Trust-in-supervisor and perceived fairness and accuracy of performance evaluations

Trust-in-supervisor and perceived fairness and accuracy of performance evaluations

301 J BUSN RES 19x5:13:301-313 Trust-in-Supervisor and Perceived Fairness and Accuracy of Performance Evaluations Janet Fulk Arthur P. Brief Steve...

909KB Sizes 0 Downloads 23 Views

301

J BUSN RES 19x5:13:301-313

Trust-in-Supervisor and Perceived Fairness and Accuracy of Performance Evaluations Janet Fulk

Arthur P. Brief

Steve H. Barr

University of Southern California

New York University

Oklahoma State University

Landy, Barnes, and Murphy’s [12] five-variable model of the correlates of subordinate perceptions of the fairness and accuracy of their performance evaluations was cross-validated successfully for a sample of telecommunications engineers engaged in research and development activities. At the same time, a trimmed three variable path model consisting of two of Landy et al’s correlates (supervisor’s knowledge of subordinate performance, and development of action plans related to performance weaknesses) in combination with a measure of the psychological context (trust-in-supervisor) explained an even larger proportion of the variance in fairness perceptions for these engineers. It was suggested that perceptions of fairness and accuracy in performance evaluation may depend as heavily on the level of trust in the on-going superior-subordinate relationship as on characteristics of the performance appraisal process itself. According to Lawler [14], employee perceptions of the fairness and accuracy of their performance evaluations play a major role in the ultimate success of a performance appraisal system. The importance of this perceptual variable is often overlooked in studies of performance appraisal, which have tended to focus primarily on rating formats and instrumentation. Yet, there are several theoretical arguments for the importance of these perceptions. Equity theory [l] suggests, for example, that individuals who perceive inequitable treatment may alter their levels of input to the jobs, particularly when the appropriate adjustment is to decrease investments [3]. Also, fairness and accuracy perceptions are likely to influence goal acceptance, which Locke and Bryan [15] contend is a major condition for the success of goal-setting in energizing high levels of performance. Recently, Landy, Barnes, and Murphy [12] conducted an investigation directed toward identifying appraisal processes and characteristics which might account for fairness and accuracy perceptions. In a survey of managers in a manufacturing firm, they found that of eleven appraisal system characteristics investigated five were particularly important, accounting for approximately 29% of the variance in perceived fairness and accuracy:

An earlier version of this paper was presented at the 1981 annual conference Management. Address correspondence to Janet Fulk, Annenberg School of Communications, ern California, Los Angeles, CA 90089-0281. Journal of Business Research 13, 299-313 0 Elsevier Science Publishing Co., Inc., 1985 52 Vanderbilt Ave., New York, NY 10017

