Ana Isabel Boa-Ventura e A. J. Ferrer Correia
Universidade de Aveiro


This work intends to characterize usersí profiles on the exploration of a hypermedia document. For this purpose, fourteen advanced university students interacted with a hypermedia program written in SuperCard about Solid/Liquid Phase Diagrams; their navigation was registrated and the data obtained this way were then exported to a spreadsheet and adequately treated to allow for the calculation of certain navigational parameters: linearity, thorougness and expectability. In the same application, the authors developed a special browser which gives a real-time feed-back on the extension of the program covered by navigation; at this level, this tool was used for research purposes only. We found that almost every student had a strongly linear exploration but for a similar value of linearity, the thorougness of the exploration is very different. In the same way, the latter varies a great deal for similar values of expectability of the navigation. The present study is a pre-experimental of a broader research on the recognition of navigational patterns. The authors intend to use the special browser developed in the present study not as a real-time feed-back for the user and thus evaluate whether the students diversify their modes of exploration according to the feed-back they get from this aid to navigation.


When, in multimedia applications, video sequences, photos, text and sound are put together in order to be used interactively, there should be a greater concern with helping userís navigation more efficiently. On the other hand, the fact that multimedia is, now, starting to be available over networks, that is, to a much larger number of people and along much larger distances, emphasises the need for a "universal" language. The expression "universal" is to be taken with care: the several meanings the same icon may have to different people are a proof of how far we are from a universal multimedia language, one that can be understood by people from different cultures.

Ted Nelsonís idea of hypertext and his suggestion of associating ideas in a multiple and flexible way, should be the rocky and solid basis for any hypermedia program. A network structure is one possible metaphor to represent nodes of information and links between them; in theory, nodes are both semantically and syntactically discrete [3].

Some factors inherent to the hypermedia philosophy are themselves responsible for the disorientation the user may experiment when exploring a hyperdocument; if we can forgive its consequences in fields like marketing or presentation, we can not as easily do it in what concerns Education. A number of factors may induce disorientation:

If an appropriate application of hypertext is to be made in Education, a reexamination of the concept of learner control should be made in educational research [1]

Modes of navigation

Different types of navigation, according to a greater or lesser branching degree, retracing of steps, etc., are possible in hypermedia exploration. The most direct and unsophisticated mode of exploration one can conceive is a linear navigation (Fig. 1) which leads to a poor exploration in terms of visited nodes. Such a linear exploration subintends an incorrect management of resources. The use of C.A.L.ís shells would allow resources savings and serve such as well this linear exploration.

Fig. 1

Increasing branching in the exploration of the hypermedia document (i. e., repeated visits to each node with different exits on each occasion), leads to a more complex, arborescent exploration, as described in Fig. 2.

Fig. 2

In Education, the two extreme situations described above may happen and if an incomplete exploration of nodes may, in certain contexts, be enough, that is not always the case. Therefore, itís easy to conclude that, concerning Hypermedia and educational material, research is needed on:

Navigational parameters analysed

The treatment of data in the present study was based on the analysis of the following parameters: In this context, linearity means the degree of approach to a "linear" path, that is, one where the user starts at one end of the network and follows a certain "branch" until he/she reaches the end of the exploration. Expectability measures the overall probability of the links selected in the navigation. This probability is assessed by attributing to each type of jump a value between 0 and 1, and depends on the nature and objectives of the application. Arborescence measures the overall tendency to repeatedly visit each node, exploring different exits each time; this parameter has not been implemented in what concerns the referred application, due to the fact that the branching in this case was not complex enough to allow any analysis of this nature. Thoroughness measures the degree in which the exploration was completed; quantitatively, it is a simple percentage of visited screens, relative to the total number of screens in the program.

The types of "jumps" ó links effectively selected by the user when leaving a particular screen ó are listed in Fig. 3. This classification has been made taking into account the authors own experience as hypermedia navigators: especially considered were the jumps more likely to occur and which could have intrinsic value as clues to any attempt of defining a profile. The following five types, corresponding to these criteria, were identified:

Fig. 3

Considering linearity as a measure of the non-branching steps taken by the user, we measure this parameter as the ratio between the linear extension and the global extension of the path. The linear and the global extensions are the diagonals of the hypercube which sides are the number of the relevant types of jumps. In Fig. 4 a simplified 3D representation, where only ascending, descending and erratic jumps are represented, is given.

