BACKGROUND OF THE ATTITUDE CONCEPT, A META-ANALYSIS OF RECENT ATTITUDE-TOWARD-SCIENCE RESEARCH, AND A COMPARISON OF THAT META-ANALYSIS WITH ANOTHER METAANALYSIS by
William Edward Craychee B.S., California State Polytechnic University, 1980
A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Master of Arts School of Education
This thesis for the Master of Arts degree by William Edward Craychee has been approved for the School of Education by
William L. Goodwin
Date TZ/n) t. I
Craychee, William Edward (M.A., Curriculum and Instruction)
Background of the Attitude Concept, a Meta-Analysis of Recent Attitude Toward Science Research, and a Comparison of That Meta-Analysis with Another Meta-Analysis
Thesis directed by Professor Glenn E. McGlathery
Among the priorities for science education in the United States, according to the National Assessment of Educational Progress (1979), is the determination of the attitudes toward science with which students are leaving the schools. Are students leaving the schools liking the sciences? In line with this priority, some researchers of the 1980s conducted complex multivariate survey studies of wide ranging student samples. Other researchers continued experimental manipulation of specific variables, sometimes on the recommendation of models developed by the survey researchers.
One of the questions that concerned researchers and educators during the 1970s and 1980s involved a suspected lack of cumulative knowledge growing from the Attitude research. A type of survey study called metaanalysis was developed to quantitatively compare the results of the wide variety of studies being produced. It was hoped the meta-analysis would show us the direction, if any, that research was moving.
This paper reports the results of a meta-analysis of survey and experimental research conducted by this author. These results are compared with the results of another meta-analysis of the same field of research conducted by different authors. The results indicate that attitude-toward-
science research is somewhat disconnected. Researchers may not be communicating as they should.
A history of the attitude concept indicates that attitude research has, over the years, received more or less attention, depending upon the needs of the time. However, it has never been considered to be of major importance. Therefore, it has never received the attention that more popular achievement research has, for example. It should not be surprising, then, that this field is still in its infancy, not having, as yet, even standardized its variables or instruments. Possibly, with the current concern over a lack of new scientists, attitude-toward-science research will continue to receive the attention it needs to become a valid part of the educational research community.
The form and content of this abstract are approved. I recommend its publication.
I would like to acknowledge the assistance given me by my wife Jeanne P. Craychee: for her moral support, her assistance in keeping the verbiage down, and her much needed assistance with Word Perfect, version 5.0.
Purpose of This Study..........................................1
Transactions: The Learning Environment.......................4
Research Assessment ...........................................6
Scientific Attitude ...........................................7
H. LITERATURE REVIEW .............................................10
History of the Attitude Concept...............................10
Some Theoretical Conceptions of Attitude......................17
Probabilistic View .........................................18
Latent Process View.........................................19
Paradigmatic View.......................................... 21
Attitude Distinctions....................................... 22
Belief, Opinion and Knowledge...............................22
Expressive and Ego Defensive Functions........................27
Other Variables................................................ 28
HI. THE META-ANALYSIS...............................................36
Definition of Meta-Analysis.....................................36
The Common Metric.............................................37
Recommended Characteristics of a Meta-Analysis................38
Design of the Study.............................................40
Type of Studies Used..........................................40
Studies in Common.............................................48
Learning Environment (LE) Variables...........................55
Grade Level Differences...............................59
Student Variables ......................................64
Implications for Research and Teaching. ................69
Implications for Teaching.............................74
Limitations of This Study...............................75
A. SUMMARY OF PERTINENT INFORMATION OF
STUDIES USED IN THIS META-ANALYSIS......................82
B. STUDY DESCRIPTIONS.......................................89
C. THE STEM-AND-LEAF DIAGRAM............................. 112
1. Science education priorities for research................2
2. Major variables being investigated with
potential interconnections depicted......................32
3. The five commonplaces of classroom instruction ............33
4. Framework for the model developed by Haladyna
and Shaughnessy (1982).................................. 34
5. Sources of studies....................................... 42
6. Subject matter breakdowns...............................42
7. Breakdown of studies by grade level........................42
8. Predictor variables.....................................46
9. Percentages of predictors used by Craychee
and Haladyna & Shaughnessy (1982)....................... 47
10. Strengths of association (percentages) between predictor variables and attitudes toward the subject matter of science across three dimensions
11. Correlations between attitude and student variables .... 52
12. Correlations between attitude and teacher variables .... 54
13. Correlations between attitude and LE variables.........56
14. Correlations between attitude and achievement
15. Correlations between attitude-toward-science and
predictor variables by grade level..........................61
16. Correlations between attitude toward math
and science and predictor variables by grade level..........62
17. Strengths of association between predictor variables
and attitudes toward the subject matter of science across three dimensions of schooling showing the combined data of Haladyna & Shaughnessys (1982) and Craychees analyses.........................................64
Purpose of This Study
The 1976-1977 National Assessment of Educational Progress (NAEP,
1979) considers the following questions as relevant ones for science education:
Do students enjoy their science classes? Do they find their teachers enthusiastic? Do they feel that the facts and methods they learn in science are useful? Do they pursue science related activities outside of science classes? (p. 5)
In other words, are students coming out of science classes with positive attitudes toward science: do they like science? And, is the current educational establishment producing people who at least recognize scientific reasoning? Yager (1978) says that:
In science teaching our social responsibility is to cause some changes in the what, why, and how of science learning in the current population. We do research to assess our effectiveness in meeting this social responsibility, (p. 99)
Yager presents an outline of priorities for science education research which, he tells us, "...suggests a multiplicity of descriptions and interactions potentially worthy of research." (p. 102) His scheme is illustrated in Fig. 1.
1. Teacher Characteristics 4. Pedagogy 7. Student Attitudes
2. Student Characteristics 5. Classroom Climate 8. Scientific Literacy
3. Social Imperatives 6. Implementation
Figure 1. Science Education Priorities for Research.
From "Priorities for research in science education: a study committee report" by R. E. Yager, 1978, Journal of research in Science Teaching. 15(2), p. 102. Copyright 1978 by the National Association for Research in Science Teaching.
A description of Yager's constructs will help familiarize us with the
Antecedent characteristics are the characteristics that everyone involved in the educational milieu possesses prior to exposure to the milieu. Yager calls them entry conditions.
Teacher characteristics. The teacher is a most important element in any teaching situation. According to Yager (1978), teacher characteristics are suspected as being highly predictive of both techniques and effectiveness of techniques used by teachers; they are also important predictors of outcomes. Researchers should recognize the antecedent characteristics of teachers.
These characteristics are personal (ambient) representing experience from elsewhere: elsewhere meaning anywhere at any time, from the remote past to something that just happened five minutes before. How does the teachers entry ambiance affect the student?
Yager offers the following as supporting evidence for his hypothesis
that teacher characteristics are fertile research variables:
Independently [of this committee], but simultaneously, the Research Committee of the National Association for Research in Science Teaching (NARST) assigned the highest priority to empirical tests of the relationships between teacher attitudes and behaviors and student outcomes. (1978, p. 102)
According to the 1976-77 NAEP report (NAEP 1979, p. 62): "The most vital factors in a science classroom are the behavior of the teacher and the teachers attitudes toward science."
Student characteristics. The student is the human target of education, not a device which, when properly tuned, can assimilate information identically. Students often possess unique differences which challenge educators. These differences include interests, previous experiences, abilities, and attitudes. According to Yager (1978): "Among students the wide diversity of antecedent backgrounds, interests, competencies, and expectations present a multiplicity of bases for studies." (p. 102) Again, Yager counsels researchers to concentrate on student characteristics that can be addressed in the school. The schools cannot change the sex of a child, for example (at least not legally). He recommends "...the relationships among the levels of thought of the students, their previous experiences, and how those levels of thought can be changed with school experiences" (p. 102) as fertile research avenues. The NARST Research Committee also gives high priority to the investigation of student characteristics.
Social imperatives. Social imperatives define the necessity of science
educational goals. Social imperatives like, the need for energy, food shortages, acid rain, ozone depletion, AIDS and so on require scientifically trained minds to grapple with them effectively; or so goes a common prejudice of the western philosophical tradition. Nevertheless, social imperatives are new types of problems that crop up without precedent and require minds that can "think on their feet" as it were. According to Yager, became attitudes mediate knowledge and action, the understanding of attitudes is also somewhat of a social imperative. (Notice that Yager is not a behavioralist.)
Transactions: The Learning Environment
"Transactions are the actions and concurrent emotional conditions intended to provoke learning." (Yager 1978, p. 103) They include the emotional climate and the activities of the classroom. They are what the participants (students, teacher[s]) are doing. Classroom environments and teaching methods (pedagogy) are "amenable to control." (p. 103)
Researcher's should focus on how changes, or a lack of changes in pedagogy and/or transactions affect learning. But, according to Yager, few studies address this issue.
While the commonplace pre-post appraisal of long term instruction may be adequate to describe the overall effects of instruction, it is too gross to reveal the moment to moment incremental learnings which integrate to the total. More detailed and individualistic studies, perhaps in the form of case histories, could pinpoint the moment to moment gains and losses, (p. 103)
Yager warns that there are too few studies that point to the
interactions among the student and teacher characteristics that make up the
classroom climate. (For an interesting treatment of interactions between
these two variables see Lawrenz, 1975.)
Classroom climate is important because "...it establishes (or at least modifies) the affective orientation and attitudes students carry from the school regarding science." (Yager 1978, p. 104) The classroom is the place where most educational activities are implemented in schools. The classroom climate is part of the learning environment in which the implementation occurs. Understanding these relationships can point to promising implementation schemes for dealing with new social imperatives.
There are two important subdivisions in this category: they are scientific literacy and the attitudes of the student toward science. "When and how such attitudes are formed...and how they may be modified, if at all, by school experiences is essentially an unexplored subject." (Yager 1978, p. 104) To summarize, according to Yager:
The diversity of experiences, of student characteristics, and of classroom climates, as well as the potential effectiveness of peer interaction are among the variables requiring more careful investigation. Potential sources of theoretical premises and investigatory procedures may be found in psychology, sociology, and anthropology, (p. 104)
Is current research looking at these variables? According to Yager (1978, pp. 106-107):
Careful, critical, and comprehensive summary papers are needed to summarize the existing knowledge relevant to research in science education, A base broader than the science education literature must be examined so that results and propositions from psychology, sociology, anthropology, and general pedagogy can serve as points of departure for new studies.-.Finding order and common direction for research efforts in science education is a worthy goal for the decade ahead.
Also, Schibeci (1984, p. 28) suggests that:
One line of investigation which could be pursued by those interested in meta-analysis would be a reliability trial. Two people (or two groups of people) would conduct, quite independently, the meta-analysis of the same broad area of concern. The results could then be compared.
So, in addition to the objectives presented above by Yager (1978) I have
decided to attempt a comparative study, using a meta-analysis by Haladyna
and Shaughnessy (1982) as a comparison. In their analysis, 49 studies
involving attitude-toward-science as a dependent variable were used. This
study is contained in CHAPTER III. CHAPTER II examines the history, the
theoretical basis, and the psychological roots of the attitude concept.
Particularly applicable as rationale for this project is the following from
Haladyna and Shaughnessy (1982, p. 549):
While student achievement is the concern in most educational research, there has been a steady increase in interest in student attitudes as an
important outcome of schooling. For this reason, a meta-analysis of research on attitudes toward the subject matter of science seems particularly salient at this time.
Schibeci (1984, p. 26) adds:
The science education community, for one, appears to regard the affective domain as important. Some crude indicators of this interest are the numbers of papers presented at conferences and dissertations in this area.
Prerequisite to any intensive discussion of attitude is an adequate purveyance of the meaning of the term. For the purposes of this paper we will be discussing what has come to be known in the literature as the "attitude-toward-science" of a student, as opposed to the "scientific attitude" of a student. There is a basic and important difference between the two constructs.
The scientific attitude as it appears in the science education literature embodies the adoption of a particular approach to solving problems, to assessing ideas and information or to making decisions. Using this approach evidence is collected and evaluated objectively so that the idiosyncratic prejudices of the one making the judgement do not intrude.
No source of relevant information is rejected before it is fully evaluated and all available evidence is carefully weighed before the decision is made. If the evidence is considered to be insufficient then judgement is suspended until there is enough information to enable a decision to be made. No idea, conclusion, decision or solution is accepted just because a particular person makes a claim but it is treated skeptically and critically until its soundness can be judged according to the weight of evidence which is relevant to it. A person who is willing to follow such a procedure (and who regularly does so) is said by science educators to be motivated by the scientific attitude. (Gauld 1982, p. 110)
Dieterich (1967, pp. 23-24) divides "scientific attitude" into twenty
1. Skepticism. Not taking things for granted.
2. Faith in the possibility of solving problems.
3. Desire for experimental verification.
5. A liking for new things.
6. Willingness to change opinions.
8. Loyalty to truth.
9. An objective attitude.
10. Aversion to superstition.
11. Liking for scientific explanation.
12. Desire for completeness of knowledge.
13. Suspended judgement.
14. Distinguishing between hypotheses and solutions.
15. Awareness of assumptions.
16. Judgement of what is of fundamental and general significance.
17. Respect for theoretical structures.
18. Respect for quantification.
19. Acceptance of probabilities.
20. Acceptance of warranted generalizations.
"This term has connoted a students approach to thinking about science." (Haladyna & Shaughnessy, 1982, p. 548) One derives the impression that a person in possession of the "scientific attitude" is advanced cognitively, is a person for whom the art of asking after the truth (in the ideal sense) has become a way of life. I am reminded, in this context, of the following from Karl Popper:
So my answer to the question How do you know? What is the source or the basis of your assertion? What observations have led you to it? would be: T do NOT know: my assertion was merely a guess. Never mind the source, or the sources, from which it may spring--there are many possible sources, but I may not be aware of half of them; and origins and pedigrees have in any case little bearing upon truth. But if you are interested in the problem which I tried to solve by my tentative assertion, you may help me by criticizing it as severely as you can; and if you can design some experimental test which you think might refute my assertion, I shall gladly, and to the best of my powers, help you to refute it. (Popper 1963, p. 27)
It is not surprising that the attitudes expressed above are considered
valuable by science educators. Reid and Hodson (1987) tell it thus:
In other words, they are the attitudes associated with the successful study and practice of science. They are also the necessary learner attitudes underpinning the principal psychological stance taken in this book-that the major cognitive task facing the teacher is to bring about a shift from personally held beliefs to the accepted scientific paradigm. If this is the case, then teachers will need to foster in pupils a set of attitudes--such as open-mindedness, self- criticism... (p. 54)
"Attitude-toward-science", on the other hand, is an affective objective.