of the Academy University

of

of South-

0148.2963/85/$3.30

302

J BUSN RES 1985:13:301-313

J. Fulk et al.

existence of a formal evaluation program, performance evaluation frequency of at least once per year, supervisor’s knowledge of evaluatee’s performance level, evaluatee opportunity to express feelings when evaluated, and development of action plans in relation to performance weaknesses. The study reported here provided a cross-validation and a constructive replication of Landy et al.‘s research. Cross-validation was accomplished by applying Landy et al.‘s parameter estimates to data gathered from a markedly different sample. The constructive replication involved development and testing of a refinement of Landy et al’s model. The refined mode1 differs from Landy et al.‘s original model in two ways. First, the revised version incorporates possible relationships among the five evaluation process variables. Thus, it includes indirect as well as direct effects on fairness and accuracy perceptions. Second, it includes an additional variable (not originally studied in Landy et al.) which reflects the quality of the overall relationship between evaluator and evaluatee. This refined conceptualization is shown in Figure 1. It includes the five direct paths proposed by Landy et al. as well as the additional relationships we propose. Existence of a Formal Evaluation Program (1) logically precedes the three variables describing characteristics of that program: Frequency of Evaluation (3), Development of Action Plans Related to Performance Weaknesses (4), and Opportunity to Express Feelings (6). The latter is also shown to be at least partly dependent on the Frequency of Evaluation (3), on the expectation that frequent evaluation provides more forums for interaction, feedback, and expression of feelings. Finally, a high level of Supervisor Knowledge of Subordinate’s Performance (2) is seen to facilitate the Development of Action Plans Related to Performance Weaknesses (4). Constructive developmental activity by the supervisor in weaker performance areas presupposes that the supervisor is aware of and understands the subordinate’s current performance deficiences. Figure 1 also incorporates the variable Trust in Supervisor (5) as an important factor in perceived fairness and accuracy. As used here, trust in supervisor primarily reflects respondent perceptions of being able to communicate openly with the supervisor on job-related problems without fear of negative repercussions. Thus, trust is taken to be a relatively enduring characteristic of the superior-subordinate relationship; over time, trust is built as superior and subordinate interact on job related issues. Although previous performance appraisal experience may have contributed to the level of trust reported currently, such experience is only one of a great many interactions and experiences which are contributory factors. Trust is therefore taken to be a general relational characteristic rather than primarily the product of any one type of interaction. As a general characteristic, it is a salient element in the psychological context of superior-subordinate interactions. Recent reviews described trust as an important factor influencing vertical dyadic perceptions of communication and interaction [cf. 91. Roberts and O’Reilly [22], for example, found that trust in supervisor was positively related to perceived accuracy of communication, satisfaction with communication, and desire for interaction. Perceptions related to the sensitive interactions involved in the provision of performance feedback are likely to be particularly dependent on the level of trust in the psychological environment [7]. Ilgen, Fisher and Taylor [8] describe trust as an important factor influencing acceptance of feedback, which they define in a fashion highly similar to perceived fairness and accuracy of performance appraisal

r

fairness

in

Figure 1. A model of perceived

(2)

and accuracy.

(5)

Trust

(4)

performance

Plans related

supervisor

W

of 11

1

to

1

Lli-’

evaluation

Frequency

knowledge

I_)

I

Supervisor’s

IiA

(1)

program?

Is there a

I

A

(7)

1

A

fairness and accuracy

Perceived

r

1

304

J BUSN RES 19x5:13:301-313

J. Fulk et al.

as “the recipient’s belief that the feedback is an accurate portrayal of his or her performance” (p. 356). As used here trust in supervisor also shares some commonalities with Schein and Bennis’ [25] notion of psychological safety, which Nadler [19] describes as particularly important when feedback disconfirms the recipient’s existing beliefs. Thus, research reports in a variety of areas suggests that trust is an important influence on fairness and accuracy perceptions in performance appraisal. We also propose an indirect relationship of Supervisor Knowledge of Subordinate’s Performance (2) to Perceived Fairness and Accuracy (7) through Trust in Supervisor (5). Since trust implies a supervisor role as helper in the performance evaluation situation [7], it is assumed that greater supervisor knowledge of subordinate performance provides a stronger basis for a constructive helping role. While the direct contribution of supervisor knowledge to trust has not been investigated in the literature, it appears reasonable to hypothesize that supervisor knowledge of performance generates subordinate confidence in the supervisor as rater of the performance. A final path in Figure 1 shows a direct positive relationship of Trust in Supervisor (5) to subordinate perceptions of Opportunity to Express Feelings (6). Research by Mellinger [16] indicates that distrust of an individual in a work situation leads to concealment of feelings when communicating about work-related issues. Similarly, Roberts and O’Reilly [22] and O’Reilly and Roberts [20] found distrust in supervisor to be related to withholding or distortion of information. In summary, this refined path mode1 is an extension of Landy et al.‘s research in two ways. First, we examine more complex relationships among the five process variables and perceived fairness and accuracy. Second, trust in supervisor has been added to the model to reflect an aspect of the psychological context of the interaction.

Method Sample and Procedure The sample consists of 198 research and development engineers in the telecommunications division of a large electronics firm. Organizational policy dictated that each engineer be evaluated by his or her immediate superior annually on a global performance scale; this contrasts with the sample studied by Landy et al. [12] which was comprised of exempt managerial and professional employees of a manufacturing organization who were operating under an existing management by objectives system. All but two of the respondents were male. The average age was 37 years and the average length of service with the company was 10 years. Educational levels ranged from bachelor’s degree through Ph.D. The engineers were involved in largely applied research and worked in 5-6 person teams. Team leaders were not included in the sample, and none of the respondents had subordinates reporting to them. The questionnaire was identified as a confidential university survey pertaining to engineers and scientists, and anonymity was guaranteed. The questionnaires were administered individually through the company mail and each respondent sealed and returned his/her own questionnaire in a self-addressed envelope directly to the university. The 198 responses represent an 81% response rate.

