You are on page 1of 6

Measuring Student Attitudes: Semantic

Differential or Likert Instruments?

R. A. SCHIBECI
School of Education, Murdoch University, Murdoch, Western Australia

Student’s systematic study of science usually begins in high school. There is much
evidence that science in the primary school is given a much lower priority than other
curriculum areas, and is often poorly dealt with; this is true in Australia (Appleton, 1977),
the U.K. (Bainbridge, 1980) and the U.S. (Klopfer, 1980). It is in the high school that
science teachers are in a position to nurture and sustain an interest in science. Objectives
in the affective domain are commonly found in modern science curricula. This interest
in affective domain objectives has logically stimulated interest in the measurement of
the achievement of these objectives-particularly those which could be labelled “attitude”
objectives.
A variety of methods is available for the assessment of attitudes. Gardner (1975), in
a review of the research on attitudes to science, noted that the following methods had
been used to assess attitudes: differential (Thurstone) scales, rating scales, summated
rating scales, semantic differential scales, interest inventories, preference rankings,
projective techniques, enrollment data, and anthropological observation.

Liked and Semantic Differential Instruments

The most popular of these methods has been the summated rating method generally
known as the Likert scale. For example, most of the attitude scales reported in Shaw and
Wright (1967) are Likert scales. In a review of 1976 research in science education,
Renner, Abraham, and Stafford (1 978) cited the Likert instrument developed by Moore
and Sutman (1970) as a popular instrument in attitude research studies. More recently,
Fraser (1978) has reported the development and validation of another Likert instrument
to assess high school student attitudes to science.
The instrument developed by Fraser (1978), the Test of Science-Related Attitudes
(TOSRA) may be used to illustrate the development of a Likert scale. TOSRA has seven
subscales: Social implications of science, Enjoyment of science lessons, Career interest
in science, Leisure interest in science, Attitude to scientific inquiry, Adoption of scientific

Science Education 66(4): 565-570 (1982)


0 1982 John Wiley & Sons, Inc. CCC 0036-8326/82/040565- 08$01.80
566 SCHlBECl

attitudes, and Normality of Scientists. In developing items for this instrument, a pool
of items was initially examined by a group of science educators. This pool was reduced
to a pool of 14 items per subscale, which was further reduced to ten items per subscale
after field-testing with high school students.
In the final version, subscales have ten statements (items). For example, one of the
items in the Enjoyment of Science lessons subscale is “Science lessons bore me.” The
respondent is asked to indicate whether he or she strongly agrees (SA), agrees (A), is
undecided or neutral (N), disagrees (D), or strongly disagrees (SD) with the statement;
the response is then scored 1 , 2 , 3 , 4 ,or 5, respectively. The scores for this and the other
nine items are added (hence the term “summated rating” is applied to this method). Total
possible scores for this subscale range from 10 (highly negative) to 50 (highly positive).
Similarly, scores are computed for the other six scales. The seven scores for the TOSRA
subscales should not be added together, since they are presumed to assess seven different
science-related attitudes. In TOSRA, the ten items of the Enjoyment of Science Lessons
subscale are distributed throughout the instrument. Five of the ten items are stated in
a positive way, the remainder in a negative way; this strategy is adopted to minimize the
possibility of a response set.
The semantic differential (SD) technique has also been used quite widely. Examples
of studies in science education which have used this method are easily located: Welch
( 1973) noted that the technique had been used to assess student attitudes to physics as
part of the national evaluation of Harvard Project Physics; Milson (1979) used the
technique to assess the attitudes of below average students with reading difficulties;
Robinson (1980) used the SD method to assess the attitudes of students to a “human
sciences” course. The SD instrument examined in this report is based on previous work
by the author (Schibeci, 1977). Student reactions to eight stimuli (concepts) were sought:
Science in society, Science lessons, Science career, Science hobbies, Scientific attitudes,
Scientists, Science teacher, and School.
Respondents are asked to provide reactions to the concepts by means of a set of bipolar
adjective pairs. These were: exciting-boring, worthless-worthwhile, easy-hard, dull-
interesting, important-unimportant, complicated-simple, useless-useful, enjoyable-
unenjoyable, gloomy-joyful, clear-unclear. These scales were chosen from a pool reported
by McCallon and Brown (1969) and Yamamoto, Thomas, and Karns (1969), and
Schibeci (1977). Each adjective pair is separated by a five-point scale, usually five boxes.
Students check one of these five boxes which are then assigned scores of 1, 2, 3, 4,and
5 , respectively. If all the adjective pairs relate to the same dimension (factor analysis is
one method of validating this), a total score is computed for each concept. If there is more
than one dimension used (such as Potency and Activity in addition to the more common
Evaluation dimension), separate scores may be computed for each dimension.
The development of a new Likert scale requires lengthy procedures. An initial large
pool of items is gradually refined until a set of reliable, valid set of items remain. Of course,
if one uses a published Likert scale for which sound reliability and validity data are
available, the task is simpler. In that case, the instrument needs to be checked for suit-
ability if the sample to be tested differs in any significant way from the sample on which
the validity data were gathered.
The development of an SD instrument is a simpler task because of the large amount
of research which has accumulated on semantic differential adjective pairs (scales). For
example, the scale “good-bad’’ has been firmly established as an evaluative scale. There
is a large body of evidence regarding other scales: see, for example, Osgood, May, and
MEASURING STUDENT ATTITUDES 567

