toys in the attic:
ideological furnishings for the homeless mind


daurril library: talcott parsons

Values and Value-Orientations 479

 

The Theory of Action and Its Application

 

4.5   An Empirical Study of Technical Problems in Analysis of Role Obligations

 

SAMUEL A. STOUFFER

 

                The approach in this volume toward a system of categories which may unify theories of culture, society, and personality is necessarily at a rather high level of abstraction.  In its present tentative form and in the numerous further revisions which may be expected, the system will be appraised from at least two standpoints.  One involves the correctness of the reasoning wherever logical inferences are made. 

The other involves the usefulness of the system, assuming that it passes muster logically.  It is with the

usefulness of such a theoretical scheme that the ensuing comments are concerned. 

 

                A theoretical system  for example, an elaborate mathematical model in economic theory  can be logically correct without necessarily being fertile in generating middle-range hypotheses capable of empirical verification.  If such a theory organizes hitherto disconnected clusters of ideas into a single integrated system, it can be useful merely in providing a new context for evaluating the separate parts more critically at a theoretical level.  But all of us would be quite disappointed if the type of thinking which has gone into this volume were to have only this as its end result. 

 

                As a minimum, it is hoped that the effort to tie together the significant ideas about culture, society, and personality will provide broad orientations which map, as it were, areas for suggested research, even if it does not directly generate specifically deducible,  testable propositions at a middle-range level. 

This hope is based on the fact that the thinking of Professor Parsons and colleagues is decidedly not arm-chair thinking alone.  The streams of influence which lie behind it represent decades of theory and also

empirical research in the three disciplines of anthropology, sociology, and psychology.  I have been impressed with the concern for empirical referents which the authors have constantly manifested during their work, even if this concern is not always visible in the resulting document. 

 

480 The Theory and Its Application

 

                One cannot yet say that the present approach to a system has matured enough to give assurance of its power to go further than signaling areas for further research.  That is, it may be asking too much to expect now a large number of deductions of specific propositions as necessary logical consequences of the basic postulates.  This would be ideal, of course.  But we should not underestimate the possible value of the more modest objective, which is furthered by the anchoring of the basic postulates in experience and at the

same time by the fact that these postulates are not ad hoc, but have logical interconnections. 

 

                The fact that big ideas of Darwin, Marx, or Freud do not generate by mathematical or logical deduction propositions in social science, like the ideas of Newton and Einstein in physics, is not an argument against the signaling value of abstract ideas in social science.  Rather, these orientations encourage us to dig in one part of a field rather than another and are indispensable because we cannot dig everywhere at once.  The theories will have been useful if the digging uncovers something, even if they did not tell us too explicitly what we might find. 

 

                At the same time, the history of social science, like the history of medicine, ought to warn us against an excess of optimism about the necessary usefulness of any new system of highly general orienting ideas.  Many proposed systems have had little or no impact.  Some - for example, ideas like those of Benjamin Rush in medicine - may have been positively harmful.  Just as medicine has profited most from middle-range propositions, tested in empirical research though not deducible, as yet, from any single Newtonian formula, so it seems likely that social science now desperately needs middle-range testable propositions. 

 

                If the concepts central to the system of unified theory proposed in this volume are to be useful, even for suggesting if not deducing middle-range propositions, it would seem preferable that they be clear and unambiguous.  The word preferable is used instead of the word essential - some of Freud's murkier concepts, for example, may prove to have been very useful.   It would be as silly to demand immaculate precision in concepts at the present stage as it would be to demand correlations of 1.00 in all empirical research.  Nevertheless, one suspects that there is a higher probability of a general theory having some impact on the vital task of setting up testable middle-range propositions if the concepts are capable of some kind of clear and, where possible, operational definition. 

 

                One of the significant ideas in the system outlined in this volume is the concept of role.  (See, for example, the discussion of roles in the General Statement in Part I.)  This is not a new concept, but its possible utility in unifying personality and societal theory has perhaps not before been seen so clearly.  Attractive as the concept is, in the abstract, there has been as yet relatively little study of the technical problems involved in using it empirically. 

 

An Empirical Study 481

 

                The work of Professor Parsons and his colleagues has inspired the following modest pilot study, written in collaboration with Jackson Toby, which offers hope as to the possibility of operational definitions of certain types of role obligations and, at the same time, provides specific warnings as to the immensity and complexity of the task.  The study was published in the American Journal o/ Sociology, March 1951.  It is reproduced here with the Journal's permission, and with the hope that it will stimulate others to do such jobs better.

