ORGANIZATIONAL BEHAVIORAND HUMAN PERFORMANCE25, 79--96 (1980)
The Behavioral and Affective Consequences of Performance-Relevant Situational Variables LAWRENCE H . PETERS, E D W A R D J. O ' C O N N O R , AND CATHY J. R U D O L F
The University o f Texas at Dallas Results of two studies are reported. Four of the situational variables identified as relevant to performance settings in Study I were experimentally manipulated in order to create facilitating and inhibiting treatment conditions in Study II. Results indicated that performance, affective responses, and the r e l a t i o n s h i p b e t w e e n t h e s e two variables t e n d e d to be a f f e c t e d by performance-relevant situational characteristics. Implications for this neglected research area are discussed.
Situational variables have long been recognized as important determinants of both the behavioral and affective responses of persons in task settings. Much of the current empirical research in this area, however, is heavily guided by the assumption that situational variables primarily affect behavior indirectly through their direct impact on motivation (cf. job design research). Several conceptual models, however, have recognized the importance of situational variables as partial, but direct, determinants of performance. For example, in their excellent review of research on motivation, Campbell and Pritchard (1976) listed seven hypothesized partial determinants of performance, one of which was "facilitating and inhibiting conditions not under the control of the individual" (p. 65). In like manner, Campbell, Dunnette, Lawler, and Weick (1970), Dachler and Mobley (1973), Schneider (1978), and Terborg (1977) all have proposed conceptual performance models which take into account the notion that characteristics of work situations directly affect performance. UnfortunatelY, the hypothesized direct impact of facilitative/inhibiting situational variables on performance has not, as yet, been tested. This is surprising since other hypothesized partial determinants of performance (e.g., ability, motivation) are expected to relate to performance only in work settings in which the variance in performance is not restricted due to the nature of such situational variables (cf. Schneider, 1978; Terborg, 1977). This research was supported by an Organized Research Award, A6495-24, University of Texas at Dallas, awarded to the first two authors. Portions of these data were presented at the Annual Meetings of the Academy of Management, Atlanta, GA, 1979. Requests for reprints should be addressed to the first author, School of Management and Administration, The University of Texas at Dallas, Richardson, TX 75080. 79
0030-5073/80/010032-14502.00/0 Copyright~) 1980by AcademicPress, Inc. All rights of reproductionin any formreserved.
80
PETERS, O'CONNOR, AND RUDOLF
The ultimate purpose of the present research was to test the assumed impact of facilitating/inhibiting situational characteristics on both affective and behavioral work outcomes. However, before this could be done, it was necessary to first define the domain of potentially relevent situational characteristics believed to have a direct effect on work outcomes. As Schneider (1978) recently pointed out, not all situational variables are relevant for explaining the variance in particular work outcomes. Two studies which address these issues are reported below. The first study attempted to define the domain of situational variables directly relevant to performance. The second study, using a subset of those situational variables identified during the first investigation, tested hypotheses derived from conceptual performance models regarding the impact of these situational variables on work outcomes. STUDY I While several taxonomies of situations do exist (e.g., Fleishman, 1972), none were developed explicitly to define the dimension space of facilitating and inhibiting situational variables. Schneider (1978) recently suggested that an appropriate strategy for defining such a dimension space might involve starting with the behavior of interest and working "backwards" in order to identify situational variables relevant to that particular behavior. The present study followed this approach in defining a situational dimension space of potential relevance to performance settings. Method
Subjects and Procedures Participants were 62 full-time employees from the Dallas business community who were also evening students in the School of Management. They ranged in age from 21 to 63, with an average age of approximately 33 years. There were 39 males and 23 females. These individuals were employed on a wide variety of both managerial and nonmanagerial jobs. Participants provided information about their jobs by responding to an open-ended questionnaire. On this questionnaire, each participant described, in detail, a particular instance on their job in which a specific situational condition negatively affected their performance. In addition, they described their affective responses to this critical incident. Thus, starting with a particular instance of poor performance on their jobs, all participants worked backwards to identify particular situational conditions which they believed explained their poor performance at work. Results and Discussion Responses to the open-ended questionnaire were summarized on 3 x 5 cards and then sorted into categories based on similarity of content by the first and second authors, independently. While completing this sorting
PERFORMANCE-RELEVANT SITUATIONAL VARIABLES
81