Miron ( I 975). In using the SD technique, scales may be selected from those used in
published studies. Maguire (1973) noted that scales:
must be representative o f the attitudes in the domain; they must be well defined for the population
of interest; they must be appropriate to the concepts of the domain; and they must be polar opposites
(p. 297).
The concepts to be assessed should cover the area of interest, and should be such as to
provoke “a variety of associations both within and among subjects” (Maguire, 1973, p.
298).
Given these descriptions of the two attitude measuring techniques, one may ask whether
the methods are equally suitable. Some studies comparing these two attitude assessment
techniques have appeared. McCallon and Brown (1971) assessed the attitudes of 68
college students to mathematics using both Likert and SD procedures. They reported
a high positive correlation ( r = 0.90) between the total score on the semantic differential
instrument and the score on a Likert instrument called the “Mathematics Attitude Scale.”
More recently, Schofield and Start ( I 978) in a study of mathematics attitude and
achievement among 3 17 college students (student teachers), reported a moderately high
correlation ( r = 0.70) between the Likert scale measuring interest and liking for math-
ematics and the corresponding semantic differential concept (“Mathematics”).
Very little information is available, however, comparing these two methods with
samples of students who are not college students. The purpose of this paper is to provide
a comparison of the two approaches to attitude measurement using a sample of high school
students in grades 8-1 0.

Method

Instruments
The two instruments described above, TOSRA and the SD instrument, were field
tested. The first five concepts were chosen to enable a direct comparison with the first
five TOSRA scales. The attitudes of students to the remaining three concepts were also
of general interest. TOSRA is based on Klopfer’s (197 1) conceptually distinct categories
for the effective domain in science education; the conceptual structure of TOSRA has
been empirically (Schibeci & McGaw, 1981).

Sample
TOSRA and SD were taken by a sample of students in 1 1 different schools. The schools
were chosen to cover a variety of socio-economic and geographic areas in the metropolitan
area of Perth, Western Australia. The schools included seven government coeducational
high schools, two independent Catholic schools (one boys’ and one girls’) and two inde-
pendent non-Catholic schools (one boys’ and one girls’). In each school, two classes from
each of grades 8,9, and 10 took part in the study. Approximately half of the students in
each class were chosen at random to take one instrument; the remainder took the second
instrument. A subset of students in each class was chosen at random to take both in-
struments.
The number of students who took the instruments were: TOSRA, 1049 students; SD,
1 116 students. A total of 362 students from this pool took both TOSRA and SD. The
number of students who took both tests at each grade level (with the percent male in
parenthesis) was: I18 (61.9), 99 (61.6), 145 (51.0). Large samples were used for each
568 SCHlBECl

of the two instruments (TOSRA and SD) to enable validity data to be gathered. These
data are currently being analyzed; some of this analysis has been published (Schibeci
and McGaw, 1981).
Administration
The author administered the instruments to 60 of the 66 classes involved in the study.
A teacher who was not the regular science teacher administered the tests to the remaining
six classes. Students were told that their responses were confidential, and that teachers
would be given average responses for a whole class, but would not see any individual
student’s replies. The two instruments were presented as surveys of high school student
attitudes to science. It was emphasized that they were not to be used for grading pur-
poses.

