33.8 Rational-Cognitive Aspects Of Choice And Learning

Several reviews of learner control in instruction (Hannafin, 1984; Milheim. & Martin, 1991; Steinberg, 1989) have identified two kinds of cognitive traits, prior knowledge and ability, which may explain some of the negative results of providing learners with choices during instruction. The relationships of learner control with achievement and with ability are presented in turn, together with some hypotheti-cal instructional prescriptions that could take advantage of these relationships.

33.8.1 Prior Knowledge

The review presented earlier in this chapter amply demon-strates that learners given control over their instruction too often make suboptimal choices. One possible explanation for these findings is that individuals do make appropriate decisions, but within their own perceptions of the problem at hand, not according to some optimal outside decision rules. This view suggests that an increase in an individual's accuracy of perception of his or her learning state in relation to the learning task should result in the individual's making more appropriate choices. Students are therefore expected to make instructional choices that are rational only to the degree they have accurate information about their current learning state. This suggests an approach based on learner prior knowledge or achievement.

There is substantial evidence that, left on their own, both children and adults very often overestimate how much they know about a given topic, and, indeed, those with more knowledge are often better able to judge their knowledge level than people fairly ignorant in that area (Flavell, 1979; Lichtenstein & Fischhoff, 1977; Nelson, Leonesio, Shimamura, Landwehr & Narens, 1982).

This finding that students are generally poor at estimating their current state of knowledge has been found also in computer-based contexts. Lee and Wong (1989) found stu-dents unable to predict their own learning of both general and specific types of knowledge. Additionally, Garhart and Hannafin (1986) and Relan (1991) found little correlation between self-rating of knowledge and performance on several tests. Garhart and Hannafin (1986) use this finding to explain plausibly why many students under learner-con-trolled conditions tend to terminate instruction prematurely.

It could very well be, then, that people often really don't know what they don't know, and that those who know very little know even less about what they don't know. (Apologies for the last sentence.) If this is the case, one might predict that students with higher levels of knowledge would make better (more judicious) instructional decisions than those with lower knowledge levels. Evidence for this phenome-non is provided by Seidel et al. (1975) and Fredericks (1976). In these learner-control CBI studies, high perform-ers were much more able than low performers to estimate their performance capabilities prior to their taking quizzes on the lesson material.

The notion that poor performers are incapable of judging how much they know has implications for the idea of instructional support, as well. That is, if students are unable to estimate their current state of knowledge, they may also be unable to assess whether they need additional instruction when given the chance to choose more. This would imply that a pretest given prior to instruction could predict the success of students given learner-controlled instruction, an extension of the achievement-treatment interaction para-digm of Tobias (1976, 1981), but here the instructional support is controlled by the learners. Such an interaction has indeed been found in studies by Gay (1986), Ross and Rakow (1981), and Ross, Rakow, and Bush (1980), although neither of these last two studies occurred in a computer-based environment. In all cases, college students scoring higher on a pretest performed as well under learner-control as similar students under program control. This was not the case for low-prior-knowledge students who per-formed much worse under learner control. It is plausible that low-prior-knowledge students were not as able to judge the instructional support they needed as were higher-prior-knowledge students.

Additional evidence within a CBI context is provided by Tobias (1987a), who found that knowledgeable students (as measured by a pretest) opted to, see more review material than did less knowledgeable students. He states, "…the presence of instructional support is no guarantee that less- knowledgeable students will use it frequently or effectively to improve learning" (p. 160). This is echoed by Judd et al.(1970), who found a similar result in their early study, namely, that students who needed additional instructional support tended to avoid seeking it.

