Obtaining more out of less text in CBI: Effects of varied text density levels as a function of learner characteristics and control strategy

Obtaining More Out of Less Text in CBI: Effects of Varied Text Density Levels as a Function of Learner Characteristics and Control Strategy

Steven M. Ross Gary R. Morrison Jacqueline K. O'Dell

Steven M. Ross, Gary R. Morrison, and Jacqueline O'Dell are at Memphis State University, Memphis, TN.

Learning from expository text in computer-based instruction [CBI] was examined as a function of text density level [high, low, or learner-controlled] and learner characteristics in achievement, attitudes, and reading ability. Low-density text was found to be as effective for learning as high-density text, but was more popular among students and also reduced completion time. The highest achievement was obtained in the learner-control condition, with low-ability readers showing a greater tendency to select high-density text than did high-ability readers. Reading ability was also positively related to achievement in learning from high-density computer text and to preferences for learning from CBI relative to print. Based on the results, use of low-density text is suggested as a standard feature or learner-control option for CBI lessons.

ECTJ, VOL. 36, NO. 3, PAGES 131-142 ISSN 0148-5806

A critical process in developing lessons for computer-based instruction (CBI) is to determine the manner in which information will be displayed on the screen. Unfortu- nately, the computer 's special display capabilities are ignored in many commercial CBI products, much as if the monitor screen were simply an electronic version of a printed page (Bork, 1987; Burke, 1981; Kel- ler, 1987). The present research addressed this issue in reference to the specific problem of displaying instructional text or "microtext" in CBI lessons. Despite the advantages of creating response-sensitive and highly flexible displays, microtext has the disadvantages of: (a) restricted page size, (b) relatively poor resolution, (c) restricted forward and backward paging, and (d) provision of limited cues regarding lesson length.

Comparisons between microtext and print on difficulty and speed of reading have yielded mixed results (Barbe & Milone, 1984; Fish & Feldmann, 1987; Han- sen, Doring, & Whitlock, 1978; Heppner, Anderson, Farstrup, & Weiderman, 1985; Morrison, Ross, & O'Dell, 1988; Muter, Latremouille, & Treurniet, 1982), but most instructional designers strongly advocate minimizing the amount of microtext displayed using such formats as wide margins, double spacing, and fade-out of unneeded information (Allessi & Trollip, 1985; Bork, 1987; Grabinger, 1983). The trade-off, however, is that when there is less text per

132 ECTJ FALL1988

screen, a greater number of screens are required to present the same amount of information.

These considerations suggested possible benefits of creating "low-density" microtext by limiting content to only the essential words or phrases needed to convey main ideas. In the process, unnecessary mod- ifiers, articles, and phrases would be de- leted. Such properties comprise what reading researchers have referred to as text "microstructure" (Davison & Kantor, 1982) or "texture" (Amiron & Jones, 1982). In prior studies, Reder and Anderson (1980; 1982) found that reading summary (low- density) versions of textbook chapters was as effective for learning as reading the com- plete chapters. The summaries were interpreted as helping students to isolate central ideas without the distraction of attending to unimportant details. Comparable results were obtained by the present authors, using both CBI and print media, in teaching a college-level statistics unit (Morrison, et al., in press). A limitation of the latter findings, however, was the very small n employed of only eight subjects per cell.

The present study was designed to ex- tend the above research in several ways. First, Morrison et al.'s (in press) comparison between low- and high-density text was re- plicated using a much larger sample and measures of both immediate and long-term achievement. Use of low-density text was hypothes ized to result in comparable achievement but shorter lesson completion times. Further, to the extent that microtext is less legible or readable than print, any benefits of low-density text were expected to be more evident under CBI. An important feature of the present methodology was that microtext and print presentations were not laid out identically (as in Fish & Feldmann, 1987), but rather used page and screen designs that appeared most common and appropriate for each medium. The decision was to sacrifice internal validity in comparing media in order to create more realistic contexts (higher external validity) for evaluating density variations within each medium.

A second research focus was on the effects of a learner control (LC) condition that allowed subjects to select the text density

levels they preferred on each lesson. Find- ings regarding the effectiveness of learner control as an adaptive strategy have been inconsistent, but are more frequently negative than positive (cf. Carrier, Davidson, & Williams, 1985; Carrier & Sales, 1987; Gray, 1987; Newkirk, 1973; Seidel, 1975). An overall conclusion is that many students, especially low achievers, lack the knowledge and motivation to make appropriate decisions regarding such variables as the difficulty and number of examples, sequencing of content, and types of learning aids used (Carrier & Sales, 1987; Gay, 1986; Ross & Rakow, 1981, 1982; Seidel, 1975; Tennyson, 1980). In contrast to these forms of learner control, the present density variable concerned what was mainly a contextual property of the lesson, with the basic lesson content left unchanged. Making effective choices (i.e., ones that accommodate learning preferences and styles) was there- fore assumed to be less dependent on students already being knowledgeable or skilled in the subject area.