Results and Discussion


TOSRA yielded seven scores, one for each subscale; the SD instrument yielded eight
scores, one for each concept. Pearson product-movement correlation coefficients were
computed for the various components of the two instruments; these are given in Table
1. Five concepts in the two instruments had been chosen for direct comparison. The
correlation coefficient for each of these pairs was: Science in society/Social implications
of science, 0.30; Science lessons/Enjoyment of Science lessons, 0.52; Science career/
Career interest in science, 0.37; Science hobbies/Leisure interest in science, 0.48; Sci-
entific attitudes/Attitude towards scientific inquiry, 0.06. Each of these correlation
coefficients is significantly different from zero 0, < 0.01) except for the last. However,
none of these values is as high as values reported in previous studies: for example, those
reported in McCallon and Brown (1971), and Schofield and Start (1978).
Discussions with students who responded to each instrument indicated that the SD
format allows an assessment of general attitudes to concepts such as “school” and
“science”. The Likert format used in TOSRA, on the other hand, allows a more fine-
grained assessment of specific attitudes. For example, the “Social implications of science”
scale in the TOSRA instrument requires students to respond to a set of ten statements,
the meanings of which are quite clear. The SD concept “Science in society,” on the other
hand, allows a wide diversity of interpretations among students. The highest correlation
coefficient for the corresponding scales, Science lessons/Enjoyment of science lessons,
may be explained (in part at least) by the unambiguous nature of the concept for the
students in the sample.
An advantage of SD instruments from the researcher’s point of view is that they can
be fully described within the confines of a journal article: that is, all the concepts and
adjective pairs can be listed. This enables others to use the same scales. With Likert scales,
limited journal space normally precludes the complete listing of the whole instrument:
authors must content themselves with providing illustrative examples, and inviting in-
terested readers to contact them for a copy of the instrument.
The differences between the two techniques are worth bearing in mind. Firstly, the
SD format (because of its general format) is less overt; Likert statements more clearly
indicate their intent. Secondly, the SD format allows a set of concepts to be compared
on a set of standard scales (bipolar adjective pairs). Likert scales, on the other hand,
cannot be so compared: the wording of statements may bias student responses (Simpson,
Rentz & Shrum, 1976).
MEASURING STUDENT ATTITUDES 569

TABLE I
lntercorrelations for t h e Seven TOSRA Subscales and the Eight SD Concepts
( n = 362)

TOSRA s u b s c a l e

1 2 3 4 5 6 7
1. Science i n s o c i e t y 0.30 0.35 0.36 0.42 0.16 0.25 0.22

2. Science l e s s o n s 0.34 0.52 0.40 0.48 0.18 0.26 0.23

3. Science c a r e e r 0.23 0.33 0.37 0.41 O.Ol* 0.13 0.11*

4. Science hobbies 0.31 0.43 0.38 0.48 0.22 0.24 0.22

5. Scientific
attitudes 0.19 0.33 0.30 0.36 0.06* 0.13 O.ll*

6. Scientists 0.39 0.44 0.46 0.52 0.23 0.30 0.28

7. Science t e a c h e r 0.20 0.40 0.27 0.31 0.08” 0.15 0.15

8. School 0.18 0.26 0.22 0.27 0.12* 0.20 0.17

<
Notes: 1. All the correlation coefflcients are significantly different from zero, p 0.01, except those
marked with an asterisk (’); for which p >0.01.2. The TOSRA subscales are: 1. Social implications
of science, 2. Enjoyment of science lessons, 3. Carerr interest In science, 4. Leisure interest in science,
5. Attitude to scientific inquiry, 6. Adoption of scientific attitudes, 7. Normality of sclentists.

It seems that semantic differential and Likert data may not be used interchangeably
as suggested by some authors. Heise (1977) in a review of the use of semantic differential
in attitude research concluded:
Most studies provide confirmation that the SD can be used to measure attitudes. Too little meth-
odological research is available to decide whether SD ratings always provide as sensitive a measure
of attitudes as is given by traditional scales (p. 247).
If one consults the general literature, differences between the techniques are not
thoroughly discussed. For example, Henerson, Morris, and Fitz-Gibbon (1978) in a book
called How to Measure Attitudes included a chapter describing three procedures: the
“ordered” scale, as well as the Likert and SD procedures. There is no indication of any
differences among these: the reader is given the impression that the methods are equally
suitable.
The data presented here indicate that SD may not be as sensitive as Likert data, and
that it is unlikely that the two methods can be used interchangeably with high school
students. Much educational research tends to be conducted with college students. For
example, West and Robinson (1 980) found that of the 88 empirical articles they selected
at random from three widely read educational research journals 74 involved college
students. It is necessary to ensure that conclusions drawn from college samples be applied
with caution to high school samples. This is particularly true of validation data for attitude
instruments where the college sample is a selected group; students in grades 8-10 represent
a much wider cross section of the general population.
High school student’s general attitudes to science can be readily measured with the
semantic differential. On the other hand, if more specific attitudes are to be assessed (such
570 SCHlBECl