Krathwohl, Bloom, and Masia (1964) tell us that affective objectives:
...emphasize a feeling tone, an emotion, or a degree of acceptance or rejection. Affective objectives vary from simple attention to selected phenomena to complex but internally consistent qualities of character and conscience, (p. 34)
Simply, a persons attitude-toward-science is a measure of how well that person likes science.
Because this is the focus of the present study, I devote the next chapter to examination of attitude as a psychological concept.
LITERATURE REVIEW History of the Attitude Concept
Lemon (1973) opens his book with the following sentence: "Attitude is one of the most ubiquitous of all the terms used in social science," (p. 1) He goes on to say that the popularity of the term is due, in part, to its highly generic nature: "Unfortunately for the conceptual state of the field, it seems that this is precisely why it has proved so attractive..." (p. 1) People involved in attitude research can shape it into any form that suites them. As a result there are many definitions of attitude currently in use. Such a widespread usage of so many ideas under one roof, as it were, can be seen as having detracted "...from the operational clarity of attitude and rendered it a pot pound term with no generally accepted definition." (p. 1) As Fishbein and Azjen (1975) point out:
Clearly, if attitude research is to reach the goal of providing a cumulative and systematically integrated body of knowledge concerning attitude phenomena, research must be more than quasi-methodological treatments of pseudo-problems created by conceptual ambiguities or procedural weaknesses, (p. 516)
However, as Allport (1935) has pointed out:
It is not difficult to trace the common thread running through these diverse
definitions. In one way or another each regards the essential feature of attitude as a preparation or readiness for response. The attitude is incipient and preparatory rather than overt and consummatory. It is not behavior, but the precondition of behavior. It may exist in all degrees of readiness from the most latent, dormant traces of forgotten habits to the tension or motion which is actively determining a course of conduct that is under way. (P- 8)
As we shall see later, the behavioralists might take issue with Allport
over the assertion that attitude is not behavior, but that is another issue.
The concept of attitude is a psychological one. Allport (1935, p. 4)
quotes Spencer, a psychologist who wrote in First Principles (1862):
Arriving at correct judgments on disputed questions, much depends on the attitude of mind we preserve while listening to, or taking part, in the controversy: and for the preservation of a right attitude it is needful that learn how true, and yet how untrue, are average human beliefs.
Another "ancient" psychologist quoted by Allport (1935) was Alexander Bain
who wrote in 1868:
The forces of the mind may have got into a set track or attitude, opposing a certain resistance as when some one subject engrosses our attention, so that even during a break in the actual current of our thoughts, other subjects are not entertained, (p. 4)
However they might disagree upon the nature of attitudes in so far as they appear in consciousness, all investigators, even the most orthodox, came to admit attitudes as an indispensable part of their psychological armamentarium. Titchener is a case in point. His Outline of Psychology in 1899 contained no reference to attitude; ten years later, in his Textbook of Psychology, several pages are given to the subject, and its systematic importance is fully recognized, (p. 5)
Attitude research has had its ups and downs over the years. It
flourished during the first thirty years of the 20th century, but, according to
Probably the concentration on attitude work had passed its peak by the late
1930s, and by 1950 the focus of action had definitely shifted to group dynamics, (pp. 136-137)
McGuire attributes this decline in attitude research both to exciting new developments in group dynamics and to shortcomings in the field of attitude research. He elaborates:
One defect of attitude research between 1920 and 1945 was that theorizing became top heavy with conceptual elaboration...A second cause [was],..The intrinsic interest of such attitudes as pacifism, religiosity, ethnic prejudice, etc., [which] attracted workers with little interest in the general theoretical issues so vital to basic research, and as a result something of a hiatus opened up between the empirical workers on attitudes and the mainstream of developing psychological theory...Third...was the premature aspiration for quantitative precision that was imposed on the field by the fine work on attitude scaling done by methodologists attracted to the area [e.g. Guttman (1950); Lazarsfield (1950); Likert (1932); Thurstone and Chave (1929)]. McGuire (p. 137)
McGuire attributes the resurgence of attitude research in the late 1950s and 1960s partly to a man named Carl Hovland "...whose part in making attitude-change research exciting is analogous to the role [Kurt] Lewin played in group dynamics." (1969, p. 138) It is interesting to note that Hovland did research in Army morale during the war. "After the war, a continuous stream of psychologists worked around Hovland at Yale in the Attitude Change and the Communications Research Projects..." (p. 138) Another researcher, Leon Festinger, surfaced at about the same time as Hovland. According to McGuire:
It is not often that one field has had the good fortune of simultaneously receiving the attention of two sets of good researchers whose styles differ so greatly as do the current Festinger and Hovland types of attitude workers. (P- 137)
The differences in the methodologies of the two workers are described in the next section. The differences are still basic today.
Ten of the last fifteen years of my life I spent roaming the United States with associates performing geophysical experiments which allowed us to "see" via instruments through the surface of the earth into the interior. Our method required that we charge the area of interest, either directly or by induction, depending on the method, with an electromagnetic field. This field was our independent variable; we could adjust its strength to suit our needs. Once the volume of earth was electromagnetically altered to specifications (this occurred when all the transmitting equipment was operating correctly) a receiving device would measure the amount of specified field remaining at either one or a variety of sequential and calibrated times. These values were normalized, via generally agreed upon mathematical algorithms, to familiar (standardized) dimensional quantities. The values of the measured normalized quantities were then compared with theoretical normalized quantities with the same dimensions. The theoretical quantities used for comparison were chosen in advance. The choice was based on the prediction of a model, which was derived from previous experience with the same phenomena in a similar geological environment, and derived via mathematical calculations. If the measured quantities did not correspond to the theoretical quantities, either the equipment was checked, if we had reason to suspect the independent variable, or the chosen model was reevaluated: maybe we had
misinterpreted the geology. The normalized quantities were our dependent variables, and were usually a measure of either the inductance, or the capacitance of the volume of earth measured. Depending on the environment we were attempting to assess, and/or the equipment currently available, we might opt for a high power, "brute force" transmission of field, or we might decide to use a very sensitive receiving antenna to measure the dependent variable, after having only weakly manipulated the independent variable. Although what we theorized would be the optimum survey configuration was necessarily edited to accommodate environmental and instrument availability constraints our choice of the strength of the independent variable and/or the sensitivity of our measurement of the dependent variable was usually based on eminently practical considerations. These types of constraints as well as the methodological approaches that are their consequences unite the physical and social sciences. Consider the following from Kiesler, Collins and Miller (1969):
There are two major portions of an experiment between which effort can be divided-the independent manipulation and the dependent measurement. If an experimental test fails because it is insensitive, the insensitivity may stem from weak manipulations of the independent variable or, instead, from unreliable dependent measures. Since the day to day realities of the working researcher inevitably impose a compromise with ideal practice, every experiment necessarily represents an accord between effort and a thoroughly adequate test of a theoretical proposition. Thus the practical experimenter often must decide whether to spend time strengthening his independent manipulation or refining his dependent variable, (p. 22.)
In attitude research, a basic philosophical distinction involves the emphasis placed on dependent variable measurement vs. independent variable manipulation. In a discussion of two prominent 1950s attitude
researchers Karl Hovland and Leon Festinger, McGuire (1969, p. 140) explains that,
It is interesting to note the contrasting stances regarding interaction effects that seem to be implied by the two styles. A devotee of the Hovland style seems to regard progress of the science as positively related to how high an order of interaction effects one needs to test ones current hypothesis. The Festingerian, on the other hand, seems to regard interaction effects as at best a necessary evil, resort to which is indicative of ones failure to tease out and manipulate the crucial variable in the situation, of ones failure to narrow down the experimental situation to the theoretically relevant issue.
In the attitude research represented in CHAPTER III of this report, the
Hovlanders are represented by the survey takers, the users of complex
statistical algorithms (factor analysis, canonical analysis), they
...exercise creativity by searching out eclectically the various theoretical formulations that suggest relevant independent variables to account for parts of the total variance in the phenomenon under study, (p. 139)
The Festingerians, on the other hand, are the experimenters, the creators of
controlled environments that..."utilize extremely clever and elaborate
manipulations of the independent variable. (p. 139) The Festingerians
create very specific environments where their specific questions can be posed.
In attitude-toward-science research, as we shall see in CHAPTER III, the
Festingerians commonly rewrite science curricula and compare, via some
measure, changes in attitude after exposure to the curricula, or differences in
post-course attitude between students receiving the curriculum "treatment" and
students who were exposed to their normal curriculum. The meta-analysis in
CHAPTER III, contains 69% survey studies, which may possibly indicate a
Hovlanderian predisposition currently dominating the field. Haladyna and
Shaughnessy (1982) conclude their meta-analysis (which I use as the model for
the meta-analysis presented in CHAPTER III) with remarks that are essentially meant as compliments to the field of attitude research. They feel that research has greatly improved between 1960 and 1980. Their reasons are blatantly Hovlanderian.
Studies published in the more recent decade are better designed, research questions more clearly stated, samples better described, and instruments more completely described and validated. The most significant change has been in statistical analysis. Virtually all older studies involved simple comparison of group differences which involves analysis of variance or t-tests, while the balance of these studies involved correlations. A few of these earlier studies employed techniques which involve tabulation instead of statistical analysis. Recent studies have used regression analysis and other multivariate techniques with much success. Certainly researchers are encouraged to continue to use these multivariate techniques when they are appropriate. While it can be argued that such results are difficult to communicate, the results of such analyses provide more substantial findings than we realize with univariate analyses, (p. 558)
All the types of data presentations mentioned above by Haladyna and
Shaughnessy (1982) are represented in CHAPTER III. Generally, the studies
conducted during the 1970s contain fewer examples of sophisticated statistics
than do the studies conducted during the 1980s, though there are exceptions
(see APPENDIX B where the studies are described in detail).
Another difference between the Festingerian and Hovlanderian schools is that the Festinger researchers tend to use small samples, emphasizing individual interactions with the subjects; while the Hovlanderian uses a larger sample and sacrifices rapport and subject involvement for economy of data acquisition time.
Since the Hovlanders tend to control extraneous variables by deliberately manipulating them as orthogonal independent variables, their designs are usually more complex; the Festinger group is more inclined to eliminate extraneous factors by holding such variables constant through elaborately staged experimental situations that allow only the theoretically crucial
factor to vary. Looked at negatively, these different control factors result, in the Hovland case, in effects that take the form of high-order interactions that are hard to interpret; and in the Festinger, in complex experimental procedures that are hard to replicate. (McGuire 1969, p. 140)
These methodological differences are reflected in the research sample
included in the following meta-analysis (CHAPTER III): the surveys tend to
use large samples and analyze test scores with multivariate statistics, while the
experiments are of smaller scope and almost exclusively use the F-ratio to
report significance of results.
The distinction between the two methodologies has been somewhat exaggerated in the discussion above. McGuire (1969) reminds us that researchers usually dont fall exclusively into the Hovland or the Festinger style categories.
We did not sharpen the contrast between them to show that a choice must be made, or to clarify the issue before voting on which is the better.
Rather we attempted to show that there are two rather different current approaches to the study of attitudes, each accepted by the establishment and each productive, (p. 140)
We will see evidence in CHAPTER IV that there is some active communication between the two approaches. Such communication can have the effect of optimizing the potential of both methods.
Some Theoretical Conceptions of Attitude
Lemon (1973) (and others) outlines some of the major philosophic schools of attitude conceptualization. It is instructive to summarize them.
In the probabilistic view, attitude is simply behavior. Certain patterns
emerge in the behavior of an individual, and these patterns can be used to
predict future occurrences of the behavior. Liska (1975) explains it clearly:
The probability conception of attitude is convenient and obviously quite simple. For one thing, this type of definition anchors the attitude concept firmly to observable events. By defining attitude as the probability of particular types of responses, the need to worry about the nature of "underlying mechanisms" operative within the individual organism is absent. (P- 29)
But, there is disagreement whether a discrepancy between attitude and the behavior observed by researchers is due to some dishonesty or ineptitude in the actor or to some problem with the measurement techniques used by the researcher. Ibis is a crucial point; and, one of the answers given in response is that consistent behaviors (behaviors repeated over and over again in similar situations at different times as measured by different researchers) can point to properties that will predict them. The implication here is that if behavioral consistency is real, then, regardless of the bias of the researcher, or the "dishonesty" of the actor (these sources of error will cancel statistically) it is measurable. In other words, if we can see it and it lasts, then its there.
Liska (1975) compares the probability paradigm to a physical law like the inverse square law. "Many theorems and consequences can be deduced from this general law-propositions concerning the tides, the trajectories of cannon balls, and the paths of artificial satellites." (p. 29) But these types of theories, while possibly being excellent predictors of physical phenomena, say
nothing about what causes the phenomena. The probability theorist may react askance to such a notion. It is enough to predict. Whether the theory is "true" or not is irrelevant. The Ptolemaic conception of the solar system is worthy because of its predictive powers. And I believe this is a good point. The question still exists, however: Do probabilistic theories actually predict human behavior and/or attitudes the way the inverse square law predicts the trajectory of a cannon ball?
