villanova & roman 1993.pdf

29
A META-ANALYTIC REVIEW OF SITUATIONAL CONSTRAINTS AND WORK-RELATED OUTCOMES: ALTERNATIVE APPROACHES TO CONCEPTUALIZATION Peter Villanova Appalachian State University Marie A. Roman Northern Illinois University This article reviews current conceptualizations of constraints and reports the results of a quantitative review of published research on situational constraints. The results of the meta-analysis are consistent with those of previous reviews that used a traditional narrative review method; con- straint scores bear a weak relationship with performance measures (- .14) and a fairly robust relationship with self-reported affect (- .32 with satisfac- tion, .39 with frustration). It is concluded that this weak pattern of findings suggests the need for alternative conceptualizations of constraints. Toward that end, a variety of conceptual framing schemes from which to generate alternative measurements of constraints and hypotheses about constraint- outcome relationships are presented. Situational factors, as facilitators or inhibitors of performance, have only re- cently received some of the attention lavished on traditional ability and mo- tivation predictors of performance. As a result, expectations regarding pro- gress in understanding situational factors must be moderated by the relative paucity of research on situational factors. Nonetheless, systematic study of situational factors holds much promise for providing additional understanding of performance dynamics. Also, because human resource management profes- sionals often undertake situational changes with the aim of influencing worker contributions to effectiveness there is a corresponding need for practice to be informed by research. Our purpose in writing this article is to stimulate different ways of thinking about situational constraints, their causes, and how they influence perfor- mance. The article is divided into three sections. We begin with a review of Direct all correspondence to: Peter Villanova, Management Department, Appalachian State University, Boone, NC 28608. Human Resource Management Review, Copyright 0 1993 Volume 3, Number 2,1993, pages 147-175 by JAI Press, Inc. All rights of reproduction in any form reserved. ISSN:1053-4822

Upload: azn1nfern0

Post on 17-Dec-2015

238 views

Category:

Documents


2 download

TRANSCRIPT

  • A META-ANALYTIC REVIEW OF SITUATIONAL CONSTRAINTS AND WORK-RELATED

    OUTCOMES: ALTERNATIVE APPROACHES TO CONCEPTUALIZATION

    Peter Villanova Appalachian State University

    Marie A. Roman Northern Illinois University

    This article reviews current conceptualizations of constraints and reports the results of a quantitative review of published research on situational constraints. The results of the meta-analysis are consistent with those of previous reviews that used a traditional narrative review method; con- straint scores bear a weak relationship with performance measures (- .14) and a fairly robust relationship with self-reported affect (- .32 with satisfac- tion, .39 with frustration). It is concluded that this weak pattern of findings suggests the need for alternative conceptualizations of constraints. Toward that end, a variety of conceptual framing schemes from which to generate alternative measurements of constraints and hypotheses about constraint- outcome relationships are presented.

    Situational factors, as facilitators or inhibitors of performance, have only re- cently received some of the attention lavished on traditional ability and mo- tivation predictors of performance. As a result, expectations regarding pro- gress in understanding situational factors must be moderated by the relative paucity of research on situational factors. Nonetheless, systematic study of situational factors holds much promise for providing additional understanding of performance dynamics. Also, because human resource management profes- sionals often undertake situational changes with the aim of influencing worker contributions to effectiveness there is a corresponding need for practice to be informed by research.

    Our purpose in writing this article is to stimulate different ways of thinking about situational constraints, their causes, and how they influence perfor- mance. The article is divided into three sections. We begin with a review of

    Direct all correspondence to: Peter Villanova, Management Department, Appalachian State University, Boone, NC 28608.

    Human Resource Management Review, Copyright 0 1993 Volume 3, Number 2,1993, pages 147-175 by JAI Press, Inc. All rights of reproduction in any form reserved. ISSN:1053-4822

  • 148 HUMAN RESOURCE MANAGEMENT REVIEW VOLUME 3, NUMBER 2. 1993

    current conceptualizations of situational constraints. The article then pro- ceeds to a quantitative review of published research on situational constraints. This section quantitatively summarizes research on situational constraints, including the reliability of measures used to represent constraints, and ob- served relationships between variables that comprise the nomological network from which construct validity evidence for situational constraints is inferred. Although the published research on constraints constitutes a relatively small body of literature, the review serves as a first attempt at quantitative summa- ry of empirical evidence relevant to the hypothesized relationships between constraints and other constructs. Finally, in the last section of the article, we introduce some alternative conceptualizations of constraints with the aim of stimulating alternative ways of thinking about constraints and the conduct of research involving the constraint concept.

    CURRENT CONCEPTUALIZATION OF CONSTRAINING FACTORS

    Attempts to improve performance efficiency through the manipulation of situa- tional factors served as a catalyst to the scientific study of work behavior and stimulated work redesign interventions (e.g., Gilbreth 1909). However, subse- quent progress in conceptualizing situational factors that influence perfor- mance was not commensurate with the scope of activity undertaken to change situational factors in specific circumstances. And, although situational factors were widely acknowledged as having significant practical implications for work behavior, no model of performance explicitly incorporated situational factors as a determinant of performance until the seminal work of Campbell, Dunnette, Lawler, and Weick (1970). In their model, Campbell et al. include situational constraints as one component of task demands that interact with individual differences, training and development experiences, and reward structures to influence job behavior, and subsequently job performance. In 1976, Campbell and Pritchard characterized situational performance determi- nants as facilitating and inhibiting conditions not under the control of the individual. This latter characterization was more inclusive as it allowed for the possibility that situational factors could promote or impede performance.

    However, aside from the recognition afforded situational factors in these models, no noticeable progress was made in developing taxonomies of situa- tional factors. Situational factors were roughly categorized as stemming from different levels of aggregation, the task characteristics, work setting, or organi- zation. Further description of the dimensions that constitute them, whether these dimensions were generalizable across levels of aggregation or other achievement-related contexts, or whether they were situationally specific, re- mained open to question. More often than not, situational factors were chosen on the basis of their availability and perceived relevance to the phenomena studied.

    Kahn and Quinns (1970; see also Kahn, Wolfe, Quinn, Snoek, & Rosenthal 1964) resource inadequacy concept was identified as one of several factors

  • ALTERNATIVE CONCEPTUALIZATIONS 149

    thought to contribute to work stress. However, it did not receive the systematic attention extended to other concepts associated with this framework (e.g., role conflict and ambiguity). Even though understanding of job stress has been limited by the absence of a widely accepted definition of what is meant by stress (Ivancevich, Matteson, Freedman, & Phillips 1990), the idea that inade- quate resources deemed necessary for successful work performance might serve as a putative cause of stress is both intuitively and logically appealing. That is, if individuals value achievement and the path to achievement is some- how made more cumbersome by the absence of necessary supports, it seems reasonable to include inadequate resources as a stressor responsible for some of the stress symptoms (strains) that might be observed (e.g., decreased satis- faction, higher rates of turnover). However, measures of resource inadequacy used in stress research have been of limited breadth, including only a small sample of the potential domain of resources that workers in these studies might find necessary for successful execution of their work roles. For example, although these studies have included occupations as diverse as health, blue- collar, and clerical, the items used to measure resource inadequacy have been limited to including work load, insutlicient job knowledge, and inadequate time. Thus, although the resource inadequacy concept was an early attempt to provide operational statement to situational factors that affect work outcomes, the necessary conceptual and taxonomic work of detailing its component parts was not pursued with sufficient rigor to make it a promising vehicle for under- standing how situational factors impact on performance.

    More recently, the work of Peters and OConnor (1980; Peters, OConnor, 8~ Eulberg 1985) has provided operational examples of situational constraints that influence performance buttressed by more systematic taxonomic efforts. In a retrospective account of the ideas which occurred to them early in their research, they stated:

    u . . . examples of constraining work factors came easily to mind. Factors

    such as broken or improper machinery and inadequate supplies quickly suggested themselves as types of situational factors we had in mind. To be sure, persons who were both able and motivated should not produce as much output as their ability and motivation levels suggested . . . (Peters, OConnor, & Eulberg 1985, pp. 80431).

    To identify constraints in the work environment, Peters and OConnor (1980) asked 62 persons employed in a variety of jobs to provide accounts of bad performance caused by situational factors. These responses were then content analyzed and eight resource variables necessary for task accomplishment were identified as potentially applicable across a variety of work settings. Further, these resource variables were thought to vary along three dimensions: (a) availability, (b) quantity, and (c) quality.

    Although described by the authors as tentative, the Peters and OConnor (1980) taxonomy has emerged as the most widely accepted method of represen- ting constraints in the work setting. Laboratory studies of constraints typically

  • 150 HUMAN RESOURCE MANAGEMENT REVIEW KILUME 3. NUMBER 2. 1993

    involve the manipulation of three or four of these resource variables. For exam- ple, job-related information, required materials, and work environment might be manipulated so that they are more or less available, of more or less suffi- cient quantity, and of more or less sufficient quality (cf. Peters, Chassie, Lind- holm, OConnor, & Kline 1982; Peters, Fisher, & OConnor 1982b; Peters, OConnor, & Rudolf 1980). These variables would be manipulated simul- taneously to test main effect hypotheses from this framework (e.g., situational constraints impede performance and reduce satisfaction). However, such tests are not very interesting from a theoretical viewpoint because situations which impede performance can be engineered easily, might overstate the impact of constraints in actual work settings, and serve only to affirm a tautology. Peters et al. (1985) correctly concluded that future research on the impact of con- straints should largely be investigated in field settings.