To examine this issue and explore a related one regarding media selection, two forms of learner control were manipulated along with standard high- and low-density treatments. One form (partial LC) allowed subjects to select preferred text density levels on each lesson in the print and CBI instructional units; the other (full LC) allowed subjects to first choose the medium for instruction (CBI or print) and then the lesson density levels within that mode. A third major interest was the relationship of individual differences in reading ability, prior achievement, and attitudes to: (a) suc- cess in learning from conventional (high- density) CBI text, (b) preferences for high- versus low-density text, and (c) preferences for learning from CBI as opposed to print.

METHOD

Subjects and Design

Subjects were 221 undergraduate teacher education majors enrolled in one of the required core courses for certification (educational psychology, instructional media, educational foundations). Participation was voluntary and was compensated by points

TEXT DENSITY IN CBI 133

credited to course grades. Subjects were randomly assigned to seven treatment groups arranged according to a 2 (presentation mode: computer or print) x 3 (density condition: high, low, or LC) factorial design with one outside condition (full LC).

To provide a larger n for analyzing individual differences in learning from high- density microtext, 75 subjects from the orig- inal pool were preassigned to that treatment, with the remainder preassigned in equal numbers to the six other treatments (n = 27 in each). In several instances, however, a greater number of subjects in the high-density CBI group attended a given instructional session than the associated software and materials could accommodate. When this occurred, the required number of individuals were randomly reas- signed to the low-density CBI treatment. This adjustment and uncontrolled absences resulted in final n's of 65 and 32 in the two respective treatments. In the remaining five treatments, final n's ranged from 24 to 27.

At the completion of the learning task, subjects completed an attitude survey and an immediate achievement posttest. Ap- proximately three weeks later, they received a delayed achievement posttest.

Materials Materials used in the study are described below in the order in which they were administered to subjects. All materials, except the statistics pretest, reading test, and delayed posttest, were adapted from Morrison et al. (in press).

Pretask Attitude Survey. This measure consisted of a brief questionnaire to determine subjects' attitudes toward the subject to be taught and toward CBI. Ratings were recorded on a five-point Likert-type scale, with 5 representing the most positive reac- tion. The first item concerned the desirability of mathematics-related subjects, the second how the respondent generally performed in those subjects, and the third whether learning mathematics from a computer would be preferable to learning it from a textbook. For full-LC subjects, an additional item indicated that during the instructional phase they might be asked to

study a lesson from either printed materials or from a computer lesson, and to check which of the two modes they would prefer.

Pretest. The pretest consisted often items on the material covered in the instructional unit on central tendency. Five of the questions involved defining central tendency measures, three calculating them, and two identifying the relative locations of the mean and the mode on a frequency poly- gon. The internal consistency reliability of the test, computed by the KR-20 formula, was found to be .68.

Nelson-Denny Reading Test. Subjects took the "comprehension and rate" section of Form D of the Nelson-Denny Reading Test (Brown, 1976). Extensive information supporting the validity and reliability of the test is reported in Brown (1976). Comprehen- sion was measured by having examinees read eight paragraphs and answer multiple-choice questions on each. Reading rate was measured by asking them to record the number of the line they had reached following the initial 60 seconds of reading. Both raw scores and percentile ranks were determined.

Instructional Unit. The learning material was an introductory unit on central tendency prepared by Morrison et al. (in press). The content was adapted from self-instructional learning modules used in an undergraduate statistics course. For research purposes, the unit was organized into five sections ("lessons") covering the mean, the median, the mode, uses of central tendency measures in different distributions, and positions of central tendency measures in different distributions. Emphasis was on teaching facts and conceptual information that students would need to recall for solving and interpreting problems.

A conventional (high-density) print version of the lesson, patterned after the origi- nal text (Ross, 1982) was initially prepared. Its total length was 18 pages and 2,123 words. Within each lesson the basic instructional orientation involved defining the main concept or idea and then illustrating its application with several numerical examples.

134 ECTJ FALL1988

Following Reder and Anderson's (1980; 1982) procedure, the low-density version was developed by (a) defining a set of general rules for shortening the material, (b) having at least two judges discuss the rules and rewrite the materials accordingly, and (c) reviewing the material and making changes until consensus was achieved that all criteria were satisfied. The rules employed were:

1. Reduce sentences to their main ideas. a) Remove any unnecessary mod-

ifiers, articles, or phrases. b) Split complex sentences into single

phrases. 2. Use outline form instead of paragraph

form where appropriate. 3. Delete sentences that summarize or

amplify without presenting new information.