...attitudes always produce pressure to behave consistently with them, but external pressures and extraneous considerations can cause people to behave inconsistently with their attitudes. Attitude or change in attitude tends to produce behavior that corresponds with it. However, this correspondence often does not appear because of other factors that are involved in the situation. (Freedman, Carlsmith & Sears 1970, pp. 385-386)
A criticism of the behavioralist, operationalist (positivist) view, however,
concerns the exceeding complexity involved in measuring (in defining!) every
"bit" of behavior. According to Mcguire (1969, p. 144):
The positivist approach, however, seems to offer a false economy, for it is extremely demanding with the regard to the number of relationships that must be defmed...a total of m x n relationships must be established. For example, with six antecedent variables and eight consequent measures, the positivistic approach would require defining only 14 constructs but 48 relationships.
Latent Process View
Another type of theory is the "latent process" or "hidden mechanism" theory. McGuire (1969) describes this approach as the most popular. Liska (1975, p. 30) tells us: "...hidden mechanism theories account for the behavior of phenomena on the basis of the action of some internal process which is not immediately observable." These processes mediate observed behavior
patterns. They are hidden processes. Allport (1935), tells us that attitude (the latent process) is:
...a mental and neural state of readiness, organized through experience exerting a directive or dynamic influence upon the individuals response to all objects or situations with which it is associated, (p. 8)
Allport has left the world of observable behaviors, postulating processes that
can only be indirectly inferred.
Attitude is...conceived as a mediating construct which relates social situations with responses in such a way that the influence of a number of antecedents is seen as being channeled through a single mediating process, which then generates an appropriate set of responses. (Lemon 1973, p. 13)
The mediating construct is inferred from patterns in successions of
antecedents and responses, though it does not completely correspond to any of
them. The mediating construct is economical. McGuire (1973) observes:
This approach does require intervening an additional construct defined only indirectly in terms of observable, but it allows a great saving in the number of relationships that need to be defined, (p. 146)
Fishbein and Azjen (1975) describe the formation of a mediating response with the example of a child who eats M & M candies. The presence of the M & Ms In the childs mouth produce overt responses such as salivating, sucking, and swallowing. At about the same time as (or just before) the overt responses are occurring, the child experiences what Fishbein and Azjen (p. 25) describe as "...an implicit response with a positive evaluative component," (I think they mean the child likes the candy.) According to mediational theory, there is a tendency for implicit responses like enjoying the candy in your mouth, to become associated with the candies themselves, even before theyve been eaten. Who cant remember a time when the thought of
some food delight caused a moment of excess saliva production? But, there is an additional bonus associated with this type of conditioning. If, for example, each time that I go to a certain restaurant, the food produces "an implicit response with a positive evaluative component," I should develop a positive attitude toward the restaurant. This is the principle of higher-order conditioning. Attitude-toward-science research is designed to communicate information to educators so that they can produce positive implicit evaluative responses toward science in students.
Another approach mentioned by Lemon (1973) is the paradigmatic approach.
This involves selecting one particular response which is made to a particular antecedent, and regarding this as the definitive response which defines the attitude under consideration...Some criterion is selected (e.g. church attendance, membership in the Communist Party) as the behavioral criterion for the attitude in question and all other attitude measures are judged in terms of their correspondence with this criterion, (p. 12)
The main objection to the paradigmatic approach involves the question of how
the criterion is selected. Why would church attendance, for example, be
considered as the main criterion of a religious attitude? McGuire (1969)
The question arises: is it possible? Is there one A [antecedent] and one R [response] so fundamental and central that any other antecedent variable, insofar as it is related to any other consequent, can be adequately defined as approximating the two paradigmatic variables? (p. 147)
Nevertheless, according to Lemon (1973), much of the then (to 1973) current
experimental social psychological research into attitude change was based on
the paradigmatic approach. Indeed, as we shall see, much of the current experimental research is used to test theoretical paradigms. In paradigmatic experimental research, for example, a relationship between a dependent and an independent variable is postulated (or inferred from theory or previous results). The independent variable is manipulated and the changes (if any) that occur in the dependent variable (the criterion) are measured using with some instrument. Differences in the scoring of the criterion between pretest and post-test are attributed to the experimental manipulations, assuming the experiment is valid. If changes do occur, then the paradigm is supported.
The paradigmatic approach is distinctly Festingerian and, therefore its conclusions are generalizable only in narrowly defined specific situations.
Various authors have differentiated attitude with respect to other similar concepts. Koballa (1988) distinguishes attitude from belief, opinion, value, behavioral intention and behavior. Fishbein and Azjen (1976) distinguish attitude from belief, behavioral intention and behavior. McGuire (1976) distinguishes attitude from knowledge, values and opinions. McGuire tells us that Allport distinguished attitude from opinion, interest and value.
Belief. Opinion and Knowledge
According to Fishbein and Azjen (1976):
Whereas attitude refers to a persons favorable or unfavorable evaluation of an object, beliefs represent the information he has about the object...For example, the belief "Russia is a totalitarian state" links the object "Russia"
to the attribute "totalitarian state." (p. 12)
Koballa (1988) explains that belief has the property of intensity.
For example, one person may be absolutely certain that acid rain is an environmental hazard, whereas someone else may believe that acid rain is only a possible environmental hazard, (p. 118)
Opinions have been placed between attitude and belief. McGuire (1969) calls opinions beliefs that follow from expectations or predictions as to how some event will unfold, whereas attitudes are beliefs that follow from what one wishes would unfold. He goes on to say that the distinction between attitude and opinion is close to a hairsplitting one, and warns the researcher to beware of too much conceptual elaboration.
According to Koballa (1988, p. 118) "...attitudes (feelings) are dependent upon our beliefs (knowledge)..." McGuire (1969) explains that while attitude has cue and drive potential, knowledge has only cue potential.
It informs us but doesnt drive us. It does provide us with any information we may decide to incorporate into our beliefs and opinions, though it is certainly not an essential part of either opinion or belief, as association with the opinions and/or beliefs of the inadequately informed has demonstrated to this author.
Value has been called a sort of broad attitude, inhabiting the
continuum: opinion, attitude, interest, and value. Values have also been
regarded as components of attitudes. McGuire (1969) explains that:
An attitude toward some state of affairs is defined as a composite of the valence (positive or negative) of all the values or goals to which that state
of affairs is perceived to have positive or negative instrumentality, (p. 151)
Koballa (1988) defines values as culturally derived. He uses respect for human life and distaste for communism as examples of culturally derived values. He explains that, for example the differences between the value systems of poor people with little leisure time and the value systems of richer people who supposedly have more time for the creative pursuits might "...in part explain the behavior of students in science class." (p. 119) Value, then, is a variable that attitude-toward-science researchers might want to explore. However, in the studies analyzed in CHAPTER III, value was not used as a variable. Haladyna, Olsen, and Shaughnessy (1982), Haladyna and Shaughnessy (1982), and Shaughnessy, Haladyna, and Shaughnessy (1983) for instance, dont specifically mention value, though they would no doubt include it in their exogenous student category with variables like family background and hours of TV watched. Haladyna and Shaughnessy (1982) view exogenous variables as being virtually untouchable by educators; and although they may indeed point to reasons for a certain attitude, they are not easily manipulated by the educator, so they are not real promising as research variables.
Koballa (1988, p. 120) tells us that, "Whereas attitudes are feelings toward a target and beliefs are cognitive links between the target and various attributes, behavioral intentions are a persons intention to perform a specific behavior." Fishbein and Azjen (1976) call behavioral intention a special case of belief "...in which the object is always the person himself and the attribute
is always a behavior." (p. 12) However, just because people intend to behave in such a way, doesnt necessarily mean they will.
Overt behaviors include several types of data. Fishbein and Azjen (1976) remind us that responses to surveys and questionnaires should be considered as overt behavior. They are, after all, observable acts. Though, usually the researcher is interested in a type of overt behavior other than questionnaire answering behavior and designs strategies or develops instruments that suggest its determinants. According to Koballa (1988) the ultimate goal of attitude-toward-science research should be the promotion of science related behaviors. Whether or not science related behaviors can be accurately inferred from answers to questionnaires is another issue.
According to the functional view, attitudes are best understood in terms of the functions they serve for the person. These functions have been subdivided various ways by different authors. Fishbein and Azjen (1976) mention two such divisions: 1) object appraisal, social adjustment, and extemalization; 2) instrumental, adjustive, ego-defensive, value-expressive, and knowledge. Gardner (1976) divides functions into: utilitarian (adaptive), economy (knowledge), expressive (self-realizing), and ego-defensive functions. It is helpful to discuss these, and explore any connections they might have with models used in studies included in CHAPTER III.
Attitude-toward-science research has a functional basis. Assuming that achievement in science is a goal to be achieved and positive attitudes toward science facilitate the attainment of that goal, attitudes toward science are functional. That researchers have generally accepted this idea is clear from the number of studies using attitude-toward-science itself as a valued goal, assessing various environment-attitude connections and experimentally manipulating the environment in an attempt to facilitate a positive attitude-toward-science in students. In a discussion of functional notions McGuire (1969) says that the instrumentality of positive attitudes towards achievement of certain goals is easily assimilated into a stimulus-response (SR) learning theory approach. Gardner (1976) is an interesting example of a study that uses an SR theory as a framework for the study. In the theory, the needs of the individual interact with the press of the environment. According to Gardner:
Needs are motivated personality characteristics, representing tendencies to move in the direction of certain goals; press are environmental influences which support and encourage the expression of needs, (p. 113)
In Gardners paradigm, manipulation of the learning milieu is thought to have
the property of optimizing the needs/press center of gravity.
Psychologists have played with the idea that humans possess a basic, primitive need to know. (McGuire, 1969) Such a drive (McGuire lists curiosity, need to explore, and orient) may be primary: hardwired into our central nervous systems, or acquired by reinforcement (learned). Research indicating that a knowledge function might be reinforced (see McGuire p.
159) is good news for attitude-toward-science researchers. Possibly, systematic conditioning can re-instill a fading knowledge function. Campbell (1972) in his study of 1300 7th- through 12th-graders, found that curiosity toward science decreased from grade 7 through 12. Additionally, he found that, as the complexity of a scientific concept increased, the curiosity of his subjects also decreased. His findings agree with similar findings cited by McGuire (1969), who believes this area to be an important one for attitude change researchers. Unfortunately, the studies represented in CHAPTER III (with the exception of Campbell  and, possibly Gardner ) are not concerned with the knowledge function.
Expressive and Ego-Defensive Functions
According to the expressive function, a person "...adopts attitudes to bolster or justify his behavior." (McGuire 1969, p. 159) Thus, if a person is forced to comply with a new norm, his attitude will change to correspond with the new norm. Theorists cited by McGuire agree that "...bolstering is a common tactic in the face of psychological conflict." (p. 160) However, I dont
believe that attitude-toward-science researchers would use the expressive function to instill attitudes toward science. They should recognize it as a disrupting influence to positive attitude change.
The ego-defense function has been touted as a sort of knee-jerk phenomenon. A traditional psychological interpretation involves the example of a person who is anti-Semitic, though, not because of any unpleasant encounter with a Jew, but because of repressed oedipal hostility toward his father. For some reason, whenever this person sees a Jew he experiences a negative "knee-jerk" reflex sort of reaction. Whatever the cause, the persons prejudice springs from some unsatisfied "inner need." This need should be considered as a possible entry condition that may need to be dealt with prior to effective attitude-toward-science indoctrination.
The studies in CHAPTER III generally do not address these types of entry conditions of the student into a curriculum or learning environment.
The unfulfilled emotional needs of the student are usually included in a larger student variable category, like family background or SES. One reason for this lack of attention is the lack of accessibility of the students family life.
One of the goals of attitude-toward-science research is the understanding of behavior. How do attitudes interact with other variables to produce behavior? According to Fishbein and Azjen (1975) there are two interpretations that can be given here. According to the first interpretation,
the relationship between attitude and behavior is moderated by other variables. For example, a students attitude-toward-science may show little relation to behavior if the student lacks sufficient skills to achieve in science. The students ability or lack of ability interacts with his attitude in determining his behavior. Several researchers included in CHAPTER HI have used sex and/or grade level as interacting variables moderating either achievement in science or attitudes toward science. This is usually accomplished by reporting results separately for boys, girls, and grade level.
According to the second interpretation, other variables act
independently of attitude to produce behavior.
Among the variables suggested are other attitudes, competing motives, verbal, intellectual, and social abilities, individual differences, actual or considered presence of other people, normative prescriptions of proper behavior, alternative behaviors available, expected and/or actual consequences of various acts, and unforseen extraneous events. (Fishbein and Azjen 1975, p. 344)
According to Fishbein and Azjen (1975), most of these variables have received little systematic study. However, one study reported by them measures the independent effects of: 1) how we perceive the consequences of behavior; 2) how we feel our behavior is evaluated by others; 3) the effects of extraneous events on our behavior; and 4) attitude on three religious behaviors (i.e. church attendance, contributions, and participation). Attitude (4), evaluation (2), and extraneous influence (3) were shown to be significantly correlated (p < .01) with church attendance and contribution behaviors. The perceived effects (consequences) of behavior category correlated significantly at p < .05 with attendance and participation behaviors. However, results such
as those presented above do little more than show that "...attitudes and other variables may or may not be related to another behavior." (p. 346) They call for the development of a framework for relevant "other variables." Within their framework intentions are seen as antecedents of behavior. Understand the intentions and predict the behavior. But how does one measure intention?