 

SAMUEL A. STOUFFER AND JACKSON TOBY

 

Role Conflict and Personality 1

 

                A convenient way to examine the informal social controls operating in a given institution is through the study of role conflict.  In an earlier statistical analysis of an example of role conflict, stress was laid on the concept of variability and implications for the theory of role of different classes of variability.2 

 

                The present paper also is concerned with role conflict.  But it seeks to provide a link between the study of social norms, with which the former paper was primarily concerned, and the study of personality.  Specifically, when there is a lack of consensus in a group regarding the "proper thing to do" in a morally conflicting situation, is there a tendency for some individuals to have a predisposition or a personality bias toward one type of solution and for other individuals to have a predisposition toward another type of solu-

tion?  If such a predisposition exists, there should be a tendency to carry over certain types of behavior from one role conflict to another with some consistency. 

 

                1 The research here reported was conducted with the assistance of the Laboratory of Social Relations, Harvard University.  Special acknowledgment is due Paul F. Lazarsfeld, Talcott Parsons, and Gordon W. Aliport.  Professor Lazarsfeld proposed the applicability of a new form of latent distance structure and himself carried out the computations reported in the note appended to this paper.  A pretest of the present study was the subject of a paper by the authors at the American Sociological Society in December 1949, at which the paper's discussion by Professor Leonard S. Cottrell contributed to the present formulation.

 

                2 Samuel A. Stouffer, "An Analysis of Conflicting Social Norms," American Sociological Review, XIV (December, 1949), 707-717. 

 

482 The Theory and Its Application

 

                An especially common role conflict is that between one's institutionalized obligations of friendship and one's institutionalized obligations to a society.  The obligations of friendship in Western culture, to use the terminology of Talcott Parsons, are

particularistic rather than universalistic,

affectively toned rather than affectively neutral, and

diffuse rather than specific.3 

A universalistic obligation is applicable to dealings with anybody (for example, obligation to fulfill a contract) ; a particularistic obligation is limited to persons who stand in some special relationship to one (for example, the obligation to help a relative or a close friend or neighbor). 

Diffuseness of particulartistic obligations provides flexibility in the definition of these roles.  That is, the content of an individual's particularistic obligations (toward a friend, a brother, a grandchild) depends in part on the intimacy of the relationship itself.  The greater the affection, the greater the sense of obligation. 

On the other hand, universalistic obligations are defined more rigidly, for they regulate behavior toward all human beings - regardless of affective involvement.  Hence, in any specific situation involving conflict between duty to a friend and duty to society, we would expect that some individuals are more prone

to regard the particularistic obligation as taking precedence than others, because there is variability from individual to individual in the intimacy of friendships. That is, respondents tend to project into the hypothetical situations reference friendships drawn from their own experience.  A description of an institutionalized social norm must not only take into account, then, the beliefs and behavior of a modal member of the group but must also observe the individual variability in the perception of obligations.  This variability - or "social slippage" - was a major concern in the earlier analysis of role conflict cited above. 

 

                In the present paper we shall deal with several situations involving conflicts between obligations to a friend and more general social obligations.  If, as our conception of the intrinsic variability of particularistic obligations would lead us to expect, some people are more likely than others to choose in a variety of situations the particularistic horn of the dilemma rather than the universalistic, we should be able to devise a scale to measure such a tendency.  With such a scale people should be ranked along a single dimension

according to their probability of possessing the attribute or predisposition of choosing one type of solution rather than the other. 

 

                What we have to present here is only a crude beginning; indeed, only a classroom example.  Yet it should prove instructive in a number of respects to those who may wish to carry on further research with needed refinements.  Our data are based on a short paper-and-paper questionnaire completed by 648 undergraduate students at Harvard and Radcliffe in February 1950.   No claim is made for the representativeness of the sample, since almost all were members of a single course in Social Relations.

 

                3 See, e.g., Talcott Parsons, Essays in Sociological Theory (Glencoe, Illinois: The Free Press, 1949), chap. viii.

 

An Empirical Study 483

 

                Four little stories were presented, as follows:

 

                1. You are riding in a car driven by a close friend, and he hits a pedestrian.  You know he was going at least 35 miles an hour in a 20-mile-an-hour speed zone.  There are no other witnesses.  His lawyer says that if you testify under oath that the speed was only 20 miles an hour, it may save him from serious consequences.

 

What right has your friend to expect you to protect him?

Check one:

[]  My friend has a definite right as a friend to expect me to testify to the lower figure.