process, each of these individuals attempted to identify the situational events which negatively affected performance, rather than the agents who were responsible for the presence of those events. Thus, responses such as " M y supervisor didn't give me the information I needed on time" or " M y co-worker gave me the wrong information" were placed in a category which reflected a lack of "job-related information" since it was the lack of that timely or appropriate information which seemed to have resuited in low performance. This classification system, therefore, focused on specific, performance-relevant situational variables which might be mediated by any of a variety of persons. The result of these two sortings yielded similar eight category dimension spaces with few disagreements which were easily resolved. The critical incidents provided by the 62 participants were then independently sorted into this derived dimension space by six advanced graduate students. The items sorted back into their appropriate categories 89% of the time. The percentage of accurate retranslations ranged from 83 to 100% across the eight categories. As such, the retranslation verified the usefulness of this situational taxonomy in capturing the items on which it was derived. Each of the eight categories, with its corresponding definition, is given in Table 1. As shown in the table, each of the identified categories represented a needed input or resource variable which may either inhibit or facilitate successful task performance. The diversity of situational resource variables within this derived set reflects the variety of jobs, both managerial and nonmanagerial, under investigation. While it is clear that some situational resource variables might be applicable only to managerial jobs (e.g., budgetary support) or to nonmanagerial jobs (e.g., tools and equipment), the particular configuration of situational variables relevant to any particular job would be expected to be situation specific. Participants in Study I consistently attributed poor performance to one or more of the situational resource variables identified. More specifically, poor performance was consistently attributed to either (1) the needed resource being inaccessible (e.g., the information needed to do my job is not always available), (2) not receiving enough of the needed resource (e.g., I never have enough supplies to do my job), or (3) receiving a needed resource which was of poor quality (e.g., my equipment requires frequent repair). Thus, for any situational resource variable, it would appear that task settings may vary along three dimensions (accessibility, quantity, and quality) which together define the extent to which performance will be facilitated or inhibited. While it might be argued that other organizational variables, such as supervision or co-worker relationships, also may inhibit or facilitate successful performance, data from this initial investigation suggest that the
82
PETERS, O'CONNOR, AND RUDOLF TABLE 1 SITUATIONAL RESOURCE VARIABLES RELEVANT TO PERFORMANCE
I. Job-related information. Refers to the information (from supervisors, peers, subordi-
nates, customers, company rules, policies, and procedures, and so forth) needed to do the job assigned. 2. Tools and equipment. Refers to those specific tools, equipment, and machinery needed to do the job assigned. 3. Materials and supplies. Refers to those materials and supplies needed to do the job assigned. 4. Budgetary support. Refers to the financial resources and budgetary support needed to do the job assigned--the monetary resources needed to accomplish aspects of the job to include such things as long distance calls, travel, job-related entertainment, hiring new and maintaining/retaining existing personnel, hiring emergency help, and so forth. This category does not refer to an incumbent's own salary, but rather, to the monetary support necessary to accomplish tasks which are a part of the job they have been assigned. 5. Required services and help from others. Refers to the services and help from others needed to do the job assigned. 6. Task preparation. Refers to the previous personal preparation, through previous education, formal company training, and relevant job experience, needed to do the job assigned. 7. Time availability. Refers to the availability of the time taking into consideration both the time limits imposed and the interruptions, unnecessary meetings, non-jobrelated distractions, and so forth, needed to do the job assigned. 8. Work environment. Refers to the physical aspects of the immediate work environment which are needed to do the job assigned---characteristics which facilitate, rather than interfere with doing the job assigned. For example, a helpful work environment is one that is not too noisy, too cold, or too hot; that provides an appropriate work area; that is well lighted; that is safe; and so forth.
effects of such variables may be reflected in and assessed at a more task-specific level by the current situational taxonomy. That is, for example, supervision might be considered poor because the supervisor withholds information, fails to order needed supplies, or assigns tasks for which subordinates are unprepared. Responses from several participants during this investigation were consistent with this position. In effect, broad concerns such as "poor supervision" might, in the future, be made more explicit by use of the proposed situational taxonomy as an aid in identifying those specific and direct causes of poor performance (see also, Wexley & Yukl, 1977, p. 165). It also should be noted that respondents in Study I were asked to describe aspects of their work setting which directly affected their performance. By contrast, variables in other situational taxonomies have been identified which are believed to primarily affect performance, indi-
PERFORMANCE-RELEVANT SITUATIONAL VARIABLES
83
rectly, through motivation. Current work in job design (cf. Hackman & Oldham, 1975) and organizational reward systems (cf. Lawler, 1977) involve two such situational taxonomies which are specifically relevant to explaining motivation. The adequacy of taxonomies which are aimed at identifying situational variables underlying motivation depends upon explanations concerning why particular job characteristics are valued by persons or how such characteristics influence their choices regarding effort expenditures. The current results, however, because they focus directly on performance rather than on motivation, reflect a taxonomy of work settings which does not fully depend upon assumptions about the nature of man or the process by which persons choose effort expenditure levels. Rather, they reflect aspects of work situations which, when severe, should have an immediate negative constraining influence on performance by making it difficult for individuals to adequately do their jobs regardless of their personal characteristics (e.g., growth need strength) or belief states (e.g., expectancy beliefs). In addition, however, it is recognized that such situational constraints on performance may also tend to reduce the motivation levels of workers, over time, by lowering their expectancy beliefs that effort will lead to successful