Since it is often impossible or impractical to measure a persons intention immediately prior to his performance of the behavior, the measure of intention obtained may not be representative of the persons intention at the time of the behavioral observation. Intervening events that may lead to changes in intentions will therefore also have to be taken into consideration, (p. 382)
They propose a model where behavioral intention is predicted by attitude and subjective norm. In other words, my intention to engage in an activity or behavior depends not only upon how / feel about the activity or behavior but upon how my significant others feel about the activity or behavior. The importance of either attitude or subjective norm can be adjusted in this model. Sometimes we do things more because we want to, other times because we have to, and other times, we are somewhere in between.
Koballa (1988) found clear support for Fishbein and Azjens hypothesis, namely: "...(1) the behavioral intentions of preservice teachers are quite predictable from the simultaneous consideration of their attitudes and subjective norms." (p. 501) However, Koballa found that measuring preservice teachers attitudes toward science alone does not provide enough information in itself to enable a behavioral prediction.
In educational attitude research the presence of "other variables" is fundamental. Attitude-toward-science, specifically, is normally modeled as interacting with student variables such as self-concept, motivation, family variables etc., with classroom variables like class emotional climate, type of teacher, and types of students. In Fig. 2 a framework developed by Simpson, Troost (1982) shows affect toward science interconnected with four broad categories of variables, "Conceptualization of how the categories of variables interconnect and interrelate is based on an extensive review of the literature and on prior work of the investigators." (p. 772)
Another framework developed by Hermann (1988), (see Fig. 3), shows how attitude along with a host of other variables act as governing causes and "...influence the meaning of human experience and the manner in which people think, feel, and act within their environment." (p. 697) We might also want to consider the ego-defensive and expressive attitude functions as primary governors. Secondary governors (heredity, family etc.) are more structural. They more or less outline the subjective norms surrounding the educational environment. The primary governors (learner, teacher, curriculum), then, fueled by various combinations of governing causes, and regulated by various combinations of secondary governors (possibly in a
Figure 2. Major variables being investigated with potential interconnections depicted. From "Influences on Commitment to and Learning of Science among Adolescent Students" by R. D. Simpson and K, M. Troost, 1982, Science Education. 66(5), p. 772. Copyright 1982 by John Wiley and Sons Inc.
way analogous to how a governor regulates engine RPM under the effects of different loadings), plunge into the milieu where learning occurs. "A learning environment, or milieu, is the product of [the] interaction of learner, teacher, and curriculum." (Germann 1988, p. 698)
Haladyna and Shaughnessy (1982); Haladyna, Olsen and Shaughnessy (1982); and Shaughnessy, Haladyna and Shaughnessy (1983) propose a model
learning environment learnerteachercut iculum rapport
affecting actions of governors
World Views Belief System Life Goals Life Style Needs Skills Attitudes
GOVERNORS GOVERNORS MILIEU
Peer & Colleagues_____
Govemnent Officials Pifolisfiers S Authors,. Life Experiences Educational Experiences
where attitude-toward-science is a function of three classes of variables:
Figure 3. The five commonplaces of classroom instruction. From "Development of the attitude toward science in school assessment and its use to investigate the relationship between science achievement and attitude toward science in school", by P. J. Germann, 1988, Journal of Research in Science Teaching. 25(8), p. 697. Copyright 1988 by the National Association for Research in Science Teaching.
student, teacher, and learning environment. Each of the three classes of variables are differentiated with respect to whether they are functions of interactions outside the school environment: the exogenous variables or interactions inside the school environment: the endogenous variables. Fig. 4 shows the schematic of the process. Student attitude-toward-science is
The Schooling Process
-5* T eacher
Figure 4. Framework for the model developed by Haladyna & Shaughnessy. From "Attitudes toward science: a quantitative synthesis" by T. Haladyna and J. Shaughnessy, 1982, Science Education. 66(4), p. 550. Copyright 1982 by John Wiley and Sons Inc.
governed by a combination of all three variables, each of which may add more or less a portion of endogenousness and/or exogenousness depending upon their place along the exogenous-endogenous continuum.
Note that all three of the above models contain regulators which allow for variations in both the combinations of variables and in the location (within the class or without) of the variables. Thus they can account for a
wide variety of behaviors, paradigms, and mediating processes. They can provide useful schematic representations of what will probably occur in a wide variety of educational situations. That these three models are similar is not entirely coincidental. Although each was developed by different sets of investigators each was developed to describe the same processes. Such an agreement among independent researchers is considered valid support of a theory in scientific circles.
THE META-ANALYSIS Definition of Meta-Analvsis
Meta-analysis has been defined by Glass (1982, p. 100) as "the statistical analysis of the summary findings of many empirical studies." Glass further characterizes meta-analysis as "undeniably quantitative" (p. 100), its purpose to derive quantitative patterns within myriad otherwise incomprehensible data. It is weakly related to survey research. According to Glass, "research review and integration is a process of surveying and analyzing in quantitative ways large collections of studies." (p. 94) Haladyna (1981, p. 1) calls meta-analysis the statistical analysis of statistical analyses: where relationships between dependent and independent variables are compared across studies. Steinkamp and Maehr (1983) characterize the procedure as an outgrowth of an ever increasing concern by researchers that they understand the results of large bodies of research. According to Tamir (1985) there are two types of metaanalysis. The first kind deals primarily with experimental and quasi-experimentai studies and calculates "effect sizes" based on treatment results. The "effect size" is defined as the ratio of the difference between treatment and control group means and the within groups standard deviation: ES =
(experimental group mean control group mean) / within group standard deviation. Its convenient; its easy; but its only use is with experimental studies. Also, it is subject to change if the standard deviations of the two groups scores are different: which one does the analyst use (control or treatment)? The latter form of meta-analysis is used to synthesize the results of correlational studies.
How can the results of correlational studies be compared with those results obtained by examining experimental studies? Surely we would benefit if such a communication were possible.
The Common Metric
As if the problems involved in comparing experimental and correlational studies were not enough, Haladyna (1981) decries the miasma of report formats used from study to study even of the same type. "In fact, there is a beleaguering variety of statistical methods and omissions of crucial data." (p. 5) The meta-analysis that follows this introduction finds the state of affairs described by Haladyna and Shaughnessy (1982) to be the case for the studies it examines also. According to Haladyna (1981) a "common metric" is needed that is capable of tying together both descriptive and experimental studies. He introduces a metric that he proposes is the common metric needed. He calls it the "percentage of accounted variance" (PAV) and explains that it can do the following things: compare any experimental or correlational result involving the interaction of two or more variables, require only a good general statistics textbook as a reference and be recovered from a
variety of statistical result presentations. Haladyna and Shaughnessy (1982) explain it further:
The squared product-moment correlation coefficient represents the percentage of criterion variance accounted by the predictor variable. For example if a correlation is 0.50, the squared correlation is 0.25 indicating that about 25% of the criterion variance is explained by the predictor. This extends to the multivariate case where the squared multiple correlation coefficient (R2), provides the same metric, percent of accounted variance.
When an independent variable has two or more levels, the results can be treated in a least squares regression framework where the sums of the squares can be computed. The ratio of the sums of squares between and the sums of squares total is identical to the correlation ratio, (p. 553)
The reader is referred to Haladyna (1981) for an elaboration of the details.
This author found the method useful in several cases.
Recommended Characteristics of a Meta-analvsis
Haladyna (1981) offers the following suggestions for optimizing metaanalysis results:
1. Identify the variables. This is probably the most critical part of the study, since misinterpretation or lack of cognizance of the variables involved would render the results difficult, if not impossible, to interpret.
2. Check the null hypothesis. Was it accepted or rejected? Or, if the study doesnt specifically mention a null hypothesis, are the reported results significant? If a null hypothesis is accepted, or the results are reported as insignificant then the PAV value is defined as 0%. This is because insignificant variance is regarded as error or noise and is, therefore not related to true association between variables. (Observed variance is the sum
of true variance and error variance.)
3. Know the unit of analysis of the study Haladyna (1981) pointedly illustrates how results can change from those obtained examining individual means to those obtained (from the same data) when the means of a group of individuals (the class, for instance) are compared. For example, in a study comparing attitudes toward mathematics with student, teacher, and learning environment (LE) variables, correlations between predictor and criterion changed dramatically when the unit of analysis was changed. Teacher quality correlated .48 with attitude toward mathematics when the scores of the individuals (n = 764) were used, but correlated .78 when the scores were analyzed as class means (n 34) (p. 11).
4. Compute the PAY for each result. If PAV is "the common metric" as Haladyna and Shaughnessy (1982) suggest, then if each study is carefully analyzed, and the results of the analyses are carefully checked, and the restrictions on PAV in terms of the null hypothesis and the unit of analysis are taken into account, then all the PAV data are comparable.
5. Present the results in a meaningful display. Haladyna and Shaughnessy (1982) like the stem-and-leaf method described by Tukey (1977) and used by Uguroglu and Walberg (1979), Walberg and Ahlgren (1970) and Steinkamp and Maehr (1983) to present raw correlations. I used the stem-and-leaf in my analysis, and found it useful.
A description of stem-and-leaf diagrams and how to interpret them is
provided in APPENDIX C.
Also found in my analysis is a presentation of PAV similar to that used by Haladyna and Shaughnessy (1982).
Design of the Study
Types of Studies Used
In the analysis that follows, 32 studies involving attitude-toward-science as either a dependent variable or an independent variable (or both) are used (see APPENDIX A). Eleven of the 32 studies involve attitude-toward-science as either both predictor and criterion or predictor only of achievement in science classes. In addition to the 32 studies mentioned above, 3 additional studies of attitude toward mathematics are included, for a total of 35 studies. Twenty two (63%) of the studies are surveys where one or more groups were asked to complete some kind of test or questionnaire. Five of the survey studies are longitudinal. Eight (23%) of the studies are experiments or quasiexperiments where a treatment is introduced and criteria, either before and after, or just after treatment, are measured.
After the studies were compiled each was analyzed according to the format of APPENDIX B. The studies were numbered 1-35 approximately alphabetically (#25 is not alphabetically positioned), and data about each study was compiled under 9 common categories (see APPENDIX A for the number and brief description of each study). In the studies where the data
were synthesizable to PAV, PAV between predictor and criterion was determined. This was difficult because of the variety of data presentation formats used. Methods explained in Haladyna (1981) (and described earlier) were used to accomplish the task.
The problem with non-standardization of data acquisition and presentation formats in studies, and the apple and orange problem with causal and non causal studies is not peculiar to Haladyna and Shaughnessy (1982). Willson (1983) had the same difficulties, and resorted to a theoretical framework to splice the formats into one. Steinkamp and Maehr (1983, p. 371) report that some of their correlations were "algebraically derived." I have spent the majority of my time dealing with the same difficulties. If only one significant conclusion comes from this humble analysis I hope it is that attitude research should present data and conclusions in some standard format. Indeed, one of the functions of meta-analysis, is to make such suggestions.
In studies where attitude-toward-science scores were presented in an intercorrelation format with the predictor scores, the correlations were consolidated in the stem-and-leaf format described in APPENDIX C.
Figs. 5, 6, and 7 compare study breakdowns by journal, subject matter, and grade level respectively, and compare the breakdowns of studies in my meta-analysis with those of Haladyna and Shaughnessys (1982) meta-analysis.
Craychee Haladyna and
n=35 n=49 n=28
1. Journal of Research in Science Teachine 18 52% 6 12% 21%
2. Science Education 8 23% 5 11% 18%
3. School Science and Mathematics 6 17% 7 14% 25%
4. Other Refereed Journals 3 8% 10 20% 36%
5. Fugitive Studies 0 0% 21 43%
Figure 5. Sources of studies.
Craychee Haladyna and
General Science 2,3,5,8,9,11,12,13,14,16,19,20, 220.127.116.11.18.104.22.168.34.35 23 6 8
Biology 6,15,17t18,22,24,25,3^ 8 5 3
Chemistry 4,15,33 3 2 5
Physics 2,15,17,32,33 5 4 9
Other 1,10 2 0 1
Subject Area Study designations (from APPENDIX A) Totals
Underlined number indicates high school study.
Figure 6. Subject matter breakdowns.
Craychee Haladyna and
n= : 63 n= 125
1 0 0% 0 0%
2 0 0% 5 4%
3 11,30,31 3 5% 4 3%
4 2,8,23,30 4 6% 10 8%
5 1,2,24,35 4 6% 10 8%
6 1,2,26 3 5% 17 14%
7 2,5,8,9,16,17,23,24,27,31 10 16% 13 10%
8 2,3,5,9,12,16,22,27 8 13% 11 9%
9 5,8,18,22,23,24,25,27,32,33 10 16% 8 6%
10 5,20,21,25,28,29,32,33,35 9 14% 15 12%
11 5,7,21,24,25,31,32,33,34 9 14% 16 13%
12 4,10,14 3 5% 16 13%
Grade Number of Study Total
n exceeds the number of studies for both analyses because several studies spanned more than one grade level.
Figure 7. Breakdown of studies by grade level.
In Fig. 5 data are reported as the quantity of studies per journal for Haladyna and Shaughnessys (1982) analysis and mine. Data were also normalized to percent total studies per journal as a rough way to facilitate pattern recognition.
Referring to Fig. 5 we can see that Haladyna and Shaughnessy (1982) use half fugitive (43%) and half archival (57%) studies, while my study includes only archival journal studies. Of the 28 archival studies analyzed by Haladyna and Shaughnessy, 36% are miscellaneous refereed journals; while of the 35 archival studies I analyzed only 8% are miscellaneous. Over half of the studies analyzed in this report came from the Journal of Research in Science Teaching, whereas Haladyna and Shaughnessy show a more even split among studies that come from the three named journals. Probably, the method I used to gather studies was less systematic than that of Haladyna and Shaughnessy. I used the method quoted below by Steihkamp and Maehr (1983, p. 371), "Especially fruitful for our purposes was examination of studies quoted in the text or appearing in bibliographies of selected studies." I started with a few studies I found in the Education Index, and using the bibliographies of those, rapidly built my collection. I conducted my search at the University of Colorado at Boulder which has an excellent collection of periodicals. Also helpful, but less so than the approach described above, were the ERIC system and the Social Science Citation Index.