[]  He has some right as a friend to expect me to testify to the lower figure.

[]  He has no right as a friend to expect me to testify to the lower figure.

 

What do you think you'd probably do in view of the obligations of a sworn witness and the obligation to your friend?

Check one:

[]  Testify that he was going 20 miles an hour.

[]   Not testify that he was going 20 miles an hour.

 

                2. You are a New York drama critic.  A close friend of yours has sunk all his savings in a new Broadway play.  You really think the play is no good.  What right does your friend have to expect you to go easy on his play in your review?

Check one:

[]   He has a definite right as a friend to expect me to go easy on his play in my review.

[]   He has some right as a friend to expect me to do this for him.

[]   He has no right as a friend to expect me to do this for him. 

 

Would you go easy on his play in your review in view of your obligations to your readers and your obligation to your friend?

Check one:

[]   Yes.

[]   No.

 

                3. You are a doctor for an insurance company.  You examine a close friend who needs more insurance.  You find that he is in pretty good shape, but you are doubtful on one or two minor points which are difficult to diagnose.

 

What right does your friend have to expect you to shade the doubts in his favor?

Check one:

[]   My friend would have a definite right as a friend to expect me to shade the doubts in his favor.

[]   He would have some rights as a friend to expect me to shade the doubts in his favor.

[]   He would have no right as a friend to expect me to shade the doubts in his favor.

 

484 The Theory and Its Application

 

Would you shade the doubts in his favor in view of your obligations to the insurance company and your obligation to your friend?

Check one:

[]   Yes.

[]   No.

 

                4. You have just come from a secret meeting of the board of directors of a company. 

You have a close friend who will be ruined unless he can get out of the market before the board's decision becomes known.  You happen to be having dinner at that friend's home this same evening.

 

What right does your friend have to expect you to tip him off?

Check one:

[]   He has a definite right as a friend to expect me to tip him off.

[]   He has some right as a friend to expect me to tip him off.

[]   He has no right as a friend to expect me to tip him off.

 

Would you tip him off in view of your obligations to the company and your obligation to your friend?

Check one:

[]   Yes.

[]   No.

 

                The problem is: do the answers to these questions indicate the existence of a unidimensional scale, along which respondents can be ordered according to the degree to which they are likely to possess a trait or bias toward the particularistic solution of a dilemma?  For simplicity,

we label for a given item the response, "My friend has a definite right . . . ," as particularistic;

the response, "He has no right . . . ," as universalistic.

If a person marks, "He has some right . . . ," we label the response

particularistic if in the second part of the question he says he would favor the friend in action; and universalistic, if he says he would not favor the friend.

 

                There was a considerable spread among the four items in the percentage giving particularistic responses:

                Item                                        Per cent

                1 (car accident)                    26

                2 (drama critic)                     45

                3 (insurance doctor)           51

                4 (board of directors)          70

 

Such frequencies suggest the hypothesis of a distance or cumulative scale. 

 

                Following Louis Guttman's scalogram method, the responses to all the items were cross-tabulated and scale patterns arranged according to nearest scale type, as shown in Table 1.  While the reproducibility (0.91) and the distribution of cutting points suggest the admissibility of the hypothesis that these items form a Guttman scale, the items are too few in number for us to speak with confidence, especially in the presence of two sets of rather numerous non-scale responses (+-++ and  -+-+).  Rigor would require ten or more items to start with, in order to determine scalability, although we might in the end select fewer items for subsequent use. 

 

An Empirical Study 485

 

Table 1

 

SCALOGRAM PATTERN FOR RESPONDENTS TO FOUR ITEMS ON ROLE CONFLICT

 

                                Scale      Particularristic                                      Universalistic

                Scale      pattern   response to item no.                           response to item no.

                type        l 2 3 4     1              2              3              4              1              2              3              4              "Error"

 

                5              ++++       66           66           66           66                                                                           0

                                +-++        52                           52           52                           52                                           12

                                + +  +      15           15                           15                                           15                           15

                                +++-        8              8              8                                                                              8              8

                                + - + -      5                              S                                             5                              5              10

                                + + --       6              6                                                                              6              6              12

 

4              -+++        ..              95           95           95           95           0

                                +++ -       ..              16           16           ..              16                                           16           16

 

                3              -+++                                        80           80           80           80                                           0

                                 --+-                                         14           ..              14           14           ..              14           14

 

                2              ---+                                                          71           71           71           71           ..              0

                                -+-+                         66                           66           66           ..              66           ..              66

                                +--+         13                                           13           ..              13           13           ..              13