Perceived Fairness and Accuracy

J BUSN RES 1985:13:301-313

305

Measures The five performance evaluation process variables and the perceived fairness and accuracy variable were assessed using the measures employed by Landy et al. [12]. Trust in supervisor was measured by Roberts and O’Reilly’s [22] three-item scale. This scale has been shown to have acceptable internal consistency reliability in more than ten samples (e.g., [18]). Coefficient alpha for the current sample was 0.71.

Analyses Landy et al. derived beta weights for the regression of fairness and accuracy on the five performance evaluation process variables. Following Kerlinger and Pedhazur [lo], the cross-validation was conducted by applying these weights to the current data and calculating an estimated multiple correlation. This was compared to the multiple correlation obtained by Landy et al. The refined model of Figure 1 was tested by the method of path analysis [4]. In this method, the relationships expressed in the path diagram are translated into a series of regression-type equations which specify the hypothesized relationships among the variables. Using these equations, ordinary least-squares regression procedures are carried out; the standardized partial regression coefficients derived are the obtained path coefficients. These coefficients are tested for significance using normal procedures for testing regression weights. Following this procedure, the “theory trimming” method of Heise [6] was applied. This method involves identification and deletion of nonsignificant paths-in effect creating a new model which is more consistent with the data (because hypothesized relationships which were not supported by the data have been eliminated). Path coefficients are re-calculated (as above) for this new “trimmed” model and tested for significance. Finally, as recommended by Kerlinger and Pedhazur [lo], the degree of consistency of the (trimmed) model with the data was assessed by attempting to reproduce the original correlation matrix by standard operations upon the coefficients of the (trimmed) path model. The original and reproduced correlations were then examined and their degree of similarity assessed.

A Note on Methodological Limitations Sheridan and Vredenburgh [27] note that “path analysis of cross-sectional data provides weaker inferences of causality since the interpretation of causal effects is induced only from relationships posited by theory” (p. 8). While the research and theory discussed earlier suggests the relationships expressed in the models tested here and by Landy et al., inferences must be made with caution. To use Landy et al.‘s terminology, the factors studied here must be considered correlates of perceived fairness and accuracy. An accumulation of findings using a variety of methods including some designed to specifically address issues of causality are required before strong causal conclusions can be drawn. A second issue in path analysis is the potential effect of unmeasured variables on the analysis. Certainly we do not propose to have included in our analyses all

306

.I BUSN

RES 198S:13:301-313

Table 1. Original Model*

J. Fulk et al.

Variable

Intercorrelations

and Reproduced

Correlations

Variables Deleted from Trimmed

Variables Trimmed 4

Model Variable

1.

Is there a program?

3. 6.

Frequency of evaluation Opportunity to express feelings Supervisor’s knowledge Plans related to performance Trust in supervisor Perceived fairness and accuracy

2. 4. 5. 7.

I

3

6

2

for Trimmed

Retained Model 5

in 7

0.21*

-

0.20* 0.13 0.12 0.18*

0.26* 0.07 0.16” 0.01

0.33* 0.25 0.44*

0.36* 0.39*

0.36* 0.20

0.39 0.14 -

0.52* 0.45* 0.47*

0.19*

0.12

0.34’

0.51*

0.44’

0.47*

-

“Original correlations shown below the diagonal;

reproduced

correlations

for the tnmmed

model shown above the

diagonal. *p < 0.05.

potentially relevant factors in the rating system and context. A brief glance at Landy and Farr’s [ 131 extensive review of appraisal-relevant variables indicates the impossibility of including all pertinent variables within a single study. Thus, we cannot rule out the possibility of effects from omitted variables. Recently, for example, Berger and Gerhart [2] reported finding that two contextual variables had indirect but not direct effects on perceived accuracy: rating level and performance-contingent pay. Landy, Barnes-Farrel, and Cleveland [ll] found no direct relationship of rating level to perceptions of fairness. Thus, indirect effects of these two variables (through, for example, endogenous variables such as trust in supervisor) should probably be investigated in future studies.

