An Attitude Scale Based Upon Behavior Situations
Arlyn C. Rosander
University of Chicago
There is no reason why the Thurstone technique of constructing social attitude scales should be limited to those in which short statements of opinion are employed as the scale elements. Pictorial representations, such as cartoon, might be collected and scaled in a similar manner. Short statements of experiences peculiar to some social object, issue, or condition could be dealt with in the same way, In this way we could scale the experiences of an individual toward intoxicating liquor, for example, and study the relation between his experience score and his attitude score. Again, one might use common situations with regard to a single issue, such as the social equality of white people and Negroes, and try to scale the various degrees of response to these situations. It is this latter type of social attitude scale which we have been investigating and which we shall now describe in some detail. We wish to make clear at the beginning that this scale is a paper and pencil variety, and not the kind that deals directly with overt behavior.
SELECTING THE ATTITUDE VARIABLE
Now some attitude variables lend themselves to the opinion type element much more readily than to the behavior situation type of element; such variables would be the League of Nations, the Monroe Doctrine, the Federal Constitution, war debts, and municipal ownership of public utilities. These are variables in which our experience is highly opinionated, intellectual, and impersonal.
Clearly the behavior situation type of attitude scale demands an issue or symbol in which there exists an abundance of personal face-to-face relationships. This suggests such symbols as races, nationalities, religious sects and denomination, divorce, intoxicants, social and economic position of women, and possibly war and patriotism.
For our study we selected the social equality of the Negro and the white as the attitude variable. This was done for several reasons. In the first place, this variable stands for common experiences and prejudices; this makes it possible to obtain numerous situations involving Negro-white relationships, and quite a gradation of response within each situation. This guarantees scaling on a wide range of attitude magnitude, and increases the reliability of the completed scale.
In the second place, most people have had actual contact of one kind or another with the Negro; this makes it easier to react vicariously to the situation.
In the third place, we were fortunate in having at hand a revised form of the Negro opinion scale constructed by Hinckley. This made it possible to give the two types of scales simultaneously to discover the degree to which they measure the same thing. In order to make a fair comparison possible, all factors in the construction of the two scales must be held constant, or accounted for, except the type of scale element. This required that the Thurstone method of construction be followed exactly. Seven steps were involved in the construction of this behavior' situation scale: the collecting and editing of scale elements, the preliminary sorting, the final sorting, the scaling, the selection of parallel forms, the determining of the reliability, and the determining of validity. These steps will now be discussed in the order given.
THE COLLECTING AND EDITING OF SCALE ELEMENTS
The behavior reaction element .for paper and pencil scales represents an attempt to project the individual into a real life situation and have him select the responses which he would make if he actually faced that situation. Both situation and response are actual not fictitious, simple not complex. Only the bare essentials of both have been retained; to include everything involved in a situation would destroy the values of this type of element for measurement purposes. To what extent this simplification invalidates the whole procedure remains to be seen. It is imperative also to obtain a wide variety of common situations and responses in order that one may secure a reliable sampling of the reaction repertoire of the subject being measured.
Situations, and associated responses, were collected by drawing
( 5) upon personal experiences and observations, by reading books on race relations, and by interviewing adults from both the northern and southern States. Early in the study we discovered that reactions to different situations might reveal quite antagonistic attitudes unless the variable measured was carefully stated. People might prefer Negro cooks, maids, or chauffeurs to white workers of the same occupations, but this preference could hardly be considered a pro-Negro attitude of a general type; it may be nothing more than a way of expressing a feeling of social superiority. Hence we avoided all situations characterized by intimate personal service.
No attempt was made to invent either situation or response. We tried to obtain situations that are fairly common. The situations are those known to exist; they are expressed in as few words as possible. There are as few as two, and as many as 13, reactions to a single situation. All reactions are stated in behavioristic terms in order to make a clear-cut break with the opinion scale. In the final master list there were 24 situations and 175 reactions in all. We give here a list of several situations and the behavior reactions associated with them.
6. You are reading in a public library. A Negro comes in and sits down beside you.
a. You rise and go to another table.
b. You keep right on with your reading unconcerned.
c. You complain to the librarian.
d. You leave the library at once.
7. In a community where you live a Negro attacks a white girl.
a. You join a mob intent on killing him.
b. You pay no attention to the incident.
c. You demand that all Negroes be driven out of town.
d. You try to break up the mob.
13. A Negro is put to work in the same department with you so that you have to associate with him every day.
a. You quit at once.
b. You protest to the employer.
c. You treat him coldly but civilly.
d. You threaten the Negro with violence if he does not leave.
e. You treat him in every way as a co-worker.