    Below, we describe a meta-analysis conducted on the findings reported in the published literature. Meta-analysis is a method of statistically combining the results of separate studies in order to facilitate conclusions about the strength of a relationship between two variables. Several different procedures exist and their suitability depends on the circumstances surrounding a specific problem area. However, some conventions regarding the appropriate methods of meta-analysis are better established today and encourage its more wide- spread use than had formerly been the case. Initial information sought from a meta-analysis includes: (a> an estimate of the average magnitude of an effect observed across studies, typically in the form of a correlation coefficient; (b) a test or series of tests involving the representativeness of the estimated average effect size as it applies to a set of studies; (c) an investigation, if necessary, of potential factors associated with the studies that may account for differences in observed effect sizes, and (d) an estimate of the fail-safe N, or the number of studies finding nonsignificant tests of the hypothesis that need to be observed in order to threaten the validity of a positive conclusion regarding the hypothe- sized relationship.

    Specifically, the meta-analysis focuses on the relationship between con- straints and outcome variables including: (a) job performance, (b) job satisfac- tion, (cl frustration, (d) turnover and intentions to leave, and (el organizational commitment. In addition, mean correlations among these variables are also estimated based on the reported relationships appearing in the literature ger- mane to situational constraints.

    QUANTITATIVE REVIEW OF SITUATIONAL CONSTRAINTS RESEARCH

    Method

    Literature Search. The search strategy consisted of a number of compo- nents. First, the Social Sciences Citation Index for the period of January 1980- June 1987 was searched, guided by the keywords situation, constraints,

  • ALTERNATIVE CONCEPTUALIZATIONS 151

    and conditions. Next, a search was made in Psychological Abstracts for the period (January 1987-December 1992) employing the key phrases and words situational constraints, performance obstacles, and stressors. Third, the reference sections of selected articles were examined for additional studies. Only English language sources were considered.

    Studies were included in the meta-analysis if they met the following criteria: (a) the study examined the effects of situational factors on employee behavior and attitudes such as job performance, job satisfaction, frustration, and com- mitment to the organization; (b) the study provided adequate description of the sample, measurement of constructs, analyses, and findings. Studies which reported only the presence of a significant finding without reporting the effect size or significance level were excluded from the meta-analysis. Applying these criteria resulted in the retrieval and coding of 15 studies (see Appendix for a list of the 15 studies).

    Coded Features. The following study characteristics were coded: (a) research design (experiment/simulation versus survey), (b) research setting (laboratory/ classroom versus field), (c) nature of respondent sample (students, managers, or subordinates), (d) sample size, (e) reliability of measurement operations, (f) criterion measurement strategy (objective, supervisor ratings, or self-ratings of performance), and (g) zero-order correlations between the variables of inter- est. These include: (a) ratings of situational constraints, (b) performance rat- ings, (c) job satisfaction, (d) frustration, (e) turnover, and (f) organizational commitment.

    Data were coded by the first and the second author. Any inconsistenc- ies were discussed until consensus was reached, thereby resulting in complete agreement.

    Procedure. An estimate of the combined, sample-size weighted effect size of all studies dealing with a particular hypothesized bivariate relationship was de- termined using the weighted effect size combinatorial procedure suggested by Hunter and Schmidt (1990). Findings which were reported as t or F values were converted to zero-order correlations (Rosenthal 1984). When studies re- ported more than one effect size pertaining to a bivariate relationship, the effect sizes were averaged to arrive at a single estimate. This was considered appropriate in light of the desire to maintain independence of the observations for combinatorial purposes and also because all of the multiple effect sizes from single studies were generated under similar study conditions. The variance of the observed coefficients and the variance due to sampling error were com- puted using formulas from Hunter and Schmidt (1990). Because of missing information and the large range of criterion reliabilities, the variance of the coefficients attributable to other artifacts such as range restriction or criterion unreliability was not estimated. However, this omission may not seriously affect the accuracy of the variance estimates reported given that previous studies have found that variance attributable to sample size differences ac- counts for 90% of artifactual variance (e.g., Pearlman, Schmidt, & Hunter

  • 152 HUMAN RESOURCE MANAGEMENT REVIEW VOLUME 3, NUMBER 2.1993

    1980). In the case of one study (Peters et al. 1982) only significance levels were reported. Thus, the zero-order correlation estimates derived from these values are conservative estimates.

    Results

    The sample sizes of the fifteen studies examined here ranged from 70 to 5080 with field studies generally employing larger sample sizes than did labo- ratory studies which never exceeded 120 subjects. Subjects in the studies in- cluded individuals employed in a wide variety of managerial and non- managerial positions.

    Table 1 details measures of the constructs examined in the 15 studies in- cluded in the meta-analysis. Information includes the instruments used to measure the constructs, the mean reliability estimates, and the range of re- liability estimates of the measurement instruments. The mean reliability of each measure used in the research reviewed here exceeds .70.

    Tables 2, 4, 6, 7, and 9 provide summaries of the published studies used in the meta-analysis including the setting and design of the studies, the sample size, the number of effect sizes examined in each study, the source (e.g., super- visory, self-ratings) and the form of the performance appraisal data (low or high judgment measures), and the mean effect size for each study. Also in- cluded in each table are the estimates of effect sizes generated from the pub- lished tests of the hypothesized relationships, the variance of observed effect sizes across studies, variance of the sampling error, the 95% credibility inter- val, a test for homogeneity of effect sizes, and a file drawer estimate of studies with nonsignificant findings needed to return the aggregated probability level across studies to .05.

    Table 2 reports the estimates of effect size among published studies for the hypothesized negative effect of situational constraints on performance. The obtained mean effect size across studies was found to be -.14. Note that while the 95% credibility interval centers in negative values, it includes the value .OO, suggesting that the calculated corrected mean effect size is an estimate of the mean of several population parameters or subpopulations identified by the operation of moderators (Whitener 1990). This is reflected in two passive obser- vation field studies, Peters, OConnor, Eulberg, and Watson (1988) and Olson and Borman (1989), both with large sample sizes (971 subjects and 5080 sub- jects, respectively), that found no relationship between situational constraints and performance ratings (7. = .OO). Also, in a study of time pressure effects on performance, Peters, OConnor, Pooyan, and Quick (1984) found a positive cor- relation between ratings of constraining situations (time pressure and di& culty) and performance ratings (study effect size = .30).

    The test for homogeneity of effects sizes (X2 (10, N = 9273) = 289.75, p < .Ol) indicates that it is not likely that each observation in this sample of effect sizes was drawn from the same population. Additionally, sampling error ex- plains only 3.3% of the observed variance in effect sizes. Application of the Hunter and Schmidt (1990) critical ratio of expected variance due to sampling

  • ALTERNATIVE CONCEPTUAL~~TIONS 153

    TABLE 1 Reliability and Types of Measures Used in Situational Constraints Research

    Constructs and Measures Mean Reliability Range

    Constraints, Peters and OConnor Goal Difficulty Rating

    .81(4) .67-.91

    Time Pressure Ratings Kahns Job Related Tension Index Composite Tailored to Study

    Satisfaction Index of Organiza~onal Reactions (Smith, 1976) Job Descriptive Index (Smith, Kendall, & H&in, 1969) Job Diagnostic Survey (Hackman & Oldham, 1975) Hoppack (1935) Cammann et al. Bullock Scale (Robinson, Athanasiou, & Head, 1969)

    Frustration Peters et al. (1980) 3 item scale Composite Tailored to Study

    Turnover Company records @ 12 & 18 months subsequent to

    study Intent (1 item)

    Performance Supervisor Ratings

    overall quality quantity

    Self ratings Objective indices

    Commitment Porter, steers, Mowday, and Boulian (1974) Mowday, Steers, and Porter (1979) Alutto, Hrebiniak, and Alonso (1973)

    .78(14) .49-.91

    .84(8) .53-.91

    .87(5) .83-.a9

    Note. All reliability estimates are coefficient alphas with the exception of one reported for perfor- mance which is an inter-rater reliability. Numbers in parentheses refer to the number of observed reliability estimates used to the compute the mean.

    error to observed variance (75%) suggests that substantive moderators as op- posed to additional artifacts, may explain considerable variability among ob- served effect sizes across studies. Because the data were consistent with the Xz test, the credibility interval application and 75% variance heuristic, moderator analysis was conducted on the hypothesized relationship between situational constraints and performance ratings.