Fig. 4

When the number of erratic jumps is very small, the two diagonals almost superimpose to one another (Fig. 5); this would happen with a strongly linear path. The value for linearity is, in this case, close to 1.

Fig. 5

On the other hand, when the number of erratic jumps is considerably high, the two diagonals are very different (Fig. 6); that would be the case of a strongly chaotic path. The value for linearity is, in this case, closer to 0.

Fig. 6

  • Experimental

  • Sample
    A total of fourteen students of an Advanced Chemistry course from the University of Aveiro formed the sample of this study. At the time the study took place, 14 was the total number of students in the discipline.
  • Design and Procedure
  • In the present research a program to teach Solid/Liquid Phase Diagrams as a topic of Physical Chemistry for University Chemistry Students was selected for entirely pragmatic reasons: a group of students was learning this subject at the time the research took place. As it was presumed that, on a first approach, the user would be unfamiliar with the topology of the document, it was decided to organise a first exploration with a time limit of 20 minutes. During this first exploration, the students were expected to explore especially the "Instructions" and "Introduction" parts; a short break of 15 minutes was arranged before the second exploration, which had no time limit.

    To trace the userís navigation, a disk file was created for each student, and for each exploration, this process being transparent to the user. During both explorations, the kind of data being registered was: visited screens, type of links and time in each screen (except for what concerned "Instructions" and "Introduction"). For this purpose, scripts of most of the objects had been modified, allowing not only the registration of time and title of screen visited, but also the kind of link the user selected when going from one screen to another (according to a classification described below). All these data were then exported to Excel and quantitatively treated according to the parameters referred to before.

  • Real-time construction of a special browser
  • The program "Phases" explores the phase-diagrams of solid-liquid binary systems at different temperatures. The user can visualise what the areas of the diagrams represent in terms of the crystal structure of an alloy. He/she may also study the variation of temperature during a cooling process.

    When the user enters the Browser at any point of the exploration of the program, he/she will get an active map of the program which, as mentioned before, allows direct access to any part, just by selecting the corresponding graphic.

    The user will also be able to gain access to what we think is a new and useful tool, by selecting a specific button on the Browser: he/she will then enter a special browser (Fig. 7), graphically similar to the one previously mentioned but which gives the user a real-time idea of the extension of the program already explored, as well as a visual and immediate feedback on the number of times each screen has been visited.

    This last data is achieved by the use of different shades of the same colour: lighter green means less number of visits and darker green means larger number of visits, according to a predetermined scale of shades1. Close to each graphic representing a screen, the exact number of visits can be read.

    We have selected some of the most significant results weíve obtained in the research, concerning the special Browser2.

  • Data analysis
  • Real-time special browsers which were particularly meaningful are presented in the next pages: Teresa, Nuno, Anabela and Carlos show quite different ways of navigating. Their browsers were selected to be presented as examples of the first or second exploration, according to the relevance of data. Teresaís, Nunoís and Anabelaís browsers are shown as examples of the first exploration; Teresaís, Nunoís and Carlosís are shown as examples of the second exploration.
  • First exploration
  • On her first exploration (Fig. 7), Teresa systematically consulted two screens of each topic: there are four of them.


    Nuno did a quite different type of exploration, having started by a reasonable exploration of the first3 topic, and decreasing the detail of the navigation from topic to topic: he didnít visit the fourth (last) topic at all (Fig. 8).

    Fig. 8

    Anabela did a very complete exploration of the first topic and didnít do a single visit to any other topic (Fig. 9).

    Fig. 9

  • Second exploration
  • Carlos explored reasonably the second topic, didnít touch the first one and visited quite incompletely the third and fourth topics (Fig. 10). The thoroughness of his exploration was quite small - 23.8 %.

    Fig. 10

    After a first well-balanced exploration of every topic (V. Fig. 7), on her second exploration, Teresa explored almost totally the second topic and covered 71.4% of the total number of screens (Fig. 11).

