Novice Linking in Hypermedia Environments

Stephen W. Harmon, Ed.D.
Assistant Professor of Instructional Technology 
Georgia State University
Sharon Dinsmore

The University of Houston-Clear Lake

Technological innovation has necessarily kept pace with the expansion of readily accessible information. To be sure, the information explosion is directly attributable in large part to technological innovation. New technologies allow us to deal with more massive and disparate sources of information than ever before, and to do so more easily than we were able to deal with much smaller, more localized information just a few years ago. When microcomputers first became popular, almost everyone who owned one was a computer programmer; users almost had to be programmers in those days. But as time went on and computers became easier to use, they became more complicated to program. The number of amateur computer programmers dropped correspondingly. This decrease continued until the advent of object-oriented, natural language authoring systems in the late 1980's. Almost overnight, virtually anyone who had access to a microcomputer and wanted to create computer programs could - and many of them did. However, it is this very ease of programming that may be leading educators into trouble.

The number of new computer programs available, particularly educational programs, has increased dramatically over the last five years. Since most software authors want their programs to be in the vanguard of educational theory and since one of the first widely distributed authoring systems dealt with an exciting new concept in cognitive theory, most of these new authors began making what they said were cutting edge programs. If you have followed computer-based education over the last few years you must be aware by now that it is hypermedia of which we speak. The promise of these hypermedia authoring systems was that "any expert on any subject should be able to produce . . . attractive, accessible, and powerful instructional applications with a moderate investment of time and energy" (Morariu, 1988, p. 17) With the advent of HyperCard in 1987 and Apple Computer Inc.'s decision to bundle it free with every new Macintosh computer they sold, the hypermedia bandwagon was off and rolling. What few people realized was that it was rolling down a pretty steep hill. Today, it is in some danger of rolling out of control.

Hypermedia is a combination of text, sound, still pictures, graphics, and motion video arranged non-linearly in linked nodes so that a user can go directly from one node to any other. It should not be confused with multimedia, which combines text, sound, still pictures, graphics, and motion video but is unconcerned with the way nodes are linked together (Reeves & Harmon, 1991; Halasz & Conklin, 1989). What makes the concept of hypermedia exciting for educators is that it closely models semantic network theory (Jo nassen, 1989). Semantic network theory holds that people store information in linked web-like node structures (Anderson, 1973). People create networks by relating different nodes of information to each other in what becomes a meaningful way for them . For one person a node for the color green might be linked to nodes for grass, money, elves, and men from Mars. For another person green might be linked to eggs and ham, cheese, and a golf tournament. Cognitive psychologists and educational researchers believe that the best way to help learners acquire new knowledge is to present the knowledge in structures that the learner can easily relate to existing network within his or her own memory (Jonassen, 1989). As Dede and Palumbo (1991, p. 16) put it, "our adeptness in quickly storing and retrieving large amounts of information seems to stem from this property of associativity."

So it is no wonder that developers of computer programs systems have embraced hypermedia wholeheartedly. In fact, development of hyperme dia systems has far outstripped research on them (Heller, 1990; Jonassen & Grabinger, 1990; Tsai, 1988). Most developers assume that because hypermedia ought to be good for instruction it is good for instruction. Those of us who have been involved in educational research for any length of time recognize this as a potentially dangerous assumption, just as surely as we recognize that common sense is neither. As Heller (1990, p. 436), so ably points out, "the problem areas within hypermedia and HAI [hypermedia-assisted instruction] are well documented but there needs to be more research done on how one comes to understand the extent and interconnections of the material represented in a hypermedia system." One method of arriving at some understanding of the innate instructional ability of hypermedia systems is to examine the ways in which learners actually use them. This study looks at how learners used a hypermedia system to form semantic associations among nodes of information that were not formally linked by the system designers in advance. Hypermedia proponents assume that learners will form semantic networks similar to those formed by experts, and many of today's educational hypermedia systems are based on this assumption (Fischer & Mandl, 1990). This study begins to examine the validity of that assumption.

Methodology
The goal of this study was to evaluate how subjects, who were novices at learning with hypermedia systems, would carry out a task of linking nodes of information in a hypermedia system which presented a relatively unfamiliar topic. We constructed a very simple to use hypermedia system which provided subjects the opportunity to read about and watch video clips of topics related to the conflicts between the Israelis and Palestinians over the land now called Israel. The hypermedia system was designed to present the topics as separate entities with no pre-imposed structure so that it would not suggest any relationships between topics.