    Two methodological features of the studies in the meta-analysis were identi- fied and examined for their potential moderating influence on the relationship between situational constraints and performance: (a) the setting in which the

  • 154 HUMAN RESOURCE MANAGEMENT REVIEW VOLUME 3, NUMBER 2,1993

    TABLE 2 Estimates of Effect Size among Published Studies for the Hypothesized Negative

    Effect of Situational Constraints on Performance

    Study Setting and

    Design

    Number Mean of Effect Performance Inokx Effect

    N Sizes (Source) Size

    Peters et al. (1988) Field Study 971

    Steel et al. (1987) Field Study 368

    Peters et al. (1984) Field Study 164

    Steel and Mento Field Study

    (1986)

    OConnor et al. Field Study

    (1984)

    Peters et al. (1982) Lab Experiment

    Peters et al. (1980) Lab Experiment

    Parasuraman and Field Study

    Allutto (1984)

    Spector et al. Field Study

    (1988)

    Jamal (1984) Field Study

    Olson and Borman Field Study

    (1989)

    438

    1450

    120

    70

    217

    181

    214

    5080

    7

    2

    3

    Over-all

    (Supervisor)

    Quality, Quantity,

    Over-all, and

    Objective

    (Supervisor and Self)

    Over-all

    (Supervisor)

    Over-all and Objective

    (Supervisor and Self)

    Over-all

    (Supervisor)

    Quantity

    Over-all & Quantity

    Over-all

    (Supervisor)

    Over-all

    (Supervisor)

    Over-all

    (Supervisor)

    Over-all

    (Peer and Supervisor)

    .oo

    -.26

    .30

    -.21

    -.12

    -.87

    -.29

    -.09

    -.05

    -.42

    .oo

    Mean effect size across studies = - .14; Variance of observed effect sizes across studies: ~2 = .03; Variance of the sampling error: (T, 2 = ,001; 95% Credibility interval: -.47 to .19; Test for homoge- neity of Effect Sizes: X2 (10, N = 9273) = 289.64,p < .Ol; File drawer Estimate: N = approximately 66 studies of ns findings (r = .OO) needed to return the aggregated p level across studies to .05.

    study was conducted (lab/classroom versus field studies), and (b) the source which provided ratings of constraints (the supervisor or the subordinate).

    A summary of the data relevant to the potential moderating effect of study setting and constraint rating source is contained in Table 3. The subset method of moderator analysis was used as recommended by Hunter and Schmidt (1990, p. 292). It is important to note that all lab studies involved manipulation of constraints and all field studies involved ratings of situational constraints. Also, measures of performance in both the lab studies and the field studies included hard criteria (e.g., output quantity, cash shortages, and loan growth) as well as soft criteria (e.g., over-all performance ratings).

    As shown in Table 3, the average correlation of situational constraints and performance measures for studies done in the lab differs from that computed for field studies. However, the variance in one subset (studies done in the lab) is greater than the total variance in effect size. This is perhaps due to the small

    ChrisHighlight

    ChrisHighlight

  • AL~R~TIVE ~ONCEPTUALI~TIONS 155

    number of studies done in the lab and their correspondingly smaller sample sizes, both of which may contribute to large second-order sampling error (Hunter & Schmidt 1990). There are too few published studies to allow a more careful examination of this hypothesis.

    The picture for source of constraint ratings is quite similar. While there is a sizable difference in mean effect size between the subsets, the observed vari- ance is larger in the one subset than the total variance effect size. The increase in observed within group variance may also be related to the differences in the relationship between constraint ratings and objective criteria (r = 0 for those two studies which reported those correlations) and the relationship between constraint ratings and performance ratings (non-zero).

    Table 4 reports the estimates of effect size among published studies for the hypothesized negative effect of situational constraints on job satisfaction. The obtained mean effect size across studies was found to be -.32. The 95% credi- bility interval includes the values of -.46 to -.18. This credibility interval suggests then, that one would expect to find a non-zero constraint-satisfaction relationship when conditions in an organization similar to those in the studies reviewed predominate (Schmitt & Klimoski 1991).

    The test for homogeneity of effect sizes (X2 (9, iV = 9273) = 32.46, p < .Ol) indicates that it is not likely that each observation in this sample of effect sizes was drawn from the same population. Additionally, sampling error explains only 28.6% of the observed variance in effect sizes. As for performance, modera- tor analysis was conducted on the hypothesized relationship between situa- tional constraint and job satisfaction ratings.

    One methodological feature of the studies in the meta-analysis was identi- fied and examined for its potentially moderating influence on the relationship between situational constraint ratings and job satisfaction ratings: the setting in which the study was conducted (lab/classroom versus field studies).

    Table 5 shows the summary of moderator analysis for the situational

    TABLE 3 Summary of Moderator Analyses for the

    Constraint-Pe~o~~ce ~lationship

    Study setting

    Lab Studies Field Studies

    i; = -49 i; = -.05 CT: = .08 CT: = .Ol a,2 = .004 02 = e .OOl

    Source of constraint rating

    Supervisor Subordinate

    i = -.17 P = -.04

    02 = .12 02 = ;

    .oos

    CT2 = ,004 ue = .OOl

  • 156 HUMANRESOURCEMANAGEMENTREVIEW VOLUME 3. NUMBER 2. 1993

    TABLE 4 Estimates of Effect Size among Published Studies for the Hypothesized Negative

    Effect of Situational Constraints on Satisfaction

    Study Setting and

    Design

    Number Mean of Effect Effect

    N Sizes Satisfaction Measure Size

    Peters et al. (1988) Field Study

    OConnor et al. Field Study (1984)

    OConnor et al. Classroom (1982) Survey

    Phillips and Freed- Classroom man (1984) Survey

    Peters et al. (1982) Lab Experiment

    Parasuraman and Field Study

    Hansen (1987)

    Peters et al. Lab Study

    Parasuraman and Field Study

    Alutto (1984)

    Spector et al. Field Study

    (1988)

    Jamal (1990) Field Study

    971 1

    1450 1

    237 2

    82 1

    120 2

    215 1

    70 2

    217 1

    181 1

    215 1

    Satisfaction With

    Work

    Satisfaction With

    Work

    Satisfaction With

    Work and General

    Satisfaction

    General

    Satisfaction

    Satisfaction With

    Work and General

    Satisfaction

    Over-All Satisfaction

    Satisfaction With

    Work and General

    Satisfaction

    Satisfaction With

    Work

    Over-All Satisfaction

    Over-All Satisfaction

    -.32

    -.28

    -.36

    -.55

    -.23

    -.13

    -.14

    -.20

    -.47

    -.33

    Mean effect size across studies = -.32; Variance of observed effect sizes across studies: cr2 = .007; Variance of the sampling error: (T, 2 = .002; 95% Credibility interval: - .46 to - .18; Test for homoge- neity of Effect Sizes: Xz (9, N = 3758) = 32.46, p < .Ol; File drawer Estimate: N = approximately 97 studies of ns findings (r = ,001 needed to return the aggregated p level across studies to .05.

    constraint-job satisfaction relationship. The mean correlation for studies con- ducted in the lab or classroom does not appear to differ greatly from the mean correlation for studies conducted in the field. Also, the observed variance in the subgroups is similar to that of all 15 studies. This finding, coupled with the indications of the credibility interval, suggest that study setting does not have a moderating effect on the relationship between situational constraints and job satisfaction. Perhaps an unidentified moderator is responsible for the variance observed in these effect sizes.

    Table 6 reports the estimates of effect size among published studies for the hypothesized positive effect of situational constraints on frustration. The mean effect size across studies was estimated to be .39. While the test for homoge- neity of effect sizes (X2 (7, N = 3246) = 13.52, p < .lO) indicate that observa- tions in this sample of effect sizes may not be drawn from the same population, it is important to note that the 95% credibility interval t.30 - .48) suggests

  • 157

    TABLE 5 Summary of Moderator Analysis for the

    Constraint-Satisfaction Relationship

    Study setting

    LablClassroom Studies Field Studies

    i = -.33 r = -.29 $72 = .015

    ai = .006 cr,2 = AI04

    0,2 = .OOl

    that a moderator is not likely to produce zero test-criterion relationships under given conditions. Thus, moderator analyses of the relationship between situa- tional constraints and frustration were not undertaken.

    The estimates of effect size among published studies for the hypothesized positive relationship between situational constraints and turnover is shown in Table 7. The obtained mean effect size across studies was found to be -21. The test for homogeneity of effect sizes resulted in a nonsignificant chi-square value Xz (5, N - 3248) = .97, p < .95. Since the chi-square test has high statistical power, a nonsignificant finding is strong evidence that there is no true variation across studies (Hunter, Schmidt, & Jackson 1982; Sackett, Har-

    TABLE 6 Estimates of Effect Size among Published Studies for the Hypothesized Positive

    Effect of Situational Constraints on Frustration

    Study Setting and

    Design

    Number Mean of Effect Frustration Effect

    N Sizes Measure Size

    Peters et al. (1980)

    O%onnor et al. (19841

    OConnor et al. (19821

    Peters et al. (19823

    Peters et al. (1980)

    Parasuraman and Alutto (1984)

    Spector et, al. (1988)

    Field Study 971 1

    Field Study 1450 1

    Classroom Survey 237 1

    Lab Experiment 120 1

    Lab Study 70 1

    Field Study 217 1

    Peters et al. 3 item Scale

    Peters et al. 3 item Scale

    Peters et al. 3 item Scafe

    Peters et al. 3 item Scale

    Peters et al. 3 item Scale

    4 item Scale

    .42

    .36

    .49

    .30

    .25

    .41

    Field Study 181 1 Peters et al. 3 item Scale

    .51

    Mean effect size across studies = .39; Variance of observed effed sizes across studies: u2 = .003; Variance of the sampling error: W, 2 = .OOf; 95% Credibility interval: .30 to .48; Test for homoge- neity of Effect Sizes: X2 (6, N = 32461 = 13.52, p < .05; File drawer Estimate: N = approxima~iy 61 studies of ns findings fr = .OO) needed to return the aggregated p level across studies to .05.