performance (Peters & O'Connor, in press). In all instances, the 62 participants expressed negative affective reactions to the inhibiting work settings they described. The most common of these reactions were frustration and dissatisfaction. Several persons indicated that these feelings were so strong that they chose to either (1) leave the work setting or, (2) lower their commitment to those goals for which situational variables inhibited goal accomplishment. It is important to note that these data were generated by persons who were attempting to explain their own performance in situational terms. Given that situational variables were identified to explain p o o r performance, these data may reflect ego-defensive reactions (Vroom, 1964) or external attribution tendencies (Jones & Nisbett, 1971) rather than accurate reports of environmental circumstances. As such, these results should clearly be regarded as tentative but nonetheless suggestive of potential situational conditions which might help explain variance in performance. S T U D Y II
Having identified a set of situational variables which seemed potentially relevant to successful task performance, it was then possible to test hypotheses derived from conceptual performance models (Campbell et al., 1970; Campbell & Pritchard, 1976; Dachler & Mobley, 1973; Schneider, 1978; Terborg, 1977). Such models indicate that inhibiting situational conditions should have a direct inhibiting effect on perfor-
84
PETERS, O'CONNOR, AND RUDOLF
mance. Persons who are both willing and able to accomplish their assigned tasks may be prevented from doing so by situational variables beyond their control. Further, based on the conjunctive nature of these conceptual performance models, an additional hypothesis is suggested concerning the influence of situational inhibitors on performance. It follows from these models that situational inhibitors should have a stronger effect on persons with high levels of ability and motivation than on those with low levels of ability and motivation (Schneider, 1978). In effect, situational inhibitors are expected to lower the "ceiling" on potential performance among those persons who, due to high ability and motivation, would be expected to perform at a level above the situationally determined ceiling while affecting their low-ability, low-motivation counterparts less severely. As a result, inhibiting situational conditions are expected to substantially restrict the top end of the performance distribution while having a less severe impact on the bottom end of that continuum. The variance in observed performance in work settings in which such constraints are present, therefore, is expected to be restricted. It was, therefore, hypothesized that (1) performance should be higher and (2) the variance in performance should be greater in work settings which do not contain the previously identified inhibiting situational conditions than in settings which do. An additional issue raised from the data gathered in Study I concerns the extent to which affective responses vary as a function of differences in the facilitating/inhibiting situational conditions across work settings. Study I participants reported negative affective reactions (frustration and job dissatisfaction) to the perceived presence of inhibiting conditions. Based on these data, it was hypothesized that both frustration and dissatisfaction would be greater among persons working in a setting which contained inhibiting situational characteristics than for those in situations in which these factors were absent. Finally, there are a great number of studies which purport to test hypotheses concerning the relationship between performance and affective responses to work (of. Locke, 1976). A study by Herman (1973) is particularly relevant to the present investigation. Herman argued that job satisfaction is causally related to performance, but that strong positive relationships between these variables will be observed only in work situations in which persons can express their feelings in terms of behavior. Thus, consistent with the present focus, she stated, " I f situational constraints are active the performance variance due to individual differences . . . should suffer from gross restrictions of range. Under such circumstances significant attitude-performance relationships cannot be expected" (p. 211"212). This hypothesis appears consistent with a somewhat similar systematic
PERFORMANCE-RELEVANT SITUATIONAL VARIABLES
85
restriction of range argument presented by Schneider (1978). While Schneider did not address the question of whether or not attitudes precede performance, he did argue, based on the conjunctive nature of his performance model, that any variable expected to correlate with performance will do so only to the degree that variance in the latter is not constrained. Thus, it was hypothesized that the relationship between performance and affective responses would be significantly stronger for persons in a work setting in which inhibiting situational conditions were absent than for individuals in work settings in which such conditions were present. This second study was designed to test the above hypotheses in a controlled laboratory experiment in which situational variables previously identified in Study I as potentially relevant to performance situations were manipulated. A review of the literature suggests that the present study may be the first attempt to experimentally explore the effects of specific, performance-relevant facilitating/inhibiting situational conditions on the behavioral and affective responses of persons in task situations. Method
Subjects Participants in this study were 13 graduate and 57 undergraduate students enrolled in organizational behavior courses at a southwestern university. Fifty-one of the seventy participants were employed on a fulltime basis. The sample was composed of 27 males and 43 females with an average age of 29 years. Volunteers participated in exchange for extra credit toward course grades.
Task The task used in this study involved building three-dimensional models from standard Erector set parts. This task was chosen because (1) both quantity and quality of performance could be assessed, (2) it easily allowed for the manipulation of those situational variables identified in Study I, and (3) few persons were expected to be highly experienced in performing it.