16. You buy tickets to a theater but find on attending the performance that Negroes have seats adjacent to yours.
a. You refuse to sit there.
b. You leave the theater at once.
c. You complain to the manager.
d. You pay no attention to the Negroes.
23. You attend a conference at a hotel which will not allow the Negro delegates to register.
a. You demand that the Negroes be allowed to
b. You advise the Negroes to withdraw.
c. You protest to the manager.
d. You propose the conference meet where Negroes are welcome.
THE PRELIMINARY SORTING
The next step was to prepare these situations and reactions so that the regular sorting technique could be applied to them. Each situation was numbered and every reaction under it was lettered; one situation and one associated reaction were printed on a card which carried a serial number.
In order to determine the number of divisions into which these reactions might be sorted, some preliminary judging was undertaken. Several persons familiar with the technique were asked to sort the cards into as many equal divisions as they judged to exist. Since some of the judges used 12 divisions, and others 11, the higher number was used. Subsequent sorting showed that this step was justified. The use of preliminary sorting is very useful in planning for the final sorting and as a check-up on the gradation of statements, since there is nothing to justify a fixed number of divisions, particularly when one is dealing with behavior situations and not just statements of opinion.
THE FINAL SORTING
Directions to sorters were given orally and usually individually or in groups of not more than three. The directions given below apply to the sorting of statements of opinion in the usual Thurstone technique; for behavior reactions similar directions were used, but some very slight modifications were made since 12 piles or divisions were used and terms descriptive of behavior had to be substituted for those descriptive of opinion.
1. There are a number of statements of opinion on the issue ranging all the way from extremely favorable to extremely antagonistic opinions.
2. There are 11 piles into which these statements are to be sorted; adjacent piles are to represent noticeable differences of opinion according to your judgment.
3. You sort the statements as though someone else were making the statement and you were judging his position by what he said.
4. Avoid agreeing or disagreeing with the statements; we are not concerned with your personal beliefs.
5. There is no right or wrong way of sorting the statements; we want your personal judgments uninfluenced by what other people believe.
6. Once you have sorted all the statements into the 11 piles, go through the statements in each pile to determine if they are of the same intensity; if not, make shifts as you judge them necessary.
7. You are to pay no attention to the logic, soundness, or truthfulness of the statements; take them at their face value.
8. Sort the statements as you will without reference to the number of statements placed in each pile; in some piles there may be many statements while in others there may be but a few.
9. The statements in the two end piles will be those which you judge the most extreme in the pack of statements which you have; do not leave these end piles empty because you can think of more extreme statements than any you find in the pack.
10. Do the best that you can giving a reasonable amount of care to your judging. Groups of judges doing careful work independently agree much more than most people realize. Avoid being extremely critical or extremely careful; such extremes may lead to confusion.
11. One does not have to assume a highly critical attitude, or a critical attitude at all for this type of work. Criticisms, however, are always welcome but should not be allowed to interfere with following the foregoing directions.
Fifty judges or sorters were used for the final sorting. These sorters were graduate and undergraduate students of the department of psychology and the division of social sciences of the University of Chicago. Instructions were given individually and orally, not by written statement. The judges were supervised but not influ-
8) -enced; if they seemed to understand the directions as manifested by
their sorting, they were let alone. Only one sorting was discarded; in this one
most of the cards were placed in four piles.
THE SCALING OF THE REACTIONS
Each sorter put each reaction in some one of the 12 piles. For every reaction there were 50 rankings; one for each of the judges. These rankings or sortings made possible a frequency distribution from which the scale value of the reaction was computed.
Using the Thurstone special calculation paper, the frequencies were tabulated, the proportions computed and accumulated, and a cumulative proportions curve drawn. Where this curve crossed the 50 percentile line, there was read the scale value of the reaction estimated to tenths of the distance between two adjacent piles. We give in Table 1 the data for reaction "e" of situation 24 and for
reaction "f" of situation 11. Reaction 24e has a scale value of 2.5 and a quartile deviation of 1.4; the frequency distribution shows
( 9) considerable agreement among the sorters. Reaction 11 f has a scale value of 8.0 and a quartile deviation of the same amount, 8.0. This latter reaction is a good example of the extremely ambiguous statement. Twenty-two of the judges placed it below the middle point whereas 28 placed it above this position. In well-edited lists these statements are seldom found, and of course are never used in the final scale.
