Address:

Learning Geography in Secondary School through a Hypermedia System

Francesco Natale, Anna Troise, Francesco Antinucci, Luisa Berlinguer

Istituto di Psicologia Consiglio Nazionale delle Ricerche Roma, Italia

Email address

Table of contents

Abstract
What is a hypermedia system
Hypermap
Testing HYPERMAP
Subjects and procedure
Results
Discussion
References
Hardware

Abstract

The paper proposes a definition of hypermedia, different from the customary one of "hypertext whose nodes belong to media other than text", as a specific communicative system that allows the structure of the communictive means to match the instrinsic structure of the knowledge field it purports to communicate. It is argued that such a system may induce a shift in the modality of learning, from an "explicit" to an "implicit" mode. Subsequently, HYPERMAP, a hypermedia system for the learning of descriptive geography designed and built in this fashion, is described. First results of testing HYPERMAP in a real classroom situation in comparison with traditional classroom teaching are reported. These results appear not only to show marked positive effects on learning but also to confirm a shift towards an implicit learning modality.

What is a hypermedia system.

Hypermedia is a communicative system that is made possible by multimedia technology. The essential feature of such a system is that it allows the construction of communicative structures that are highly similar (isomorphic) to the structure of the knowledge fields they communicate about. Communication is a fundamental step in the process of learning. In order to acquire and remember, first of all one has to understand. In traditional learning environments most communication is in language form and most of the language form is written and textual. The textual form has its own structure, namely a pan-linear one. In order to appropriately convey a given knowledge field in textual form a mapping operation has to be performed. One has to map the structure of the knowledge field into the pan-linear structure of the textual form. The problem is that most knowledge fields do not have a linear structure: when one access in his mind a knowledge field one does not go back and forth through the pages of a book (with the possible exception of historical narratives and similar fields). The difficulty of performing this mapping operation is the difficulty experienced by anyone who has attempted to write a textbook. Conversely, the main difficulty from the learner's point of view is to perform the reverse mapping operation in order to reconstruct the structure of the knowledge field from the pan-linear structure of the text. These processes can be illustrated by the following diagram (where KS stands for "Knowledge Structure" and CS for "Communicative Structure"):

Let's suppose, for example, that the field of knowledge is biological taxonomy. Presumably, the way in which this knowledge is organized in the mind of somebody who knows biological taxonomy is some sort of hierachical tree-structure relating species and genera (KS) . If one has to communicate this field in the form of a textbook, one will have to map this hierarchical tree- structure into a linear succession of chapters (CS). In doing so he will be forced to choose, for example, whether to "scan" the tree (so to speak) at each level of nodes and, hence talk first about all classes, then all orders, then all families, then all genera, then all species, or to scan it along each branch and talk about, for example, the order of primates, the familiy of lemurs, the genus Lemur and the species Ring-tailed lemur, and so on. The learner, on the other hand, will have to start from this linear succession of chapters (CS) and try to reconstruct from it the hierarchical tree-structure (KS). Since KS is, for any knowledge field, given, the only way to simplify the mapping operation is to try to change CS in such a way that it is more similar, or, if possible, identical (isomorphic) to KS. In this way the communicative structure would become "transparent" with respect to the knowledge field. Obviously, this cannot be done through the textual medium which has only a pan-linear structure. Let's give some concrete, though very simple, examples of how this change can be effected and, at the same time, convey an intuitive appreciation of the remarkable difference it causes on the cognitive process of understanding. Let's suppose that the "knowledge field" is a very simple one, like variation of a quantity (say, output production) through time (say, successive years), than one can compare the difference between communicating this information through: (a) a textual description; (b) a line graph.

(a) "In 1983 production was 420 units and it went up to 650 units in 1984 with an increment of 230 units. It then went up to 1000 units in 1985 with an increment of 450 units from the year before, to 1500 in 1986 with an increment of 500 units, to 2200 in 1987 with an increment of 700 units....One can thus observe that not only production has been constantly increasing from 1983 to 1990 but also increments have been increasing themselves every year during the same period and that, furthermore, the rate of increase has been itself constantly increasing . . ."