Results Cross-validation The zero order correlation matrix appears in Table 1. In the Landy et al. study, the regression of the five process variables on perceived fairness and accuracy produced multiple correlations of 0.54 and 0.51 in the derivation and hold-out samples, respectively. (Neither R changed when corrected by a shrinkage formula.) In the non-MB0 sample used in the current study, the application of Landy et al.‘s weights produced an estimated multiple correlation of .56 (and a shrunken R of 0.54). These estimates compare very favorably to Landy et al.‘s, and thus it must be concluded that the model cross-validated well.

Refined Model The path analytic results are reported in Figure 2. As expected, a significant path joined frequency of evaluation with opportunity to express feelings (P,, = 0.25, p < 0.05). However, the path from the latter to perceived fairness and accuracy

Perceived

Fairness

and Accuracy

J BUSN RES 1985:13:301-313

307

only one significant path-linked was not significant (P76 = 0.06, ns). In addition, existence of a program to a program characteristic-frequency of evaluation in this case the suc(PSI = 0.21, p < .05; P,, = 0.05, ns; P,, = 0.05, ns)-and ceeding path to perceived fairness and accuracy was nonsignificant (P,, = 0.02, ns). Clearly, these results do not support the expected logical ordering among existence of a program and the three focal program characteristics. More promising results were obtained for the remaining hypothesized relationships. The paths from supervisor knowledge of subordinate performance to development of action plans and from there to perceived fairness and accuracy were paths significant (Pd2 = 0.35,p < 0.05; P,4 = 0.026, p < 0.05). The hypothesized among supervisor knowledge, opportunity to express feelings, trust in supervisor, and perceived fairness and accuracy were all significant (P,, = 0.39, p- < 0.05; Ph5 = 0.36,~ < 0.05; P,5 = 0.28,~ < 0.05; P,, =. 0.28, p < 0.05). Thus, overall, the complete refined model received some support, and this suggests the need for the trimmed model presented below.

Trimmed Model The results for the trimmed model are reported in Figure 3. Comparisons of the magnitudes of the path coefficients to those in Figure 2 shows little change, indicating that the trimming procedure did not eliminate any substantial paths. Using only three predictors, this trimmed model obtained an R of 0.65, explaining 42% of the variance in perceived fairness and accuracy. For the reproduction of the correlation matrix, Table 1 shows the original correlations below the diagonal and the reproduced correlations above the diagonal. With one exception, the correlations reproduced almost perfectly. For the trust in supervisor/plans related to performance relationship, the reproduced correlation was a nonsignificant 0.14. The original correlation was 0.20 (p < 0.05)-a difference of .06. According to Kerlinger and Pedhazur’s [lo] rule of thumb, this difference may be considered marginal. However, the higher original correlation may have been spurious since both variables are specified as having a common cause (cf, [lo]). When this common cause (supervisor knowledge) is partialled out, the original correlation drops to a nonsignificant 0.07. Thus, the data appear to support the trimmed model on an overall basis.

Discussion Landy, Barnes, and Murphy’s [12] model of the correlations of an individual’s perceptions of the fairness and accuracy of his or her performance evaluation was cross-validated successfully using a similar data collection procedure in a different sample. The 31% of variance in fairness and accuracy perceptions which was explained by the model in this organization compares favorably with the 29% and 26% of variance explained by Landy et al. Of course, both studies are limited in that they used same-source survey data collected at a single point in time. Also, a more complex model depicting possible relationships among Landy et al.‘s hypothesized correlates and incorporating trust in supervisor as a salient variable received partial support. This led to the development of a less ambitious but more parsimonious “trimmed” model which includes only three predictors: 1)

I

Is there

a

P&

Pm

= 0.35”

P‘J1 = 0.05

P6,

PSI

= 0.28”

= 0.08

= 0.21

Figure 2. Path analytic results (*p
knowledge

Supervisor’s

(I)

program?