 

                                 

                1              -+--                          21                                           21           ..              21           21           21

                                +---          6              ..                                                              6              6              6              6

                                ----                                                                           114         114         114         114         0

 

                                                171         293         336         458         477         355         312         190         233

 

                Reproducibility = 1 - ~33/ (4~648)] = 0.91

 

The pure Guttman model can be viewed as the limiting case of a more general latent distance model which Paul F. Lazarsfeld has introduced.4   It seems worth while, therefore, to examine the applicability to these data of the Lazarsfeld latent distance model, which postulates a latent continuum with as many ordered classes as there are items, plus one.  The model assigns to each item a probability that a positive (e.g., particularistic) response to that item assigns the respondent to a particular segment of the hypothetical latent continuum.5 

 

                4 Stouffer, Guttman, Suchman, Lazarsield, Star, and Clausen, Measurement and Prediction, Vol. IV of Studies in Social Psychology in World War II (Princeton University Press, 1950).  Guttman's theory and procedures are described in Chapters 2 to 9, Lazarsfeld's in Chapters 10 and 11.  Chapter 1 provides an introduction to both methods. 

 

                5 Latent structure theory postulates that all the relationship between any two manifest items can be accounted for by the Joint correlation of the items with the latent structure.  In other words, within any segment of the latent structure the correlation between two manifest items is zero. 

 

486 The Theory and Its Application

 

                For reasons of space, the arithmetic in testing the applicability of the latent distance model to our data will not be exhibited here.  However, a brief technical summary of the results appears at the end of this paper.  Al-though the procedure used is still too new to have developed wholly satisfactory acceptance standards, the outcome was quite encouraging. 

 

                Actually, an additional precaution was taken. Experience with projective material has taught us to expect considerble differences when we ask, "What do you think about something?" from results if we ask, "What do you think somebody else would think about something?"  Especially when we are seeking by crude question and answer procedures to learn something about social norms, it is very important to know what, if any, differences are produced by such shifts imposed on the point of view of the respondents.  Hence, only a third of our 648 respondents were asked questions in the form exhibited

above.

                For a third of the subjects the stories were rewritten so that the friend of the respondent, not the respondent himself, faced the role conflict.  To illustrate with the motor car example:

 

                Your close friend is riding in a car which you are driving, and you hit a pedestrian. He knows that you were going at least 35 miles an hour in a 20-mile-an-hour zone.  There are no other witnesses.  Your lawyer says that if your friend testifies under oath that the speed was only 20 miles an hour it may save you from serious consequences.

 

What right do you have to expect him to protect you?

Check one:

[]   I have a definite right as a friend to expect him to testify to the lower figure.

[]   I have some right as a friend to expect him to testify to the lower figure.

[]   I have no right as a friend to expect him to testify to the lower figure. 

 

What do you think he would probably do in view of his obligations as a sworn witness and his obligation as your friend?

Check one:

[]   Testily that you were going 20 miles an hour.

[]   Not testify that you were going 20 miles an hour.

 

                For still another third of the respondents, a third version was presented.  In this case neither the respondent nor his friend faced the dilemma, but two hypothetical people, Smith and Smith's friend, Johnson.  Again to illustrate with the motor car example:

 

                Smith is riding in a car driven by his close friend, Johnson, and Johnson hits a pedestrian.  Smith knows that his friend was going at least 35 miles an hour in a 20-mile-an-hour speed zone.  There are no other witnesses.  Johnson's lawyer says that if Smith testifies under oath that the speed was only 20 miles an hour, it may save Johnson from serious consequences. 

 

An Empirical Study 487

 

What right does Johnson have to expect Smith to protect him?

Check one: 

[]   Johnson has a definite right as a friend to expect Smith to testify to the lower figure.

[]   He has some right as a friend to expect Smith to testify to the lower figure.

[]   He has no right as a friend to expect Smith to testify to the lower figure.

 

If Smith were an average person, what do you think he would probably do in view of his obligations as a sworn witness and his obligation to his friend?

Check one:

[]   Testify that Johnson was going 20 miles an hour.

[]   Not testify that Johnson was going 20 miles an hour.

 

                The different forms of the questionnaires were interleaved and handed out at random. In testing for the goodness of fit of the latent distance scale, separate tests were applied to each of the three types of items.  The model seemed to fit about equally well in all three cases, and the rank order assigned to particular scale patterns was very much the same, except for a few scale types containing a negligible number of cases.  As would be expected, the rank-order groupings derived from the latent distance model is very close to the rank-order grouping obtained by scoring to nearest scale type in scalogram analysis.6 

For purposes of subsequent analysis the rank groupings for each of the three forms were constituted as in Table 1.  The extent to which the three forms agreed with one another can be seen from Table 2. 