Manipulations Four of the eight situational variables identified in Study I, judged to be relevant to the laboratory task, were simultaneously manipulated to result in either facilitating or inhibiting work conditions. The manipulation of each variable in the inhibiting condition was designed to hinder, rather than to prevent, good performance. It was felt that too strong of a manipulation, one which would make high performance impossible, would (1) not be representative of typical work situations, and (2) make the research trivial. Data supporting the assertion that the subjects were not highly
86
PETERS, O'CONNOR, AND RUDOLF
constrained in either experimental condition are reported below. The strength of the overall manipulation, thus, was based on the combined effects of four separate manipulations, each of which was designed to be substantive, but not overwhelming. Specific manipulations are described below. 1. Job-related information. Professionally prepared photographs of correctly completed models were used as "blueprints" to describe to the subjects the models they were to build. Four photographs were taken of each completed model, representing front, side, top, and perspective views of the model to be constructed. In the facilitating condition, subjects were provided with all four views of each model to be built. Thus, complete information was provided. In the inhibiting condition, however, participants were not given the perspective view to help them visualize the completed work. It should be noted that while subjects in the inhibiting condition were not given complete information, they had all the information necessary to do the task. Thus, the differences between conditions involved both the amount and quality of job-related information provided. 2. Tools and equipment. In order to make the tight screw-to-nut connections required to build the models, it was necessary to use both a wrench and screwdriver. In the facilitating condition, persons were provided with tools explicitly made for use with Erector set parts. Additionally, they were provided with an extra screwdriver (4.5 in., 11.5 mm.) which, depending upon personal preference, might feel more comfortable than the standard Erector set tool. In the inhibiting condition, persons were not given standard Erector set tools. Rather, they were given a small screwdriver (3.0 in., 7.5 mm.) and a small adjustable wrench (2.75 in., 7.0 mm.) which, while adequate, were not as convenient as those provided to subjects in the other condition. Thus, the conditions differed with regard to the amount and quality of the tools provided. 3. Materials and supplies. Materials and supplies were manipulated by separating, or not separating, each of the 15 different types of parts into individual compartments in a large parts box. Thus, in one condition, parts availability was facilitated, while in the other, persons had to sort through all available parts to find the ones needed. It should be noted that subjects had the same number and type of parts regardless of experimental condition. The difference between conditions, therefore, involved the accessibility of needed materials. 4. Task preparation. In both experimental conditions, subjects were provided with the same training and allowed to finish one practice trial. Task preparation was manipulated by allowing persons in the facilitating condition to have two additional practice trials, thus allowing them to become more familiar and comfortable with the model-building task. With the exception of the additional practice trials in the facilitating condition,
P E R F O R M A N C E - R E L E V A N T S I T U A T I O N A L VARIABLES
87
procedures were identical across experimental conditions. Thus, the difference between conditions involved the amount of practice or task preparation provided.
Procedures Treatment conditions were randomly assigned to groups of from one to three persons who were then run simultaneously. Subjects, in both the training and work areas, were separated by partitions to prevent differential social facilitation effects (Zajonc, 1965) associated with the varying number of subjects run simultaneously. Subjects were trained using standardized taped messages which emphasized the importance of learning to build the models correctly. Following training, and the additional two practice trials in the facilitating condition, subjects were taken to individual work areas. Instructions, given by taped messages, stressed the importance of both quantity and quality of performance during the task session. After a 40-min work period, subjects completed a postexperimental questionnaire, were debriefed, asked not to discuss the study, and were dismissed. Measures Performance. Completed models were evaluated in terms of both quantity of work done (i.e., the number of pieces used and the number of screw-to-nut connections made) and quality of work done (i.e., percentage of correct pieces used, percentage of correct screw-to-nut connections made, and percentage of such connections which were tight). Test-retest reliabilities were assessed for both the quantity and quality of performance dimensions by having 20 completed models scored on two occasions. Test-retest reliabilities ranged from 0.77 to 1.00, with a median value of 0.88. Based on the pattern of substantive correlations within each of these sets of variables, overall quantity of work and overall quality of work variables were created using standardized linear composites based on z score transformations of each variable. For each measure, high scores indicated better performance. Satisfaction. Two separate measures of satisfaction were employed in the present study. The first was a modification of the Work Itself scale from the Job Descriptive Index (JDI) (Smith, Kendall, & Hulin, 1969). The modification involved rewriting the instructions to have subjects describe the laboratory task, rather than their job situation. The second measure was a similarly modified form of the General Satisfaction scale from the Job Diagnostic Survey (JDS) (Hackman & Oldham, 1975). Reliabilities, using coefficient a, were .76 and .75 for the JDI and JDS satisfaction measures, respectively. Frustration. Since a literature search revealed no standardized frustra-
88
PETERS, O'CONNOR, AND RUDOLF
tion instrument appropriate to the present context, a short, three-item direct rating scale was developed. The items were: (1) Trying to get this " j o b " done was a very frustrating experience, (2) being frustrated comes with this " j o b , " and (3) overall, I experienced very little frustration on this " j o b " (reverse scored). Subjects responded to each item on a 7-point Likert scale, and responses were summed so that a high score indicated more experienced frustration. The ~ coefficient was .76 for this measure.