Fig. 6 shows how the studies breakdown by subject. Haladyna and
Shaughnessy (1982) report subject matter breakdown for high school studies only, since science isnt split into separate disciplines usually until high school I report all the studies (all grade levels) included in my analysis in the left totals column of Fig. 6. As can be seen, however, the majority of the studies concerned with general science are not high school studies. When these are subtracted and the results (center column) are compared to Haladyna and Shaughnessys, (right column) a similarity is evident. Both samples contain more general science studies, followed, in number, by biology studies. My analysis contains even fewer high school chemistry studies (two) than biology or general science, while Haladyna and Shaughnessys analysis contains more. Also, for both analyses, high school studies examining physics (there are 13) rank second to the high school studies examining issues in general science. It is hoped that this similarity points to a possible generality of both samples to a similar universe of content.
Fig. 7 expresses the breakdown of studies by grade level. As in Fig. 5 both raw data and normalized percentages are used to compare the two analyses. The two sets are similar, though Haladyna and Shaughnessy (1982) analyze more 6th- and 12th-grade studies than I.
Only 4 out of the many instruments used in the analyzed studies (see APPENDIX B) were used in more than one study. The Science Attitude Inventory (SAI) was used in studies 2, 3, 14, 15, and 20. (see APPENDIX A) The Inventory of Affective Aspects of Schooling was used in studies 8 and 23,
Fig. 9 both of which are authored by the Haladyna and Shaughnessy (1982) team. The instrument used in the most studies is the Learning Environment Inventory (LEI). It was used by studies 8, 14, 15, 16, 23, and 32.
Fig. 8 compares the independent variables of the two analyses. The variables used are after Haladyna and Shaughnessy (1982, p. 555) since it was hoped their use would facilitate ease of comparison. For purposes of clarification, by independent variable I mean a variable that has been shown at some previous time to correlate with a criterion that researchers (as agents of the society) are interested in predicting. Based on a pattern of previous correlations, it can be said that the presence in a subject of a predictor variable (e.g. high grades) with such and such characteristics predicts, to a certain extent, whether some criteria (e.g. landing a job) may be realized (Smith & Glass 1987, p. 208). The PAV metric is an estimation of the predictive association between the criterion variable and the predictor variable.
The variables shown in Fig. 8 were picked by Haladyna and Shaughnessy (1982) because they are most prevalent in the literature. The characteristics of the bar graph (Fig. 9) indicate that the studies I analyzed used generally the same proportions of predictor variables as the studies Haladyna and Shaughnessy analyzed. Even though the absolute numbers are rarely the same, similarities can be noted. For example, sex
Variable Student Variables Number of Studv % Craychee % Haladyna & Shaughnessy
Sex differences 2 3 12 16 18 14% 35%
Interactions w/ sex 1 8 9 11 17 20 22 23 24 29 30 34 34% 18%
IQ 11 34 6% 12%
Achievement 1 4 7 9 12 17 20 21 25 27 29 33 34% 14%
Grade Level (time) 1 2 5 8 21 23 24 27 30 31 34 35 29% 29%
Grades 3 3% 8%
Academic Plan 18 20 22 9% 6%
Family Background 8 12 23 34 11% 6%
SES 8 12 35 11% 6%
Student Attitude 4 7 9 19 21 27 33 20%
Activities 14 29 6% 10%
Personality 7 8 14 25 29 14% 8%
Class 7 8 12 14 23 28 29 20% 4%
Effort 8 14 23 9% 4%
Quality 7 8 14 23 29 14% 4%
Attitude 7 13 29 33 11%
Learning Environment Variables
Program (treatment) 4 6 10 11 18 20 22 26 29 30 31 31% 45%
LEI instrument 8 14 15 16 23 32 17% 8%
Other 12 28 6%
The number of each study in the Craychee analysis using each predictor, and the percentage of studies in both Craychee and Haladyna & Shaughnessy (1982) using each predictor.
Figure 8. Predictor variables.
IIM Haladyna & Shaughnessy
Sex Differences Sex Interaction IQ
Achievement Grade Level Grades Academic Plan Family Backgrnd SES
0 10 20 30 40 50
0 10 20 30 40 50
0 10 20 30 40 50
Figure 9. Percentages of predictors used by Craychee and Haladyna & Shaughnessy (1982).
variables and program were used more than any other variables in the studies analyzed by both Haladyna and Shaughnessy (1982) and me. On the other hand, I analyzed more studies where the relationship between achievement and attitude was explored than did Haladyna and Shaughnessy. Generally, though, the two samples appear roughly comparable.
Studies in Common
Of the 35 studies analyzed here 3 were also analyzed by Haladyna and Shaughnessy (1982). Ayers and Price (1975) is a good example of a study that is nonsynthesizable. Data were presented only as percentages of students responding to various questions, though the study does find that attitudes decline with time. Walberg and Ahlgren (1970), on the other hand, is reported by Haladyna and Shaughnessy (1982) as being nonsynthesizable while I report it as showing that the attitude of students accounts for over 25% of the variance in LE. In other words, the attitudes of the students, among other variables, can be used to predict the classroom climate. Walberg and Ahlgren use a procedure called canonical analysis. Canonical analysis amalgamates variables into combinations (batteries) of variables with the highest possible correlation. Various batteries are then compared. Theoretically there are as many canonical variates as there are single variates involved. However, usually only a few combinations prove to be significant. Apparently, Haladyna and Shaughnessy feel that an intercorrelation between attitude and LE is not retrievable from the analysis. In any case, we should view the results of the Walberg and Ahlgren study as reported in my analysis with this caveat in
Hofman (1977) was synthesized by Haladyna and Shaughnessy (1982) and me. Although they dont mention Hofman by name, Haladyna and Shaughnessy do report (p. 556) that three experimental studies involving the SCIS (Science Curriculum Improvement Study) program and primary grade pupils report results less than 6%, which was what I recovered from the F-ratio data reported by Hofman (see APPENDIX A).
Fig. 10 shows how the synthesizable studies are distributed by PAV over student, teacher, and LE variables for this analysis (part A) and Haladyna and Shaughnessys (1982) analysis (part B). The use of the categories student, teacher, and learning environment (LE) for the variables used in this analysis is after Haladyna and Shaughnessy.
Of the 17 synthesizable studies represented in the student variable category, 12 have to do with attitude and achievement associations. Six (1,4,7,12,17,33) of these involve zero-order correlations between attitude and achievement variables, where no direction of causation is implied. (For a key relating study numbers to study authors refer to APPENDIX A). Six (3,19,21,25,27,29) are multiple regression (or ANQVA) studies where attitude was reported as a predictor of achievement. The remaining 5 studies (5,8,9,23,24) involve attitude as a criterion, predicted by curiosity, various
Student Teacher LE
Variables Variables Variables
25+5 8 21 (33) (14) (15) 28 (32) 25 +
20 8 9 28  8 20
15 12 15
14  22 14
13 3 17  13
9.5  9.5
8 9 27 29 8
7 4 19 25 (16) 7
<6 7 24 11 12 18 <6
Mean % = 14% Mean % = 24% Mean % = 16%
Part A: Craychee analysis.
Numbers indicate study codes (from APPENDIX A.) Underlined number indicates experimental study. (Parenthesized) number indicates group unit of analysis. [Bracketed] number indicates attitude toward math study. Bold number indicates attitude/achievement study.
Student Teacher LE
Variables Variables Variables
25+XXX XXX 25 +
20 XX XX 20
15 XXX X XXXXX 15
14 X X 14
13 X X XXX 13
12 XXX 12
11 X XXX 11
10 X X X 10
9.5 X 9.5
9 X 9
8.5 X XX 8.5
8 X 8
Part B: Haladyna and Shaughnessy (1982) analysis.
Figure 10. Strengths of association (percentages) between predictor variables and attitudes towards the subject matter of science across three dimensions of schooling.
student variables, sex, various student variables, and sex, respectively. It can be seen that the attitude/achievement studies are scattered all over the figure. Four studies (5,8,9,24) that deal with student variables predicting attitude are scattered around also. Shaughnessy et al. (1983), (study 23, Fig. 10), is the only non attitude/achievement study reporting midscale. It suggests that student variables are moderate predictors of attitude toward mathematics. That means that there are only 4 studies dealing with student variables predicting attitude-toward-science (5,8,9,24). Thats 11.4% of the studies I examined. A comparison of Fig. 10 parts A and B reveals somewhat similar data distributions in both analyses. Haladyna and Shaughnessys (1982) data show a tri-modality, with clusters near the top of the scale, near mid-scale, and near the bottom of the scale. My data are clearly bi-modal. Also noticeable in my data are the weak results reported for the two experimental studies reporting in this area which is in agreement with Haladyna and Shaughnessy. The cluster of mostly attitude achievement studies at the bottom of the scale is also in agreement with results reported by Haladyna and Shaughnessy. There is no generally agreed upon relationship between attitude-toward-science and achievement in science. Schibeci (1984) reports that a variety of attitude achievement associations were found in the studies he examined. Steinkamp and Maehr (1983) report about a 10% PAV for the attitude achievement relationship.
Fig. 11 shows an average correlation of 0.2 over all three categories displayed. The correlations presented in Fig. 11 are all intercorrelations
Ln w to o i- to U> Lri
.7 8 .7
.6 258 .6 8
299 .5 56 3 0236
335578 .4 000111246799999 .4 001226678
011668 3 00022223334567888 .3 0000111133444556666888999
00111122555669 .2 111122334455556668 .2 0022233344669
111234457778899 .1 2233558 .1 22334455667889
13566778 .0 2234556778899 .0 112577888
22258 -.0 1123457788 -.0 2344666789
24589 -.1 2 -.1 1
07 -.2 2226 -.2 16
-.3 -3 248
-.4 2469 -.4 2
-.5 -.5 134
Undifferentiated N = 64 Mean = .166 S = .192
8 Studies (1,4,5,7,12, 14,19,33)
Boys N = 95 Mean = .206 S = .255
5 Studies (1,8,9,17,23)
N = 95 Mean = .255 S = .251
5 Studies (1,8,9,17,23)
Figure 11. Correlations between attitude and student variables.
where no direction is implied between attitude and the student variable (please refer to APPENDIX C for an explanation of stem-and-leaf diagrams). The studies that reported their data by sex reported higher correlations than those which did not divide the data on the basis of sex. Also, girls are seen to display a higher mean correlation (.255) than boys (.206). If the data are representative of the population this means that girls attitudes toward science are more contingent upon personal variables than are the boys. This could be interpreted as an indication of societal constraints operating against girls becoming scientists, so they require more motivation than boys.
The two teacher variable columns of Fig. 10 contain some differences. Haladyna and Shaughnessys (1982) data are spread thinly, over roughly the upper half of the graph, while my data cluster at 20%. All 5 studies (in my analysis) involve teacher variables as predictors of attitudes. The association between teacher variables and attitude-toward-science is quite high. Twenty percent of the variance in attitude-toward-science can be attributed to teacher variables. In four of the five studies, the teacher variables reporting PAVs are endogenous. Only Lawrenz (1975) (study 14) reports the exogenous ones.
Fig. 12 is interesting. The studies which report correlations between attitude and teacher variables by sex report average correlations of .35 for boys and .36 for girls. However, in the data which are not broken down by sex the correlations show a definite bi-modality. The lower correlations are from Gardner (1976) (study #7) and are correlations between the Physics
.6 .6 2 .6
.5 0 0 3 133555
.4 01123456 .4 001145677899 .4 000123444569
.3 1235 3 133666789999 3 0356689
.2 139 .2 00224679 .2 00222378
.1 00022333444455559 .1 22669 .1 359
.0 011122345666777799 .0 .0
-.0 11223889 -.0 -.0
-.1 1337 -.1 -.1
-.2 -.2 -.2
-.3 -3 -3
-.4 -.4 -.4
-.5 -.5 -.5
Undifferentiated Boys Girls
N = 63 N = 41 N = 36
Mean = .123 Mean = .348 Mean = 355
S = .188 S = .120 S = .126
5 Studies (7,14,19,28, 29) 3 Studies (9,8,23) 3 Studies (9,8,23)
Figure 12. Correlations between attitude and teacher variables.
Attitude Inventory and the Physics Classroom Inventory (PCI). The PCI asks questions about how the student feels about the teacher. Gardners is the only study included in Fig. 12 which deals with the content of Physics.
Lawrenz (1976, p. 513) has indicated that attitude and classroom variables for physics students arent associated as strongly with achievement as they are with other science students. These data fit his conclusion.
Learning Environment (LEI Variables
Haladyna and Shaughnessy (1982) investigated a lot more studies involving LE variables than I did. Their graph contains over twice (22) the data points as mine (10). Nevertheless the data may indicate two broadly distributed clusters: one near the top and one near the bottom, resembling the student variable data distribution somewhat. Like Haladyna and Shaughnessy (1982), I have included under the rubric "LE" variable, studies involving experimental teaching procedures and surveys where either the LEI or some other LE instrument was administered to students. In my data the LE surveys cluster generally near the top (8,15,28,32). Only one is found in the middle (23).