The principle discrepancies are due to differences in frequency of responses to Items 2 and 3 respectively, but the groupings in Table 2 do not differ from one form to another more than would be expected by chance, according to the chi-square test. Incidentally, it is of some interest to note that the reproducibility of each

form is in the neighborhood of 0.90.

 

                6 In scoring to nearest scale type by scalogram procedure, the objective is to arrange the scale patterns to minimize error."  Thus ++-+ is grouped with ++++, on the assumption that only the response to the third item is an error.  If it were grouped with -+++, we should have to assume two errors, in the first and third items, respectively.  However, there are some items which might be grouped in different ways with

the same amount of error. For example, - + - + would be grouped with - +++ if we assumed that the third item was an error, but would be grouped with - - -+ if we assumed that the second item was an error. 

Such doubtful cases arc resolved by the latent distance analysis, which in the present example usually gave clear and consistent information. 

 

488 The Theory and Its Application

 

This is, of course, much too small a set of items about which to make any serious claims either to rigorous scalability or to generality, but the results encourage one to believe that we can develop good measures of individual predisposition to a bias in a particularistic or universalistic direction.  We must note that a scale such as this is not an unequivocal measure of particularism-universalism.  Since friendship obligations are diffuse and affectively toned as well as particularistic, and societal obligations are specific and affectively neutral as well as universalistic, we have scaled a predisposition for diffuse, affectively toned obligations over specific, affectively neutral obligations as well as a predisposition for particularistic over universalistic obligations.  But this fusion of variables in our situations does seem to generate a unidimensional scale, the dimension involved being the degree of strength of a latent tendency to be loyal to a friend even at the cost of other principles.  The rank groupings would represent ordered degrees of probability of taking the friend's side in a role conflict.7 

 

Table 2

 

SCALE PATTERN GROUPINGS SHOWN SEPARATELY FOR THREE FORMS

OF QUESTIONNAIRE

 

                Scale      Scale      Form A:  Form B:  Form C:

                type         patterns Ego faces              Ego's friend          Smith laces

                                1234       dilemma                faces dilemma     dilemma

                                                20           20           26

                +              ++           9              23           20

                                                6              4              5

                5                              2              3              3

                2              3              0

                {               1              3              2

                +

                40           56           56

                                                38

                4              { i++++t  7

 

45

                25           32

                6              3

                31

                24           29           27

                3  {          ++           6              5              3

                I1+

                30           34

                r               -~ -+        23                           31

                2  j           +-~~++   25                           15

                                                4                              4

                52           50

35

 

 

 

30

17

26

5

48

                                                6              6              9

1              {               ~              1              2              3

                                                42           37           35

                49           45           47

                216                         216

                Reproducibiuty     0.92        0.91        0.90

I'

 

                7 Of course, we shall eventually be interested in finding out whether a more abstract scale - for example, one of universalism-particularism alone - would stand up and, if it did, more about its genesis.

 

An Empirical Study 489

 

                Ideally, having assigned each of the 648 individuals to one of five scale types or rank groupings, we would like to see how these groupings relate to behavior in a new, nonverbal situation of role conflict. 

Such a design would be very costly and complicated but must be carried out sooner or later if we are to have full confidence that our scale is not an artifact - that it does not, for example, arise merely from differences in imaginative ability, a possibility which was suggested by Leonard S. Cottrell in his discussion of the first draft of this paper.  As a simple but decidedly inferior procedure, we investigated the relationship between the scale and other verbal responses relative to role conflict.  We selected some academic situations not too far removed from the experience of college students.  The problem was to see whether respondents who were near the particularistic end of the scale, for example, tended to have a higher probability of giving particularistic responses in these academic situations than other respondents. 

(The scale itself involved no academic situations.)

 

                Consider the following story:

 

                You are employed by Professor X to mark examination hooks in his course.  Your close friend makes somewhat under a passing grade.  If you give him a special break you can boost him over the passing line. He needs the grade badly.

 

What right does your friend have to expect you to give him a special break?

Check one:

[]   He has a definite right as a friend to expect me to do this for him.

[]   He has some right as a friend to expect me to do this for him.

[]   He has no right as a friend to expect me to do this for him.

 

Would you give him this special break in view of your obligations to the university and your obligation to your friend?