4. Present information in "frames" con- taining limited amounts of new information.

The completed low-density lesson consisted of 1,189 words, a 56% savings relative to the high-density version, and 15 pages, a 17% savings. CBI versions of the high- and low-density lessons were prepared directly from the print materials. Word counts for corresponding low- and high-density versions were identical across print and computer modes. Due to the much smaller display area of the computer screen, it was not possible (or considered desirable) to dupli- cate the print page formats. Computer frames were thus designed independently, using what were subjectively decided to be th~ most appropriate and realistic screen layouts on the basis of current models (Hannafin & Peck, 1988; Hartley, 1987; Heines, 1984; Morrison & Ross, 1988). This orientation emphasized organizational de- vices such as headings, liberal white space, and standard uses and locations of verbal and symbolic prompts . Each screen provided both back- and forward-paging options.

The final versions of the low- and high- density CBI lessons consisted of 49 and 66 frames, respectively. Figure I shows one of the high-density frames, along with its parallel low-density version. It should be noted

that the low-density version presents the same key information, but with little elab- oration or supporting context.

Posttask Attitude Survey. A six-item printed attitude survey was given to all subjects at the completion of the lesson. Items consisted of statements about the learning experience to which subjects indicated levels of agreement or disagreement on a 5-point Likert-type scale (e.g., 1 = strongly dis- agree, 5 = strongly agree). Abbreviated de- scriptions of the statements are: "Lesson moved quickly," "Material was interest- ing," "Was easy to learn," "Explanation was sufficient," "Text layout was easy to read," and "Prefer this method over lec- ture." Internal consistency reliability for the survey, computed by Cronbach's alpha formula, was r = .63.

Achievement Posttest. The achievement posttest (print format) consisted of three sections designed to assess different types of learning outcomes. The first section was labeled a Knowledge subtest, since it as- sessed recognition or recall of information exactly as it appeared in the text. The first 17 knowledge items were multiple-choice questions, each consisting of a statement defining one, all, or none of the three central tendency measures (mean, mode, or median). Those that described central tendency measures included the exact key words contained in both low- and high- density lessons.

Eight additional questions asked the student to determine relative placements of the mean and the median in distributions that were exact replications of examples that appeared in the lesson. On four of those items, subjects were asked to write a brief rationale for their answers.

The Calculation subtest contained five problems requiring computation of different central tendency measures from new data not used in lesson examples. The Transfer subtest consisted of 13 items that involved interpreting how central tendency would vary with changes in distributions or individual scores. Items of this type were not included in the lesson, nor were the underlying principles needed to answer them explicitly stated. To illustrate, one of


The median corresponds to the middle frequency score in a ranked set of data

Half the scores will be higher Half will be lower

X f Hi 50%

M e d i a n . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Lo 50%

If N=40 (40 scores), median = 20th score If N=17, median = 8.5 highest score

Median corresponds to the 50th percent i le

Higher than half the scores Lower than half

The median, another measure of central tendency, is the number that corresponds to the middle frequency (that is, the middle score) in a ranked set of data. The median is the value that divides your distribution in half; half of the scores will be higher than the median, and half will be lower than the median.

X f Hi 50%

Median .................................... Lo 50%

It is important to remember that the median is the halfway point in the distribution--in terms of frequencies. For example, if N=40 (meaning that you have 40 scores), the median will be your 20th score (in terms of rank); if N=17, the median will be your 8.5 highest score, etc.

Another way of defining the median is to say that it corresponds to the 50th percentile.

In any distribution, the median will always be the score that corresponds to a percentile rank of 50; it is higher than half the scores, and lower than half the scores.

FIGURE 1 Sample Low-Density and High-Density Displays.

136 ECTJ FALL1988

the transfer problem statements was as fol- lows:

On a math test, the mean, mode, and median are all 75%. Later, a student's test that had been scored as 60% is rescored and found to actually be 74% (the only score of 74 in the class). How will this adjustment affect the median, mean, and mode? Provide a brief explanation of your answers.

Scoring rules on objective items and calculation problems awarded one point for a correct answer. On interpretative items, one point was awarded for a correct answer and an additional point for a correct explanation. A summary of subtest and total test l eng ths , m a x i m u m poin ts , and KR-20 internal consistency reliabilities is as fol- lows: Knowledge (25 items, 29 points, r = .81); Calculation (5 items, 5 points, r = .68); Transfer (13 items, 20 points, r = .82); and total test (43 items, 54 points, r = .90).