All three columns of Fig, 13 may be bi-modal. Certainly the studies that dont differentiate with respect to sex show a bi-modal distribution, with clusters at the top and bottom of the scale, maybe indicating the existence of positive and negative types of environments. The studies that differentiate by sex hint at the same bi-modal distribution, though it is shoved in toward the middle of the scale. The negative mode data (in the undifferentiated column)
.7 .7 .7
.6 112222223335568 .6 .6
.5 0012244677 .5
.4 0001245669 .4 003 .4 03668
.3 112344799 3 00011122344456788 3 1112225566899
.2 12245555688 .2 012223346788899 .2 0001113344446677899
.1 23367 .1 111249 .1 23359
.0 089 .0 0033347 .0 12579
-.0 8 -.0 -.0 14
-.1 05 -.1 03445679 -.1 11256
-.2 022 -.2 02678 -.2 245
-.3 1 -.3 35 -3 1236
-.4 189 -.4 -.4
-3 1334677 -.5 -3
Undifferentiated Boys Girls
N = 82 . N = 65 N = 59
Mean = .221 Mean = .221 Mean = .225
S = .390 S = .221 S = .225
7 Studies (8,12,15,16, 28,32) 3 Studies (8,23,28) 2 Studies (8,28)
Figure 13. Correlations between attitude and LE variables.
are mainly from Lawrenz (1976) (study #15). This study pits attitude against LEI variables for 12th-grade chemistry, biology, and physics students.
The bi-modal distribution, if it really exists, might be related to the data cluster inhabiting the top of Fig. 10. This is so because in the PAV presentation we lose the directionality characterizing correlations. So, extreme correlation data points, positive or negative, inhabit the upper portion of a graph that portrays them as percentages. Similarly, correlation data points inhabiting the central regions of a stem-and-leaf diagram tend to hang out in the lower areas of a graph portraying them as percentages. It should be noted, also, that the correlations presented in Fig. 13 contain no experimental (treatment) data, mainly because experimental data is usually reported in an F-ratio format. Evident from Fig. 13, the mean correlations in all three columns (undifferentiated, boys, and girls) are essentially the same indicating that LE is equally associated with attitude-toward-science for girls as well as boys.
Haladyna and Shaughnessy (1982) report that studies involving LEI scales report high contributions of LE variables to attitudes. Schibeci (1984, p. 38) reports that "Positive associations have been reported between measures of classroom climate and attitudes in science..." My data indicate some strong (positive and negative) associations, but, also, some rather weak ones. The average value of all the correlation data points presented in Fig. 13 is 0.222, indicating that a rather lot of data points live near the center of the stem-and-
The highest percentage of studies examined by me (51%) and by
Haladyna and Shaughnessy (1982) (35%) involved studies where attitude was
reported broken down by sex, or where attitude and some other variable
(grade level, LE) are reported broken down by sex. Schibeci (1984 p. 33)
quotes Gardner (1976) as saying: "Sex is probably the single most important
variable related to pupils attitudes to science." Schibeci agrees:
His assertion appears to be still true today. Of the myriad variables which are possible influences on attitudes, sex has generally been shown to be a consistent influence. (1984, p. 33)
Nevertheless, Schibeci also reports on studies which show either negligible or no influence of sex on attitude. Steinkamp and Maehr (1983, p. 384) report that the influence of gender on affect is negligible in the studies they examined. The PAVs reported by Haladyna and Shaughnessy (1982) for the influence of sex on affect are low: ranging from 0 to 12.8%, with an average value of 3.2%. (p. 555) Of the studies I examined, results are similar. Ayres and Price (1975), for example report (p. 318), that there "...are no major differences in the science attitudes of males and females." Baker (1985, p. 109) reports that his findings indicate that females possess a more positive attitude-toward-science than males. Keeves (1975, p. 449) reports correlations of +0.25 and -0.24 (6.3% and 5.8% PAV) between sex and two measures of attitude, one administered at the beginning of the school year and one at the end. His results seem quite unusual. Lin and Crawley (1987) found sex to be significant but only at around the p = 0.47 confidence level. (Most authors
would consider p = 0.47 to be insignificant.) Mason (1989) reports that the interaction of sex with attitudes exists but is not significant. And, finally Doran and Sellers (1978) report a correlation of -0.04 between gender and science self concept.
In the words of Schibeci (1984, p. 33): "How may we explain this apparently inconclusive set of results?" Schebici explains that the apparently contradictory findings are actually the results of several factors which are: 1) a discrepancy between the relationship of sex to attitude in the physical vs. the biological sciences, 2) problems with inadequate instrumentation, and 3) there is a possibility that sex alone is not a significant influence on student attitudes.
In Figs. 11, 12, and 13, stem-and-leaf diagrams of attitude correlations with student, LE, and teacher variables respectively, are divided by sex. In two of the figures, the mean correlations are slightly higher for girls than they are for boys. The association of attitude with student variables is much stronger for girls than boys, however. Fig. 14 shows how attitude and achievement (for a few of the attitude achievement studies) interact by sex. Again, the mean correlation is higher for girls than for boys, suggesting that non-achieving girls could benefit from interventions involving affective variables more so than non-achieving boys.
Grade Level Differences
Schibeci (1984, p. 35) tells us that "Many studies have reported a decline in attitudes with increasing grade level." With this in mind I constructed a stem-and-leaf diagram of attitude-toward-science correlations
.7 .7 .7
.6 .6 .6
.5 569 .5 .5
.4 01355 .4 00168 .4 0002277
.3 0113568 .3 2333455678 .3 00002334566678
.2 011266667889 .2 000012234555 .2 222223333449
.1 113378999 .1 0001666779 .1 003556789
.0 .0 1 .0 14779
-.0 -.0 1224669 -.0 26777
-.1 4 -.1 3 -.1 00367
-.2 -.2 2 -.3 3 -.2 4
Undifferentiated Boys Girls
N = 39 N = 49 N = 58
S = .141 S = .186 S = .179
10 Studies 4 Studies 4 Studies
(1,4,7,12,19, 21,25,27,29,33) (1,9,17,29) (1,9,17,29)
Figure 14. Correlations between attitude and achievement variables.
broken down by grade level (Fig. 15) and of attitude-toward-science and mathematics correlations broken down by grade level (Fig. 16). Mean correlations in Fig. 15 show that attitude grade level associations increase from primary to middle school and then decrease somewhat in high school, In Fig. 16, with the two attitude toward mathematics studies added, mean correlations rise with time.
vo Tt ^ *-j p p rH rn Tt in
.6 2588 .6
5 0355566 .5 125
A 0011244556666889999 .4 000111234778
000124 .3 000111122235666888899 3 01122234466799
06 .2 00001224444455666666789 .2 0122344566999
0222333344569999 .1 12223455556678889 .1 11223459
023345779 .0 123334479 .0 2477789
111566788 -.0 123334479 -.0 246778
115 -.1 0122579 -.1 3466
6 -.2 12566 -.2 2247
8 -.3 22335 -.3 16
-.4 4 -.4 9
-.5 1 -.5 3
Primary N = 47 Mean = .082 S = .151 1 Study (8)
Middle N = 129 Mean = .204 S = .259 2 Studies (8,9)
Secondary N = 75 Mean = .168 S = .250 1 Study (8)
Figure 15. Correlations between attitude toward science and predictor variables by grade level.
.7- .7 8 .7
.6 .6 2588 .6
5 5 03355566 .5 00125
.4 02 A 0001112344455566666889999 .4 000000111233344677778899
.3 00000111223346678 .3 000011111111222334555566666677888889999 .3 00112222233344445566666677888899999
.2 000011222222223333345666779 .2 000001112224444444556666667888999 .2 00011112223333344566888999
.1 02223333445669999 .1 12223455556678889 .1 11223459
.0 023345779 .0 123334479 .0 2477789
-.0 111566788 -.0 123334479 -.0 246778
-.1 115 -.1 0122579 -.1 3466
-.2 26 -2 012566 -.2 222478
-.3 48 -.3 22335 -.3 16
-.4 -.4 245 -.4 69
-5 -5 1 -5 34
Primaiy Middle Secondary
N = 81 N = 173 N = 133
Mean = .146 Mean = .209 Mean = .277
S = .171 S = .244 S = .241
3 Studies (1,8,23) 3 Studies (8,9,23) 3 Studies (1,8,23)
Figure 16. Correlations between attitude toward math and science and predictor variables by grade level.
Recall that one of our objectives is the comparison of the results arrived at by my meta-analysis with the results obtained by Haladyna and Shaughnessy (1982). One of the advantages of presenting different data in a standardized format (eg. PAV) is that it can be combined. And, if there are any persistent trends (modes, asymmetry), we will notice them. And if our data are really measurements of the universe of content, any stable trends that show up in the data should represent properties of that universe (assuming validity). And, if there arent any trends forthcoming, we should see that too. Recall that my sample is newer than Haladyna and Shaughnessys sample which spanned 1960-1970. Mine spans the 1970s and 1980s (6%, late 60s, 49%, the 70s, and 45%, the 80s). Let us combine the results of my metaanalysis with the meta-analysis of Haladyna and Shaughnessy. This will give us the broader data base from which to draw our conclusions. And then we can compare the final statements of some of the other meta-analysts whove reported in this area during the 70s and 80s.
Fig. 17 is illustrative here. It contains the strength of association data from both Fig. 10, A and B (data points from both analyses) presented on the same graph. Let us compare Fig. 17 with Fig. 10 (p. 50). In all
25 + XXXXXXX X xxxxxx 25 +
20 XX XXXX XXX 20
15 XXXX X xxxxx 15
14 X X XX 14
13 XXXX X XXX 13
12 XXX 12
11 X XXX 11
10 X X X 10
9.5 XX 9.5
9 X 9
8.5 X XX 8.5
8 XXXX 8
7 XXX X 7
<6 XX XXX <6
Student Vars Teacher Vars LE Vars
Figure 17. Strengths of association between predictor variables and attitudes toward the subject matter of science across three dimensions of schooling showing the combined data of Haladyna and Shaughnessys (1982) and Craychees analyses.
three variable categories we can see that there are more studies (45) reporting high (13% or more PAV) associations than low associations (29). This trend is foreseeable in Part B of Fig. 10, though it cannot be seen in Part A. It becomes more obvious when we combine the two data sets. Im not sure what more can be said. It is a blurry picture.
Thirth four of the studies from the two meta-analyses report in the student variable category. This quantity of data reflects the popularity of these variables in the research. And, clearly these data reveal strong
associations of attitude with student variables.
Campbell (1972) (study # 5) is an interesting effort where scientific
curiosity a la the affective levels described by Krathwohi et al. (1964), are
compared from grades 7 through 12. They have the following tidbit for us:
The results show that students seem to become more aware of what to be curious about as they take more and more science courses, but unfortunately show lower and lower levels of involvement as they take these courses, (p. 144)
Is it possible that, somehow, science education is actually destroying the Knowledge Function, natural and innate (theoretically) to humans? Or do large groups of humans characteristically lose interest with age, possibly under the influence of puberty? Campbell goes on to explain how his results agree with the results of previous researchers. Handly and Morse (1984) (study #9) and Haladyna et al. (1982) (study #8) create mega variables (batteries) where a variety of similar characteristics are included in one battery.
However, Haladyna et al. report high associations while Handly and Morse report low ones. Handly and Morse (1984) is a longitudinal study that shows a weak association between the attitudes toward science and the self concept and gender role perceptions of 7th- and 8th-graders. As the children advance in grade level (or age), Handly and Morses data suggest that attitudes become more fixed, less associated with self-concept. Haladyna et al. (1982) show an increase in percent association (PAV) between student variables and attitude-toward-science from grades 4 to 7 to 9. It is difficult to ascertain which researchers are correct.
The teacher category (Fig. 17) is almost too scant of data to evaluate. But the data suggest strong associations between teacher characteristics and student attitude-toward-science. Haladyna et al. (1982) show teacher influence to be strongly related to attitude in their stepwise regression models (from 23.5% for 4th graders to 34% for 9th grade boys). Using the same methodology as Haladyna et al. (1982), Shaughnessy et al. (1983) obtained the same results with teacher variables and attitude toward mathematics. Talton and Simpson (1987) present product moment correlations of teacher as a classroom variable with attitude-toward-science criteria. Their data consists of 3 measurements of attitude-toward-science over a school year. The correlations are .45, .44, and .50 respectively. And Lawrenz (1975) uses canonical analysis to examine the association between two variable associations or groups that he calls student and teacher characteristics. His ultimate canonical correlation between the two mega-variables is .61 and represents a PAV of 37%. Haladyna and Shaughnessy (1982) note the strong association between attitude and teacher initiated activity, teacher dominance and aggression, and teacher use of intrinsic motivation in their meta-analysis. Schibeci, also found some strong associations between attitude-toward-science and teacher variables in the studies he examined. Haladyna et al. (1982, p. 685) report: "Clearly the most striking conclusion in [this] study deals with the consistently high relationships of the endogenous teacher quality variables to
It is somewhat surprising, considering the apparent strong association between attitude and teacher variables, that there are not more than the nine studies found reporting on the subject. One of the reasons for this paucity of teacher variable studies might be the fact that many studies include teacher variables as a subset of the LE. Nevertheless, one derives the definite impression that the effect of the teacher as the major variable is in need of more attention.
The LE scale is scattered with data. Learning environment variables are popular. Thirty-one studies report in this area. And nearly half (45%) of the data points represent attitude LE associations at or above 15% PAV. Obviously learning environment is associated with attitude. Haladyna and Shaughnessy (1982) report that:
Several learning environment variables have been found to be highly related to attitudes, including satisfaction, speed, apathy, favoritism, goal direction, and disorganization. In four studies, multiple correlation yielded effect coefficients ranging from 16.8 to 75% These studies reveal the potency for learning environment variables as predictors of attitudes toward science, (p. 557)
Three of the LE studies used the group unit of analysis, which as Haladyna (1981) shows, tend to report higher PAVs than studies that examine individuals. Haladyna et al. (1982, p. 685) suggest that "Learning environment may be a phenomenon that affects classes more than individuals..."
One study, (Lin & Crawley 1987), which is of Taiwanese students from different geographical areas of Taiwan, shows rather low correlation between Taiwanese students attitude-toward-science and geographic location. (Study numbers refer to APPENDIX A.) Keeves (1975) uses zero order correlations to derive only a 2% association between attitude and class.