Check one:

[]   Yes.

[]   No.

 

                The same scoring system was used as in the scale items.  Among those with Scale Type 1, only 7 per cent responded particularistically in this situation, but the percentage rose to 49 among those in Scale Type 5:

 

                Scale type             Per cent 

 

5                              49

4                              25

3                              31

2                              30

1                              7

 

490 The Theory and Its Application

 

                Another situation presented was the following, scored similarly to the others:

 

                You are in charge of the reserve desk at a library.  A certain reserve hook is in heavy demand. A close friend is pressed for time and can only use the book at a certain hour.  He has suggested that you hide the book for a while before his arrival so that he will be sure to get it.  He needs it badly.

 

What right does your friend have to expect you to hide the book?

Check one:

[]   He has a definite right as a friend to expect me to hide the book for him.

[]   He has some right as a friend to expect me to do this for him.

[]   He has no right as a friend to expect me to do this for him.

 

Would you hide the book for him in view of your obligations to the library and your obligation to your friend?

Check one: 

[]   Yes.

[]   No.

 

                Variation in proportions responding particularistically was from 16 to 70 per cent: 

 

                Scale type             Per cent

 

                5                              70

                4                              55

                3                              58

                2                              46

                1                              16

 

                The following story, almost identical with that used in the paper published in 1949 in the American Sociological Review, also was presented and scored according to the methods used in the present paper.

 

                You are proctoring an examination in a middle-group course.  You are the only proctor in the room.  About half-way through the exam you see a fellow student, who is also your close friend, openly cheating.  He is copying his answers from previously prepared crib notes.  When he sees that you have seen the notes as you walked down the aisle and stopped near the seat, he whispers quietly to you, "O.K., I'm caught. That's all there is to it."

 

Under these circumstances, what right does he have to expect you not to turn him in?

Check one: 

[]   He has a definite right as a friend to expect me not turn him in.

[]   He has some right as a friend to expect me not to turn him in.

[]   He has no right as a friend to expect me not to turn him in.

 

Under these circumstances, what would you probably do in view of your obligations as a proctor and your obligation to your friend?

Check one: 

[]   Report him.

[]   Not report him.

 

An Empirical Study 491

 

                Variation was from 6 to 50 per cent, in proportions responding particularristically:

 

                Scale type             Per cent

                5                              50

                4                              35

                3                              28

                2                              25

                1                              6

 

                These items, like the items included in the scale, were asked in three alternate forms. 

A respondent, for example, who had the Smith-Johnson form of the scale items also had a Smith-Johnson form of the new academic items.  There was considerable variability in patterns of relationship, but the upward progression was present on all forms on each item, as is shown in Chart I. 

 

                80           8C           8C

4C

                Library    p    ~

                20           2C           Book       2C

                0              0              C

                12345    12345   12345

                SCALE  TYPE

 

                                When responden~ faces the dilemma

                                When respondent's friend faces the dilemma

     When '1Smith11 faces the dilemma

 

Chart I. SCALE SCORES AS RELATED TO THE PROPORTION "PARTICULARISTIC" IN CERTAIN

ACADEMIC SITUATIONS

 

                An important element of a friendship relationship is what Parsons calls an "other-orientation" rather than a self-orientation, such as is institutionalized in our society in a business transaction.  Though other-orientation is institutionalized, it is probably not an absolute value.  While the individual is supposed to subordinate his own interests to those of his friends under many circumstances, there are limits to the sacrifices which one may legitimately expect of a friend.  These limits tend to be vague and undefined, per-

haps so that they may vary with the intimacy of the friendship.  This introduces another source of behavioral variability: the respondent's perception of the risk to himself by defying universalistic norms and coming to

friend's aid. It was of interest, therefore, to vary the cheating situation asking the respondent to imagine an analogous setting with much greater to the proctor: 

 

492 The Theory and Its Application

 

                Consider the same cheating situation as above, with an additional element.  Suppose now there is another proctor (an extremely conscientious fellow!) in the examination room with you and that you would be running a fifty-fifty risk of personal exposure by him to the authorities for failing as proctor turn in a cheater. 

 

                The list to be checked was the same as before,

 

                How the increase in risk reduced the particularistic responses is shown Table 3. 

 

Table 3

 

PERCENTAGE "PARTICULARISTIC" WHEN RISK VARIES

 

                Scale      In both    In low-risk             In neither

                type         situationS              SituOtion only       SituOtion               Totol

                5              20           30           50           100

                4              16           19           65           100

                3              10           18           72           100

                2              11           14           75           100

                1              2              4              94           100

 

                We had hoped to make a further study of high and low risk to see differences in predispositions might be related to other factors in this specific cheating situation, such as students' perceptions of the severity of penalties, of fellow students' attitudes, and of the cheater's own probable reactions. 