The s ubjects were asked to spend one hour browsing the system and constructing links between the topics in any way that made sense to them. In addition to having the subjects browse the hypermedia system and construct links, we asked questions occasionally about the reasoning behind a linkage. This data was collected to help determine the types of links they were constructing. In addition, subjects completed a pre-treatment survey, a post-treatment questionnaire and a post-treatment individual differences measurement test.

Subjects
Students in two graduate level "Introduction to Educational Computing" courses were chosen as subjects for several reasons. First, these students already had a general knowledge of computers and a specific knowledge of how to use the treatment system, the Apple Macintosh. Second, as newcomers to educational computing they were likely to be inexperienced in learning or teaching with hypermedia systems. Third, we felt that a few students at most might already have a more than general knowledge of the chosen topic - the conflict over the Holy Lands in the Middle East. Fourth, as graduate students in the College of Education, these students have a similar education level.

Subjects were recruited by asking students in the two classes to volunteer to participate in a study of "Hypermedia and Learning". Initially, everyone in the two classes signed up, but due to scheduling conflicts, only 24 actually completed the study.

Instrumentation
The hypermedia system was created using the Apple Macintosh, HyperCard 2.0, a laser videodisc player and the ABC News Interactive Videodisc entitled In The Holy Land. So the subjects could complete the task in one hour, HyperCard was used to construct a system which provided access to a limited set of the topics (or nodes) in the videodisc and allowed the subject to browse the nodes and link them easily.

The main screen (card) of the hypermedia system displayed the 25 available nodes as icons. The icons were purposefully presented as separate entities scattered on the screen in no pre-defined order. To view a node a subject simply clicked on the appropriate icon. Each node presented one card of information with options for viewing related video-clips or occasionally for getting more detailed information. Each node card also included a button for returning to the main system screen.

( Figure 1 )

The hypermedia system also included a very simple to use mechanism for setting up links between related nodes. Buttons for setting up the link appeared on the bottom of the main screen. The subject simply pressed the "link to " button followed by the icon for the first node to be linked. Next he or she pressed the icon for the second topic to be linked and pressed a second button to confirm the link. As each link was constructed, a line showing the link between the two node icons appeared on the main screen. The system was constr ucted in this manner, so that subjects could quickly alternate between browsing and linking and also keep track of the links they had established.

(Figure 2 )

The hypermedia system was designed to record all the links that each subject made to ensure that complete, accurate data was available for analysis.

Pre-treatment Survey
A computer-based pre-treatment survey was constructed to gather general background data from each subject. In addition to demographic questions, the survey included questions regarding: hobbies, special interests, teaching specialities, knowledge of and experience with hypermedia systems, and familiarity with the conflicts over the Holy Lands. The purpose of these additional questions was to determine whether any subjects had more than general knowledge of either hypermedia learning or the Holy Lands conflicts and issues. Students completed this survey in class prior to using the hypermedia system.

Post-treatment Interview
After using the hypermedia system, subject s were interviewed to provide insight into their experience in using the system and what they "learned" from using it. A structured interview process (Bogden & Biklen, 1982) was employed as well as a constant-comparative data analysis/gathering strategy (Glaser & Strauss, 1967). Structured questions included:

1. What was your strategy for going through the system?
2. Who is right?
3.
How do you feel about using this sort of hypermedia system to learn something?
4.
 How would you describe your reading habits?
5.
What did you know about the topic (Holy Lands conflict) prior to using this system?
6.
What other comments do you wish to share regarding this system or hypermedia?

Interaction Procedure
Prior to participating in the study, all subjects took the pre-treatment questionnaire during their regularly scheduled class time. The "interaction" with the hypermedia system was conducted on an individual basis due to the nature of a videodisc-based hypermedia system and the need for the researchers to observe each subject during the treatment. Upon arrival for the treatment, the subjects were given brief instructions on how to use the customized hypermedia system, and asked to express how much they felt they knew, and to report what they actually knew about the content area. Next, the subject was asked to spend the next hour browsing the nodes and setting up links between any nodes they felt were related. Subjects were ask to verbally report (Ericsson & Simon, 1984) on the types of links they were making. They were also advised that the researcher would be available to assist them, would be observing them and would occasionally ask for a more detailed explanation of why a particular link was constructed.