*

W related (4)

performance

Plans

evaluation

of

to

Ph3 = 0.25

= 0.06

Frequency

P,,

1

*

(6)

P,,

= 0.28”

= 0.02

to

feelings

Opportunity express

P,4 =0.26*

W

W

Pn

I

J BUSN RES 1985:13:301-313

Perceived Fairness and Accuracy

P2r = 0.36”

Plans related to

309

Pa2 = 0.28”

performance

ti

(4)

Supervisor’s

PQ1

Perceived fairness

= 0.30” *

knowledge

and accuracy (7)

(2)

A

P3,

Trust in

= 0.39”

Pa3 = 0.30

*

supervisor

*

(5)

I Figure 3. A trimmed model of perceived

I fairness

and accuracy

(*p < 0.05).

supervisor’s knowledge of subordinate performance, 2) development of action plans which relate to performance weaknesses (both from Landy et al.‘s work), and 3) subordinate trust in supervisor (a new inclusion in the model to reflect the character of the superior-subordinate relationship). The trimmed model explained 42% of the variance in fairness and accuracy perceptions. Both variables retained from Landy et al. might be viewed as reflecting the degree to which the superior/subordinate dyad was willing to implement performance appraisal in a sincere fashion. Supervisor’s knowledge suggests the degree to which the supervisor is willing to obtain the appropriate information required to provide informed feedback-and to communicate that sincerity to the subordinate. Similarly, development of action plans related to performance weaknesses suggests the degree to which performance appraisal was conducted with a view toward creating constructive, relevant outputs. At the same time, the three variables which were cross-validated in the five-variable model but trimmed from the refined model (existence of a program, frequency of evaluation, and opportunity to express feelings) less clearly imply a constructive orientation to the process. The former two, in particular, appear to access formal system characteristics and do not appear to tap the processes which characterized the actual dyadic interaction in the appraisal.

310

J BUSN

J. Fulk et al.

RES

1985:13:301-313

If these results are replicated in future studies, this might suggest that these particular system features may be relevant but nevertheless insufficient conditions for perceptions of fair and accurate performance appraisals. The revised model suggests that subordinate perceptions of the supervisor’s knowledge, in addition to being directly related to perceived fairness and accuracy, acts indirectly through plans related to performance and trust in supervisor. Since supervisor knowledge seems to have many points of influence, it is important to rule out alternatives to the explanations posited here. One possibility is that subordinate perceptions of supervisor knowledge may be a surrogate for a high performance rating (i.e., a subordinate may report that the supervisor is knowledgeable of his or her performance because the supervisor assigned him or her a high rating). This would be particularly troublesome if fairness and accuracy perceptions were a function of favorability of rating. However, Landy, Barnes-Farrel, and Cleveland [ll] found that “variance in the perceptions of fairness of evaluation could not be a possible confounding of suexplained by rating level” (p. 356), and therefore pervisor’s knowledge with rating favorability probably is not a tenable alternative interpretation. It is possible that the trust variable was particularly salient here due to the nature of the sample. R & D engineers and scientists tend to perform a greater proportion of less-well-structured activities, making it difficult to develop and apply a preponderance of clear-cut, objective performance standards [21]. Because a greater proportion is not objectively verifiable in the short term, more reliance must be placed on the supervisor’s subjective judgments. This may inflate the importance of the subordinate’s confidence and trust in the supervisor. Whether the structure of the job and the specificity of its performance standards affect the importance of the trust variable for fairness and accuracy perceptions can only be determined by research directed specifically toward this issue. Nevertheless, it remains a plausible partial reason for the role of the trust variable found here.’

Conclusion

and

Suggestions

for Future

Research

Given the results for trust, it is probably worthwhile to investigate other characteristics of the supervisor-subordinate dyad which form the context for the performance evaluation process. For example, one excellent possibility lies in the concept of homophily-the similarity of attributes between the source and receiver of a communication. Communication has been shown to be generally more effective when the source-receiver pair is optimally homophilous; heterophilous communication has been shown to lead to dysfunctional consequences, such as message distortion and cognitive dissonance [23]. Although “the number of elements of similarity-dissimilarity between two people approaches infinity” [ 17, p. 3241, some

‘To

investigate dependent formed

the possibility variables.

a separate

the complete similar were

that

the

results

analysis

except

was rerun

reproduced

trust scale).

within

proportion

using a two-item

Given

in the interests

Kerlinger

variance

the similarity

of consistency

trust

and

other

in which

criterion

and accuracy the findings studies

conceptual

which

overlap

axis with

trust scale (dropping

Pedhazur’s

of results,

reflect

using principal items also loaded

were no changes

in fairness with

might

analyzed

that one of the three

to those for the full scale. There

substantial

for

the data set was factor

factor.

the complex

paths would

of

perceptions

.05.

and

between

varimax

on the fairness

item).

be retained.

the

trimmed

(R of 0.61 compared

for the complete

three-item

employ

scale.

the trust

independent

rotation. factor.