Delayed Posttest. The delayed posttest (r = .84) consisted of 13 items patterned after p re tes t i tems. Ten of the i tems tes ted knowledge definit ions and relative positions of central tendency measures; the other three tested computational skills.

Procedure

The pretask attitude survey, pretest, and reading test were administered during a regular class session. During the learning phase of the study, from 2 to 12 subjects representing a random mixture of treatments attended an individual session. The classroom used for the CBI condition contained 12 Apple IIe computers. Following introductory instructions and a review of prerequisite information, the learning materials were distributed according to treatment. Full-LC subjects received their preferred presentation mode, CBI or print, as selected on the pretask attitude survey.

Instructions for all treatments indicated that (a) five lessons would be presented on central tendency; (b) learning was to be self-paced; (c) turning back to reread pre- ceding pages (or frames) was permitted if desired; (d) it was permissible to ask the proctor any questions about the task procedure while learning; and (e) a pos t tes t would be given following the learning task.

Subjects in the two LC treatments received additional instructions indicating that, depending on how much explanation they desired, they could choose between "long" and "shor t" presentat ions in each unit. Examples of matched high- and low-density displays were shown to help them make a selection for the initial lesson.

In the CBI condition, subjects pressed a key to indicate their preferences; in the print condition they informed the proctor. Den- sity-level selection was repeated at the beginning of each of the remaining four lessons. After subjects comple ted the last lesson, their finish times were recorded and the posttask attitude survey and immediate posttest were administered. Approximately three weeks later, the subjects took the delayed posttest at the beginning of a regular class meeting.

RESULTS

Major dependent variables consisted of four achievement measures (Knowledge, Calcu- lation, and Transfer subtests ; de layed posttest), total posttask attitude score, and lesson completion time. Since all achievement measures were significantly intercor- related (r range was .43 to .70), multivariate analysis of variance (MANOVA) was used in associated analyses. Univariate ANOVAs were performed on the attitude and completion time variables, which had very low, most ly nonsignif icant correlations with other dependent measures. Regression so- lutions were used in each two-way (presentation mode by density condition) design to control for disproportionate treatment n's. Follow-up comparisons of means, where appropr ia te , were pe r fo rmed using the Tukey HSD procedure . In p re l iminary analyses, no differences were found between treatment groups in pretest performance, pretask attitudes, reading comprehension, or reading rate.

Learner Control Analyses

Partial LC versus Full LC. In an initial set of analyses, outcomes in the partial-LC and full-LC treatments were compared. Inspec- tion of LC treatment means showed them to be quite similar and directionally higher


than those for the standard high- and low- density treatments. A 2 (LC strategy) x 2 (p resen ta t ion m e d i u m ) M A N O V A on achievement confirmed the former impres- sion by failing to show any significant effects due to LC strategies. Nor were significant LC s t ra tegy effects ob ta ined in a univariate ANOVA on post task attitude scores. The ANOVA on completion time, however, did yield a significant LC strategy • p r e s e n t a t i o n m e d i u m in terac t ion , F (1,69) = 5.71, p < .02, and LC strategy main effect, F (1,69) = 7.73, p < .01.

Follow-up analyses indicated that in the print condition, no differences occurred between learner control variations, but under CBI, the full-LC group (M = 18.9 min.) took significantly less time than the partial-LC group (M = 29.0 min.). Thus, those who selected CBI completed the lesson more quickly than those who were prescribed CBI (partial LC), perhaps as a result of having greater experience and confidence in using that medium. Another explanation, sup- ported in the next analysis, is that faster readers were more apt to select CBI than print, thus giving the fuU-LC group a built- in advantage on the completion rate criterion.

Media Preferences. Media selections by the fuU-LC group were almost equally distributed between print (n = 11) and CBI (n = 13), X 2 (1) = .04, p > .05. For exploratory purposes, a discriminant analysis was performed using the subgroups as the criterion and the following as predictors: pretask attitudes, pretest, reading comprehension, and reading rate. Applying stepwise selection, only reading rate was found to be a significant discriminator (p < .01). Subjects who selected CBI had higher reading rate scores (M = 271.8) than those who selected print (M = 189.3). Given the very low sub- group n's, it seems risky to assign too much i m p o r t a n c e to this o u t c o m e w i t h o u t additional substantiation.

From a statistical standpoint, reading rate had much greater variability than did any other pretask measure, making it poten- tially stronger as a discriminator. From a theoretical standpoint , Salomon's (1983; 1984) research has suggested that CBI is genera l ly pe rce ived by s t uden t s as a

"smarter" medium than print, a quality that would be likely to make it more attractive to good students than to poor students. The reading rate correlation, while needing to be interpreted cautiously, is consistent with this pattern.