When an hour had passed, the subject was told that time was up and was interviewed. The researcher recorded the subject's oral answers to the post-treatment questions, thanked him/her for participating, and finally gave a brief synops is of the purpose of the study. After the subject left, the researcher chose an option from the system to save a list of links that the subject had created.

 Results
Twenty four subjects completed the study. Complete data sets were retained for all subjects. Subjects generally had little trouble becoming familiar with the interface. They typically were immediately able to browse from one node to the next and were able to confidently make links on their second attempt. Only two of the subjects reported any extensive prior knowledge of the content area beyond having heard of the some of the names and events included in the study. Interestingly, both of these subjects either were currently, or had at one time been career military officers. One of these subjects had read several books on the content area. Some descriptive statistics were compiled in order to give an overview of the data sets. 300 different links were mathematically possible in th e system, though not all nodes bore any semantic relationship to each other. The mean number of links made by subjects in the study was 20.71 with a standard deviation of 11.82. The range of links was from 8 to 69.

Statistics were also compiled for links by node. For the 25 different nodes combined,a mean number of 39.8 total links were created by participants, with a standard deviation of 13.99. Each node had a mean of 13.44 different links established, with a standard deviation of 3.94. Each node had a mean number of 3.8 instances in which it was not linked at all, with a standard deviation of 2.93. Table 1 presents the means and standard deviations obtained in the study.

Category
Mean
Std. Dev.
Total Links per subject
.20.7
11.82
Total Links for All Nodes
39.8
13.99
Total Links per Individual Node
.13.44
3.94
Instances Where Nodes Were Left Unlinked per Node
3.8
2.93

Table 1

Means and Standard Deviations of Links by Subjects and Nodes
As might be expected, a fairly strong significant correlation (p = .0001) was observed for total links by topic and number of different links (r = 0.673), indicating that where more links were made more different topics were linked together. Somewhat surprisingly, there was also a fairly strong positive correlation (p = .0001, r = 0.672) between total number links per card and instances of cards receiving no links. This correlation indicates that subjects linked some topics either quite a lot or not at all. One reason for this might be that some nodes were difficult to understand and subjects either failed to see that they related to anything or else failed to see how they didn't relate to everything. Table 2 below shows all of the nodes with the total number of links, number of different links, and instances of no links for each node.

Node Total # Links Links to Different Nodes Instances of No Links to Any Node

  •  Balfour Declaration 36 11 2
  •  United Nations 44 13 2
  •  Terrorism 65 19 0
  •  Security for Israel 54 22 2
  •  PLO Violence 52 16 4
  •  War and Refugees 42 17 3
  •  Deir Yasin Massacre 22 11 8
  •  Sanctuary 43 14 3
  •  The Intifadah 18 9 9
  •  Yom Kippur War 30 14 6
  •  Palestinian Homeland 59 20 2
  •  Fedayeen Attacks 20 9 7
  •  UN Resolution 242 31 7 6
  •  War of Independence 30 9 5
  •  Children of Palestine 39 14 0
  •  Camp David 28 11 6
  •  Sabra and Shatilla 18 9 9
  •  British Mandate 34 9 4
  •  The Holocaust 54 16 0
  •  Satate of Israel Proclaimed 61 16 0
  •  Children of Israel 45 15 0
  •  Arafat and the UN 41 11 4
  •  Six day War 61 17 2
  •  Munich Olympics 33 11 3
  •  The Three Religions 35 16 8
Types of Links
More interesting themes emerged from the qualitative analysis of the interview and verbal report data. Eight different types of links were identified by the subjects. This sort of emic (subject generated) categorization is of particular interest since as Franz Boas remarks "If it is our serious purpose to understand the thoughts of a people the whole analysis of experience must be based on their concepts, not ours" (Boas, 1943, p. 314). In each case the term used here to describe the type of link was used by at least one subject. In many of the instances subjects used different but equivalent expressions to identify a particular type of link. We have chosen the words we feel best describe the types of links subjects made and have occasionally changed case of the words used as links types in order to maintain consistency. To increase inter-rater reliability, each investigator coded links separately, before combining to discuss and resolve discrepant cases. In the few instances where we disagreed, the disagreement was caused by different, but synonymous terms used to describe a particular link. Note that no attempt was made to determine the frequency of occurrence of any particular link type; we were interested primarily in identifying the different types.