In Table 2 we give a distribution of the 175 reactions in the 12 piles into which they were sorted. These data show that we obtained a better distribution of reactions at the pro-end than at the anti-end. They show also that we were fully justified in using 12 divisions "not"11. While the number of reactions in piles 1, 5, 6, and 12 is not great enough for the most convenient construction of parallel forms, this deficiency does not in any sense destroy the usefulness of the list.
Using 50 sorters makes it possible to avert the use of a calculator in computing the proportions, since this number of judges leads to proportions in any one division which are twice that of the corresponding frequency; the proportions can be added mentally without a machine. Twenty-five sorters could be used to advantage in the same way; then the proportion in any division would be four times the corresponding frequency.
THE SELECTION OF PARALLEL FORMS
Once the scale values and the Q values (quartile deviations) of every reaction were computed, the next step was to construct two parallel forms. In the opinion type scale the statements are equated on the basis of equal or nearly equal scale values with careful consideration being given to the Q value which must not be too high and to the "color" of the statement.
In the behavior situation type scale, however, the problem is more complicated due to the existence of both situations and responses. Two different methods were used in constructing parallel forms: Scale I in which responses within the same situation are paired on the basis of equal or nearly equal scale values, and Scale II in which reactions of equal or practically equal scale values are paired regard-less of the situation in which they appear. In each case 22 reaction elements are employed in the final scale.
In Scale I every paired reaction used in constructing the two parallel forms is taken from the same situation; some situations were used twice so that in all there were 14 different situations and not 22, which would be true if a different situation as well as a different reaction were used. There is no reason why this should invalidate the scale; the 14 situations represent a fairly adequate sampling of common situations. Some irregularity in magnitude of pairs of scale values exists because two reactions within the same situation seldom had exactly the same scale values. After the two groups of situations were selected, the reactions within each group were arranged in random order and printed as Forms A and B. In each form there are 22 reactions.
In Scale II there are 17 situations common to the two forms, but no two situations are paired at the same scale value. By this method the corresponding scale values of the two forms are much more nearly identical than those of the parallel forms of Scale I. In each form of Scale II there are 23 different reactions, one more than the number in the forms of Scale I.
We give below a copy of Scale I Form A just as it was administered to the various groups. The numbers preceding each statement are the scale values of the reactions. Of course, these do not appear on the scale as administered.
BEHAVIOR ATTITUDE SCALE
SCALE I—FORM A
Put a check mark (√) in front of statement if you would act as stated. Put a cross (X) in front of statement if you would not act as stated.
( ) 1.
In the community where you live a Negro marries a white girl.
You do nothing about it.
( ) 2.
You are bathing at a beach. Some Negroes approach and enter the water near you.
You start a fight with them.
( ) 3.
A Negro family moves into the residential district where you live.
You invite them to your home.
( ) 4.
A Negro with a college education is a qualified candidate for Congress from your
You vote for him on the basis of race alone.
( ) 5.
Your sister takes a friendly interest in an educated and unmarried Negro boy to
whom she has been introduced.
You commend her for her broad-mindedness.
( ) 6.
A Negro family moves into the apartment building in which you live.
You act friendly toward them.
7. You stop at a hotel which you discover caters to Negroes as
well as to whites.
You remain in the hotel.
8. You are reading in a public library. A Negro enters and sits
down beside you.
You leave the library at once.
( ) 9.
A well-educated Negro applies for membership in a high school or college society
of which you are a member.
You move that the constitution of the club be amended to allow Negro members.
( ) 10.
You attend a conference at a hotel which will not allow the Negro delegates to
You propose that the Negroes attend the meetings but live in another hotel.
( ) 11.
In a community where you live a Negro attacks a white girl.
You demand that all the Negroes be driven out of town.
12. You are bathing at a beach. Some Negroes approach and enter
the water near you.
You go to some other beach.
13. In the community where you live a Negro marries a white
You fight for the maintenance of the color line.
14. The congregation of the church you attend has always been
white. One Sunday morning a Negro attends the services.
You do nothing about it.
15. A Negro is put to work in the same department with you so
that you have to associate with him every day.
You try to have as little to do with him as possible.
16. In a community where you live a Negro attacks a white girl.
You try to break up the mob which forms.
17. A Negro family moves into the apartment building in which
You treat them coolly.
18. A Negro is put to work in the same department with you so
that you have to associate with him every day.
You act friendly toward him.
19. You .attend a conference at a hotel which will not allow the
Negro delegates to register.
You advise the Negroes to withdraw.
20. You see a white girl whom you know walk down the street with
a Negro boy.