In another paper, Lee and Lee (199 1) found that, consis-tent with the prior-knowledge hypothesis, students with the lowest levels of prior achievement related to their lesson topic performed poorest on a posttest when given learner control during the beginning phase of the lesson, the phase designed for learners to acquire their initial knowledge of the content area. However, another group of learners, also with low pretest scores, who were given learner control during a review phase of the lesson, performed the best of all the students in the study, even better than both program- control groups (both in the initial knowledge acquisition and in the knowledge review phases). They clarify the distinctive findings for the two lesson phases:

In other words, the LC strategy cannot function effectively when learners have to learn new materials. It makes intuitive sense that when LC subjects have to learn new materials [initial knowledge acquisition], they would work through CAL sessions under the pervasive influence of their previous knowledge base. Learners' management of learning activities is less influenced by previous knowledge differences when they have some grasp of the target knowl-edge [knowledge review] (p. 496).

In sum, prior achievement has been found to be a major factor affecting the effectiveness of learner-controlled instruction. Additionally, unlike many learner-control studies on other factors, this area of research is well grounded in relevant theory. Generally, it seems that students with some knowledge about the topic being taught seem better able to sense at any given choice point what they need from instruction and to choose additional instructional support accordingly. Here the key instructional variable seems to be amount of instructional support. Students with low amounts of topic knowledge have inaccurate perceptions of what they know, and consequently make poor use of needed instructional support.

Three possibilities are suggested for improving the effectiveness of learner control in relation to learner prior knowledge: (a) informing learners directly of their progress (i.e., supplanting the self-monitoring function); (b) instruct-ing students to try to gauge their current knowledge (i.e., activating the self-monitoring function); and (c) training the students to better monitor their learning.

The students' continual estimation of their level of knowledge (a metacognitive strategy, according to Flavell, 1979) affects the effectiveness of their choices. Without feedback data from the instruction about their knowledge level, students with more prior learning seem better able to assess what they do and do not know, and therefore how much more or what kinds of instruction (i.e., optional material) they need to see. Hannafin (1984), Milheim and Martin (1991), and Steinberg (1989) each suggests that learner control that regularly informs learners of the state of their learning might provide an aid, perhaps in the form of coaching or advisement to the students, in deciding whether they need more instruction. Additionally, Steinberg (1989) suggests that instruction should gradually wean the student from such crutches in order to promote more internalization of the metacognitive processes. Such information supports to the learner fall under the category of "decision aids," which have been shown to be quite useful in helping people make judgments and select appropriate courses of action (Pitz & Sachs, 1984).

Studies by Arnone and Grabowski (1992), Holmes et al. (1985), Schloss et al. (1988), Tennyson (1980, 1981), and Tennyson and Buttrey (1980) support the contention that providing students with updated information as to their moment-by-moment mastery level would improve the effectiveness of learner control over providing no such information. Here the researchers provide students under learner control with such information and show beneficial effects in comparison to students not given such informa-tion. A related study by Steinberg, Baskin, and Hofer (1986) showed that providing informative feedback to students during the course of a CBI lesson increased the chances that learner-controlled memory tools would be used. That is, students were able to use the feedback information to help them decide when and how to use the memory tools.

Results are not unequivocal, however. Ross et al. (1988) did not find an interaction between student selection of density of text displayed on the computer screen (high and low densities) and student pretest scores. Additionally, Goetzfried and Hannafin (1985) did not replicate in a CBI setting the achievement-by-treatment interaction that was demonstrated by Ross and Rakow (1981) in a non-CBI setting.

An additional wrinkle is suggested in results reported by Pridemore and Klein (1991), who compared selection of feedback by students (see 32.5.5.4) under two learner-con- trolled conditions differing in the elaborateness of feedback information provided. They found generally that students in the less-elaborate condition selected less feedback than those in the more-elaborate condition. The authors suggest that the amount of information contained in the feedback message helps students decide whether choosing to see such feedback is worthwhile. This might imply that stu-dents select their instructional support only to the degree they perceive it will help them. It's possible, then, that students choose to experience more instruction, not just on their perceived learning nee I d but also on the perceived usefulness of the material to be offered. Instruction then, designed for learner control, should have as its goal the
expansion and clarification of the student's own perception of the task as well as their progress toward it, particularly for those who are deficient in the accuracy of their self-monitoring. This notion builds on the prior-knowledge hypothesis, in that because students With low knowledge of an area tend to bypass additional optional instructional support, an improvement in the accuracy of perception of students as to their own knowledge level or to the task requirements would be expected to result in better decision making during learner-controlled instruction.