Density Selections. A 2 (LC-strategy) x 2 (presentat ion mode) ANOVA was performed on the total number of low-density selections (out of a possible 5) made by LC subjects. No significant effects were found. Overall means were 3.5 for print and 3.0 for CBI. Thus, there was a general tendency by subjects to prefer low-density materials, re- gardless of presentation mode. To examine whether certain types of individuals were more likely than others to select low-density material, the number of low-density selections was regressed, using a stepwise procedure, on the four pretask predictor variables. Again, only reading rate was identified as a significant predictor, R z = .08, p < .01. As reading skills decreased, the tendency to select low-density material also decreased.

Density Condition x Presentation Mode Analyses Treatment means and standard deviations on each dependent variable are shown in Table 1. For analyses of achievement and attitudes, data from the full- and partial-LC treatments were pooled within presentation modes. Results for each variable are reported below.

Achievement. The 2 x 2 MANOVA on achievement data showed both main effects to be significant (p < .05). In follow-up univariate tests, the density condition effect was significant on all dependent measures except the Knowledge subtest. Tukey HSD comparisons showed that on the calculation subtest, F (2,215) = 2.88, p < .05, the LC group (M = 3.71) was directionally, but not significantly, superior to both the low-density (M = 3.19) and high-density (M = 3.27) groups.

On the Transfer subtest, F (2,215) = 5.53, p < .01, the LC group (M = 12.33) had a significant advantage (p < .01) over the low-density group (M = 9.58) and a near-

138 ECTJ FALL1988

TABLE 1

Treatment Means and Standard Deviat ions on Performance and Atti tude Measures

Presentation Mode and Density Condition

PRINT CBI

Measure Low-Density High-Density LC a Low-Density High-Density LC a (n = 27) (n = 24) (n = 35) (n = 32) (n = 65) (n = 38)

Pretest (10) b M 5.0 4.0 4.1 3.5 4.0 4.7 SD 2.7 2.2 2.4 2.4 2.7 2.7

Knowledge (29) M 18.1 18.1 17.9 16.8 18.0 20.4 SD 6.0 6.5 5.5 6.4 5.1 4.7

Calculation (5) M 3.3 2.9 3.5 3.1 3.4 3.9 SD 1.6 1.7 1.4 1.6 1.4 1.1

Transfer (20) M 9.2 10.0 11.5 9.9 10.8 13.1 SD 4.5 5.6 4.9 5.4 4.8 5.1

Total Posttest (54) M 30.5 31.0 32.4 29.7 32.3 36.6 SD 10.8 12.8 11.2 12.1 10.1 8.9

Posttask Attitude Total (30) M 14.5 14.9 13.9 14.2 14.3 14.4 SD 3.4 4.3 2.9 4.3 3.8 3.0

Completion Time (min.) M 20.2 22.9 21.1 21.7 27.8 25.5 SD 9.2 8.5 6.3 5.6 9.1 8.2

Delayed Posttest (13) M 6.6 6.6 7.5 7.4 7.6 9.1 SD 2.6 2.3 2.2 2.7 2.5 2.3

aFulI-LC and partiaI-LC subjects are combined within each presentation mode. bValues in parentheses indicate maximum scores.

TABLE 2 Correlat ions Between Measures for High-Density CBI Group

Measure 2 3 4 5 6 7 8 1. Pretest .21" .29** .12 .12 .18 .08 .36** 2. Pretask Attitude .30"* .29"* .21 * .23* .00 .25* 3. Posttest .34** .51"* .08 .05 .40** 4. Attitude Total .21 * - .07 .00 .31 ** 5. Reading Comprehension .38"* .00 .33"* 6. Reading Rate .07 .01 7. Completion Time .20 8. Delayed Posttest

Note: n = 61 for correlations involving delayed posttest; n = 65 for all others. *p < .05

**p < .01


significant advantage (p < .10) over the high-density group (M = 10.58).

Similarly, on the delayed posttest , F (2,186) = 5.41, p < .01, the LC group (M = 8.30) was significantly superior to the low- density group (M = 7.02), and directionally superior to the high-density (M = 7.34) group.

None of the comparisons between high- and low-density means was significant. The only significant presentation mode effect occurred on the delayed posttest, on which CBI subjects (M = 7.95) surpassed print subjects (M = 7.00), F (1,186) = 8.61, p < .01.

