education and Technology, 8(1), 319.Wynne, C. F., Stewart, J., & Passimore, C. (2001). High school students use of meiosis whensolving genetics problems. International Journal of Science Education, 23(5), 501515.Yu, Q. (2002). Model-based reasoning and similarity in the world. In L. Magnani & N. Nersessian(Eds.), Model-based reasoning. Science, technology, values (pp. 275286). New York: KluwerAcademic/Plenum Publishers.Downloaded by [Laurentian University] at 02:55 11 October 2014