However, it is not known at this point whether students even need to be aware that self-monitoring of knowledge is important in learner-controlled instruction as a type of learning strategy (Garner & Alexander, 1989). It might be that simple directions to the student to think about what or how much they know might be enough to dislodge them from more habitual "mindless" activity. If we could some- how activate the learner's own untapped self-monitoring skills, it is speculated, then, that it may be unnecessary to inform them directly of their mastery using some decision superstructure (e.g. -, Bayesian probabilities; see 22.4.2.3; Tennyson & Rothen, 1979). This approach, however, has not been explored in learner-controlled CBI contexts.

In addition to supplanting a student's monitoring activi-ties, or activating existing monitoring strategies, instruction might attempt to actually improve the student's conscious use of metacognitive strategies. This would involve some type of strategy training (see Garner & Alexander, 1989, for a review of some of these training approaches). Tobias (1987a) supports metacognitive strategy training, indicating that many students might need to be taught when and why to use various instructional supports. Kinzie (1990), too, advocates an approach to help the learners become better managers of their learning with the suggestion that perhaps,

... students should be given the training to become self-managers as well as instructional assistance in self--management, and that those without a strong knowledge base should be assisted in making the links that will help establish the structure for new knowledge.

However, at this point, metacognitive strategy training has not been much investigated in a learner-control CBI context.

33.8.2 Learning Strategies and Ability

In addition to the prior-knowledge hypothesis and related issues just discussed, there is another explanation for the general ineffectiveness of providing instructional options. This notion begins with the suggestion that individuals have developed either good or poor strategies for dealing with learning problems. The metacognitive self-monitoring processes mentioned in the previous section on prior knowledge in fact represent a subset of a larger collection of cognitive processing strategies most often called learning strategies. Jonassen (1985) reviews some of the research on learning strategies, and describes four classes of strategies, all of which have clear implications for learn-er-controlled instruction:

• Metacognitive strategies are those processes by which students tell themselves how much they know. It is often described as "self-monitoring," and reflects a sense of both knowledge and ignorance.
• Information-processing strategies make up the largest group of learning strategies. These strategies include developing readiness, reading/viewing for meaning, recalling material, integrating it with prior knowledge, expanding or elaborating on the material, and finally reviewing what has been learned. These strategies seem to correspond to what Merrill (1984) calls "con-scious cognition" processes.
• Study strategies (occasionally called study skills) are explicit techniques to help learners actively process information. These consist of such activities as note taking, outlining, underlining, and the identification and noting of patterns in the new material.
• Support strategies relate to the mental climate or attitude at the time of learning, such as the degree students can internally motivate themselves and stay on-task during the instruction. Jonassen (1985) says these last strate-gies are a sine qua non for learning, and are required in order for the other strategies to be effective.

When many people using both good and poor strategies are averaged in a study, a less-than-ideal picture is painted of the effectiveness of decision making as a whole. Some researchers suggest that the use of such learning strategies as Jonassen (1985) presents is linked closely with the con-cept of general intelligence (Snow & Yalow, 1982). It is not unreasonable to imagine that higher-ability students might have a greater repertoire of strategies to draw on when faced with a learning problem. In fact, as Snow and Yalow (1982) point out, very often the concept of ability is equated with the capacity to learn.

If indeed we can infer that (a) higher-ability students consciously or unconsciously bring to bear the mental resources appropriate to the learning task and avoid using inefficient ones, (b) lower-ability students somehow either lack or don't know how or when to activate their learning strategies, and (c) the success of learner control depends to a large degree on students judiciously applying their mental resources to the learning problem, then we can begin to explain the mixed results of learner control of instruction as being to a degree a function of learner ability, with higher-ability students capitalizing on learner control and lower-ability students left floundering.