Comparative
The first two of these eight link types are similar and opposite links. Subjects would identify a similar link when one node shared characteristics with another node. For example, several subjects linked a node about "Children of Israel" with a node about "Children of Palestine" because they were both about children. Subjects would identify an opposite type link when they felt nodes were opposites of each other. For example, many subjects linked "Terrorism" to "Sanctuary" because they felt they were diametrically opposed to each other. (Interestingly, some subjects linked the same two nodes because they felt terrorism caused the need for sanctuary.) Since these two categories appear to us to be different sides of the same coin, we treat them together and suggest that they be simplified into one category called comparative links. Nineteen of the twenty-four subjects made at least one link of this type.

Causal
The third t ype of link identified by subjects is causal. Subjects would call a link causal when they felt that one node was the result of another node. For instance, one common causal link subjects made was between the node for "State of Israel Proclaimed" and the "War of Independence." They felt that the War of Independence was a direct result of the proclamation of the state. Another was linking "PLO Violence" to "Security for Israel" because the violence created (or caused) the need for security. However, as with comparative linking of "Terrorism" to "Sanctuary," several subjects linked these two nodes together for a different reason. All of the subjects made at least one link of this type.

Sequential
The fourth type of link is sequential. Subjects identified a link as sequential when they felt that one node occurred chronologically soon before or soon after another, or even when they felt that two nodes occurred at the same time. For example, some subjects linked the "Six Day War" and "U.N. Resolution 242" because resolution 242 was passed immediately after the Six Day War. Again though, different subjects would make the same links for different reasons. At least as many subjects identified the link between "State of Israel Proclaimed" and the "War of Independence" as sequential as did causal. These two links types do seem logically related. Eight of the twenty-four subjects made at least one sequential link. Interestingly, the two subjects who had more extensive prior knowledge than the rest of the sample made no sequential links.

Associative
The fifth type of link identified by subjects is associative. Subjects used this word to describe links they made where one node explicitly mentioned another, or when they felt that one node was vaguely related to another and were unable to express a more concrete relationship. Subjects often identified "The Children of Israel" and "The Holocaust" as an associative link because the text for "The Children of Israel" n ode mentions that the children of Israel "are well schooled in the horrors of the Holocaust." They would identify "Arafat and the UN" and "PLO Violence" as being associated, but were not clear how. Twenty-two of the twenty-four subjects made at least one associative link.

Exemplary
The sixth type of link is exemplary. Subjects called links between nodes that illustrated each other exemplary. For instance, some subjects identified the link between "The Munich Olympics" and "Terrorism" as exemplary because the events at which occurred at the Munich Olympics were an example of terrorism. Similarly, they identified the link between "PLO Violence" and "Terrorism," as exemplary, saying terrorism was an example of PLO violence. Likewise they identified the link between "PLO Violence" and "The Munich Olympics" as exemplary saying the acts at the Munich Olympics were an example of PLO violence. Subjects rarely had trouble expressing this type of link. Eightee n of the twenty-four subjects made at least one exemplary link.

Componential
The seventh link type is componential. A link between two nodes was said to be componential when one node was a component of another one. For example, some subjects called the link between "UN Resolution 242" and the "Security for Israel" componential because one part of resolution 242 guarantees Israel's security. Similarly they called the link between "PLO Violence" and "Terrorism" componential because terrorism was one part of the grand scheme of PLO violence. Five subjects made at least one componential link. Both subjects who had greater prior content knowledge made links of this type.

Accidental
The eighth type of link identified by subjects is accidental. Subjects occasionally slipped when they were clicking the mouse or else misread a title when linking two nodes. They invariably noticed when this had occurred and were quick to call it to researcher's attention. Accidental links were the least common type of li nk observed in this study. This link type is interesting because of its potential to lead to serendipitous learning. Although no direct evidence for it was obtained in this study, it is possible that subjects could come to a new understanding of a relationship initially made accidentally.

Although subjects weren't always able to classify every link that they made, of the 497 total links made by subjects, all were either immediately placed into one of the seven categories (combining similar and opposite into comparative) by the subjects themselves or in retrospect by the researchers. It is important to remember that these seven link types may not represent every possible link type. They only represent every link type made in this study with this particular set of nodes. Further research with different subjects and subject matter is needed to find the limits of the utility of this categorization.