Anybody will find grasping and understanding through (b) much easier than through (a). In the first case one must decode verbal symbols extracting their referents, keep in memory the successive pieces thus processed, mentally compare them and, eventually, "assemble" the overall structure. Nothing of this sort happens in the second case: one looks at the graph and "sees" this structure: increase in production, rates of increments, changes in the rate of increments, and so on are directly "perceived" as they are in their reciprocal relationships and do not have to be mentally reconstructed or assembled. The crucial limitation of the linear structure of the textual form also clearly emerges from the arbitrariness involved in the mapping process. Rather than (a) above, we could have (a'):

(a') "Production steadily increased in the years from 1983 to 1990, going from 420 to 650 to 1000 to 1500 . . . . to 9500, respectively in each year. During the same period increments have been themselves constantly growing , going from 230 units between the first and the second year to 450 between the second and the third, to 500 between the third and the fourth . . . . They have been especially strong during the last three years . . ."

There is no principled way of choosing between these two (or among the several) possible textual forms: since the structure of the field to be communicated is not linear, forcing it into a linear form will always involve non-univocal, arbitrary choices. On the other hand, since the line graph matches the structure of the knowledge field, its form is univocally determined and non- arbitrary. Examples can be easily multiplied by changing the knowledge field. Suppose this is of spatial nature, then one can compare the textual form with another kind of communicative form: the so- called "plan" or "map", as in (c) below.

There is no need of formulating here the textual form: anybody can do it for himself and compare its effects on understanding with (c). They will be similar to, though larger than, those noted for (a) and (b) above, since the knowledge field is in this case slightly more complex and richer. The important point is that here too the structure of the communicative form (c) matches (is isomorphic to) that of the knowledge field, and hence one can see and grasp it directly, rather than having to reconstruct it mentally from the linear succession of a textual description.

One could object that such transparent, isomorphic communicative forms (such as the line graph or plan showed above), though much more appropriate than the textual form, cannot be applied to much larger, richer and/or more complex knowledge fields: it is our contention that this is precisely what hypermedia systems allow us to do and should be used for. Our definition of hypermedia is, therefore, a structural one and not a content one: hypermedia in not hypertext whose "nodes" or documents belong to communicative media other than text, as commonly accepted in the literature (Nielsen, 1990). Rather, it allows the very structure of communication to be taken from media other than the linear text (such as 2D- or 3D-graphic design) and modelled on the structure of the knowledge field. How this can be done, we have tried to show by building a hypermedia system for a relatively large and rich knowledge field: that of descriptive geography. Since geographical knowledge is intrinsically structured in a spatial way (i.e., phenomena are organized in terms of relative locations, distances, dimensions, etc.), the communicative structure we choose to match this organization is that of a map. The result is HYPERMAP (Antinucci, 1992; Berlinguer, Meloni, Troise, 1992), the system we describe below together with the results of its testing. There are reasons to believe that a change in the structure of communicative form along the lines proposed here might strongly affect learning. When we talked above about differences in the cognitive process of understanding we used (informally) terms such as "decoding symbols", "referring", "mentally reconstructing and assembling" for the textual form, as opposed to "seeing", "perceiving", "grasping", etc. for the transparent isomorphic form of the line graph, plan, map. Psychologists (especially developmental psychologists) have long known the existence of two different modes of cognizing: a symbolic one and a sensorimotor one (Piaget, 1951; 1971; 1974). It is customary, however, to pay attention only to the first mode and associate "knowing", "thinking" with the symbolic mode. Many scientists and researchers in fields outside of psychology have repeatedly pointed out and stressed (often reflecting on their own experience) the role of what as been called "non-verbal", "non-symbolic" or "visual thinking" (Arnheim, 1969; Penrose, 1989; Ferguson, 1992). Correlated to these two different modes of cognizing, two different forms of learning have been traditionally recognized. One, variously termed "explicit", "declarative", "relational", etc., and the other, called "implicit", "nondeclarative", "habituational", etc. (a somewhat similar contrast is expressed by the phrases "learning that" and "learning how"). The prototypical cases of learning where these two forms can be easily contrasted are, for example, learning biological taxonomy vs. learning to ride a bycicle. Notice, in fact, that the first is explicit and step-by-step conscious: I can state declaratively what I learned and how I learned it . The second, on the other hand, is implicit and largely unconscious: I learn but I cannot describe what I learned or the process through which I learned. Traditionally, it has been assumed that these two forms are dependent, as in the example quoted above, upon the nature of the object that is to be learned: notions (biological taxonomy) vs. abilities (riding a bycicle). Both the literature on non-verbal, non-symbolic cognitive processing (such as that quote above) and recent research on memory and learning (Squire, 1992; Berry and Broadbent, 1988) tend to show that this concept is wrong and that a most important factor (though not the only one) in determining the form of learning is precisely what we have called the communicative form. Which of the learning mechanisms is activated does not depend on the "content" but rather on the form through which it is represented: implicit learning, if this form allows (or favours) sensorimotor processing; explicit learning, if it requires symbolic processing. Notice that this phenomenon is well known for a class of "objects" of learning that, though notion-like, have some perceptual and motor component. One example is learning to operate and program a videotape recorder. If the form in which this knowledge content is communicated is a booklet of instruction, I'll be forced to use the explicit learning mode. If this form, is whatching somebody who actually operates the VTR or (even better) observing the results of my own trying to operate it directly, then learning will be channelled through the implicit mode. The problem is that almost all notional fields of knowledge are today communicated through the "booklet" form. Changing the communicative form along the lines described above through the use of hypermedia systems, might, then, shift the learning process to the implicit mode with respect to knowledge fields that have always been the object of explicit learning only, thus making their acquisition more similar to the learning of abilities. The consequences of this shift might have great relevance for teaching and instruction: implicit learning tends to be much more powerful, effortless and less sensible to background and individual variations. First results obtained in testing learning through our hypermedia system HYPERMAP in a real school environment seem, as we shall see below, to support this statement.