An opposing viewpoint that higher ability will predict use of better learning strategies comes from Clark (1982). From a review of aptitude-treatment interaction studies, be first hypothesizes that high-ability students would profit most from activating or cueing methods, that is, techniques that prompt the student to adopt appropriate mental strate-gies from their repertoire of strategies for a given problem. Second, he suggests that low-ability students would do best under the supplanting or modeling methods, which are techniques that do not rely on the student to use his or her own mental resources, but rather explicitly guide the stu-dent through the optimal learning strategies. But regardless of what high-ability students would need, he suggests that they would prefer to choose supplantation or modeling, while low-ability students would prefer activating or cueing methods. Each group does so because that is the method perceived to be the lowest "mental workload" for the student. In this case, he proposes, neither group would select an appropriate strategy.

There is the additional question, however, of what the patterns of "optimal" choices would look like in a learner-controlled lesson. Would the best students generally choose more options, regardless of the specific type of instructional event put under their control? In this case, "more" of any-thing would be perceived by students as being "better." Or would their effective strategy use lead them to select only those specific types of options they feel would produce the greatest benefit? In this case ' we would be able to see only some kinds of options being chosen by higher-ability students, while by others, perhaps not at all. Lower-ability students would perhaps manifest converse types of options--selection patterns, or maybe even random patterns.

It is possible to examine the notion of ability being related to overall amount of options selection only in those studies that offer a variety of options to the student. Other-wise, for studies that offer only one type of selectable instructional event, it is difficult to conclude anything about selective strategy choice and ability level. Of studies that do offer several types of options to students, Carrier et al. (1985) did find a strong positive relationship between a measure of general ability and general amount of options selected, regardless of the type of instructional event. This was not replicated in a follow-up study, however (Carrier et al., 1986), which included some additional motivational feedback in one of the treatments. Perhaps the presence of encouraging feedback (which did increase overall options selection in the study) was a more salient factor affecting decisions by the students to choose or to skip over material, so much so that ability affects were minimized. Other stud-ies, too, found little or no relationship between overall level or frequency of options selection and ability measures. Snow (1979), for example, found near-zero correlations between standard ability and achievement measures and frequency of choice of instructional options. Reinking and Schreiner (1985), too, found no differences between low- and high-reading ability groups in any type of options selected. Another study by Morrison et al. (1992), although more indirect in its implication about ability level relationships, reports no association between amount of instructional sup-port selections made by students under learner-controlled conditions and their posttest performance (posttest perfor-mance also being assumed to be generally related to student ability level). From these findings, it is difficult to conclude that higher-ability students make indiscriminately more frequent use of any and all instructional options that they might be offered.

Connections between ability and selective use of specific types of options seem a little more evident in the literature. In a study that looked for a possible curvilinear connection between options use and ability, Carrier and Williams (1988) found that students with the highest ability levels chose medium frequency levels of instructional options. The suggestion they made was that high-ability students do not act compulsively, indiscriminately selecting all options presented to them, but rather act more reflectively, choos-ing some as needed, but skipping over others deemed not useful. Another study by Sasscer and Moore (1984) found that when students in a TICCIT lesson were given the option of terminating the lesson, the dropout rate was relat-ed to the types of options chosen. The students who left the lesson early typically chose the "easier" kinds of options in the lesson. Snow (1979) found that aptitude measures of fluid-analytic ability and perceptual speed predicted the choice activities of successful college students in a BASIC programming task. The best choice activities he described as indicating a reflective and thoughtful * style, and were more frequently selected by high-ability students. (Some caution is urged in reading this study, however, as the data analysis presented is sketchy and contains too few subjects to trust unequivocally the stability of the multivariate analysis employed.)

A couple of other studies are worth mentioning, although they offer somewhat qualified tests of associations between ability and type of options selected. For example, a study by Kinzie et al. (1988) found that students higher in reading ability selected a high proportion of options to review material than did lower-ability students. This was the only type option offered to subjects in the study, but because it was highly germane for the particular lesson, higher-ability students seemed perhaps better able to gauge the benefits of frequently selecting it. Additionally, if we use posttest performance as a type of surrogate ability measure, we find in a reanalysis of data presented in Seidel et al. (1975, p. 29, Table 6) that while low-posttest performers selected overall more options, high performers selected proportionally more of certain types of options-namely, options to take quizzes-than did low performers. This was not the case for options to recap or review materi-al presented in the lesson. This finding seems not to have occurred for Holmes et al. (1985), however, who found no relationship between pretest scores (here taken as a surrogate ability measure) and particular strategy use-in their case, either opting to take unit tests before proceeding through the lesson, or going through instruction first before taking the tests.