Higher and Lower Order Links
Apart from the seven link types, two other major themes emerged from the study. The first deals with levels of links and the second with video and node size. In looking at link types on another level, two other classes of links emerged from this study, direct and indirect, or higher and lower order links. Direct or lower order links follow the pattern of A is related to B. Indirect or higher order link types follow the pattern of if A is related to B and B is related to C then A is related to C. Lower order links were the most common type observed in this study. All of the links mentioned above can be classified as lower order links. All of the subjects made lower-order links, while only three made higher-order links. Higher order links occurred only among those subjects who indicated having a greater conceptual knowledge of the topic area prior to beginning the study. The one subject who made higher-order links who did not indicate having a broad prior knowledge of the content area, did indicate having a good knowledge of, and a stro ng interest in, history in general.

One example of a higher order link was made by a subject when he in rapid succession linked "Arafat and the UN" to "PLO Violence", "PLO Violence" to "Terrorism," then "Arafat and The UN" to "Terrorism." In another example that same subject liked "Sanctuary" to "PLO Violence" because in his words "countries around the world give sanctuary to the PLO." He then linked "Sanctuary" to "Terrorism" because since PLO violence is most often expressed as terrorism, "those same countries are supporting terrorism when they give sanctuary." Interestingly, this subject had read several books about the content area. This subject made 69 links overall, 4 standard deviations above the mean. Higher-order linking appears to indicate a greater degree of assimilation of information than direct linking, hence the higher and lower order distinction.

Video Nodes
One final interesting theme to emerge from the study is that the video segments appeared to be detrimental to purposeful linking. Subjects were frequently observed watching a video segment, then appearing to have forgotten what they were doing before they watched the video. In fact, subjects almost invariably had a hard time making any links at all for a few minutes after watching a video segment. This phenomenon is also reflected in the fact that the students who made the most links watched the least video and vice-versa. One possible explanation for this observance is that the video segments were too long in length. They typically lasted between forty-five seconds and two minutes. The video segments may have actually represented several nodes worth of information. After watching a video segment, subjects appeared to be unclear as to which node they were on, perhaps because they had just encountered several nodes disguised as one. Three subjects actually expressed aloud being confused by the video. One subject remarked "I'm not watching any video because there's too much to do. I can't make connections and watch video at the same time, that's too much stimulation." Twenty-three of the twenty-four subjects looked at video at least once; Seventeen looked at video only once.

General Conclusions
The results of our research indicate that subjects gave seven types of reasons for creating their links. We consider this to be a positive finding. The variety of link types that subjects created without prompting from us is a positive indication of the diverse goals which similar hypermedia systems can be designed to address. In addition, because comparative, associative and causal linkages were among the types of links subjects created, hypermedia systems organized in similar ways may be useful for developing critical thinking skills. One of the goals mentioned by many educators and education reform advocates is that to survive and prosper in the 1990's, and beyond, American students need to develop critical thinking skills. Hyper media enthusiasts have been claiming that hypermedia has the potential to help students develop those skills. While we cannot claim to have proven this for all hypermedia systems, we believe that our subjects did demonstrate the use of critical thinking skills in establishing those links. In Bloom's taxonomy of learning (1956), comparison, association, and causation are analytical skills. Since the subjects in this study generated links which they subsequently classified as being of one or the other of these types, it would seem that they used these analytical thinking skills to arrive at the linkages. Because subjects were able to relate previously unrelated pieces of information in a non-random fashion, they must have to some extent analyzed that information in the process. When these analyses led them to create However, we feel that our findings were definitely influenced by the fact that we forced active linking of information. We also believe that the subject and nodes of information chosen for our system could have influenced the type of links subjects chose to create.

Dansereau (1978, & Dansereau et al., 1979) advocates a general schema training approach called "networking." In networking he suggests that learners be trained to recognize six types of links between nodes of information. These are: Part links, Type links, Leads-to links, Analogy links, Characteristic links, and Evidence links. Learners read a passage of text, then create a "node-link map" on paper. They can then relate or link information nodes on the map by classifying them as one of these link types. Links represent the way the ideas represented by the nodes are interrelated. Clearly this strategy relates closely to the link types identified in this study. It may be that experts in a content domain or in working in hypermedia environments would exhibit link types more closely related to those suggested by Dansereau. Although McKeachie (1984) suggests that this networking strategy is dif ficult and time-consuming to learn and employ, it would be well worthwhile to examine how well it works in hypermedia environments. This study suggests that training in networking is highly applicable to work in hypermedia environments, and should be researched. Subjects in this study had no training in networking. Whereas Dansereau's link types are etic, the types identified in this study are emic, coming from the subjects, not the researchers. At the time analysis of the data for the this study was conducted, neither researcher was familiar with Dansereau's work.