Questions were designed on these points, but they were not satisfactory. - main problem which emerged, however, and which negated much further intensive cross-tabulation, was the sizable differences in response depending on whether we asked the cheating question involving little risk to the proctor

be/ore or alter the cheating question involving risk to the proctor.  Actually, in a random half of the cases the little-risk situation was presented first; in the other half the higher risk situation was presented first. 

 

                For each form (ego as proctor, ego's friend as proctor, Smith as proctor) we have, then, two reports.  There are six replications in all.  Results are graphed in Chart II.  The reader will observe that the form in which ego is proctor stands up well.  We get about the same picture, irrespective of the order of presentation of the low-risk and high-risk situations, respectively.  But the results are chaotic for the forms in which ego is the cheater or in which the actors are third persons. 

 

An Empirical Study 493

 

                The reasons for this result are not immediately obvious.  Further trials and study are required before reaching a conclusion.  One plausible suggestion is that a paper-and-pencil test like this requires a good deal of imagination on the part of a respondent and that the act of imagination is made easiest when ego himself is pictured as confronting the dilemma.  By increasing the salience, one reduces the temptation for casual or careless checking.  However, this speculation is inadequate to explain why, on the two aberrant forms, the prior presentation of the high-risk situation produced a higher particularistic response to the two items than the prior presentation of the low-risk situation.

 

Ego as Proctor

Friend as Proctor

 

 

SCALE SCORE

Low risk SitUa#I'on presen~ed first

Hi~h risk situation presented fIrst 

 

Chart II. SCALE SCORES AS RELATED TO THE PROPORTION "PARTICULARISTIC" IN THE

 

CHEATING SITUATIONS-SHOWING VARIATIONS RELATED TO DIFFERENT FORMS OF QUESTIONNAIRES

 

494 The Theory and Its Application

 

The systematic study of the extent to which identification, salience, ego defenses, and so forth, modify questionnaire responses is still in its infancy.  Hence, the superior results shown in Chart II on the form in which ego himself faced the dilemma should not tempt us to hasty conclusions.  After all, as Table 2 shows, all three forms yielded about the same pattern of distribution of scale types, and as Chart I shows, all three scales showed about the same general relationship in the specific academic situations, including the cheating situation.8  

 

                Our study suggests that it is possible to classify people according to a predisposition to select one or the other horn of a dilemma in role conflict.  As more studies are made - not only with pencil-and-paper tests, but also with role-playing in experimental and real-life situations and with other procedures

- information exceedingly important to social science can be derived.  We must anticipate the possibility, as Edward A. Suchman of Cornell has suggested in a letter to the writers, that tendencies of a respondent to adopt more stereotyped roles in hypothetical than in real life situations will complicate prediction. 

 

                Studies in this field will help sociologists in developing theories of institutionalization and social psychologists in developing theories of personality and, indeed, can serve as a crucial link between the two bodies of theory.  The importance of such a link, employing such variables as particularism-universalism, affectivity-affective neutrality, specificity-diffuseness, self-orientation~collechvity-orientation, has been in the forefront of the thinking of Talcott Parsons and his associates, who have been working on a new schema

looking toward unification of social science theory.  The immensity of the technical task involved in making such concepts amenable to measurement in the years of patient work which lie ahead is at least suggested by the experience of our present study. Indeed, one of the most important values of this paper should be its service as a brake on the enthusiasm of those who may anticipate quick and easy progress in moving from highly abstract concepts in social science to empirical operations. 

 

                Such studies as ours can also be applied in practical research if sustained effort is devoted to technical developments.  Leadership, for example, involves skill in the solution of role conflicts.  Classic examples are the foreman in industry or the noncommissioned officer in the army.  If such thoughtful ob'

servers as Chester I. Barnard are correct, skill in handling role conflicts is also an essential at the high executive levels.9  Eventually, we may have role-playing situational tests, involving nonverbal as well as verbal behavior, which will be useful in the selection and training of leaders.  The present study represents only a primitive effort to formulate some of the problems of definition and measurement. 

 

                8 In the high-risk cheating situation (not shown in Chart I), when the two sequences of presentation are combined there is also relatively little difference among the three forms, all showing a definite correlation with the scale types. 