as ‘Adoption of scientific attitudes’), Likert instruments appear t o b e more appro-


priate.

References

Appleton, K. Is there a fairy godmother in the house? Aus. Sci. Teach. J., 1977, 23, 37-42.
Bainbridge, J. ‘Tchirrip . . . Thchichirrip . . . tseep’: an alarm call for primary school science. School
Sci. Rev., 1980,61,623-638.
Fraser, B. J. Development of a test of science-related attitudes. Sci. Educ., 1978,62, 509-515.
Gardner, P. L. Attitudes to science: a review. Stud. Sci. Educ., 1975, 2, 1-41.
Heise, D. R. The semantic differential and attitude research. In G.F. Summers, (Ed.). Attitude
measurement. London: Kershaw, 1977,235-253.
Henerson, M. E., Morris, L. L., & Fitz-Gibbons, C. T. How to measure attitudes. Beverley Hills:
Sage Publications, 1978.
Klopfer, L. E. Evaluation of learning in science. In B. S. Bloom, J. T. Hastings, & G. F. Madaus.
(Eds.). Handbook on formative and summative evaluation of student learning. New York:
McGraw-Hill, 197 1,559-641.
Klopfer, L. E. Science education in the 1980’s. Sci. Educ., 1980,64, 1-6.
Maguire, T. 0. Semantic differential methodology for the structuring of attitudes. Am. Educ.
Res. J . , 1973, 10, 295-306.
McCallon, E. L., & Brown, J. D. A semantic differential instrument for measuring attitudes toward
mathematics. J . Exper. Educ., 1971, 39, 69-72.
Milson, J. L. Evaluation of the effects of laboratory-oriented science curriculum materials on at-
titudes of students with reading difficulties. Sci. Educ., 1979 63, 9-14.
Moore, R. W., & Sutman, F. X . The development, field test and validation of an inventory of sci-
entific attitudes. J . Res. Sci. Teach., 1970, 7, 85-94.
Osgood. C. E., May, W. H., & Miron, M. S. Cross-cultural universals of affective meanings.
Urbana: University of Illinois Press, 1975.
Renner, J . W., Abraham, M. R., & Stafford, D. G. A summary of research in science education
1976, N.Y.: John Wiley, 1978.
Robinson, James T. Student attitudes toward science courses in test schools using human sciences.
J . Res. Sci. Teach., 1980, 17, 231-241.
Schibeci, R. A. Attitudes to science: a semantic differential instrument. Res. Sci. Educ., 1977,
7,149- 155.
Schibeci, R. A., & McGaw, B. Empirical validation of the conceptual structure of a test of
science-related attitudes. Educational and Psychological Measurement, 198 1,41, 1 195-
1201.
Schofield, H. R., & Start, K. B. Mathematics attitudes and achievement among student teachers.
Aus. J . Edu., 1978, 22, 78-82.
Shaw, M. E., & Wright, J. M. Scales for the measurement of attitudes. New York: McGraw-Hill,
1967.
Simpson, R. D., Rentz, R. R., & Shrum, J. W. Influence of instrument characteristics on student
responses on attitude measurement. J. Educ. Res., 1980, 73, 271 -275.
Welch, W. W. Review of the research and evaluation program of Harvard Project Physics. J . Res.
Sci. Teach., 1973,10, 365-376.
West, C. K., & Robinson, D. G. Prestigious psycho-educational research published from 1910
to 1974: Types of explanations, focus, authorhship, and other concerns. J . Educ. Res., 1980,
73,271-275.
Yamamoto, K., Thomas, E., & Karns, E. School related attitudes in middle school age students.
Arner. Educ. Res. J., 1969, 6 , 191-206.
Accepted for publication 16 October 198 1

You might also like