The

Given

The

results

all original model

still

trust

and items

these results, were

highly

correlations explained

a

to .65 for the three-item

trust scale are presented

here

Perceived Fairness and Accuracy

J BUSN RES 1985:13:301-313

311

initial suggestions for investigation might include such factors as background and training, sex, race, age, cognitive style, expressive vs. instrumental orientations toward work [28], degree of interpersonal orientation in task activities, preferences for working with people vs. things, commitment to the oganization, and linguistic style. At the least, similarity on such factors could facilitate rapport and hence, the potential for openness to feedback. Also, as Festinger ([5], quoted in Salancik and Pfeffer, [24], p. 228) has noted “people evaluate information sources in terms of personal relevance, using similar others for comparison: the more similar someone is, the more relevant his or her views for understanding one’s own world.” The rationale for inclusion of the trust variable was based partially on research on communication and feedback. It is worthwhile to investigate fairness and accuracy perceptions in direct relation to feedback and communication variables under different conditions of trust. This combined approach has the advantage of tapping both context and process-since communication/feedback variables may be seen as internal to the performance appraisal process, while trust is a more enduring characteristic and therefore is partially external to the process. Also, since good communication and feedback skills are neither common nor easy to practice in high stress situations [19]-as performance appraisal can be [7]-contextual conditions such as trust may partially compensate for these limitations. Research which included both sets of variables would allow testing for the possibility of either a compensatory or interactive effect. Landy and Farr’s [13] review describes a host of contextual factors presumed to affect the validity of ratings. A few possibilities which seem particularly worthy of investigation in relation to perceived fairness and accuracy include: subordinate self-esteem, types of information required on a rating form, opportunity for subordinate to share a self-evaluation, degree to which the performance evaluation is linked to important organizational rewards, and level of secrecy surrounding the evaluation. In addition, recent work by Saiancik and Pfeffer [24] on social cues and Schriesheim [26] on work group factors-both in relation to dyadic supervisorsubordinate interactions-suggests that social factors also may exert influence on the interpretations which individuals place on cues from the supervisor. Given the potentially important role of supervisor knowledge of subordinate performance, future research ought to be directed at exploring further its nature and determinants. For example, Landy and Farr [13] argue that “relevance rather than frequency of contact [between rater and ratee] appear to be the critical factor [in determining rater knowledge of ratee]” (p. 77). A highly useful approach to future research might involve the use of a behavioral criterion in addition to fairness perceptions. This criterion might be grievances or appeals filed regarding the appraisal. Also, as noted earlier, future research of a longitudinal or experimental nature would assist in developing sound causal inferences. In conclusion, it is important to reiterate that while the three process variables trimmed from the original model were not strong correlates of perceived fairness and accuracy in this study when considered in combination with trust, this does not imply that such factors were unimportant to the total performance appraisal system. They may have been important for other criteria of interest in performance evaluation (e.g., validity of written ratings by the supervisor). This study focused only on fairness and accuracy perceptions. While such perceptions are important

312

.I BUSN

RES lY85:13:301-313

J. Fulk et al.

factors in system success [14], the trimmed variables should be investigated in relation to other performance evaluation outcomes as well. Similarly, the role of trust for other facets of performance appraisal should be investigated. This paper has benefited from the comments of Chester Schriesheim, Mary Ann Edward E. Lawler, III. and two anonymous reviewers to whom we owe our thanks.

Von

Glinow,

References 1 Adams, J. S. Toward an understanding Psychology 67: 422-436 ( 1963). 2.

Journal of Abnormal

of inequity.

Berger, C. J., Gerhart, B. A. Direct and indirect effects outcomes on ratee perceptions of performance appraisal Indiana: Institute for Research in the Behavioral, Economic, Purdue University, 1981.

and Social

of rating level and reward accuracy. West Lafayette, and Management Sciences,

3. Carrell,

M. R., Dittrich. J. E. Equity theory: The recent literature, methodological considerations, and new directions. Acudemy of Management Review 3: 202-210 (1978).