Attitudes and Completion Time The two-way ANOVA on posttask attitude total scores failed to yield any significant effects. For the analysis of complet ion times, it was decided to exclude LC data for two reasons: (a) LC represented a variable mixture of density levels across individuals, which would weaken the contrast between low- and high-density conditions; and (b) the significant interact ion be tween LC strategy and presentat ion mode, as reported earlier, would further confuse in- t e rp re ta t ions . The r e su l t an t analys is showed both main effects, but not the interaction, to be significant. Specifically, completion times were longer for CBI (M = 25.8 min.) than for print (M = 21.5), F(1,144) = 4.45, p < .05, and for high-density (M = 26.5) than for low-density text (M = 21.0), F (1,144) = 8.60, p < .01.

To correct for a possible inflation of CBI time scores due to keyboarding and wait time for prompt ing and displays, an esti- mate was made of the amount of time required merely to page through the entire CBI lesson without at tempting to read any display. Across several trials, average "paging" times were determined to be 6.45 n-tin. and 4.45 min. for the high- and low-density units, respectively. The 2 x 2 ANOVA was then repeated after subtracting these con- stants from the corresponding unadjusted time scores. Consequently, the presentation mode main effect was no longer significant (p < .20), but the significant density condition effect was maintained, F (1,44) = 5.12, p < .05.

Learning from High-Density Microtext A collateral research interest was the rela- t ionship between s tudent characteristics and learning from high-density microtext. It may be recalled that, for this purpose, more subjects were tested in the high-density CBI treatment than in other treatments to increase the reliability of findings (final n = 65). Separa te s tepwise mul t ip le re- gressions were performed on criteria consisting of immediate and delayed posttest scores, attitude total score, and completion time. Predictor measures consisted of the pretest, pretask attitudes, reading scores, and, where appropriate, other "criterion" variables. Table 2 shows the intercorrela- tions between all variables concerned.

When immediate posttest scores were t rea ted as the cri terion, reading comprehension (simple r = .51) was the first predictor entered in the equation (R 2 = .26, p < .001). Pretest score (simple r = .26), which had been expected to be the strongest predictor, was entered on the second step, but made a relatively weak contribution to the equation (AR 2 = .05, p < .05). No other predictors were entered.

Other findings were: (a) no predictor variables were significant in accounting for variance in completion time, (b) pretask attitude score was the only significant predictor of attitude total score, (c) and pretest score was selected on the first step and reading comprehens ion on the second step, with the delayed posttest as the criterion. For comparison purposes, we repeated the regression analysis on the immediate posttest for the three remaining standard (high- or low-) density treatments, recognizing that findings would be less reliable due to their relatively small n 's (from 24 to 25 subjects). For both the low-density print and low-density CBI groups, only the pretest was selected for entry (R2's = .18 and .49 respectively). For the high-density print group, attitude score was selected first (R 2 = .23), followed by the pretest. In none of these analyses was either of the reading measures selected.

DISCUSSION

Consistent with previous studies (Morrison et al., in press; Reder & Anderson, 1980;

'140 ECTJ FALL1988

1982), low-density text was found to be as effective as conventional text in teaching an expository lesson to college students. A clear benefit of low-density text was the significant reduction in lesson completion time. Also, subjects were highly receptive to low-density presentations, choosing them over high-density versions on approximately 65 percent of the LC selections. Contrary to expectations, the findings did not show relatively greater advantages of using low-density text with CBI than with print.

These results, overall, do not imply that instructional uses of conventional (elaborated) expository text should be discour- aged. Frequent exposure to such forms seems essential for developing students' skills in reading and writing. Further, re- cent findings suggest that elaborated text is advantageous for learning complex pro- cedural skills, as a result of providing more explicit explanations (Reder, Charney, & Morgan, 1986). However, where it is dif- ficult or costly to display long segments of expository text, as is frequently the case in CBI, the present results suggest consideration of low-density presentations as a means of reducing lesson length and reading time, without adversely affecting learning of important facts and concepts. Based on the LC results, low-density text may also have motivational advantages for many learners. It could thus serve as a means of making expository material in a new or dif- ficult subject area appear less threatening and easier to read.

A second research interest was the effectiveness for learning of learner control of text density. As reviewed earlier, much of the prior research on learner control has investigated conditions in which students were asked to make decisions regarding the amount or type of instructional support to be received (see Hannafin, 1984; Steinberg, 1977). In the majority of cases, the associated learning outcomes have been negative, especially for low achievers. As Tobias (1987) suggests, aside from the problem of not knowing what materials to select, low achievers often lack sufficient metacogni- five skills to use the resources effectively. In contrast with these studies, the text-density variable examined in this research con-

cerned what was mainly a contextual or "stylistic" property of the lesson. Regard- less of which density variation was received, the basic information taught and its support material (number and type of examples) remained unchanged.