There is also evidence that ability plays an important role on the attrition of students in large instructional units. An early example, the TICCIT system (Merrill, 1973) offered college students a great deal of choice in selection of both content and strategy. Results showed a high dropout rate, but positive effects on achievement for those who persisted. Those who stayed were generally higher-ability students to begin with (O'Shea & Self, 1983, p. 92).

The mixed results from these studies, while indicating the potential for ability and learning strategies to explain overall performance in learner-controlled CBI, also demon-strate that more research needs to be done. The hypothesis Of Clark (1982) that higher-ability students will probably seek more instructional support (even though they do not need it) appears supported in some studies, but not in oth-ers. There is also some evidence that high-ability learners have some capacity to choose their instructional support with some circumspection and discrimination, rather than wholesale. It might be that the specific type of tasks presented to the students need to be more precisely matched to the specific learning strategies with which they most cor-respond. Overall, though, ability measures do not seem to have the power to differentiate the more relevant learning strategies adopted by a given student at a given time.

Some types of instructional interventions do appear to work to compensate for the poor use of mental resources in low-ability learners. Jonassen (1985) presents within the four learning strategy categories listed earlier several suggestions for improving the use of strategies in computer-based instruction. Most of these approaches have yet to be tried in learner-control CBI studies, however.

Ability appears to predict, in addition to the individual's perception of need for instructional support (a metacogni-tive strategy), other types of mental learning strategies in which the student might engage. Although the relationship between ability and choice seems more tenuous than that of prior achievement and choice, there still seems cause to believe that appropriate choice strategies can be made salient to the learners when these learners lack the inclina-tion to spontaneously make their own decisions, perhaps via simple instructions or suggestions, and perhaps by changing the attractiveness of the various choices to be made. Additionally, the types of options selected appear more related to ability than quantity of options chosen.

Only two instances were found of learner-controlled CBI studies that attempted to improve students' strategy use. Elementary school students in Jacobson and Thompson's (1975) study were given prompts at various points to help them make appropriate instructional decisions. Although the instructional treatments used in the study were quite large and in many ways not comparable, the authors still conclude that such strategic prompting can help students to make appropriate decisions. In another study by Relan (1991), three types of strategy-training groups (comprehen-sive, partial, or no training) were experimentally crossed with two levels of learner control over review of material (complete or limited). She found that, for the immediate posttest at least, both strategy-training groups did improve performance, but only for the limited learner-control treatment group. She hypothesizes that the complete learner-control group with strategy training added on top was simply overloaded. The implication is that strategy training might be most effective when close attention is paid to matching appropriately that training to the context of the lesson.

Reigeluth (1979) proposed that learner-controlled instruction offer students an "advisor" option, a sort of pre-scriptive "help' feature, which would suggest to the student various so-called "optimal" strategies for how to process information or what to do next in the lesson. The potential flaw in this proposal is that students might not know how or when to access the optional advisor. Another intervention system is proposed by Allen and Merrill (1985), which pro-vides to the learners varying amounts of learning-strategy suggestions depending on their aptitudes for accomplishing the learning tasks. For students of low abilities, for example, the computer would provide explicit processing representa-tions for the students to follow; for medium-ability students, the system would "guide" the learner to use certain previ-ously learned strategies; high-ability students would be left with the most freedom to select and apply their previously acquired processing strategies without external suggestions or interference from the computer system. This type of system has not yet been tested.

The idea behind all these approaches is to promote the conscientious and mindful use of instructional options according to individual needs for instructional support. The following section shifts the examination from the rational predictors of learner choices to the emotional or affective predictors.