  • 158 HUMAN RESOURCE MANAGEMENT REVIEW VOLUME 3. NUMBER 2, 1993

    TABLE 7 Estimates of Effect Size among Published Studies for the Hypothesized Positive

    Effect of Situational Constraints on Turnover

    Study Setting

    and Design

    Number Mean of Effect Turnover Effect

    N Sizes Measure Size

    Peters et al. (1988)

    OConnor et al.

    (1984)

    Parasuraman and

    AIutto (1984)

    Spector et al.

    (1988)

    Jamal (1984)

    Jamal (1990)

    Field Study 971 1 Intentions .28

    Field Study 1450 1 Actual 18 Months .ll

    After Study

    Field Study 217 1 Actual 1 year .02

    After Study

    Field Study 181 1 Intentions .57

    Field Study 214 1 Intentions .24

    Field Study 215 1 Intentions .38

    Mean effect size across studies = .21; Variance of observed effect sizes across studies: crz = ,017; Variance of the sampling error: cre 2 = .002; 95% Credibility interval: - .03 to .45; Test for homoge- neity of Effect Sizes: Xz (5, N = 3248) = .97, p < .95; File drawer Estimate: N = approximately 25 studies of ns findings (r = .OO) needed to return the aggregated p level across studies to .05.

    ris, & Orr 1986). Thus, a search for moderators does not appear to be necessary.

    However, it should be noted that the credibility interval (-.03 to .45) in- cludes a zero value. This is quite likely due to the fact that two of the studies examined in the meta-analysis, OConnor, Peters, Pooyan, Weekly, Frank, and Erenkrantz (1984) and Parasuraman and Alutto (1984) measured actual turn- over 12 months and 18 months, respectively, after collection of constraint data. Four other studies which examined the hypothesized constraint-turnover rela- tionship, Peters, et al. (19881, Spector, Dwyer, and Jex (19881, Jamal (1984), and Jamal (1990), employed measures of intent to leave as indicators of turnover. The difference between intent to leave by an employee and the actual subse- quent behavior may account for the lack of consistent findings. Thus, analysis was conducted on the data examining the moderating influence of turnover measurement technique on the constraint-turnover relationship (see Table 8).

    TABLE 8 Summary of Moderator Analysis of the

    Constraint-Turnover Relationship

    Turnover Measure

    Intended Actual

    P = .37 i = .07

    a,2 = .007 up = .002

    u,2 = .002 a,2 = .OOl

  • ALTERNATIVE CONCEPTUALIZATIONS 159

    The results of the moderator analysis show that the mean correlations be- tween situational constraints and turnover are quite different in the two sub- groups. Additionally, the observed variance in the subgroups is less than the observed variance in effect sizes. Thus, it can be concluded that constraints appear to be more highly correlated with intentions to leave than with actual turnover. This is quite likely due to the large number of influences on an employees decision to terminate his/her employment relationship (i.e., oppor- tunity for alternative employment; cf. Hulin, Rosnowski, & Hachiya 1985).

    The estimates of effect size among published studies for the hypothesized negative effect of situational constraints on organizational commitment are found in Table 9. The obtained mean effect size across studies was found to be -.22. The 95% credibility interval, which includes the values of -.37 to -.07 suggests that one would expect to find a non-zero constraint-organizational commitment relationship given similar conditions to those in the reviewed studies.

    The test for homogeneity of effect sizes (x2 (4, N = 1098) - 28.47, p < .lO) indicates that observations in this sample of effect sizes may not be drawn from the same population. Additionally, sampling error explains only 14% of the observed variance in effect sizes. However, given the small number of studies which examined the constraint-commitment relationship, moderator analysis is not beneficial since any conclusions drawn would be tenuous at best.

    Summa/y of Results. The pattern of observed relationships between variables constituting the nomological network which help define constraints are sum- marized in Figure 1. The nomological network refers to a network of relation-

    TABLE 9 Estimates of Effect Size among Published Studies for the Hypothesized Negative

    Effect of Situational Constraints on Organizational Commitment

    Study Setting

    and Design

    Number Mean of Effect Measurement Effect

    N Sizes Method Size

    OConnor et al. (1982)

    Parasuraman and Hansen (1987)

    Parasuraman and Alutto (1984)

    Jamal (1984)

    Jamal (1990)

    Classroom survey

    Field Study

    Field Study

    Field Study

    Field Study

    237 1

    215 1

    217 1

    214 1

    215 1

    Porter et al. Scale

    Mowday et al. Scale

    Alutto et al. Scale

    Mowday et al. Scale

    Mowday et al. Scale

    -.37

    -.14

    -.17

    -.17

    -.24

    Mean effect size across studies = -.22; Variance of observed effect sizes across studies: 02 = .007; Variance of the sampling error: u,2 = ,001; 95% Credibility interval: - .37 to - .07; Test for homoge- neity of Effect Sizes: X2 (4, N = 1098) = 28.47, p < .Ol; File drawer Estimate: N = approximately 13 studies of ns findings (r = .OO) needed to return the aggregated p level across studies to .05.

  • 160 HUMAN RESOURCE MANAGEMENT REVIEW VOLUME 3. NUMBER 2,1!393

    I Situational \ 39 (7) Constraints

    ..\ \ -08 (3) -28 (3)

    16 (3)

    r-l Commitment /- 33 (3) Figure 1. Network of Observed Relationships among Scores on Measures of Constraints and Outcome Variables.

    Note: Decimal points are omitted. The number of study effect sizes used to compute average correlations between measures appear in parentheses.

    ships among measures of the constructs that is used to infer the meaning of the measures used to represent the constructs. To the extent that the pattern of relationships observed among these measures is consistent with the hypothe- sized relationships as articulated by a theory or model, greater confidence is warranted in the theory, concepts, and/or measures. However, if the relation- ships fail to conform to an expected pattern, the validity of the theory, con- cepts, and/or measures is suspect.

    The correlation coefficients between constraints and outcome variables in Figure 1 represent the mean effect size estimates reported in the tables. In addition, correlation coefficients between outcome variables in the figure rep- resent estimates of mean relationships between the outcome variables as re- ported in the 15 original studies reviewed. Some caution is required in inter- preting the displayed relationships, as they are based on a small number of observed coefficients, not corrected for measurement error. Just the same, several observations are worth expanded discussion with respect to their im-

  • ALTERNATIVE CONCEPTUALIZATIONS 161

    plications for future research involving current conceptualizations of con- straints.

    The small effect of -.14 averaged across these 15 studies and the small filedrawer estimate lead us to conclude that the constraint-performance rela- tionship has yet to be convincingly established. This conclusion agrees with that of Peters et al. (1985) who used a traditional review methodology on a smaller number of studies. Thus, the status of the constraint-performance hypothesis hasnt changed despite the fact that twice as many studies are included in this analysis as were in the Peters et al. review. The small magni- tude of this relationship is particularly striking, given that, without exception, each correlation between constraints and performance was cross-sectional and in some instances ratings of constraints and performance came from the same source (supervisor). Given the variety of potential artifacts that might inflate estimates of concurrent validity vis-a-vis estimates of predictive validity in those instance where one source provides ratings for both constraints and performance (e.g., desire for consistency), the mean effect size observed in the literature could be interpreted as an upper-bound estimate of the relationship.

    The pattern of results across studies involving self-reports of both con- straints and outcomes stands in stark contrast to the pattern observed for relationships between variables that are based on greater method- and source- independence. That is, affective outcomes such as satisfaction and frustration, which employ similar questionnaire-based measurement methods and rely on a single source for responses, demonstrate a consistent relationship to con- straint scores whereas data involving constraints and outcomes that are mea- sured through different measures and by different sources, tend to be unre- lated. This conclusion is also replicated at the individual study level; field research that has investigated the relationship between situational con- straints and performance measures that were defined through multiple opera- tions indicates that constraint scores correlate more highly with performance measures derived through methods more similar to those used to represent constraints. For example, Steel and Mento (1986) found that supervisory rat- ings of situational constraints correlated -.36 with supervisory ratings of performance and - .31 with subordinate self-appraisals whereas constraints were largely unrelated to objective indices of subordinate performance. A second study using multiple measures of performance conducted by Steel, Mento, and Hendrix (1987) largely replicated the pattern of relationships be- tween constraints and multiple measures of performance found by Steel and Mento (1986).

    No conclusion regarding the cause-effect relationship of constraints and put- ative outcomes is tenable on the basis of these studies as the exclusive use of single panel designs (i.e., cross-sectional or concurrent measurement of both constraints and outcomes) confuses inferences regarding the cause-effect rela- tionship between constraint and outcome measures. With the exception of findings based on laboratory data, the observed covariation between con- straints and outcomes in the field studies conducted to date could be attribut- able to the desire of respondents (subordinates or supervisors) to keep their

  • 162 HUMAN RESOURCE MANAGEMENT REVIEW WLUME 3, NUMBER 2. 1993

    constraint and affective self-reports consistent (cf. Salancik & Pfeffer 1977). Similar logic applies to those instances that solicit performance and constraint ratings from supervisors. Too few studies have been conducted to derive accu- rate estimates of the extent to which these relationships are inflated.

