each class to respond was considered important to avoid the selection bias that might occur if moderate to large groups of students decided not to take part. It was
Methodology 1 1 5
also of concern that students might respond in a socially desirable way, that 1s, answer in the way they believed a "good" student might answer.
Both of these potential problems were addressed by telling the students they would be given individual feedback on their responses and a booklet on learning strategies. The importance of answering honestly, if they were to get a useful profile of their
learning behaviour, was stressed. On an approximate estimation nearly
1 00%
ofstudents who attended the initial briefing returned the following week to complete the questionnaires.
Student Comprehension of Questions
The researcher personally administered all of the questionnaires and was available to answer any questions. In the event, questions were asked only by three Asian students, for whom English was a second language. The questions related to words of whose meaning they were unsure. Thewords in question were "intrinsic", "discrete" and "recite".
Pilot Study
The questionnaire was trialled on a small sample of27
to identify and correct potential problems. Subjects were asked to comment on difficulties or ambiguities in the questionnaire. Some minor adjustments resulted from the pilot study. These have been described in earlier sections of the chapter.6.7 Summary
The questionnaire had five sections, four of which were developed for this study. These included learner characteristics (mainly biographical information), study management skills, learning strategies and task representation. The last three of these represented a departure from instruments previously used to gather this type of information. First, both qualitative and quantitative methods for gathering information were used. The instrument studies used a range of written and oral techniques to allow students to describe their versions of the learning and studying processes as fully and as accurately as possible. They also contributed substantially in determining the best format for asking students about their perceptions. On the
I
The instruments themselves were designed to obtain separate measures of study management skills and learning strategies. Other inventories have tended to blur the distinctions between these two despite the fact they make different contributions to the learning process.
The learning strategy instrument is a more comprehensive and "pure" measure of learning strategies than others reviewed in the literature. Only by recognising all of the learning strategies used by a student can an accurate diagnosis be made of the student's learning profile.
The importance of the context to learning outcome has been widely discussed
(Laurillard, 1 979; Entwistle & Ramsden, 1 983; Chambers, 1 992). However, one
aspect of context has escaped scrutiny; the student's representation of the task. Although many ideas can be found in the literature to identify the contents of this measure, finding a suitable instrument on which such items could be modelled did not appear to be available. This measure was completely new.
The final format and organisation of the questionnaire resulted from information obtained from the instrument studies and a review of the literature. The instrument studies were responsible for the use of short phrases, rather than complete sentences, the terminology of many of the items, especially in the learning strategy section and the specific content of some items. The literature influenced the use of both qualitative and quantitative methods of data collection, the decision to use a classroom, rather than laboratory setting and the Likert-type format of most of the items.
The next chapter presents the results from the questionnaire described m this
1 17
Chapter 7
Results
Introduction
This chapter presents results from the analysis of the questionnaire. The results are presented in three parts; first for the principal components analyses of the three measures that were developed for the study, secondly for multiple regressions and thirdly for logistical regression results. The multiple regression results are presented in three stages. Stage one investigated the influence a number of independent variables had on the selection of learning strategies. In stage two the effects of learning strategy choice, study management skills and grade point average on learning outcome were examined. Finally, the strategy scales of the SPQ were regressed on to learning outcome for comparison with the learning strategies developed for this study.
7. 1 Principal Component Analysis
Principal component analysis was used to explore three sections of the questionnaire; learning strategies, study management skills and task representation. Principal component analysis was considered an appropriate method of identifying underlying structures, and reducing these three sections to a more parsimonious
representation of the relationships being measured (Tabachnick & Fidell, 1 989). It
had the further advantage of producing principal component scores that could be used in additional analysis of the data. The PCA results for each of these three sections are now presented.
7.1 . 1 Learning Strategies
Learning strategies form the central component in the two relationships this study examines. First, the study is concerned with factors that influence the selection of learning strategies, and secondly, the influence of learning strategies on learning outcome. The complexity of these relationships indicates the need for intrinsically multivariate techniques. The choice of principal components analysis to reduce the number of learning strategies to a smaller number of components was influenced by the large number of learning strategies, the need for uncorrelated components to be used in later regression analysis and to determine the presence of underlying relationships between the learning strategies.
The analysis was carried out in four steps. First the appropriateness of the data for
factor analytical techni�es was evaluated. A correlational matrix was computed for
all variables and the presence of correlations greater than 0.3 determined the likelihood of some underlying processes. The size of the sample needed for the
number of variables has been widely debated (O'Neil & Child, 1 984), however the
ratio of sample size to items of 7.85 to 1 meets many of the more stringent guide
lines (Tabachnick & Fidell, 1 989). The Kaiser-Meyer-Oklin of .90069 as a measure
of sampling adequacy is described by Kaiser (1 974) as "marvellous". The KMO together with Barlett' s test of sphericity (13843 .36; pS .0000) established the appropriateness of the data for principal component analysis.
In the second step, components were extracted. A scree test supported the retention of seven components (Cattell, 1 966; Tabachnick & Fidell, 1989). While a scree test is not always very exact, the large sample size, generally high communality values and the high loading of variables on each component favoured the use of a scree test for determining the number of components (Gorush, 1 983). To avoid overspecification, component loadings were set at .40. Variables that cross-loaded were assumed to load on the component for which they had the highest loading. To minimise errors in interpretation, components were described by considering loadings in descending order. This seven-factor solution extracted 49.9% of the variance. Information on the seven components is set out in Table 7. 1 .