Manipulation Checks Four items (see Table 2) were written to assess the effectiveness of the experimental manipulations. Subjects expressed their extent of agreement with each statement on a 7-point Likert scale ranging from strongly disagree (1) to strongly agree (7). Results, using one-tailed t tests, indicated that the two experimental conditions differed significantly with respect to task-related information (p < .05) and materials and supplies (p < .01). The t tests between conditions approached significance for both the tools and equipment (p < .07) and task preparation (p < .08) manipulations. In addition, the summated score over all four manipulation check items differed significantly across conditions (t = 2.54, p < .01). It should be noted that even in the inhibiting TABLE 2 MEANS AND STANDARD DEVIATIONS FOR FACILITATING AND INHIBITING MANIPULATED CONDITIONS
Facilitating condition Manipulation check items
Inhibiting condition
M
SD
M
SD
The information c o n v e y e d in the photographs was helpful in getting this task done efficiently.
5.89
1.16
5.34
1.49
1.70"*
T h e tools I used were helpful in getting this task done efficiently.
5.14
1.56
4.51
1.82
1.55"
T h e training I received prepared me to do this task efficiently.
5.66
1.21
5.17
1.54
1.46"
T h e fact that the parts were easily accessible helped me to do this task efficiently.
5.54
1.44
4.34
1.77
3.12"**
22.23
4.32
19.37
5.05
2.54***
Overall manipulation
t
N o t e . N = 35 within each condition. Total manipulation equals the s u m of the information, tools and equipment, training, and materials and supplies scores. All tests are onetailed. * p < .10. * * p < .05. *** p < .01.
89
P E R F O R M A N C E - R E L E V A N T S I T U A T I O N A L VARIABLES
condition, mean scores did not reflect severe constraints with respect to the resources manipulated. Even though the t tests were not significant in each instance, however, these data support the effectiveness of the overall manipulation. Results In order to assess the effects of differing situational conditions on both performance and affective responses, multivariate analysis of variance (MANOVA) was employed (Barr, Goodnight, Sail, & Helwig, 1976). The dependent variables were (1) quantity of work completed, (2) quality of work completed, (3) satisfaction with the task itself, (4) general satisfaction, and (5) frustration. Using Pillai's criterion (see Olson, 1976), a significant multivariate effect was observed (F(5,64) = 2.98, p < .05), indicating that the vector of dependent variable means within the facilitating condition differed significantly from that for the inhibiting condition for at least one of the dependent variables. In order to explore this significant multivariate effect, a series of univariate analyses, t tests, were employed (Spector, 1977). Table 3 contains means and standard deviations for each of these variables. Significant differences, across conditions, were observed for both work quality (p < .01) and frustration (p < .05). In addition, the mean differences for both quantity of work and task satisfaction approached traditional levels of statistical significance (p < .06). No difference was observed for general satisfaction. Thus, persons in the facilitating condition did better quality work and were less frustrated than those in the inhibiting condition. Further, facilitating condition subjects tended to do more work and reported more task satisfaction than did their counterparts in the inhibiting condition. Overall, t h e s e results p r o v i d e some s u p p o r t for the
TABLE 3 COMPARISON OF CRITERION SCORES ACROSS EXPERIMENTAL CONDITIONS Inhibiting condition Dependent variables 1. 2. 3. 4. 5.
Quantity of work Quality of work Task satisfaction General satisfaction Frustration
Facilitating condition
X
SD
.~
SD
-.384 -.815 19.286 4.648 4.581
1.99 2.71 10.15 1.30 1.55
.384 .815 23.286 4.848 3.810
1.94 1.42 10.11 1.55 1.45
Note. N = 35 within each condition. All tests are one tailed. * p < .06. ** p < .05. *** p < .01.