    The weak evidence of support for the hypothesized relationship between constraints and performance prompted a reformulation of the framework by Peters et al. (1985). Peters et al. (1985; Peters & OConnor 1988) proposed four boundary conditions (Fromkin & Streufert 1976; Lynch 1982) in the form of additional assumptions that must be present in order for constraints to corre- late with performance. First, the reformulation assumes that persons are assigned tasks that require the use of their abilities and motivation (p. 114). If performance standards are not sufficiently demanding and if performance standards are not maintained by management, then the impact of low to mod- erate levels of constraints should have little impact on performance. Second, it is assumed that organizations value individual performance such that slip- page in performance standards is kept to a minimum. Third, the model as- sumes that individuals cannot easily invoke self-handicapping explanations as an excuse for poor performance. That is, the impact of constraints on perfor- mance is not compensated for by a supervisor who evaluates individual perfor- mance. If supervisors were to adjust performance ratings based on worker accounts of constraints responsible for poor performance, then essentially, the impact of constraints on performance would be removed. Finally, given the first three assumptions are met, the reformulated framework also requires that resources vary in their availability across individuals. Should resources be so plentiful that shortage is improbable, then situational obstacles to effec- tive performance can no longer offer a plausible account for performance dif- ferences.

    In tandem, these conditions characterize what Peters et al. (1985) refer to as a high performance tension situation in which: (a) people must marshal1 their abilities and motivation in order to succeed, (b) high standards of perfor- mance are maintained by management, (cl excuse-making for poor perfor- mance is an ineffective strategy for altering performance scores, and (d) re- source availability is variable across individuals. Performance tension is a complex notion; it requires that all four conditions described above are present in a situation. Thus, each assumption is a necessary but not sufficient precon- dition for the prediction of performance scores from situational constraints scores.

    Limitation of the circumstances in which a relationship is expected to occur is one response to a series of empirical disconfirmations. Additional assump- tions can reduce confidence in prior nonsupportive research findings and si- multaneously expand the domain of questions remaining to be answered. It might in this way, contribute to the viability of a conceptual framework and maintain interest in a research program. On the other hand, the failure to accumulate empirical support for the core propositions of the framework could also suggest the need for examining constraints through different conceptual

  • ALTERNATIVE CONCEPTUALIZATIONS 163

    frames. The aim of this section is to stimulate alternative approaches to con- ceptualizing and measuring constraints.

    ALTERNATIVE APPROACHES TO CONCEPTUALIZING SITUATIONAL CONSTRAINTS

    Several articles have appeared recently to stimulate novel approaches to famil- iar research problems (Bobko 1985; Brinberg & McGrath 1985; Campbell, Daft, &, Hulin 1982; Wicker 1985). In each instance, these authors describe one or more different heuristics, including for example, the use of metaphors, the challenging of assumptions, the consideration of different contexts and levels of aggregation, the casting of stable entities into a dynamic process, and more careful scrutiny of conceptual frameworks in order to expand the scope of empirical inquiry. Several strategies can be applied to examine concepts, ex- pand their conceptual frameworks, and assess their suitability to different contexts. Contrary to conventional wisdom, concepts that explain a larger set of phenomena across a variety of contexts and that retain their distinctive attributes at higher levels of abstraction are possible (Osigweh 1989). Below, we describe alternative ways of thinking about and measuring constraints that may contribute to that aim.

    The most systematic and frequently studied conceptualization of situational constraints is reflected in the work of Peters and OConnor (1980). Our reading of this literature identified three necessary attributes that distinguish situa- tional constraints from other situational factors. Incorporated into a single statement, the three attributes define situational constraints as follows:

    Situational constraints are aspects of the immediate work situation, beyond the immediate control of persons, and that prevent people from fully utiliz- ing their relevant abilities and motivation.

    First, the content domain of constraints is limited to those conditions rele- vant to the immediate work situation. Situational factors that are removed from the immediate work situation, such as those located at the organizational level, are considered outside the taxonomic scope of the concept as currently defined (Peters et al. 1985). Second, constraints are not the products of persons, but rather reflect events that are uncontrollable and impersonal (Peters & OConnor 1980). This attribute avoids the problem associated with the concept of accident in that personal responsibility is excluded as a cause of con- straints. Third, constraints negatively affect a limited number of outcomes, primarily individual affect, motivation, and performance (Peters et al. 1985). As a result, the concept is precise at a specific level, pertains to a narrow behavior domain, and its constituent entities are distinct. According to Osig- webs (1989) classification of scientific concepts, constraints can be described as

  • 164 HUMAN RESOURCE MANAGEMENT REVIEW VOLUME 3, NUMBER 2. 1993

    high with respect to connotation (depth of description) and low with respect to extension (breadth of coverage).

    Concept attributes, such as those noted above, serve an essential function with respect to scientific concepts by delineating the characteristics necessary for representing the construct. However, our review of the published literature on constraints indicates that several distinguishing attributes of the concept have not been investigated, but rather have been assumed to be true. For example, although defined as uncontrollable, there has been no direct test of constraint controllability in field settings, nor any data bearing on whether responses differ when workers are instructed to consider only the status of uncontrollable resources versus when they are not provided specific instruc- tions to do so. If constraints are somewhat controllable, then the small effect of constraints on performance comes as no surprise. Also, the inductive method used to construct the situational constraints taxonomy coupled with the objec- tive of describing aspects of the immediate work situation appears to have resulted in a regretfully narrow application of a concept that may generalize to other levels of aggregation and a greater variety of contexts.

    Confusion about the scientific status of the constraint concept, as exem- plified by the equivocal meta-analysis findings, is in part due to the absence of information about these concept attributes. The asking of higher order ques- tions such as those involving causal relations among variables presupposes satisfactory answers to more basic questions about the phenomena and con- cepts involved. As a simple example, it seems reasonable that a good represen- tation of concept X needs to be established in a setting before proceeding to estimate the causal relationship between X and Y. Dillon (1984) notes that the logic of question sequences has not received much systematic study and that failure to do so makes some research questions posed by scientists premature relative to the current status of answers to lower level questions. Quite simply, the truth value of tests of higher order hypotheses must be buttressed by supporting tests of subordinate hypotheses. One recommendation we offer from the above discussion is that research on constraints needs to proceed more carefully and systematically with respect to testing fundamental attri- butes of the concept, perhaps following Dillons scheme for the classification of research questions. Research attempting to answer higher order questions needs to include information about more basic questions in order to support otherwise tenuous answers to higher order research questions.

    Generalizing Across Levels of Aggregation and Context

    Alternatives to the Peters and OConnor (1980) framework for conceptualiz- ing and researching constraints have begun to emerge. For example, in con- trast to the inductive method used by Peters and OConnor in order to identify constraints, Schoorman and Schneider and their associates (1988a; Moeller, Schneider, Schoorman, & Berney 1988) described a deductive method of identi-

  • ALTERNATIVE CONCEPTUALIZATIONS 165

    fying work facilitators and inhibitors on the basis of Katz and Kahns (1978) open systems model. This procedure entails identifying the functions per- formed by various organizational subsystems (e.g., supportive, maintenance, production, etc.) in the service of organizational goals and the requisite re- sources available for the subsystem to perform adequately. The approach is also novel in that it applies the concept of facilitators/inhibitor at a higher level of aggregation-the work unit as opposed to the individual. Schoorman and Schneider (1988b) note that the appropriateness of one approach or the other to study constraints might depend on the degree of interdependence between work group members in a particular work setting. When workplace technology encourages interdependent and interactive effort as opposed to independent and individual effort, the work group might be a preferable unit of analysis. The work of Schoorman and Schneider (1988a) is encouraging in that it sug- gests further generalization of the facilitation/inhibition concept to higher levels of aggregation might be feasible.

    The conceptualization of constraints as aspects of the immediate work envi- ronment does not necessarily obviate extension of the concept to other behav- ioral contexts. However, like generalization to a higher level of aggregation, generalization to a broader context requires that the context, or substantive system (Brinberg & McGrath 1985), to which the concept is generalized main- tains an essential similarity to the original context. The original context used as the basis for developing the Peters and OConnor (1980) framework was the work environment of the respondents who generated data for taxonomic pur- poses. These respondents represented a cross-section of employees engaged in different work located in a variety of settings, some of whom may have per- formed their job functions in multiple work settings. This would seem partic- ularly likely with respect to responses from managers. Thus, the attribute immediate work environment does not necessarily proscribe application of the constraint concept to other behavioral contexts that are virtually iso- morphic in function. Isomorphism of behavioral function refers to similarity of purpose with respect to behavior in a setting. One common aspect of the di- verse respondents described above is that they were describing constraints that impeded the effectiveness of behavior directed toward some tangible achievement (performance, mastery, pay). Isomorphism of behavioral function could serve as a substantive system boundary for purposes of identifying in- stances where the constraint concept might appropriately apply. Application of this notion suggests numerous settings are characterized by their emphasis on instrumental behavior including traditional vocational settings, education, and sports. The recent work of Mathieu, Tannenbaum, and Salas (1992) illus- trates one application of constraints in the context of a proofreading skills training program.

    Expansion of the context to which constraints may reasonably apply has several advantages. First, extension of a concept to multiple contexts allows for the possibility that the concept may explain a greater variety of phenomena, thereby augmenting its specific scope. Second, the extension of the concept to

  • 166 HUMAN RESOURCE MANAGEMENT REVIEW VOLUME 3. NUMBER 2.1993

    more contexts allows for more opportunities to test the concept and its atten- dant theoretical framework. Ellsworth (1977, p. 605) notes that all settings have unfavorable quirks and characteristic sources of error and that to avoid these problems, research might need to be pursued in alternative contexts. As Peters et al. (1985) noted, field research on situational constraints has consis- tently reported low levels of constraining factors, rarely exceeding the scale midpoint. This restriction in the range of constraint levels in traditional voca- tional contexts suggests a need to test the effects of constraints in settings that include more elevated levels of constraints. Future research on constraints should follow Ellsworths prescription that the choice of setting for the test of the hypothesis if X, then Y is determined by X. Settings should be chosen in such a way as to maximize the likelihood of finding a good representation of constraints.