HYPERMAP

HYPERMAP is a hypermedia system designed and build according to the theoretical hypoteses just stated, i.e. to try to implement an isomorphic, "transparent" relation between the structure of the communicative means employed and that of the field of knowledge that must be communicated. HYPERMAP domain is descriptive geography, i.e., the physical, economical, institutional, political and cultural characters of the countries of the world. The architecture of the system is a map of the entire world that can be freely and continuously navigated in all four cardinal directions (North, South, East, West) and can be zoomed at five different levels of scale: from continent to single region. Contrary to what happens in traditional atlases, at each level maps are rigorously drawn at the same scale, so that proportions and distances are correctly and continuously represented for the whole world. All information is accessed from the physical points of the map it belongs to, so that it is not only always situated in space but also automatically relevant to the scale shown. Thus, if the map is, for example, at the country scale, institutional or economical information will be given at the level of a whole nation; if the map is at the district level, on the other hand, economical or demographic data relative to the specific area will be displayed. The information is contained in documents that simultaneously combine different media (written text, films, pictures, diagrams, spoken and music pieces, etc.), each performing a different function. Care has been taken in devising a format for the combination of different media capable of maintaining a high level of attention and interest, while avoiding wrong and ineffective communicative forms, such as reading (long) written texts on the screen. Documents are also connected in a thick network of hypertextual links that enable the user to construct an unlimited number of trails following his own personal interests and associations. Traversing a hypertextual link will also automatically show, whenever relevant, the area of the map to which the new information refers. The specific hypothesis underlying the design of the system is that this field of knowledge is intrinsically structured (KS) in a spatial way; i.e., spatial relations, positions, and dimensions are essential in order to understand and master the phenomena of the field. The (multi-layered) map (CS) allows one to present directly this structure, that is, without having to translate it into an unnatural and inappropriate linear structure, as it happens in ordinary textbooks. Thus, for example, common change of scale in traditional atlases or book maps induces the very frequent error of believing that, say, France is as large as Saudi Arabia, which generates difficulties in the understanding of differences in phenomena such as demographic distribution, transport and communication infrastructure, etc. HYPERMAP communicates this information perceptually through map size (scale), navigation (distance), location and representation and not symbolically through a textual description. HYPERMAP is not a prototype only, but a system fully developed to test its validity: it is one of the largest existing hypermedia systems, including, up to now, 1491 audio files totalling approximately 16 hours, 2915 images, 1314 text files and 553 full-motion video sequences totalling approximately three and one half hour. Its database currently occupies 12 CD-ROMs. It has been developed by our Institute (funded by Progetto Strategico Comunicazione Didattica Mul- timediale e Insegnamento a Distanza) in collaboration with ENEL, the Italian National Electric Company. Its software has been implemented by Infobyte. A large effort has been devoted to allow it to run on standard hardware and software 11 and to keep it accessible to young school children. HYPERMAP is currently being systematically tested and evaluated, under the authority of the Italian Ministry of Public Education, in three junior high schools. Here below we report the first results of this experimentation.