Results 1 1 9
In the third step, orthogonal rotation with varimax was chosen for simplicity of reporting and because it was intended to use component scores for further analysis
(Tabachnick & Fidell, 1 989). Adequacy of the rotation was determined by the
presence of a simple structure (Thurstone, 1 94 7). Several variables correlated highly
with each component, and generally only one component correlated highly with each variable. Three variables had high loadings on a second component. These were item 49 which loaded primarily on component 1 , but also had a 0.45 loading
on component 5, and items 5 1 and 53 from component
6
which also loaded 0.40and 0.48 on component 1 . This cross loading is not altogether unexpected as component 5 may be described as a more advanced or complex version of component 1 . The difference in the loading on the two components was sufficiently large to justify retaining the three variables.
Finally component scores were computed for each case usmg the regression method.
Table 7.1 Principal Co m ponents Analyses of the Learning Strategy
Inve n to ry
Component 1 : Practice I 11 Ill IV V VI
Work through new examples .84 . 1 7 -.0� .08 -.03 .09 .
Work through given examples .79 .22 -.05 .07 -. 1 0 . 1 1
Complete set exercises/ assignments . 7 1 . 1 5 .04 -.00 -.03 . 14
Complete extra exercises or .69 . 1 1 -.06 . 1 0 . 0 1 .06
assignments
Find own or new examples .67 .01 . 19 .0 1 . 1 6 . 1 5
Practice on simple tasks first . 6 1 .20 -.05 .09 -. 1 5 . 3 8
Use principals with different data :54 .08 .03 .06 ··:n . .45
Change formulas into words .47 .07 .02 . 3 5 . • 27. .04
Explain ideas to self to test :45 . . 3 9 . :22 . 1 5 . 12 . 1 7
understanding
Try doing something the wrong way to -. 1 5 .06 . 1 8 . 2 9 . 3 3
see what happens
VII .05 .02 .03 -. 0 1 .00 .05 .08 :O L: ,.,,'. ,· - . 0 1 . . 0 7
Component 11: Identifying Key I 11 Ill Ideas
Identify main or -.07 .65 .03
significant points
Summarise information .03 .62 .0-l
Reduce notes to get a skeleton of . 6 1 . 1 5
main ideas
Read notes/text & make general .03 · . .• .57 . 1 2
statements
Rewrite notes into a \I 1 .54 .01 more concise form
Use headings and sub-headings ;08 . 5 2 . 2 9
Take notes point by point & learn each .28 . 5 2 . 0 9
Underline or highlight points .24 . 5 1 . 17
Identify key words & phrases . 2 4 . 4 9 .22
Break task into parts & learn each . 2 9 .45 . 10
separately
Component ID: Relate to other 11 Ill :.
contexts
Relate ideas to own experience of a -:oi . 1 0 . 7 8
problem or situation
Think how I might use the . 0 5 . 1 5 .75
information in some aspect of my life.
Apply to real world . 0 1 . 1 1 . 7 4
situations
Relate ideas to own -:16 .06 .73-
views/beliefs/ emotional reactions
Relate ideas to other � 1 0 .07 .68
subjects
Try to use information in new . 3 6 .08 .47
situations
Work out how ideas are related to :o9 . 3 6 ... -10
i
/:.. each other IV V -.03 . 1 8 .03 .03 . 2 1 -.03 .06 , 2 1 .20 -:01 . 1 9 . 1 2 . 2 3 -.03 . 1 1 . 2 5 . 1 6 . 1 2 . 1 8 -.05 IV v · . 1 6 :09 . 1 0 .07 .06 . 1 6 . 2 2 . 1 5 . 2 0 .09 . 1 5 . 2 5 · .. .. .. -.07 .26 VI .05 . 1 5 . 1 2 . 1 3 -.0 1 -.05 .07 -.03 -. 1 2 . 1 1 VI .05 .09 . 1 7 . 0 1 . 1 1 . 2 5 . 3 3 VII . 1 8 . 2 5 .00 -. 1 3 :20 . 1 6 -.02 . 1 8 .27 .07 · · .. : . · vn···••· .05 . 1 5 . 1 0 . 1 9 . 1 1 .17 ••••
•
••
•••• : 1 2 ·.•··.···•·•• .. . ·· .·•· ·. · .. ··•Components IV: Remember infonnation
Take first letter from a phrase to make a word
Make up rhymes or rhytluns
Associate new idea with something easy to remember
Memorise until I can recall without error
Recite notes out aloud
Change diagram into words
Change notes into diagrams
Component V: Background filling Read more wtclely on subject Get more background infonnation
Collect ex1.ra infonnation from variety of sources
Find out more about the context in which the ideas were developed
Talk to other people about topic/subject
Component VI: Projecting! Predicting
Practice predicting outcomes
Do something & try to explain to self
why it happens
Determine trends or patterns
Identify or simulate consequences of doing something a certain way Mentally picture what is described Get a basic overall
picture then fill in details
Component Vll: Grouping
Put information into groups �.02: .•. . 1 8 Label groups of information .07 .23
Put information into lists ;02 :: :. :;·:: .21
·-·:::-::.·
Number each item in list \()9_ · =<. . 15
. m Eigenvalues 1 4�07> 4.68 of variance 2�.4 7.8 Reliability Coefficient )88/ .83 Mean 2�77 3.67 SD 1 .06 . 74 m ••· ••••• .. . ·. · .14 .i22 ()