t 1.63' 3.14"** 1.65* .58 2.15"*
90
PETERS, O'CONNOR, AND RUDOLF
hypothesized effect of facilitating/inhibiting environmental conditions on both affective and behavioral responses. It also was hypothesized that the variance in performance would be restricted in the inhibiting, as compared to the facilitating, condition. No support was found for this hypothesis. In fact, for the quality of work measure, the variance in the inhibiting condition was significantly larger than that in the facilitating condition (Fmax(2,34) = 3.63, p < .01) ! The last hypothesis stated that the performance-affective response relationships should be significantly stronger in the facilitating, as compared to the inhibiting, condition. This prediction was tested for both composite measures of performance (quantity and quality) using each of the three affective response measures. Performance-affective response correlations are given in Table 4. Within the facilitating condition, quality of work was significantly correlated with each affective response variable. Corresponding correlations within the inhibiting condition were nonsignificant. In each case, the differences between corresponding correlations, across conditions, were significant at thep < .05 level. Further, the variance in affective measures accounted for by quality of work differed greatly across conditions. The a v e r a g e variance in the affective measures accounted for in the facilitating condition was 32%. The corresponding figure in the inhibiting condition was only 5%! This difference clearly is of practical, as well as statistical, significance. A different picture emerges for quantity of performance (see Table 4). In no instance did corresponding correlations differ significantly across experimental conditions. Thus, while strong support is provided for the hypotheses with regard to quality of work done, no support was offered for the performance quantity measure. Discussion Based on existing theoretical models of performance (cf. Campbell & Pritchard, 1976), it was hypothesized that both quantity and quality of performance, as well as the variance in these measures, would be significantly lower in the inhibiting performance setting than in the facilitating condition. Results provided support for the hypotheses in that poorer quality work (p < .05) and lower quantity of output (p < .06) were observed in the inhibiting, as compared to the facilitating, condition. However, while no difference in variance across experimental conditions was observed for quantity of work completed, the data show a variance difference opposite to that predicted for quality of work done. The significant and marginally significant effects for quality and quantity of performance, respectively, were observed in the presence of relatively weak experimental manipulations. Not only did two of the four
.09 .55*** -.29*
- . 11 .30* -.12
Inhibiting condition .80 1.23 .71
Z
Note. N = 35 within each condition. * p < .05. ** p < .01. *** p < .001. Difference between independent correlations significant at the p < .01 level. b Difference between independent correlations significant at the p < .05 level.
1. Task satisfaction 2. General satisfaction 3. Frustration
Facilitating condition
Quantity of work
COMPARISON OF PERFORMANCE--AFFECTIVE
.45** .59*** -.64***
Facilitating condition
-.20 .12 -.28
Inhibiting condition
Quality of work
TABLE 4 RESPONSE CORRELATIONS ACROSS EXPERIMENTAL CONDITIONS
2.75 a 2.23 b 1.91 b
Z
>
< >
;e >
>
.d
< >
> Z ¢3
©
92
PETERS, O'CONNOR, AND RUDOLF
manipulation check items fail to reach significance, but, in all cases, subjects in the inhibiting condition did not report that the resources provided strongly hindered their task performance. Further, only four of the eight situational variables identified in Study I were experimentally manipulated. Thus, not only were relatively weak manipulations used, but further, they involved a limited subset of the situational variables identified in Study I. This suggests that in task sltu~[tions which contain either a greater number of and/or more severe situational inhibitors, the effects on performance may be more pronounced. The results with respect to variance differences across conditions were surprising. Previous researchers who have speculated on variance effects due to situational inhibitors (Herman, 1973; Schneider, 1978) clearly expected systematic range restriction to result from the presence of such inhibitors. Such restriction was hypothesized to occur in the inhibiting situational condition since situational inhibitors were expected to have a larger negative impact on the performance of the most capable relative to the least capable workers. If the major impact of situational inhibitors falls upon the most capable workers, then, across experimental conditions, no mean or variance differences in performance should be observed among those who were poorer performers, while among those who were high performers, both the mean of and variance in performance within the facilitating condition should be significantly greater than their counterparts in the inhibiting condition. This explanation was explored by dividing the persons within each experimental condition into high- and low-performance groups based on a median split of the performance distributions within each condition. High performers in the facilitating condition were compared to high performers in the inhibiting condition with respect to both mean and variance differences in performance. In like manner, low performers within each condition were compared. Data relevant to this explanation are given in Table 5. Contrary to the prediction, low as well as high performers in the facilitating condition produced significantly better quality than did their respective counterparts in the inhibiting condition. With regard to variance, the results were once again contrary to predictions, with the variance in quality significantly expanded, rather than constricted, in the inhibiting as compared to the facilitating condition. These variance differences were observed for both high- and low-performance groups. The results for quantity are less clear. Only the mean difference in the high-performance group was significant. It may be useful to note, however, that each of the mean and variance comparisons for both the highand low-performance groups, while nonsignificant in three out of four