Based on the positive achievement results, learner control generally appears to have been used effectively by most students. This interpretation gains credence by the finding that preferences for high- and low-density material were unrelated to prior achievement, and thus did not reflect the frequent, inappropriate tendencies of low achievers to select low instructional support and high achievers to select high support (Ross & Rakow, 1981; 1982; Tenny- son, 1980). In fact, the correlations with reading rate suggested the opposite pattern. Less skilled readers were more likely to choose high textual support, whereas better readers were more likely to choose low textual support (which ostensibly could reduce their reading time, yet adequately support comprehension). Further, most subjects varied their selections across lessons, suggesting some attempt to adjust strategies as perceived needs changed. These findings reinforce the idea that learner control is not a unitary concept, but rather a collection of strategies that work in different ways, depending on what is being controlled by whom. Allowing subjects to select contextual lesson properties in accord with personal interests and performances (also see Ross, 1983) shows potential as one effective form.

In contrast to the usual finding in apti- tude-treatment interaction research (To- bias, 1976; 1987), prior achievement (pretest) was a relatively weak predictor of posttest performance for students who learned from high-density microtext. Reading comprehension, however, was a relatively strong one. In all other treatments, the pattern was the opposite, with prior achievement making strong contributions to the regression equation and reading comprehension being consistently excluded. Although small n's in the latter treatments require caution in drawing conclusions, it remains directly consistent with the present rationale that learning from high-density microtext screens appears to be most de-


penden t on reading skills of all conditions. Aside from the questionable theoretical

i m p o r t a n c e of c o m p a r i n g p r e s e n t a t i o n modes (Clark, 1985; Hagler & Knowlton, 1987), the present comparison was inten- tionally compromised by the decision to create realistic rather than parallel CBI and p r in t d i sp lays . Even wi th this bu i l t - in source of variance, results for CBI and print were almost identical. In contrast, Morrison & Ross ' (1988) CBI subjects pe r fo rmed poorly relative to print subjects on virtually the same task. Differences in the sampling pools and refinements made in the present CBI methodology would seem likely to ac- count for the variation across studies.

Print and CBI were also comparable in their appeal to s tudents , with approxi- mate ly half of the LC-media select ions favoring each. Given the absence of any strong relationships with prior achievement or atti tudes, these selections seem mainly attributable to subjects ' idiosyncratic expe- riences with CBI and attributions concern- ing its desirability and difficulty (see Salo- mon, 1983; 1984). From a practical s tand- point , however , a ccommoda t ing med ia preferences seems much less important or viable than the systematic matching of instructional strategies to the specific media that most powerfully suppor t them. The present results suggest that traditional text narrative formats do not serve such purposes for CBI. The availability of low-density text as a s tandard feature or learner-control opt ion in expository lessons deserves further consideration as a design alternative.

REFERENCES

Allessi, S. M., & Trollip, S. R. (1985). Computer- based instruction: Methods and development. En- glewood Cliffs, NJ: Prentice-Hall, Inc.

Amiron, M. R., & Jones, B. F. (1982). Toward a new definition of readability. Educational Psy- chologist, 17, 13-30.

Barbe, W. B., & Milone, M. N. (1984). Computers, common sense and reading. Paper presented at the World Congress on Reading, Hong Kong.

Bork, A. (1987). Learning with personal computers. New York: Harper & Row, Publishers, Inc.

Brown, J. I. (1976). Examiner's Manual: The Nel- son-Denny Reading Test. Chicago, IL: The River- side Publishing Company.

Burke, R. L. (1981). CAI sourcebook. Englewood Cliffs, NJ: Prentice-Hall, Inc.

Carrier, C. A., & Sales, G. C. (1987). Pair versus individual work on the acquisition of concepts in a computer-based instructional lesson. Jour- nal of Computer-Based Instruction, 14, 11-17.

Carrier, C., Davidson, G., & Williams, M. (1985). The selection of instructional options in a computer-based coordinate concept lesson. Educa- tional Communication and Technology Journal, 33, 199-212.

Clark, R. E. (1985). Evidence for confounding in computer-based instruction studies: Analyz- ing the meta-analyses. Educational Communica- tion and Technology Journal, 33, 249-262.

Davison, A., & Kantor, R. (1982). On the failure of readability formulas to define readable texts: A case study from adaptations. Reading Re- search Quarterly, 17(2), 187-209.

Fish, M. C., & Feldmann, S. C. (1987). A comparison of reading comprehension using print and microcomputer presentation. Journal of Com- puter-Based Instruction, 14, 57-61.