    Finally, constraints can be construed as both causes and effects. Brown and Mitchell (1991) investigated employee perceptions of performance obstacles during a conversion from a batch to a just-in-time (JIT) manufacturing system. Their findings indicated that the conversion to a JIT system reduced some obstacles (material delivery, resource availability, information quality) and magnified others (training, computer use, scheduling, reliance on coworkers). More studies that construe constraints as dependent variables are needed in order to better understand how tractable different constraints might be to different types of interventions.

    Modeling of Constraint Relationships

    Another way of advancing alternative conceptualizations of constraints is to focus on modeling of the constraint-motivation/performance relationship. Cur- rently constraints are thought to have a negative, linear relationship with motivation and performance. Perhaps when faced with a work situation char- acterized by a moderate level of constraint, individuals experience increased physiological arousal and attention to the task that facilitates performance. On the other hand, when constraints are severe, individual arousal becomes debilitating in that only obvious remedies to the constraining circumstances are considered and the individual becomes preoccupied with reducing anxiety. Such a model of the constraint-motivation/performance relationship is consis- tent with the Yerkes-Dodson law which states that there is a curvilinear rela- tionship between arousal and performance. Further extension of the Yerkes- Dodson law to the constraint-motivation/performance relationship would include the moderating influence of task complexity. This complicates matters further in that each task is characterized by its own optimal level of constraint induced arousal, depending on its complexity. Moreover, for highly complex tasks, the relationship between constraints and motivation/performance might deviate altogether from the inverted-U-shaped function in that a nega- tive, linear function might better characterize the relationship. This suggests that the relationship between constraints and motivation/performance might

  • ALTERNATIVE CONCEPTUALIZATIONS 167

    be more contingent on task or activity factors than previously considered. Indirect empirical support for more complex modeling of the constraint- motivation/performance relationship comes from the literature on frustration (reviewed by Spector 19781, goal setting (Locke, Shaw, Saari, & Latham 19811, and work stress (McGrath 1976). In fact, several studies have found that the relative scarcity of resources (e.g., time), that would ordinarily be charac- terized as constraining, actually serve to facilitate the rate of work (Andrews & Farris, 1972) and level of performance (Peters et al. 1984). No simple model of constraints seems capable of accounting for these diverse patterns. Moreover, as discussed by Schoorman and Schneider (1988131, a number of logical diffi- culties result from simplifying assumptions such as conceptualizing con- straints and facilitators as two independent dimensions, or conversely, as be- ing extremes along a single continuum.

    To further underscore the significance of task characteristics for the model- ing of constraint effects on outcomes, consider the concept time-span of discre- tion (TSD) introduced by Jaques (1964). TSD is defined as the longest period which can elapse before substandard performance is recognized as such. In simpler terms, it is the time before a persons sins catch up with him (McCor- mick 1979, p. 308). For some jobs, such as those involving production, the TSD is very short, perhaps a couple of hours. For other jobs, the TSD can extend to a period of weeks, months, or years, as in the case of management and executive positions. If the TSD for job performance extends beyond the measurement window of gathering constraint and performance data, the self-reports of con- straints and the supervisory ratings of performance may not be related. Some individuals who report severe levels of constraints might not find these situa- tional obstacles reflected in their performance ratings or accomplishment re- cord because the TSD may extend beyond the period covered by the latest performance review. Also, jobs with a long TSD permit an individual time to respond, perhaps effectively, to performance obstacles before the decline in performance comes to the attention of a supervisor. The idea that constraints may be characterized by different TSDs suggests careful job analysis must lie at the basis of identifying relevant dimensions along which constraints might vary. The current dimensions used to represent the level of constraints (ratings of availability, quantity, and quality summed across resources) may not be the most suitable for representing potential disruption or interference with the translation of abilities and motivation into effective performance. The TSD concept has potential implications for the dimensionality of constraints and the dynamic modeling of constraint-outcome relationships.

    Moreover, the current focus on constraint chronicity is reminiscent of the exclusive focus on chronic work stressors that characterized decades of re- search on work stress (Eden 1982). Perhaps, as persuasively argued by Bobko (19851, more attention needs to be directed at modeling dynamic events, non- linear relationships, and challenging bipolar assumptions. The constraints that have the greatest impact on performance may well be those that are variable as opposed to those that are constant since variability of resource availability and quality may be more disruptive of work continuity. Also, this

  • 168 HUMAN RESOURCE MANAGEMENT REVIEW VOLUME 3, NUMBER 2, 1993

    variability could contribute to perceptions of instability and uncontrollability in the work environment, potentially increasing strain. Should this be the case, then representing the severity of constraints by the average rating on a semantic differential scale fails to capture this variability and hence, is less able to predict important outcomes. The critical incident technique (CIT) was used fruitfully to develop the prevailing taxonomy of situational constraints. Bobko describes one such application of the CIT to measuring the underlying variation of a focal construct.

    The work of Eden (1982) on critical job events (CJE), acute stress, and strain offers yet another way of modeling the debilitating effects of disruptive events. CJE are defined by Eden as time-bounded peak performance demand made on the individual as an integral part of [the] job (p. 314). Examples include firefighters mobilized for extinguishing a fire, air crews during flight emergen- cies, and students sitting for exams. In each case, the event described is tran- sient, may recur intermittently, and may conform to a cyclical pattern. Repre- sentation of CJE and measurement of associated outcomes may be more idiographic and setting specific than more nomothetic and generalizable mea- surement strategies (e.g., questionnaires). Moreover, it could also be that mea- sures of chronic constraints, or steady state measures, are influenced to some extent by discrete events. A series of CJE co-occuring with constraining cir- cumstances and accompanied with occasional respites, might be perceived as stemming from a chronic shortage of necessary resources. The implication as drawn by Eden is clear: measures of both chronic and acute situational condi- tions may augment the prediction equation.

    Individual Differences and Constraints

    The current conceptualization of constraints would seem to ascribe a mini- mal role to individual differences in the reporting of constraint severity. One of the desirable features of a concept with descriptive depth is that it can be defined carefully and in sufficient detail to mitigate against perceptual differ- ences attributable to persons. Yet, Peters et al. (1985) were careful to point out that the systematic influence of factors other than the physical work environ- ment on persons ratings of constraints cannot be ruled out. The systematic influence of one such factor was illustrated in a study by Pooyan, OConnor, Peters, Quick, Jones, and Kulisch (1982). They found substantive differences in subordinate/supervisor perceptions of situational constraints among a sample of bank employees. Subordinates ratings of constraint severity were consis- tently higher than supervisors ratings across twelve constraint dimensions; statistically significant differences were found between seven of the twelve pairs of means. Also, interrater reliabilities of supervisor/subordinate ratings never exceeded .27. These differences in constraint levels attributable to posi- tion differences of respondents suggest that perception-based measures of con- straints might be susceptible to a host of personal and situational influences. Inclusion of person sources of variance in the constraint-motivation/perfor- mance relationship makes the modeling of the putative effects of constraints more involved but it also expands the scope of inquiry.

  • ALTERNATIVE CONCEPTUALIZATIONS 169

    Although numerous individual difference variables might influence percep- tions of and responses to constraining circumstances, we will briefly discuss one in order to illustrate their potential relevance for perception-based mea- sures of situations. Drawing on Gibsons (1979) notion of affordances, Dworkin and Goldfinger (1985) described processing biases in the cognitive appraisal of situations. Affordances refer to characteristics of the environment that facilitate particular behaviors or aims specific to an organism. For exam- ple, some surfaces afford support for resting objects; because humans possess sight, lighting affords us the ability to avoid collision with other objects; and some tools afford manipulation. Dworkin and Goldfinger define processing bias as inclinations or tendencies for individuals to differ in the extent to which they are oriented toward recognizing the availability of different affordances. Individuals placed in a situation that provides affordances consistent with their own processing bias are more apt to know what to look for, how to conduct a perceptual search effectively, and are able to discriminate quickly, effi- ciently, and accurately relevant objects from irrelevant ones (p. 484). Thus, some individuals may be more capable of identifying a greater variety of uses for an affordance if they possess traits that orient them toward that affordance. For example, Type A individuals, who are sensitive to time constraints, could be expected to make greater use of available time when working under time pressure than Type B individuals. Type A individuals would be predicted to perceive time constraints to be more severe than Type B individuals, but be- cause they are oriented to maximizing this affordance, they would also achieve more than Type Bs. Such a perception-response pattern could produce a posi- tive correlation between perceived time constraints and job performance. Fur- ther evidence for the influence of individual differences on perceptions of situa- tions comes from work on perceptions of environmental uncertainty. Downey, Hellriegel, and Slocum (1977) found that cognitive process variables (cognitive complexity, anomie, and independence of judgment) were more strongly re- lated to perceived environmental uncertainty than were perceived environ- mental characteristics (static-dynamic, simple-complex).