However, we feel that the results of our study also suggest that caution should be used before choosing hypermedia to address some educational goals. Based on our observations of the subjects, their answers to our questions, and the linkages they established, it is clear to us that there is no way to guarantee that specific learning objectives will be met with a completely non-linear, unstructured system such as that used in this study. Based on p ost-treatment interviews, we felt that some subjects left the study with an incomplete and/or inaccurate understanding of some of the facts presented in the system. (One subject left the study thinking Camp David was a concentration camp in World War II.) Because of this finding, we believe that we can only say that these systems allow learners to construct their own world views, and that there is no evidence that these newly constructed views will be consistent with the norm. More research must performed before it is clear what features hypermedia systems must incorporate in order to allow learners to achieve more specific learning objectives.

This study gave us the opportunity to observe subjects actually using a hypermedia system and to gain some new insights into potential advantages and disadvantages of non-linear, learner-controlled instructional systems. One aspect of that observation was eye-opening for us and we believe should be considered caref ully in the development and implementation of hypermedia instructional systems (HIS). To be precise, subjects occasionally mentioned during their interaction that they were avoiding particular topics (like violence and terrorism) because although they knew things like that happened, they didn't like to hear or see the details. This suggests that allowing students to choose the topics they review in a hypermedia system could actually allow them to bias their learning because they are unwilling to explore certain topics. Much emphasis has been placed on giving students access to raw information so that their learning is not biased by the opinions held by their teacher, however, we have not seen anyone mention this equally unsettling prospect of students censoring their own learning due to personal preferences. While most hypermedia proponents extol the virtues of students being able to come to their own conclusions based on raw, uncensored data, we wonder if they have considered how valid those conclusions might be if t he learner refuses to expose himself/herself to certain facts or theories.

Further, the size of a hypermedia system may turn out to be a critical factor in its success. Although much emphasis has been placed on how we can develop hypermedia systems which encompass great volumes of information, we believe that too large a system might actually hamper the hypermedia learning process. Our system used only 25 nodes of information taken from a larger system of over 400 nodes. The subjects had a difficult time reviewing that much information in the one hour allotted and some of them said they had difficulty maneuvering among that many options.

Limitations Evident within the Study
Due to nature of the topic chosen for this hypermedia system, certain types of links might not have been possible or likely. Therefore, we do not believe that our seven link types are necessarily the full extent of link types possible or that they are completely representati ve of link types likely to be found in all hypermedia knowledge construction systems.

Due to the limited time available for the study, the subjects were artificially limited to one hour to explore the topics. This limitation possibly precluded the viewing of many videodisc clips and /or the careful reviewing of data nodes. This time limitation therefore could have impacted our findings.

Subjects often found it difficult to verbalize link types. This difficulty may have had two impacts on the study, 1) in identifying all of the different link types subjects made, and 2) in interfering with the subjects' progress through the system. Being forced stop and verbalize reasons for links may have caused the subjects to progress through the system in a manner different than they might have otherwise. The verbalization may have also affected the number of links subjects made.

The video-clips available often were too big to be single nodes, often containing information about more than the topic they represente d. This was distracting to subjects in some cases and possibly caused confusion in what nodes were appropriate to link. The structure of the system utilized video-clips as part of the nodes (or topics) rather than as separate entities.

One other major consideration is that subjects had no extrinsic motivation for learning about this particular topic area. They were rewarded for participating in the study by being given extra credit in their courses, but they had no reason to want to learn about the Israeli/Palestinian conflict. Findings may have been different given a sample already studying this content area.

Suggestions for Further Research
Some of the conclusions already mentioned suggest the need for further research. For instance, investigation will be necessary to understand the optimal size of hypermedia instructional systems. Also, as mentioned earlier, research must continue into how hypermedia can be designed to ensure that specific ed ucational objectives are met by learners. There are two additional research directions which we would like to introduce here.