P E R F O R M A N C E - R E L E V A N T S I T U A T I O N A L VARIABLES
93
cases, were all in a direction consistent with the quality results discussed above. These data may warrant a "rethinking" of the hypothesized effects of situational inhibitors on both performance levels and variance. With regard to performance levels, these data indicate that inhibiting conditions tend to both' 'lower the ceiling" on performance among better performers and "lower the floor" among poorer performers. It might be that inhibiting environmental conditions, in addition to having the expected ceiling effect, may also overwhelm poorer performers, resulting in very low performance. That is, situational inhibitors may act to lower the entire range of the performance distribution. The variance data are clearly contrary to predictions among both high and low performers, and thus difficult to explain. The greater variance exhibited by the inhibiting condition subjects, particularly on the work quality variable, in part may be due to the influence of situational inhibitors on persons' motivation to perform well. While some individuals may persevere in the face of inhibiting conditions where dissatisfaction and frustration are higher, others may develop expectancies that effort does not lead to performance and, therefore, not continue to be motivated to do well. Such differential perseverance may well be due to stable individual differences. Lawler (1973), for one, speculated that it is difficult to motivate persons with low self-esteem since such persons are predisposed to believing that they cannot perform well regardless of how much effort they expend. Leid and Pritchard (1976) have reported that persons with low, as compared to high, protestant work-ethic beliefs report both lower expectancies and lower effort expenditure. Merrens and Garrett (1975), in an experimental study, have shown that individual differences in work ethics bear directly on both effort and performance. These data consistently indicate that some persons, based on stable individual attributes, are more likely to persist at tasks than others. It may well be that such individual differences are more relevant as the situation makes it increasingly more difficult to accomplish a task. If so, the present data are somewhat understandable. Future research might meaningfully be aimed at not only replicating the present variance results, but also at exploring the relations between relevant individual differences and indices of motivation across situations which differ systematically with respect to the amount and severity of situational inhibitors present. With regard to the affective measures, persons in the facilitating condition expressed significantly less frustration and tended to be more satisfied with the task itself than their counterparts in the inhibiting condition. Thus, situational variables relevant to performance seem to have affective, as well as behavioral, consequences.
94
PETERS, O'CONNOR, AND RUDOLF
The relationships between affective responses and quality of work seem strongly affected by situational conditions relevant to performance settings. As shown in Table 4, quality of work was significantly related to each affective measure in the facilitating condition, but unrelated in the inhibiting condition. In all cases, the differences between corresponding correlations were significant. These data would appear to suggest that, with regard to quality of work, situational variables relevant to performance seem to act as boundary variables (Fromkin & Streufert, 1976) defining conditions under which affective responses and performance can, and cannot, be expected to correlate with one another. While the pattern of observed correlations is consistent with Herman's (1973) predictions, it cannot be understood in terms of the restricted variance explanation she proposed. In fact, the observed variance in the quality of work measure was greater, rather than smaller, in the inhibiting as compared to the facilitating condition. It may be precisely this variance difference, however, which provides the best explanation for the pattern of affective response to performance associations. Peters and Champoux (1979) have pointed out that the magnitude of a correlation varies as a function of both the absolute slope of the regression line between two variables as well as the ratio of the variances between these variables (i.e., r = b ( c r x / o y ) ) . If we assume that when work quality is regressed on each of the affective measures, the regression effect does not change significantly across experimental conditions, then the significantly smaller variance in quality in the facilitating versus inhibiting condition discussed above combined with no significant differences in the variance in any of the affective measures across these treatment groups could easily account for the correlation differences evident in Table 4. As Peters and Champoux (1979) have noted, such fluctuations in correlations based on variance differences will not be apparent as significant interactions within regression equations. To further investigate the above explanation, three separate regression analyses were conducted. In each, quality was hierarchically regressed on one of the affective measures, the experimental manipulation, and their cross-product term. Results from two of the three regression analyses supported the variance explanation presented above. Of the three cross-product terms, only the task satisfaction × experimental manipulation interaction term was significant (p < .05) and, therefore, consistent with the correlational data. While the proposed variance explanation would appear to account for the observed correlations between quality of work and both general satisfaction and frustration, it cannot account for all of the data. It would appear that further research will be necessary in order to explain these results. The lack of correlation differences involving the affective responses and quantity of performance, across experimental conditions, appear to