    Systemic Validity of the Constraint Concept

    The recent conceptualization of stress as involving perceptions of uncertain- ty involving desired outcomes stemming from encounters with opportunities, constraints, and demands (Beehr & Bhagat 1985; Schuler 1980) suggests some commensurability between the situational constraints framework and theories of work stress. As used here, commensurability refers to the extent to which elements and patterns of relationships constituting one theoretical framework can be compared to another using similar language, methods, problem scope, and explanations. Commensurability is a prerequisite for systemic validity (Zaltman, Pinson, & Angelmar 1973). Systemic validity represents the degree to which a concept facilitates integration of different theoretical systems. Sirgy (1988) describes seven different strategies that can be used to construct unified theories from nominally disparate conceptual domains. Like concept exten- sion, theory integration is desirable in that it, too, has the potential to ex-

  • 170 HUMAN RESOURCE MANAGEMENT REVIEW VOLUME 3. NUMBER 2, 1993

    plain a broader array of phenomena, may do so more parsimoniously, and furthers the aim of scientific unity. Also, Campbell et al. (1982) and others have noted that in almost every instance, the theory or knowledge for generating resolutions to significant problems faced by the subdiscipline of industrial- organizational psychology have been borrowed or transported from other allied disciplines.

    Stress researchers have been at the vanguard of the person-situation de- bate. Perhaps more than any other body of research on organizational phenom- ena, the stress literature reflects a preoccupation with models of person- environment interaction. Given the need to identify environmental sources of stress in order to compliment work on the personal side, it is not surprising that the earliest systematic attempt to include situational constraints as a causal variable came from this body of literature (see above). Theorizing about the processes involved in individual perception of and response to situational conditions might be advanced by embedding constraints in a process model of stress such as that described by Lazarus and colleagues (1981; Lazarus & Launier 1978).

    CONCLUSION

    Our knowledge about situational constraints as a concept, its potential influ- ence on behavior, and its applicability to different contexts remains limited because systematic effort at understanding constraints as a concept, as op- posed to treatment as an artifact for which control might suffice, has occurred only recently. Similar ways of thinking about extraneous sources of variance in criterion measures contributed to the relative lack of attention given to the dynamic nature of criteria (Austin & Villanova 1992). However, once a phenom- enon is reliably demonstrated, the need for a conceptual framework to explain its occurrence becomes more urgent. Just the same, original conceptualiza- tions are frequently incomplete or erroneous and may require alternative con- ceptual schemes that hold the potential for more frequent comirmatory evi- dence than an existing framework. We hope that the foregoing discussion stimulates further thought about constraints and promotes research that ex- plores the implications of different conceptual schemes for representing con- straints and their putative effects on work outcomes.

    APPENDIX STUDIES INCLUDED IN THE META-ANALYSIS

    Jamal, M. 1984. Job Stress and Job Performance Controversy: An Empirical Assess- ment. Organizational Behavior and Human Performance 33: 1, 21.

    -. 1990. Relationship of Job Stress and Type-A Behavior to Employees Job Satis- faction, Organizational Commitment, Psychosomatic Health Problems, and Turnover Motivation. Human Relations 43: 727-738.

  • ALTERNATIVE CONCEPTUALIZATIONS 171

    OConnor, E. J., L. H. Peters, A. Pooyan, J. Weekly, B. Frank, and B. Erenkrantz. 1984. Situational Constraint on Performance, Affective Reactions, and Turnover: A Field Replication and Extension. Journal of Applied Psychology 69: 663-672.

    OConnor, E. J., L. H. Peters, C. J. Rudolph, and A. Pooyan. 1982. Situational Con- straints and Employee Affective Reactions: A Field Replication. Group and Or- ganization Studies 7: 418-428.

    Olson, D. M. and W. C. Borman. 1989. More Evidence on Relationships between the Work Environment and Job Performance. Human Performance 2: 113-130.

    Parasuraman, S. and J. A. Alutto. 1984. Sources and Outcomes of Stress in Organiza- tional Settings: Toward the Development of a Structural Model. Academy of Management Journal 27: 330-350.

    Parasuraman, S. and D. Hansen. 1987. Coping with Work Stressors in Nursing: Effects of Adaptive versus Maladaptive Strategies. Work and Occupations 14: 88-105.

    Peters, L. H., M. B. Chassie, H. R. Lindholm, E. J. OConnor, and C. R. Kline. 1982. The Joint Influence of Situational Constraint and Goalsetting on Performance and Affective Outcomes. Journal of Management 8: 7-20.

    Peters, L. H., E. J. OConnor, J. R. Eulberg, and T. W. Watson. 1988. An Examination of Situational Constraints in Air Force Work Settings. Human Performance 1: 133- 144.

    Peters, L. H., E. J. OConnor, and C. J. Rudolph. 1980. The Behavioral and Affective Consequences of Performance-Relevant Situational Variables. Organizational Behavior and Human Performance 25: 79-96.

    Peters, L. H., E. J. OConnor, A. Pooyan, and J. C. Quick. 1984. The Relationship between Time Pressure and Performance: A Field Test of Parkinsons Law. Jour- nal of Occupational Behavior 5: 293-299.

    Phillips, J. S., and S. M. Freedman. 1984. Situational Performance Constraints and Task Characteristics: Their Relationship to Motivation and Satisfaction. Jour- nal of Management 10: 321-331.

    Spector, P. E., D. J. Dwyer, and S. M. Jex. 1988. Relation of Job Stressors to Affective, Health, and Performance Outcomes: A Comparison of Multiple Data Sources. Journal of Applied Psychology 73: 11-19.

    Steel, R. P., and A. J. Mento. 1986. Impact of Situational Constraints on Subjective and Objective Criteria of Managerial Performance. Organizational Behavior and Human Decision Processes 37: 254-265.

    Steel, R. P., A. J. Mento, and W. H. Hendrix. 1987. Constraining Forces and the Work Performance of Finance Company Cashiers. Journal of Management 13: 473- 482.

    REFERENCES

    Andrews, F. M. and G. F. Farris. 1972. Time Pressure and Performance of Scientists and Engineers: A Five-year Panel Study. Organizational Behavior and Human Performance 8: 185-200.

    Austin, J. T. and P. Villanova. 1992. The Criterion Problem: 1917-1992. Journal of Applied Psychology 77: 836-874.

    Beehr, T. A. and R. S. Bhagat. 1985. Introduction to Human Stress and Cognition in Organizations. Pp. 3-19 in Human Stress and Cognition in Organizations, edit- ed by T. A. Beehr and R. S. Bhagat. New York: Wiley.

  • 172 HUMAN RESOURCE MANAGEMENT REVIEW VOLUME 3. NUMBER 2, 1993

    Bobko, P. 1985. Removing Assumptions of Bipolarity: Towards Variation and Circu- larity. Academy of Management Review 10: 99-108.

    Brinberg, D. and J. E. McGrath. 1985. Validity and the Research Process. Beverly Hills: Sage.

    Brown, K. A. and T. R. Mitchell. 1991. A Comparison of Just-in-time and Batch Manu- facturing: The Role of Performance Obstacles.Acao!emy of Management Journal 34: 906-917.

    Campbell, J. P., R. L. Daft, and C. L. Hulin. 1982. What to Study: Generating and Developing Research Questions. Beverly Hills, CA: Sage.

    Campbell, J. P., M. D. Dunnette, E. E. Lawler, and K. E. Weick. 1970. Managerial Behavior, Performance, and Effectiveness. New York: McGraw-Hill.

    Campbell, J. P. and R. D. Pritchard. 1976. Motivation Theory in Industrial and Organi- zational Psychology. Pp. 63-130 in Handbook of Industrial and Organizational Psychology, edited by M. D. Dunnette. Chicago: Rand McNally.

    Dillon, J. T. 1984. The Classification of Research Questions. Review of Educational Research 54: 327-361.

    Downey, H. K., D. Hellreigel, and J. W. Slocum. 1977. Individual Characteristics as Sources of Uncertainty Variability. Human Relations 30: 161-174.

    Dworkin, R. H. and S. H. Goldfinger. 1985. Processing Bias: Individual Differences in the Cognition of Situations. Journal of Personality 53: 480-501.

    Eden, D. 1982. Critical Job Events, Acute Stress, and Strain: A Multiple Interrupted Time Series. Organizational Behavior and Human Performance 30: 312-329.

    Ellsworth, P. C. 1977. From Abstract Ideas to Concrete Instances: Some Guidelines for Choosing Natural Research Settings.American Psychologist 32: 604-615.

    Fromkin, H. L. and S. Streufert. 1976. Laboratory Experimentation. Pp. 415-465 in Handbook of Industrial and Organizational Psychology, edited by M. D. Dun- nette. Chicago: Rand McNally.

    Gibson, J. J. 1979. The Ecological Approach to Visual Perception. Boston: Houghton Mifflin.

    Gilbreth, F. B. 1909. Bricklaying System. Chicago: Myron C. Clark. Hulin, C. L., M. Rosnowski, and D. Hachiya. 1985. Alternative Opportunities and

    Withdrawal Decisions: Empirical and Theoretical Discrepancies and an Integra- tion. Psychological Bulletin 97: 233-250.

    Hunter, J. E. and F. L. Schmidt. 1990. Methods of Meta-analysis: Correcting Error and Bias in Research Findings. Newbury Park, CA: Sage.