Duncan. 0. D. Path analysis: 72: l-16 (1966).

5.

Festinger, (1954).

6.

Heise, D. R. Problems in path analysis and causal inference. Sociological methodology. San Francisco: Jossey-Bass, 1969.

7. Huse,

April, 8.

L. A theory

Sociological

American Journul of Sociology

4.

of social comparison

E. F. Performance 3-5, 16-18 (1967).

appraisal-A

examples. processes.

new look.

Human Relations 7, 114-140 In E. F. Borgata

Personnel Administrution,

(Ed.), March-

Ilgen, D. R., Fisher, C. D.. Taylor, M. S. Consequences of individual feedback behavior in organizations. Journal of Applied Psychology 64: 349-371 (1979).

9. Jablin,

F. M. Superior-subordinate Bulletin 86: 1201-1222 (1979).

communication:

on

The state of the art. Psychological

E. J. Multiple regression in behavioral research. New York: 1973.

10.

Kerlinger, F. N., Pedhazur, Holt, Rinehart & Winston,

11.

Landy, F. J., Barnes-Farrell, J., Cleveland, J. N. Perceived fairness and accuracy of performance evaluation: A follow-up. Journal of Applied Psychology 65: 355-356 (1980).

12.

Landy, F. J., Barnes, J. L., Murphy, K. R. Correlates of perceived fairness and accuracy of performance evaluation. Journal of Applied Psychology 6: 751-754 (1978). Landy,

14.

Lawler, E. E. The multi-trait multi-rater approach to measuring performance. Journal of Applied Psychology 51: 369-381 (1967).

15.

Locke, E. A., Bryan, J. F. The directing function of goals in task Organizutional Behavior and Human Performance 4: 35-42 (1969).

16.

Mellinger, G. D. Interpersonal trust as a factor in communication. Social Psychology 52: 304-309 (1956).

17.

McCroskey. J. C.. Richmond, V. P., Daly, J. A. The development of a measure of perceived homophily in interpersonal communication. Human Communication Research 1: 323-332 (1975).

18.

Muchinsky. organizational 188 (1972).

19. Nadler,

Reading,

F. J.. Fat-r, J. L. Performance

rating.

Psychological Bulletin 87: 72-107 (1980).

13.

P. M. An intraorganizational communication questionnaire.

managerial

job

performance.

Journal of Abnormal

analysis of the Roberts and O’Reilly Journul of Applied Psychology 62: 1X4-

D. A. Feedback and organization development: MA: Addison-Wesley, 1977.

Using data-bused methods.

Perceived

Fairness

J BUSN RES 1985:13:301-313

and Accuracy

C. A., Roberts, K. H. Information filtration in organizations. 20. O’Reilly, experiments. Organizational Behavior and Human Performance 11: 253-265

313

Three (1974).

21. Pelz, D. C., Andrews, F. M. Scientists in organizations: Productive climates for research and development (Revised ed.). Ann Arbor, MI: Institute for Social Research, 1976. 22. Roberts, E. M., O’Reilly, C. A. Measuring Applied Psychology 59: 321-326 (1974). 23. Rogers, E. M., Agarwala-Rogers, Press, 1976.

organizational

communication.

Journal of

R. Communication in organizations. New York: Free

24. Salancik, G. R., Pfeffer, J. A social information processing approach and task design. Administrative Science Quarterly 23: 224-235 (1978).

to job attitudes

25. Schein, E. H., Bennis, W. G. Personal and oranizational change through group methods: The laboratory approach. New York: Wiley, 1965. 26. Schriesheim, J. F. The social context of leader-subordinate relations: An investigation of the effects of group cohesiveness. Journal of Applied Psychology 65: 183-194 (1980). 27. Sheridan, J. E., Vredenburgh, D. J. Structural model of leadership organization. Academy of Management Journal 22: 6-21 (1979).

influence

in a hospital

28. Strauss, G. Job satisfaction, motivation and job redesign. In G. Strauss, R. E. Miles, C. C. Snow, A. S. Tannenbaum (Eds.), Organizational behavior: Research and issues. Belmont, CA: Wadsworth, 1976, pp. 19-49.