 

                9 Chester I. Barnard, The Functions of the Executive (Cambridge: Harvard University Press, 1938) see especially chap. xvii. 

 

An Empirical Study 495

 

NOTE ON LAZARSFELD'S LATENT DISTANCE SCALE AS APPLIED TO ROLE CONFLICT DATA 

 

                In Measurement and Prediction, Chapter 11, pages 441-147, the reader will find a numerical example of a latent distance analysis carried out in full, on Research Branch data on psychoneurotic symptoms.  That analysis used only one computed parameter for each item.  In the present example on role

conflict data, Lazarsfeld, who kindly made the analysis, introduced more flexibility by computing two parameters for each item. T he latent structure is set up as follows: 

 

                Latent     Item        Item        Item        Item

                class       1              2              3              4

                  I             a1            a2            a3            a4

                  II            b1            a2           a1           a4

                 III            b1            b2           a3           a4

                 IV            b1            b2           b3           a4

                 V             b1            b2           b1           b4

 

Each value of a tends to be a large fraction and each value of b tends to be small.  (The example in Chapter 11 added the restriction that  ai  = 1 - bi, In the perfect Guttman scale each a = 1 and each b = 0.) 

The algebra and arithmetical routine involved will be presented by Lazarsfeld in a separate paper.  Final results, however, are shown here as Table 4, using as illustration, for reasons of space, only the form in which ego faces the dilemma.  In this table, the scale patterns are ordered as in Table 2 and do not follow precisely the rank order they would have in Lazarsfeld's schema.  The most serious discrepancy between the ordering indicated by the Lazarsfeld model and by the scalogram procedure of scoring to the nearest scale type is with respect to pattern + - - + based on only four cases (see Table 4). The Lazarsfeld

 

Table 4

 

ILLUSTRATIVE RESULTS OF FITTING LATENT DISTANCE STRUCTURE

 

(Data for form in which ego faced role conflict)

 

                Item        Per cent of each pattern in each latent class                 Totals

                1 2 3 4    1              II              III             IV             V              Total       Fitted Actual

 

                ++++       95.9        4.0          0.1                                          100         19.1        20

                + - ++      94.8        3.9          0.2          0.2          0.9          100         10.0        9

                ++ - +      91.7        3.7          3.0          0.3          1.3          100         6.5          6

                +++-        95.7        4.3          100         2.5          2

                + - + -      96.7        3.3          .               100         1.3          2

                ++ - -       92.3        5.1          2.6                                          100         0.8          1

                -+++        0.9          95.6        3.0          0.3          0.2          100         39.9        38

                -++ -        0.8          87.7        2.7          OA           8.4          100         5.7          7

                -- ++        0.9          88.9        2.7          4.8          2.7          100         22.2        24

                - - + -       0.3          34.8        1.2          1.7          62.0        100         7.4          6

                - - - +       0.3          25.3        19.9        34.7        19.8        100         25.4        23

                - + - +      0.5          52.3        41.2        3.8          2.2          100         23.7        25

                + - - +      86.2        3.6          2A           4.8          3.0          100         3.6          4

                - + - -       0.3          23.1        18.4        1.6          56.6        100         6.9          6

+ - - -       33.9        1.8                          1.8          62.5        100         1.2          1

                - - - -                        2.1          1.6          2.8          93.5        100         41.0        42

 

496         The Theory and Its Application

 

procedure would place this pattern within the top group.  By scalogram procedure, to assign this pattern to the top group would be to imply that repondents made two "errors," in both Items 2 and 3, which, indeed, may have been the case.  The present assignment implies only one error, on Item 1. The reader

will note that two-error patterns + - + - and ++ - -, with two cases and one case respectively, which could have been assigned variously by scalogram methods, belong, by the Lazarsfeld model, just where they have been put. 

 

                The picture presented in Table 4 is analogous to the picture presented in Measurement and Prediction, Chapter 11, Table 13, but it must be remembered that it has involved a more flexible basic design. 

 

                The last two columns of Table 4 show good agreement between the fitted and actual totals.   Approximately as good a fit was obtained with the other two forms of the questionnaire, and the rank ordering of the scale patterns on the basis of the percentage of a given pattern in each latent class is not markedly different.  Much further study is needed of the latent distance model used here, especially with respect to reliability of small frequencies and, as has been mentioned earlier, to the testing of acceptance standards.  The concept of a latent structure is theoretically quite appropriate to data of the type we are

likely to assemble in subsequent investigations of role, and of informal social norms generally.

Hosted by www.Geocities.ws

1