Testing HYPERMAP

Subjects and procedure

Experimental subjects were two classes of the first year of junior high schools, for a total of 43 students aged between 11 and 13 years. The two classes followed all the traditional courses, class and home activities except for geography that was completely abolished. It was substituted by a weekly one-hour long session of individual work on HYPERMAP. The students were demanded to explore, in whatever way they liked, six European countries: Switzerland, Austria, Ireland, Finland, Holland and Greece. Teachers were explicitly instructed not to interfere in any way in the experimental course and behave as if geography did not exist as a school subject matter. Experimental subjects were systematically compared with a control group, consisting of two classes of the second year of the same junior high schools, for a total of 39 students. Since they belonged to the second year, these students were on average one year older and had, obviously, a year more of schooling than the experimental ones. This unfavourable choice of the control group, which is reflected by its average higher initial knowledge, was due to the way in which traditional geography curriculum is organised in the three years of junior high school: European countries, on which the learning comparison could take place, are dealt with only during the second year. The control group followed in every respect the traditional curriculum, including teacher's oral lectures, assigned home work and class testing. Teachers were simply asked to follow their traditional program relatively to the same six countries of the experimental group. Class time allotted to geography was, as in all ordinary classes, two hours per week., i.e. twice the time the experimental group spent on HYPERMAP. Both experimental and control subjects were individually tested with the same materials and procedure at the beginning of their course, in October 1993, and then at the end of the first quarter, in February 1994. In this time lapse, each experimental subjects took part to about 15 work sessions with HYPERMAP, while the control subjects followed about 30 hours of geography classes. The individual testing procedure was articulated in two parts. The first one consisted of a questionnaire (17 items) for each of the countries asking about physical, economical and institutional characters of the country. Answers to each item were scored according to a four value scale, where 0 was incorrect or no answer, 3 fully correct answer, and 1 and 2 were intermediate. Scoring was effected blindly by two research assistants different from the testers, who had been trained to a 92% level of agreement. The second one in a task of spatial knowledge in which subjects were requested to locate correctly the name of 15 European countries on a mute map (i.e., a map stripped out of any name). Learning was evaluated on the basis of differences between pre-test and post-test performances.

Results

Both groups showed a significant improvement of their performances on the questionnaire about the six European countries (2way ANOVA, dof 1, 80, F = 71.9, p < .0001). In particular, experimental subjects' mean value passed from 38.1 to 74.7 (increment: 95 %), while control subjects mean value passed from 46.7 to 82.4 (increment: 76%). In the pre-test session control subjects showed higher scores for all the six countries tested, with differences reaching statistical significance in two cases (Switzerland and Holland). On the contrary, in post-test session, no difference between the two groups was significant, and furthermore the experimental subjects' score was higher than control ones' in two cases (Holland and Finland). Thus, both groups increased their performance but the experimental group increased it more. A remarkable effect can be seen in the distribution of performances in the two groups. While in the pre-test condition the two group show a similar inner variance (in fact, slightly higher in the experimental group), in the post-test condition the experimental group markedly reduced its internal variance while the control group increased it, as shown in the two figures below.