PERFORMANCE-RELEVANT SITUATIONAL VARIABLES
95
be consistent with the proposed variance explanation since no variance differences in quantity of performance were observed across conditions. Alternatively, the differential impact of situational inhibitors on both the variance in performance and the performance-affective response relationships across the quantity and quality data may be partially explained by the nature of the specific manipulations in this study. In retrospect, the manipulations employed seem more relevant to affecting behavior on the quality of work dimension than on the quantity of work dimension. The results concerning mean differences and variances across conditions for both performance dimensions support this contention. Thus, it may be that in task situations in which the inhibitors more strongly affect quantity of work, a moderating effect due to situational inhibitors on quantityaffective response relationships might be more likely to be observed. Implications The present results imply, as human factors psychologists have long known (Chapanis, 1976), that work situations should not be regarded as constants for which people are selected and trained. The prediction that inhibiting situational conditions Should adversely affect performance is straightforward and expressed in several theoretica4 models of performance. Thus, it should come as no surprise that su/ch conditions significantly affected quality of work and tended to do the same for quantity of work in the present investigation. What is surprising is the fact that so little work has been done to define the nature of and dimensions along which work settings might vary to produce such effects. The present study provides a first step in this direction. While the situational taxonomy developed in Study I is clearly tentative, manipulations of task environments based on variables in this taxonomy successfully affected work behavior and affective responses. As discussed above, these effects were observed using relatively weak manipulations of relatively few situational variables. To the extent that applied situations contain greater numbers of such inhibiting conditions and/or more severe ones, stronger consequences could be anticipated. If so, continued research on situational variables relevant to performance settings is clearly warranted. REFERENCES Ban', A. J., Goodnight, J. H., Sail, J. P., & Helwig, J. T. A user's guide to SAS. Raleigh: Sparks Press, 1976. Campbell, J., Dunnette, M., Lawler, E., & Weick, K. Managerial behavior, performance, and effectiveness. New York: McGraw-Hill, 1970. Campbell, J., & Pritchard, R. Motivation theory in industrial and organizational psychology. In M.D. Dunnette (Ed.), Handbook of industrial and organizational psychology. Chicago: Rand-McNaily, 1976.
96
PETERS, O'CONNOR~ AND RUDOLF
Chapanis, A. Engineering psychology, tn M. D; Dunnette (Ed.), Handbook of industrial and organizational psychology. Chicago: Rand-McNally, 1976. Dachler, H. P , & MoNey, W. H. Construct validation of an instrumentality-expectancytask, goal model of work motivation: Some theoretical boundary conditions. Journal of Applied Psychology, 1,973, 58, 397-418. Fleishman, E. On: the relation between abilities,, learning, and human performance. American Psychologist, I972, 27, 1017-1031. Fromkin, H. L., & Streufert, S. Laborato~ experimentation. In M. D. Dunnette (Ed,), Handbook of industrial and organ&ational psychology. Chicago: Rand-McNally, 1976, Hackman, J. R,, & Otdham, G. R. Development of the Job Diagnostic Survey. Journal of App#ed Psychology, 1975, 60, 159-t70. Herman, J. B. Are situational: contingencies limiting job attitude-job performance relationships? Organizational Behavior and Human Performance, 1973, 10, 208-224, Jones, E. E., & Nisbett, R. E. The actor and the observer: Divergent perceptions of the causes of behavior. Morristown, N.J.: General Learning Press, 1971. Lawler, E. Motivation in work organ&ations. Monterey: Brooks/Cole, 1973. Lawler, E. Reward systems. In J. R. Hackman and J. L, Suttle (Eds.), Improving life at work: Behavioral science approaches to behavioral change. Santa Monica: Goodyear, 1977. Leid, T: R., & Pritchard, R. D. Relationships between personality variables and components of the expectancy-valence model. Journal of Applied Psychology, 1976, 6I, 463-467; Locke, E. A. The natureand causes of job satisfaction. In M. D. Dunnette (Ed.), Handbook of industrial and organizational psychology. Chicago: Rand-McNally, 1976. Merrens, M. R., & Garrett, J. B. The protestant work ethic scale as a predictor of repetitive work performance. Journal of Applied Psychology, I975, 60, 125-127. Olson, C. L. On choosing a test statistic in multivariate analysis of variance. Psychological Bulletin, 1976, 83, 579-58& Pete~s, L. H., & O! Connor, E. J. Situational constraints and work outcomes: The influence of a frequently overlooked construct. Academy of Management Review, in press. Peters, W. S , & Champoux, J. E. The use of moderated regression in job redesign decisions. Decision Sciences, 19792 t0, 85-95. Schneider, B. Person-situation. selection: A review of some ability-situation interaction research. Personnel Psychology, I978, 31, 281-297. Smith, P , Kendall L., & Hulin, C. The measurement of satisfaction in work and retirement. Chicago: Rand-McNalty, 1969. Spector, P. What to do with significant multivariate effects in MANOVA. Journal of Applied Psychology, 1977, 62, 158-163. Terborg, J. T. Validation and extension of an individual differences model of work performance. OrganizationalBehavior andHuman Performance, 1977, 18, 188-2t6. Vroom, V: Work and motivation. New York: Wiley, t964. Wexley, K., & Yukl, G. Organizational behavior and personnel psychology, ttomewood: Irwin, 1977. Zajonc, R. B. Social facilitation~ Science; 1965, 149, 269-274. RECEIVED: April 14; 1979-