    Hunter, J. E., F. L. Schmidt, and G. B. Jackson. 1982. Meta-anulysis: Cumulating Research Findings Across Studies. Beverly Hills, CA: Sage.

    Ivancevich, J. M., M. T. Matteson, S. M. Freedman, and J. S. Phillips. 1990. Worksite Stress Management Interventions. American Psychologist 45: 252-261.

    Jamal, M. 1984. Job Stress and Job Performance Controversy: An Empirical Assess- ment. Organizational Behavior and Human Performance 33: 1-21.

    -. 1990. Relationship of Job Stress and Type-A Behavior to Employees Job Satis- faction, Organizational Commitment, Psychosomatic Health Problems, and Turnover Motivation. Human Relations 43: 727-738.

    Jaques, E. 1964. Time-span Handbook. London: Heineman Educational Books. Kahn, R. L. and R. P. Quinn. 1970. Role Stress: A Framework for Analysis. In Occupa-

    tional Mental Health, edited by A. McLean. Chicago: Rand McNally. Kahn, R. L., D. M. Wolfe, R. P. Q umn, J. D. Snoek, and R. A. Rosenthal. 1964. Organiza-

    tional Stress: Studies in Role Conflict and Ambiguity. New York: Wiley.

  • ALTERNATIVE CONCEPTUALIZATIONS 173

    Katz, D. and R. L. Kahn. 1978. The Social Psychology of Organizations (2nd ed.). New York: Wiley.

    Lazarus, R. S. 1981. The Stress and Coping Paradigm. Pp. 174-214 in Models for Clinical Psychopathology, edited by C. Eisdorfer, D. Cohen, A. Kleinman, and P. Maxim. New York: Spectrum.

    Lazarus, R. S. and R. Launier. 1978. Stress-related Transactions between Person and Environment. Pp. 287-327 in Perspectives in interactional psychology, edited by L. Pervin and M. Lewis. New York: Plenum.

    Locke, E. A., K. N. Shaw, L. M. Saari, and G. P. Latham. 1981. Goal-setting and Task Performance: 1969-1980. Psychological Bulletin 90: 125-152.

    Lynch, J. G. 1982. On the External Validity of Experiments in Consumer Research. Journal of Consumer Research 9: 225-239.

    Mathieu, J. E., S. I. Tannenbaum, and E. Salas. 1992. Influences of Individual and Situational Characteristics on Measures of Training Effectiveness. Academy of Management Journal 35: 828-847.

    McCormick, E. J. 1979. Job Analysis. New York: Amacom. McGrath, J. E. 1976. Stress and Behavior in Organizations. In Handbook of Zndus-

    trial/Organizational Psychology, edited by M. D. Dunnette. Chicago: Rand McNally.

    Moeller, A., B. Schneider, F. D. Schoorman, and E. Berney. 1988. Development of the Work-Facilitation Diagnostic. Pp. 79-104 in Facilitating Work Effectiveness, edited by F. D. Schoorman and B. Schneider. Lexington, MA: Lexington.

    OConnor, E. J., L. H. Peters, A. Pooyan, J. Weekly, B. Frank, and B. Erenkrantz. 1984. Situational Constraint on Performance, Affective Reactions, and Turnover: A Field Replication and Extension. Journal of Applied Psychology 69: 663-672.

    OConnor, E. J., L. H. Peters, C. J. Rudolf, and A. Pooyan. 1982. Situational Constraints and Employee Affective Reactions: A Field Replication. Group and Organization Studies 7: 418-428.

    Olson, D. M. and W. C. Borman. 1989. More Evidence on Relationships between the Work Environment and Job Performance. Human Performance 2: 113-130.

    Osigweh, C. A. B. 1989. Concept Fallibility in Organizational Science. Academy of Management Review 14: 579-594.

    Parasuraman, S. and J. A. Alutto. 1984. Sources and Outcomes of Stress in Organiza- tional Settings: Toward the Development of a Structural Model. Academy of Management Journal 27: 330-350.

    Parasuraman, S. and D. Hansen. 1987. Coping with Work Stressors in Nursing: Effects of Adaptive versus Maladaptive Strategies. Work and Occupations 14: 88-105.

    Pearlman, K., F. L. Schmidt, and J. E. Hunter. 1980. Validity Generalization Results for Tests Used to Predict Job Proficiency and Training Success in Clerical Occu- pations. Journal of Applied Psychology 65: 373-406.

    Peters, L. H., M. B. Chassie, H. R. Lindholm, E. J. OConnor, and C. R. Kline. 1982. The Joint Influence of Situational Constraint and Goalsetting on Performance and Affective Outcomes. Journal of Management 8: 7-20.

    Peters, L. H., C. D. Fisher, and E. J. OConnor. 1982. The Moderating Effect of Situa- tional Control of Performance Variance on the Relationship between Individual Differences and Performance. Personnel Psychology 35: 609-621.

    Peters, L. H. and E. J. OConnor. 1980. Situational Constraints and Work Outcomes: The Influences of a Frequently Overlooked Construct. Academy of Management Review 5: 391-397.

  • 174 HUMAN RESOURCE MANAGEMENT REVIEW VOLUME 3. NUMBER 2,1993

    Peters, L. H., E. J. OConnor, and J. R. Eulberg. 1985. Situational Constraints: Sources, Consequences and Future Considerations. Pp. 79-l 13 in Research in Personnel and Human Resources Management, edited by K. Rowland and G. Ferris. Green- wich, CT: JAI Press.

    Peters, L. H., E. J. OConnor, J. R. Eulberg, and T. W. Watson. 1988. An Examination of Situational Constraints in Air Force Work Settings. Human Performance 1: 133- 144.

    Peters, L. H., E. J., OConnor, and C. J. Rudolf. 1980. The Behavioral and mective Consequences of Performance-relevant Situational Variables. Organizational Behavior and Human Performance 25: 79-96.

    Peters, L. H., E. J. OConnor, A. Pooyan, and J. C. Quick. 1984. The Relationship between Time Pressure and Performance: A Field Test of Parkinsons Law. Jour- nal of Occupational Behavior 5: 293-299.

    Phillips, J. S. and S. M. Freedman. 1984. Situational Performance Constraints and Task Characteristics: Their Relationship to Motivation and Satisfaction. Jour- nal of Management 10: 321-331.

    Pooyan, A., E. J. OConnor, L. H. Peters, J. C. Quick, N. D. Jones, and A. Kulisch. 1982. Supervisory/Subordinate Differences in Perceptions of Situational Constraints: Barriers Are in the Eye of the Beholder. Presented at the Southwest Academy of Management Meeting.

    Rosenthal, R. 1984. Meta-Analytic Procedures for Social Research. Beverly Hills, CA: Sage.

    Sackett, P. R., M. M. Harris, and J. M. Orr. 1986. On Seeking Moderator Variables in the Meta-analysis of Correlational Data: A Monte Carlo Investigation of Statisti- cal Power and Resistance to Type 1 Error. Journal of Applied Psychology 71: 302-310.

    Salancik, G. R. and J. Pfeffer. 1977. An Examination of Need-Satisfaction Models of Job Attitudes. Administrative Science Quarterly 22: 427-456.

    Schmitt, N. W. and R. J. Klimoski. 1991. Research Methods in Human Resources Man- agement. Cincinnati, OH: Southwestern Publishing.

    Schoorman, F. D. and B. Schneider. 1988a. Grappling with Work Facilitation: An Evolving Approach to Understanding Work Effectiveness. Pp. 3-20 in Facilitat- ing Work Effectiveness, edited by F. D. Schoorman and B. Schenider. Lexington, MA: Lexington. . 1988b. Integration and Overview of the Research on Work Facilitation. Pp. 215-230 in Facilitating Work Effectiveness, edited by F. D. Schoorman and B. Schneider. Lexington, MA: Lexington.

    Schuler, R. S. 1980. Definition and Conceptualization of Stress in Organizations. Organizational Behavior and Human Performance 25: 184-215.

    Sirgy, M. J. 1988. Strategies for Developing General Systems Theories. Behavioral Science 33: 25-37.

    Spector, P. E. 1978. Organizational Frustration: A Model and Review of the Litera- ture. Personnel Psychology 31: 815-829.

    Spector, P. E., D. J. Dwyer, and S. M. Jex. 1988. Relation of Job Stressors to Affective, Health, and Performance Outcomes: A Comparison of Multiple Data Sources. Journal of Applied Psychology 73: 11-19.

    Steel, R. P., and A. J. Mento. 1986. Impact of Situational Constraints on Subjective and Objective Criteria of Managerial Performance. Organizational Behavior and Human Decision Processes 37: 254-265.

  • ALTERNATIVE CONCEPTUALIZATIONS 175

    Steel, R. P., A. J. Mento, and W. H. Hendrix. 1987. Constraining Forces and the Work Performance of Finance Company Cashiers. Journal of Management 13: 473- 482.

    Whitener, E. M. 1990. Confusion of Confidence Intervals and Credibility Intervals in Meta-analysis. Journal of Applied Psychology 75: 315-321.

    Wicker, A. W. 1985. Getting Out of Our Conceptual Ruts: Strategies for Expanding Conceptual Frameworks. American Psychologist 40: 1094-1103.

    Zaltman, G., C. Pinson, and R. Angelmar. 1973. Metatheory in Consumer Research. New York: Wiley.