Results on the spatial ability task of matching country names to the mute map show a trend closely similar to the one discussed above. Experimental subjects passed from a mean score of 17.1 to 22.4 (increment: 31%), while control subjects passed from a mean score of 20.4 to 23.9 (increment: 17%). The ANOVA performed on these data shows, beyond a significant improvement in performance of both groups (dof 1, 80, F = 63,6, p < .0001), an interesting effect of treatment, plotted in the figure below, which approaches statistical significance. This effect mirrors in differences between groups assessed with Tukey HSD test: they are significant in pre- test condition (p = .0004), but they are not in post-test condition (p = .21).

Thus, the increase in performance of the experimental group tends to be significantly higher than that of the control group, as shown in Figure 2 above. Furthermore the same effect of a higher reduction of inner variance in the experimental group is present also in this task: SD decreases from 8.8 to 6.8 in the experimental group, while it decreases from 7.8 to 7.1 in the control group. It is interesting to note that a difference in performance increment occurs also when considering the subset of countries not directly studied. From pre-test to post-test, the capacity of locating correctly these countries passes from a mean score of 10.5 to 12.5 in the experimental group (increment: 19.1%), and from a mean score of 12.6 to 13.7 in the control group (increment: 8.7%).

Discussion

Results presented above show a quite homogeneous pattern: in all tasks the experimental subjects improve their performance more than the control subjects. As a consequence, differences between groups found in the pre-test, as a function of older age and longer schooling of the control group, drastically diminish or disappear in post-test. Furthermore, HYPERMAP produces a most relevant effect in terms of homogeneity of performance: learning through HYPERMAP "normalises" the distribution curve of scores, while learning through traditional methods tends to increase the inner variability of the group. In other words, pre-existing extreme differences among subjects tend to be overcome with the hypermedial learning, while they are widened by traditional learning. It is important to stress that the result of the experimental group have been obtained with half of class time dedicated to the subject matter with respect to the control group and with no time at all dedicated to home work and, finally, with no teacher's tutoring whatsoever. It is, furthermore, interesting to report the personal impression of the students when we asked them, at the end of the quarter, how did they like their experience on HYPERMAP as compared to that of traditional class-work. A common report was: "It was very amusing to play with the computer , but when do we start studying?" Both the results on the increase of performance, and, especially, those on the decrease of inner variability appear to support, when coupled to the free modality of use, to the students' subjective impression of playing and exploring rather than studying and to the absence of any explicit, directive tutoring, the hypothesis that a hypermedia system like HYPERMAP shifts the learning process to the implicit mode, as stated and defined above.

References

Antinucci, F. (1992).
Ipermappa. Multimedia, 5-6: 52-54.

Antinucci, F. (1993).
Summa Hypermedialis (Per una Teoria dell'Ipermedia). Sistemi Intelligenti, 2: 227-257.

Arnheim, R. (1969).
Visual Thinking. University of California Press, Berkeley.

Berlinguer, L., Meloni, D., Troise, A. (1992).
Ipermappa. Golem, 11: 15-20.

Berry, D.C., Broadbent, D.E. (1988).
Interactive Tasks and the Implicit-Explicit Distinction. British Journal of Psychology, 79: 251-272.

Ferguson, E.S. (1992).
Engineering and the Mind's Eye. The MIT Press, Cambridge.

Nielsen, J. (1990).
Hypertext and Hypermedia. Academic Press, New York.

Penrose, R. (1989).
The Emperor's New Mind. Oxford University Press, Oxford.

Piaget, J. (1951).
Play, Dreams and Imitation. Norton, New York.

Piaget, J. (1971).
The Construction of Reality in the Child. Ballantine Books, New York.

Piaget, J. (1974).
The Origins of Intelligence in Children. International University Press, New York.

Squire, L.R. (1992).
Memory and the Hyppocampus: A Synthesis from Findings with Rats, Monkeys and Humans. Psychological Review, 99: 195-231.

Hardware:

PC 486, 8 MB RAM, 100 MB HD, SVGA colour monitor, ACPA audio board, Action Media AM/2 (DVI) board, two multidisc CD-ROM players (6 CD-ROMs each), microphone.
Software: proprietary software written in C++, running under Windows 3.1.