    In his first exploration, Nuno decreased the level of detail from the first to the last topic (V. Fig. 8); on his second exploration (Fig. 12), he insisted on the screens that lead to the best animations in the program (notice the darker shades on the terminal nodes of topic 4, on the right of the Browser) and clearly made use of the Browser for a considerable number of times. He did a quite thorough navigation, having covered 76.2% of the total number of screens, the best of all cases.

    Fig. 11

    Fig. 12

  • Monitoring the data
  • In the pictures 13 to 15, we can see the data obtained from the tracing of the userís navigation already exported to Microsoft ExcelTM. On the right part of the spreadsheet, the weight attributed to each type of jump (Pi), the total number of jumps in each category and the values for expectability and linearity can be seen; a graphic showing the number of jumps for each type is also represented.

    Carlos has an exploration based mostly on descending and ascending jumps; on the opposite, Nuno has less ascending jumps and quite a few lateral and erratic. Paulaís spreadsheet is shown here in order to explain the way expectability was determined: it is a weighted average obtained by _ Pi*Ni divided by _Ni, where Pi represents the probability of each type of jump and Ni the number of jumps of each type. The probability assigned to each type of jump is entirely subjective and based only on the experience the authors have in the exploration of hypermedia documents, as well as on the category of the document, which is clearly one of formal learning.

    Fig. 13

    Fig. 14

    Fig. 15


  • Overall results
  • In Fig. 16, we have the overall results regarding linearity (L), expectability (E) and thoroughness (T).

    Fig. 16

    We may conclude that, although thereís a clear correlation between the values obtained for linearity and expectability (which is easy to understand, taking into account the way these data were obtained), the values are not strongly clustered.

    Regarding the graphical representation T/E, we see that the values are very loosely clustered and that for a similar expectability, the covered area of the program is very different.

    The graphical representation T/L shows that the students had an exploration which was essentially very linear. For a similar linearity, however, thoroughness is very different.

    This study is only preliminar and is a pre-experimental of a broader research. The primary prediction of this study was that to similar values of linearity, the completeness of the exploration may vary a lot. While data gathered are in support of this prediction, the inherent design and other weaknesses of the study and especially the small dimension of the sample woud appear to limit any generalizations of the findings. Before the interaction with the program, some documentation about the application was distributed, which basically represented the several screens by a logical order. It seemed to us that the students tried to find in the software the order in the paper; this may have affected the way they conducted the navigation. A replication of the study with more stringent controls and with a much larger sample will be a promising direction for further research

  • Conclusions and future work
  • We believe the special Browser we developed for this application can be very useful as a software development tool, allowing the software engineer to test the applicationís interface with small populations. In this research, some of the conclusions we reached in terms of the interface design, and not described in the context of this paper were the recognition of (1) an excessive emphasis of appealing animations in just one of the topics which made it unproportionally exploited; (2) an incorrect toggle (appearing/disappearing) of one button in certain situations, since almost none of the students visited the particular screens which that button would lead to; and (3) the influence of the physical order of icons and graphics upon the order of selections done by the user ó this was particularly evident since the visual order of objects to be selected in the graphical representations contradicts the chronological order of events.

    We intend to develop this tool so it can give a graphical representation of the links chosen by the user; we believe this kind of implementation can be significantly important, especially to the researcher. The automatic creation of graphical traces of learning paths on a node-link lesson map ó which would then be subjected to analysis by graph theory [8] ó may reveal patterns of learning processes [0]. A similar study has been carried out by where the numer of visits to the same node is identified by the diameter of the circular representation of nodes in the graphical browser and where links effectively chosen by the user are also automatically represented [8].

    Besides the value the special Browser developed in this application may have as a software development tool to the software engineer or to the researcher, we believe that, by allowing a real-time visual information on the completeness of the navigation, it can truly help the user in finding other points of interest, not yet (or poorly) exploited in the hypermedia application.

    This study was the pre-experimental of a broader research concerning patterns of navigation. The next step will be the implementation of the above referred browser in an application on a Nature Reserve, developed by the authors. We intend to reach some conclusions on in what degree the user changes his/her navigation according to the information he/she obtains from this special browser?