First, extensive research into the active versus passive hypermedia instructional systems is needed. Many proponents of hypermedia in education are emphasizing the virtues of hypermedia as a teacher presentation tool. While it may be more than multimedia due to its non-linear capabilities, it still amounts to relatively passive learning for students. In other cases, hypermedia systems are designed for use by individual students but still do not promote active learning by students. We feel that hypermedia systems which promote active learning by forcing learners to construct their own links between nodes of information will prove to be the most beneficial. However, we feel that more research is needed into how active versus passive hypermedia instructional systems will impact learning.

Second, our research results indicate that subjects with a richer prior knowledge of the subje ct created more links between nodes. We expect that there will be a difference between the way domain experts and novices create links between nodes. We feel that there is a need for further research to examine the way learners at different levels construct links in hypermedia systems compared to the way experts in the field construct them. If a goal of instruction is for learners to become experts in whatever area they are studying, every stop on the path to expertise should be mapped out and road signs put up for the travellers.

Hypermedia-based instructional systems have the potential to become the vehicle that carries education into the twenty-first century. They also have the potential to become the out-of-control bandwagon that plummets off a cliff, sending computer-based instruction to a much deserved grave in a jaded public's eye. It's time to take control of hypermedia before it gets away from us. Let's send it where we want it to go, letting r esearch be our guide.

References
Anderson, J. R., & Bower, G. H. (1973). Human Associative Memory. Washington, D.C.: Winston & Sons.

Bloom, B. S. (1956). Taxonomy of educational objectives: The classification of educational goals. Handbook I: Cognitive domain. New York: Longman.

Boas, F. (1943). Recent anthropology. Science, 98, 311-314, 334-337.

Bogden, R. C., & Biklen, S. K. (1982). Qualitative research for education: An introduction to theory and methods. Boston: Allyn and Bacon, Inc.

Cetron, M. J. (1988). Class of 2000: The good news and the bad news. , 23(6), 9-15.

Dansereau, D. F. (1978). The development of a learning strategy curriculum. In H. F. O'Neill Jr. (Eds.), Learning Strategies (pp. 1 - 29). New york: Academic Press.

Dansereau, D. F., Collins, K. W., McDonald, B. A., Holley, C. D., Garland, J. C., Diekhoff, G. M., & Evans, S. H. (1979). Development and evaluation of an effective learning strategy program. Journal of Educational Psychology, 79, 64 - 73.

Dede, C. J., & Palumbo, D. B. (1991). Implications of hypermedia for cognition and communication. Impact Assessment Bulletin, 9(1-2), 15-27.

Ericsson, K. A., & Simon, H. A. (1984). Protocol analysis: Verbal reports as data.

Fischer, P. M., & Mandl, H. (1990). Toward a psychophysics of hypermedia. In D. H. Jonassen & Mandl, H. (Eds.), Designing hypermedia for learning (pp. XIX - XXV). Berlin: Springer-Verlag.

Glaser, B. G., & Strauss, A. L. (1967). The discovery of grounded theory: Strategies for qualitative research. New York: Aldine.

Halasz, F., & Conklin, J. (1989). Issues in the design and application of hypermedia systems. In CHI '89, Tutorial #26 . Austin:

Heller, R. S. (1990). The role of hypermedia in education: A look at the research issues. Journal of Research on Computing in Education, Summer, 431 - 441.

Jonassen, D. H. (1989). Hypertext/Hypermedia. Englewood Cliffs, N.J.: Educational Technology Publica tions.

Jonassen, D. H., & Grabinger, S. R. (1990). Problems and issues in designing hypertext/hypermedia for learning. In D. H. Jonassen & H. Mandl (Eds.), Designing hypermedia for learning (pp. 3-26). Berlin: Springer-Verlag.

McKeachie, W. J. (1984). Spatial strategies: Critique and educational implications. In C. D. Holley & D. F. Dansereau (Eds.), Spatial learning strategies: Techniques, applications, and related issues (pp. 301 - 312). Orlando, FL: Academic Press.

Morariu, J. (1988). Hypermedia in instruction and training: The power and the promise. Educational Technology, 28(11), 17 - 20.

Reeves, T. C., & Harmon, S. W. (1991). What's in a name: Hypermedia vs. multimedia. Interact, In Press.

Trujillo, I. (1989). Academic Computing: The los andes project. , Summer.

Tsai, C. J. (1988). Hypertext: Technology, applications, and research issues. Journal of Educational Technology Systems, 17(1), 3-14.