Two experimental studies (11, 18), also show low results. One of them (18) is a study of a program involving biology teachers attending or not attending a workshop, the other is a study of 8 year old biology students in ''traditional" schools receiving textbook instruction (the control) and students from an "open" school receiving the Science Curriculum Improvement Study (SCIS). The third experiment (22) involves middle school biology students and the Human Sciences curriculum. Based on the large variance shown in the study maybe the Human Sciences curriculum is a promising avenue for further study (though, it is somewhat dated: 1980). Possibly, though, there is another reason, a question that needs to be addressed in a study of the validity of some of these results. Keeves (1975) (study #12) is not a treatment study, but reports a low association between attitude and classroom variables.
Haladyna et al. (1982) summarize the attitude LE connection as
Although the learning environment is highly related to attitude-toward-science, there are no clear patterns other than: (a) students must sense satisfaction with the work they do, (b) the class environment must be positive, and (c) instruction needs to be organized, (p. 685)
Implications for Research and Teaching
To quote Haladyna and Shaughnessy (1982):
The results of this meta-analysis have clearly revealed the accuracy of earlier reviews that research on attitudes is somewhat disorganized and chaotic. Clearly, research on attitudes is diffuse in focus as well as emphasis, (p. 556)
And Glass (1982):
Educational research will probably continue to be an unorganized, decentralized, non-standardized activity pursued simultaneously in dozens of places without thought to how it will all fit together in the end. (p. 109)
These two quotes comment on the state of educational research
(methodological concerns). In that regard Yager (1978) calls for the
development of "...a single research style or methodology [that] is appropriate
for all studies..." (p. 103)
A lack of comparability and therefore of incremental growth in our knowledge results from the continuous generation of new data gathering devices and the plethora of local variations of those already extant, e.g., the large number of interaction scales being used, [a Festingerian orientation?]. The development and establishment of reliability and validity of high quality tests and other devices is not a simple task quickly accomplished with limited resources. Invalidated tests constructed by the investigator add to the confusion, (p. 103)
Variables ought to be standardized like machine bolts or jet engines. Clearly,
from an instrument standardization viewpoint, attitude research has not
progressed much since the 60s and 70s. Most of the instruments used in the
studies I examined, were used only in that study. LE is a possible exception.
6 studies reported in my analysis use the LEI and report strong associations. So, while researchers are creating models of the attitude-toward-science process, and people (like Haladyna and his associates and Yager and his) are calling for a plan, everyone is still measuring something slightly different. According to Schibeci (1984):
Given established methods, why do researchers continue to report studies in which instruments are used which are either clearly invalid or for which few data on reliability and validity are reported...Given the number of attitude instruments which have been published with detailed reliability and validity data, there appears to be no need for further instruments to be developed, except in special circumstances... (p. 43)
The following list of conclusions from Schibeci (1984) addresses the results of research:
Sex appears to be an important variable, both alone and in interaction with other variables.
The effect of particular science programmes [sic] on attitudes varies considerably. There is not a consistent set of results for this variable.
Home background and peer group variables are probably important, but the influences are not direct.
"Science" must be subdivided into physical and biological science. Student attitudes to biological science appear generally to be more favorable than to physical science.
Attitudes to science appear to decline as school students move to higher grades, (p. 46)
And, finally, the following from Haladyna and Shaughnessy (1982), echoes what was just reported by Schibeci:
The essence of these research findings reveals that: (a) there are small differences in attitudes for boys and girls, (b) sex interacts slightly with many variables but in no systematic way, (c) programs generally have a
variable, positive effect on attitudes, and (d) some learning environment and teacher variables have been found to be highly related to attitudes. However, the evidence is not yet conclusive as to which of these teacher and learning environment variables are the most predictive, (p. 558)
According to Schibeci (1984, p. 46): "It is disappointing that the set of
conclusions that can be drawn from such a large body of literature is so
limited." Though it does seem encouraging that two meta-analyses come up
with similar conclusions. The same conclusions can be gleaned from the
meta-analysis of CHAPTER III, also. This outcome supports a contention
that, even though research lacks focus, and results are mixed, certain trends
Roughly 30% (10) of the studies analyzed for the above meta-analysis mentioned what they implied in terms of directions for further research. And only 14% (5) of the studies were concerned with their implications for teaching.
Experiments. Only two experiments mentioned anything about future directions: one (Vanek and Montean 1977) suggests that curriculum cannot be judged on its theoretical structure alone. A teacher is also involved. The following is from that study:
...[teachers] using an activity oriented approach appear to be as effective as teachers using a traditional textbook approach when effectiveness is measured by scores on Piagetian tasks of classification, science achievement tests, and attitude scales. Longer-range research studies (greater than six months in duration) may need to be performed in order to show that participation in the EES program leads to an increase in cognitive development or achievement, or a change in attitudes, (p. 61)
The other experiment concerned with future directions (Fulton, 1975) is a study comparing the effectiveness of traditional (group) instruction and individualized methods. Fultons conclusion is restricted to the area of teaching. He says that in group instruction much more of the burden for stimulating students attitude-toward-science is on the teacher than on the student, while in the individualized approach, the student is able to take some of the responsibility. Without exception the remaining 8 experimental studies analyzed in CHAPTER III concluded with vague exhortations that attitude is important and that somehow it should be emphasized.
Surveys. The remaining studies that tied conclusions to implications for the future were survey studies. These studies, as was noted in CHAPTER II, tend to work with larger samples and more complex statistics than experimental studies. Survey results are generally intended to be generalizable to larger populations and in more diverse situations than experimental results. The sample presented in CHAPTER III is not exceptional in that regard. It is not too surprising, then, that the surveys are more concerned with directions and implications; they are designed to be. Experiments, on the other hand, are designed to answer more specific questions.
Examples of some of the survey conclusions might prove instructive here. Anttonen (1969) suggests that future longitudinal work in attitude toward mathematics/achi evement in mathematics research should measure achievement over periods of time less than the six year interval they
measured. Antonnen would agree with Yagers assessment that measurement
over such long periods of time produces results that are too gross to be
meaningful (see p. 4). Gardner (1976) in a study that attempts to measure
the relationships among student attitudes toward physics and relate those
measures to various student and teacher characteristics concludes that:
Future research in this area could fruitfully investigate three questions: (1) What are the classroom behaviors of those physics teachers whose students regard them as intellectual, well-organized, achievement-pressing and stimulating? (2) Can teachers without these behaviors be taught to modify their behavior? (3) Can physics curricula be developed (if they do not already exist) that will arrest the decline in enjoyment displayed by most students... (p. 124)
Haladyna et al. (1982) explain that:
Ultimately, there is a need to monitor such factors as attitude, other student variables, teacher quality, and dimensions of the learning environment. Instruction program improvement efforts could be educational in terms of how these factors change over time. (p. 686)
Lawrenz (1976) concludes that:
Further research needs to determine possible cause-effect relationships between environment and attitude...In addition to determining the effect of particular scales, research could be instigated to determine what types of classroom procedures would produce a particular environment, (p. 514)
Lawrenz and Haladyna et al. are, in effect, invoking experimental research to
answer questions generated in a survey. Their suggestions propose a logical
relationship between the survey and the experiment. Marjoribanks (1976)
What is required now is a set of studies which investigate relationships between attitudes and achievement at different levels of variables such as family, classroom and neighborhood environments, personality measures and other affective characteristics, (p. 659)
Napier and Riley (1985) invoke experimental research to determine the causal
link that characterizes the relationship between attitude and achievement. Stevens and Atwood (1978) also call for experimental research.
Implications for Teaching
In terms of implications for teaching, Napier and Riley (1985) describe the ideal class as one that fosters achievement, where the teacher encourages students to study science out of school as well as in, where the work is not too difficult and the atmosphere is comfortable, where the teacher is the ultimate guide with final authority, but who allows students to speak their minds and think for themselves. Doran and Sellers (1978) suggest that teachers create an environment that is warm and accepting, one where communication is the norm, where students are given the freedom and responsibility to choose and think for themselves. Talton and Simpson (1987) call for stimulating and supportive classroom environments.
Does the research presented in the meta-analysis of CHAPTER HI indicate some sort of coherent direction in attitude-toward-science research? We have looked at a lot of studies in this paper, and I think we can say that there does appear to be agreement in the research as to which variables are most important, but there isnt good agreement as to how important. The student, teacher, and LE variable categories are broad and vague. The scatter of results indicates that. Possibly with better definition the results would better show directions to follow.
What do the meta-analysts feel should be done to improve the field of
attitude research? Not surprisingly, they all say the same thing: attitude
research needs help. As Schibeci put it (1984, p. 47): "The picture which has
been painted appears to be one of unrelieved gloom." And, generally they
call for some sort of organized framework. Koballa (1986) asks: "Has the
absence of theoretical frameworks and the lack of specificity in the
identification of attitude objects hampered research?" According to Simpson
and Troost (1982, p. 780): "Finally, once we develop a model depicting
commitment to science and achievement in science among learners of this
age, it is our intention to use the theory to improve practice." And Haladyna
and Shaughnessy (1982) conclude their meta-analysis with this:
There has been limited progress in understanding the determinants of science attitudes through previous research. Programmatic research is needed which operates in the context of a conceptual framework and provides findings that are translatable to practitioners in terms of improving instruction, (p. 559)
Limitations of This Study
The first and most obvious limitation of this study is my inexperience in such matters. In first attempts one tends to take faltering steps, some of which go in useless directions. Much time is spent learning the language (statistics) and the methods; and the learning curve is steep. I have made many changes to the original draft of this paper because of what I have learned studying for it.
The second, and maybe the most important, limitation involves the
criteria I used to divide variables. They were a subjective criteria. I guessed. Division into student, teacher, and LE presented few problems (although the distinction between teacher and LE is somewhat blurred). Subdivisions may be more or less arbitrary. The subdivisions presented in Figs. 4 and 5 should be regarded as somewhat tentative.
A third limitation involves the validity of the results. We have seen that most researchers measure their variables with different instruments. And researchers are corning to the same conclusions: data are scattered. In the LE variable category of Fig. 17, for example there are high PAVs and low PAVs.
Even with these acknowledged limitations, the substantial consistency between the two meta-analyses compared renders this studys outcomes substantial and worthy of note.
Allport, G. W. (1935). Attitudes. In M. Fishbein (Ed.), Readings in attitude theory and measurement (pp. 3-13). New York: John Wiley and Sons.
Anttonen, R. G. (1969). A longitudinal study in mathematics attitude. The Journal of Educational Research. 62(10), 467-471.
Ayres, J. B., & Price C. (1975). Childrens attitudes toward science. School Science and Mathematics. 75. 311-318.
Baker, D. R. (1985). Predictive values of attitude, cognitive ability, and personality to science achievement in the middle school. Journal of Research in Science Teaching. 22(2), 103-113.
Campbell, J. R. (1972). Is scientific curiosity a viable outcome in todays secondary school science program? School Science and Mathematics. 3, 139-147.
Dieterich, P. B. (1967, February). Components of the scientific attitude.
The Science Teacher, pp. 23-24.
Doran, R. L., & Sellers, B. (1978). Relationship between students self concept in science and their science achievement, mental ability, and gender. Journal of Research in Science Teaching. 15(6), 527-533.
Fishbein, M. & Azjen, I. (1975). Belief. Attitude. Intention, and Behavior.
Reading, MA: Addison-Wesley.
Freedman, J. L., Carlsmith, J. M., & Sears, D. O. (1970). Social Psychology. Englewood Cliffs, NJ: Prentice Hall.
Fulton, H. F. (1975). A comparative study of student attitudes toward science and the ability of the teacher to make material understandable in individualized and group approaches to BSCS biology. School Science and Mathematics. 5, 198-202.
Gardner, P. L. (1976), Attitudes toward physics: personal and environmental influences. Journal of Research in Science Teaching. 13(2), 111-125.
Gauld, C. (1982). The scientific attitude and science education: a critical
reappraisal. Science Education. 66(1), 109-121.
Germann, P. J. (1988). Development of the attitude-toward-science in school assessment and its use to investigate the relationship between science achievement and attitude-toward-science in school. Journal of Research in Science Teaching. 25(8), 689-703.
Glass, G. V. (1982). Meta-analysis: an approach to the synthesis of research results. Journal of Research in Science Teaching. 19(2), 93-112.
Haladyna, Tom. (1981). A common metric for integrating research findings.(ERIC Document Reproduction Service No. ED 202 873)
Haladyna, T., & Shaughnessy, J. (1982). Attitudes toward science: a quantitative synthesis. Science Education. Â£6(4), 547-643.
Haladyna, T., Olsen, R., & Shaughnessy, J. (1982). Relations of student, teacher, and learning environment variables to attitudes toward science. Science Education. 66(5), 671-687.
Handly, H. M., & Morse, L. W. (1984). Two year study relating adolescents self concept and gender role perceptions to achievement and attitudes toward science. Journal of Research in Science Teaching. 21(6), 599-607.
Holman, H. H. (1977). An assessment of eight year old childrens attitudes toward science. School Science and Mathematics. 77, 622-670.
Keeves, J. P. (1975). The home, the school, and achievement in mathematics and science. Science Education. 59(4), 439-460.
Kiesler, C. A., Collins, B. E., & Miller, N. (1969). Attitude change: a critical analysis of theoretical approaches. New York: John Wiley & Sons.
Klopfer, L. E. (1976). A structure for the affective domain in relation to science education. Science Education. 60(3), 299-312.
Koballa, T. R., Jr. (1988). Teaching hands-on science activities: variables that moderate attitude-behavior consistency. Journal of Research in Science Teaching. 23(6), 493-502.
Krathwohl, D. R., Bloom, B. S., & Masia, B. B. (1964). Taxonomy of educational objectives. Handbook II: the affective domain. New York: David McKay Co. Inc.