Gay, G. (1986). Interaction of learner control and prior understanding in computer-assisted video instruction. Journal of Educational Psy- chology, 78, 225-227.

Grabinger, R. S. (1983). CRT text design: Psycho- logical attributes underlying the evaluation of models of CRT text displays. Unpublished doctoral dissertation, Indiana University.

Gray, S. H. (1987). The effect of sequence control on computer assisted learning. Journal of Com- puter-Based Instruction, 14, 54-56.

Hagler, P., & Knowlton, J. (1987). Invalid implicit assumption in CBI comparison research. Jour- nal of Computer-Based Instruction, 14, 84-88.

Hannafin, M.J. (1984). Guidelines for using locus of instructional control in the design of computer-assisted instruction. Journal of In- structional Development, 7, 6-10.

Hannafin, M. J., & Peck, K. L. (1988). The design, development, and evaluation of instructional software. New York: Macmillan.

Hansen, W. J., Doring, R. R., & Whitlock, L. R. (1978). Why an examination was slower on- line than on paper. International Journal of Man-Machine Studies, 10, 507-519.

Hartley, J. (1987). Designing electronic text: The role of print-based research. Educational Com- munication and Technology Journal, 35, 3-17.

Heines, J. M. (1984). Screen design strategies for computer-assisted instruction. Bedford, MA: Dig- ital Press.

Heppner, F. H., Anderson, J. G., Farstrup, A. E., & Weiderman, N. H. (1985). Reading performance on a standardized test is better from print than from computer display. Journal of Reading, 28, 321-325.

142 ECTJ FALL1988

Keller, A. (1987). When machines teach. New York: Harper & Row.

Morrison, G. R., & Ross, S. M. (1988). A four- stage model for planning computer-based instruction. Journal of Instructional Development, 11, 6-14.

Morrison, G. R., Ross, S. M., & O'DeU, J. K. (1988). Text density level as a design variable in instructional displays. Educational Communica- tion and Technology Journal, 36, 103-115.

Muter, P., Latremouille, S .A. , & Treurniet, W. C. (1982). Extended reading of continuous text on television screens. Human Factors, 24, 501-508.

Newkirk, R. L. (1973). A comparison of learner control and machine control strategies for computer-assisted instruction. Programmed Learning and Educational Technology, 10, 82-91.

Reder, L. M., & Anderson, J. R. (1980). A comparison of texts and their summaries: Memorial consequences. Journal of Verbal Learning and Verbal Behavior, 19, 121-134.

Reder, L. M., & Anderson, J. R. (1982). Effects of spacing and embellishment on memory for the main points of a text. Memory and Cognition, 1 O, 97-102.

Reder, L. M. Charney, D. H., & Morgan, K. I. (1986). The role of elaborations in learning a skill from an instructional text. Memory and Cognition, 14, 64-78.

Ross, S. M. (1982). Introductory statistics: A conceptual approach. Darwell, IL: The Interstate Publishers, Inc.

Ross, S. M. (1983). Increasing the meaningful- ness of quantitative material by adapting context to student background. Journal of Educa-

tional Psychology, 75, 519-529. Ross, S. M., & Rakow, E. A. (1981). Learner con-

trol versus program control as adapt ive strategies for selection of instructional support on math rules. Journal of Educational Psychology, 73, 745-753.

Ross, S. M., & Rakow, E. A. (1982). Adaptive instructional strategies for teaching rules in mathematics. Educational Communication and Technology Journal, 30, 67-84.

Salomon, G. (1983). The differential investment in mental effort in learning from different sources. Educational Psychologist, 18, 42-51.

Salomon, G. (1984). Television is "easy" and print is "tough": The differential investment of mental effort in learning as a function of per- ceptions and attributions. Journal of Educational Psychology, 76, 647-658.

Seidel, R. J. (1975). Learner control of instructional sequencing within an adaptive tutorial CAI envi- ronment. Alexandria, VA: Human Resources Research Organization Technical Report, 75-77.

Steinberg, E. R. (1977). Review of student control in computer-assisted instruction. Journal of Computer-Based Instruction, 3, 84-90.

Tennyson, R. D. (1980). Instructional control strategies and content structure as design variables in concept acquisition using computer-based instruction. Journal of Educational Psy- chology, 72, 525-532.

Tobias, S. (1976). Achievement treatment interac- tions. Review of Educational Research, 46, 61-74.

Tobias, S. (1987). Mandatory text review and interaction with student characteristics. Journal of Educational Psychology, 79, 154-161.

Documents

Obtaining more out of less text in CBI: Effects of varied text density levels as a function of learner characteristics and control strategy