You never speak to her after that.
21. A welt-educated Negro applies for membership in a high
school or college society of which you are a member.
You consider his application the same as you would that of any other student.
22. Your sister takes a friendly interest in an educated and
unmarried Negro boy to whom she has been introduced.
You warn her of the possible consequences of her behavior.
Both the opinion type scale and the behavior type scale are scored in the
same way. In scoring one uses only those statements or re-actions
with which the person agrees, that is, those which he has checked. Those which
he has placed a cross in front of are ignored. The scale value of each of these
checked statements or reactions is noted, and the median of the several scale
values computed ; this median value is the attitude score of the individual.
RELIABILITY OF THE SCALE
In order to test the reliability of the two scales, we administered both forms of Scale. I to 98 students at the University of Chicago; we gave both forms of Scale II to 82 students at the University of Chicago and to 88 students at a southern university. For purposes of comparison we also gave to the group of 98 and to the group of 88 the revised Hinckley opinion type scale of attitude toward the Negro. In order to compute the reliability of this shortened Negro opinion scale, we divided the scale into two equivalent scales and correlated the scores on one half with those on the other half. The results are given in Table 3. These coefficients are based upon a total of 44 reactions in the combined forms of the behavior scale and 24 statements in the entire opinion scale.
Several points are brought out in Table 3. In the first place, the reliability coefficients for both behavior scales are fairly high—as
high as those of the ordinary intelligence test but not so high as those
which we have found for the opinion scale. The usual opinion scale of 40 to 44
statements gives a reliability coefficient of .90 to .95. In the second place,
the reliability of Scale I was slightly higher for northern students than was
that of Scale II, although each form of Scale II has one more reaction than the
corresponding forms of Scale I. In the third place, the reliability coefficient
for Scale II, and for the opinion scale as well, is much lower for the southern
students than for the northern students. It is not fair to compare the
reliability coefficients of the behavior scale and the opinion scale because
they were not of the same lengths. If we correct the opinion scale coefficients
for length, we obtain a coefficient of .835 for the southern students and .945
for the northern students. We can make this comparison: that the reliability
coefficient of the Opinion Scale based upon 88 southern students (.835)
is less than that of Behavior Scale I based upon 98 northern
students (.856) after correcting for equal lengths. It seems apparent,
therefore, that a carefully constructed behavior situation type of attitude
scale should give a reliability co-efficient between .80 and .90 for two
parallel forms of 22 reactions each.
VALIDITY OF THE BEHAVIOR SCALE
In testing the validity of the behavior scale we are faced with the problem of determining to what extent it measures attitudes. We present two approaches to the problem; one in which we examine the extent to which the scale differentiates individuals known to have significantly different attitudes, and the other in which we determine to what extent measurements made by this scale correspond to similar
( 14) measurements made by another scale constructed on a different basis. In the first approach we present the mean scores made by northern and southern students since we would expect their attitudes to be different. In the second approach we present the correlation between scores on the Negro behavior scale with scores on the Negro opinion scale for the same population.
The mean score of 82 northern students on Scale IIA was 7.0 and on Scale IIB was 7.3. The corresponding scores for 88 southern students were 4.3 and 4.2. This shows clearly that the northern students and the southern students are on opposite sides of the mid-point which is 6.0. This shows, too, that the southern students are farther down the scale than the northern students are above, and that the former are more strongly anti-Negro than the latter are strongly pro-Negro.
The correlation between opinion scores and behavior scores for the same group of students is shown in Table 4. Here we show not
only the raw correlations but also the corrected coefficients based upon Spearman's formula for the correction of attenuation. This corrected coefficient is that which we would obtain if we were able to eliminate all chance errors from our measurements.
Here again we find differences between the northern and southern students, the northern students showing the greater consistency between scores on the opinion scale and scores on the behavior scale. The main point, however, is the high degree of commonality which exists between the opinion scale and the attitude scale. There seems no reason why we cannot obtain correlations .between opinion and behavior ranging from .80 to .90, especially as these types of scales are further improved.
We may use these data to support the validity of the opinion scale, but there is more to the behavior scale than just that. The behavior scale is so much more specific that one obtains a sharper picture of an individual's attitude pattern toward the Negro than he obtains from the more or less general statements of opinion which appear in the opinion type scale. This type of paper and pencil scale brings us more closely to the potential behavior of the individual than does the opinion type scale, and is therefore likely to be much more predictive of actual behavior. This however is a hypothesis which needs to be investigated further.
University of Chicago