Lawrenz, F. (1975). The relation between science teacher characteristics and student achievement and attitude. Journal of Research in Science Teaching. 12(4), 433-437.
Lawrenz, F. (1976). The prediction of student attitude-toward-science from student perception of the classroom learning environment.
Journal of Research in Science Teaching. 13(6), 509-515.
Lemon, N. (1973). Attitudes and their measurement. New York: John Wiley and Sons.
Lin, B., & Crawley, F. E. (1987). Classroom climate and science-related attitudes of junior high school students in Taiwan. Journal of Research in Science Teaching. 24(6), 579-591.
Liska, A. E. (1975). The consistency controversy: Readings on the impact of attitude on behavior. New York: John Wiley and Sons.
Mason, C. L., & Kahle, J. B. (1989). Student attitudes toward science and science related careers: a program designed to promote a gender free learning environment. Journal of Research in Science Teaching. 26(1), 25-39.
Maijoribanks, K. (1976). School attitudes, cognitive ability, and academic achievement. Journal of Educational Psychology. Â£Â£(6), 653-660.
McGuire, W. J. (1969). The nature of attitudes and attitude change. In Lindzey, G. & Aronson, E. (Ed.). The Handbook of Social Psychology (p. 138-202). Reading, MA: Addison-Wesley.
NAEP. (1979, October). Attitudes toward science (Report No. 08-S-02). Denver: National Assessment of Educational Progress.
Napier, J. D. & Riley, J. P. (1985). Relationship between affective determinants and achievement in science for seventeen-year-olds. Journal of Research in Science Teaching. 22(4), 365-383.
Popper, K. R. (1963). Conjectures and refutations. New York: Basic Books Inc.
Reid, D. J., & Hodson, D. (1987). Science for all. London: Biddles Ltd.
Schibeci, R. A. (1984). Attitudes to science: an update. Studies in Science Education. 11. 26-59.
Shaughnessy, J., Haladyna, T., & Shaughnessy, M. J. (1983). Relations of student, teacher, and learning environment variables to attitude toward mathematics. School Science and Mathematics. 83(1), 21-37.
Simpson, R. D., & Troost, M. K. (1982). Influences on commitment to and learning of Science among adolescent students. Science Education.
Smith, M. L., & Glass, G. V. (1987). Research and evaluation in education and the social sciences. Englewood Cliffs, NJ: Prentice Hall, Inc.
Steinkamp, M. W., & Maehr, M. (1983). Affect, ability, and science achievement: a quantitative synthesis of correlational research. Review of Educational Research. 53(3), 369-396.
Stevens, J. T., & Atwood, R. K. (1978). Interest scores as predictors of science process performance for junior high schools. Science Education. 62(3), 303-308.
Talton, E. L., & Simpson, R. D. (1987). Relationship of attitude toward classroom environment with attitude toward and achievement in science among tenth grade biology students. Journal of Research in Science Teaching. 24(6), 507-525.
Tamir, P. (1985). Meta-analysis of cognitive preferences and learning.
Journal of Research in Science Teaching. 22(1), 1-17.
Tukey, J. W. (1977). Exploratory Data Analysis. Reading, MA: Addison-Wesley.
Uguroglu, M. E., & Walberg, H. J. (1979). Motivation and achievement: a quantitative synthesis. American Educational Research Journal. 16(4), 375-389.
Vanek, E. P., & Montean, J. J. (1977). The effect of two science programs (SES and Laidlaw) on student classification skills, science achievement, and attitudes. Journal of Research in Science Teaching. 14(1), 57-62.
Walberg, H. J. (1969). Predicting class learning: An approach to the class as a social system. American Educational Research Journal 6(4), 529-542.
Walberg, H. J., & Ahlgren, A. (1970). Predictors of the social environment of learning. American Educational Research Journal. 7(2), 153-167.
Willson, V. L. (1983). A meta-analysis of the relationship between science achievement and science attitude: kindergarten through college. Journal of Research in Science Teaching. 20(9), 839-850.
Yager, R. E. (1978). Priorities for research in science education: a study committee report. Journal of Research in Science Teaching. 15(2), 99-107.
SUMMARY OF PERTINENT INFORMATION OF STUDIES USED IN THIS META-ANALYSIS
Author (date)/ dependent variables/ independent variables Type of Study/unit of analysis/ Description of how PAV was determined sample/ content
Anttonen (1969)/ math achievement/ attitude toward math longitudinal survey/student/ Zero-order correlations between attitude scores and math GPA were summed over elementary and secondary levels: (.311 + .432) / 2 = .372. Rsquared(lOO) = 14% PAV. 607 5th & 6th/ math
Ayers & Price (1975)/ attitude toward science/ student variables survey/student/ results nonsynthesizable 455 4th-8th/ gen sci
Baker (1985)/ attitude toward science/ science grades survey/student/ ANOVA of science grades (A, B, C, or D) x attitude. Sum of Squares (SS) between grade groups (16.20) / SS total (124.62) = 13% PAV. 98 8th/ gen sci
Butzow & Linz (1977)/ attitude toward science/ science achievement quasi-experiment/college student/ Zero-order correlations of total attitude with final grade: r = .26; PAV = 6.8%. 103 college/ chem
Campbell (1972)/ scientific curiosity/ effect of Krathwohl levels experiment/student/ between groups SS (10907) / total SS (34792) = 31% data presented in ANOVA format 1229 7-12th/ gen sci
APPENDIX A (continued).
Fulton (1975)/ attitude toward science/ curriculum experiment/student / results not significant 40 8th/ biology
Gardner (1976)/ attit. toward science (PAI)/ science achievement (PPI) survey/student/ zero-order correlations between PAI/PPI scales = 0.20, PAV = 4%. 1014 11th/ physics
Haladyna & Shaughnessy(1982) attitude toward science/ student, teacher, learning environment (LE) survey/student/ Regression analysis: attitude toward science predicted by student vars, 25% PAV; teacher vars, 22% PAV, and LE vars, 22% PAV. Outcomes differentiated by grade and sex. 1965 4,7,9 / gen sci
Handley & Morse (1984)/ attitude toward science/ self concept and gender role longitudinal survey/student/ Sums of correlation values of self-concept with gender variables by grade and sex: male/female = 5.4%; self/ teacher = 22%; and overall self concept = 8%. 155 7th, 8th/ gen sci
Hess & Shringley (1981)/ metric knowledge and attitude/treatment experiment/pre-teachers/ not related to science students 141 teachers/ metrics
Hofman (1977)/ attitude toward science/ treatment experiment/student/ PAV of 3.3% averaged from SS of the 3 statistically significant PTOA categories. 79 8-yr olds/ gen sci
Keeves (1975)/ attitude toward science/ student and LE variables longitudinal survey/student/ zero-order correlations between attitude/achievement (15% PAV) and attitude/class structure (2% PAV) 242 8th/ gen sci
APPENDIX A (continued).
Koballa (1986)/ behavioral intentions of of elera teachers survey/individual teachers/ results not significant 76 teachers/ gen sci
Lawrenz (1975)/ student attitude/ teacher characteristics survey/class/ canonical correlation (a stepwise regression procedure) of student and teacher characteristics cited in text of study at 0.61. PAV = 37%. (Note: LEI used but results not reported.) 236 teachers/ secondary
Lawrenz (1976)/ attitude toward science/ LE survey/class/ cited in text: LE accounts for about 30% of the variance in the mean SAI scores. 75(?) sec/chem, biol, physics
Lin & Crawley (1987)/ attitude toward scientific inquiry/LE survey/class/ Zero-order correlations of LEI scales with TOSRA attitude toward scientific inquiry scale. 80 mid sch classes/ g sci
Marjoribanks (1976)/ physical and biological science achievement/ school related attitudes survey/student/ Sum of zero-order correlations between academic achievement and school related attitudes for boys and girls respectively: (.40 + .40 + .33 + .40) / 4 = .365 which is a PAV of 13%. 450 12yr old/ phys, biol
Mason & Kahle (1989)/ attitude towards science/ treatment experiment/student/ [SS(between)/SS(total)](10Q) = 1.2% PAV calculated from SAQ test means and F-ratio, after a method by Haladyna 550 sec./ biology
(1981). teachers oo
APPENDIX A (continued).
19 Napier & Riley (1985)/ science achievement/ student motivation
20 Novick & Duvdvani (1976)/ attitude toward science/ student, LE variables
21 Oliver & Simpson (1988)/ sci & math achievement/ attit. toward science and mathematics
22 Robinson (1980)/
attitude toward science/ treatment
23 Shaughnessy et al. (1983)/ attitude toward mathematics/ student, teacher, LE
Results of regression analysis reported in text of study: subject motivation to do science/science teaching teaching accounted for 7% of the variance in achievement.
survey/student/ results nonsynthesizable
PAV reported by the author in the text of the study.
4 attitude variables accounted for 20% of the variance in achievement for llth-grade and 30% for 12th-grade.
The average PAV is 25% for the two grade levels.
Between columns SS / total SS was calculated from ANOVA presentation. PAV of 14% was obtained by averaging the PAVs of the 3 significant SQ scales.
Zero-order correlations between attitude toward math and student, teacher, and LE variables revealed PAVs of 13%, 18%, and 9.5% respectively.
3135 17yr olds/ gen sci
850 sec/ gen sci
402 middle/ gen sci
2081 4,7,9/ gen sci
APPENDIX A (continued).
24 Silberstein & Tamir (1981)/ attitude toward using animals in experiments
25 Doran & Sellers (1978)/ science self concept/ achievement and mental ability
26 Simmons & Esler (1975)/ attitudes toward science/ treatment
27 Stevens & Atwood (1978)/ science process performance/ attitude toward science
28 Talton & Simpson (1987)/ attitude toward science and science achievement/ attitude toward science, teacher, LE
Between sex PAV=4% for llth-graders tested, calculated from t-value between boy and girl groups and test mean scores, after Haladyna (1981).
Multiple regression analysis results cited in text of study: variances between science self-concept and biology achievement = 7.8%; process achievement = 6.8%, and mental ability = 5.3%. The average of these three student variable values is 6.6%.
experiment/student / results nonsynthesizable
PAV of 8% calculated for 7th-, 8th-, and 9th-grade students using SII test mean scores and t-values.
Results showed negative effect of treatment on attitude.
Teacher characteristics correlated .46 with attitudes 21% PAV) (from product-moment correlation table), cited in the text of the study as accounting for and LE was 37% of the variance in attitude.
577 5,7,9,11/ biology
320 hi sch/ biology
132 6th/ gen sci
1070 7,8,9th/ gen sci
1560 10th/ gen sci
APPENDIX A (continued).
29 Towse (1983)/ experiment/student/
science interest/ science performance Correlations of science interest vs. science performance averaged over sex and treatment groups r = .27, which is a PAV of 7.3%. 647 sec/ gen sci
Vanek & Montean (1977)/ attitude toward science/ treatment experiment/student/ results nonsynthesizable; and insignificant effect noted by author (p. 61). 110 3,4th/ gen sci
Vargas-Gomez & Yager (1987)/ attitude toward science/ curriculum survey/student/ nonsynthesizable 150 3,7,11th/ gen sci
Walberg & Ahlgren (1970)/ social environment of learning/LE and student variables survey/class/ Canonical correlations of student and LE predictors with LEI scales average .65 as presented in table. PAV = 43%. 144 hi sch/ physics
Ward (1976)/ attitude toward science/ student achievement survey/class/ Partial correlations between attitude and achievement controlled for class size and teacher attitude of .59. (PAV = 34%) Result of partial correlation implies that the relationship between attitude and achievement is independent of both class size and teacher attitude. -240 sec class/ bio,chem,phys
Wynn & Bledsoe (1967)/ interest in science/ survey/student/ results not significant 325 hi sch/
student variables gen sci
APPENDIX A (continued).
35 Yager & Yager (1985)/ survey/student/
attitude toward science results nonsynthesizable
2500 9,13,17yr/ gen sci
APPENDIX B STUDY DESCRIPTIONS
Anttonen, R. G. (1969). A longitudinal study in mathematics attitude. The Journal of Educational Research. 62(10), 467-471.
Type of Study: Longitudinal: cohort survey design
Dependent Variables: achievement over time subdivided by grade and sex Independent Variables: attitude toward mathematics
Instruments: in-house attitude composed of 94 yes or no questions (rel = .95); the Iowa Test; GPA
Sample Description: 607 students from the fifth and sixth grades of the Roseville elementary schools of St. Paul, Minn. Unit of analysis = student Methods: The students were tested once in the spring of 1960 and again 6 years later.
Data Analysis: Pearson product-moment correlation coefficients were generated among the elementary and secondary attitude scores (first and second measurements of independent variables); quantitative thinking and mathematics GPA which were measured at the end of the survey. Reported results are significant. The study reports an intercorrelation between elementary attitude score and math GPA of .311 (9.6% PAV), and between secondary attitude score and math GPA of .432 (18.7% PAV). (p. 468) The average PAV, then, is 9.6 + 18.7/2 =14.2%. These values are for the follow up study only. The correlation matrix was presented for all the students once, and was further reproduced four times broken down by grade and sex. Conclusions: Sex by grade breakdowns showed low correlations. Correlations between secondary attitude and achievement and were higher than those for elementary students."Although the conclusions of this study are limited, the study does point to future work in the area of mathematics attitude" (p. 471). (19)SubjecÂ£ Grade: Math, elementary
Ayres, J. B., & Price, C. (1975). Childrens attitudes toward science. School Science and Mathematics. 75. 311-318.
Type of Study: multiple-group, single-observation survey design Dependent Variables: attitude toward science Independent Variables: grade level, sex difference
Instruments: Science Attitude Inventory (SAI) developed by Ayres for this study; no reliability data provided.
Sample Description: The 455 4th through 8th grade students, 232 males, 233