AN EXPERIMENTAL TEST OF, AN EXPECTANCY THEORY 1 HYPOTHESIS: _ THE MULTIPUCATIVE COMBINATION OF _ EXPECTANCY, lNSTR-UMENTALITY, ANO VALENCE Thesis fOT the Degree Of M. A. . MICHIGAN STATE UNIVERSITY . - JOHN MICHAEL RAIISCHENBERCER - - ' -1975 g' ' “ III III III III III III IIII II IIII IIIIIIIIIII III I.“ JI-IESIS ABSTRACT AN EXPERIMENTAL TEST OF AN EXPECTANCY THEORY HYPOTHESIS: THE MULTIPLICATIVE COMBINATION OF EXPECTANCY, INSTRUMENTALITY, AND VALENCE BY John Michael Rauschenberger Expectancy theory models of work motivation specify multiplicative relationships among three variables defined as expectancy, instrumentality, and valence. An experimental test of this assumption was conducted to determine if hypothesized interaction effects among these variables, with the presence of main effects, existed. Each of 120 college-aged male undergraduate students at a large mid- western university received either high or low treatments of expectancy (positive feedback on a test vs. negative feedback on a test), instrumentality (a modified piece-rate vs. a salary), and valence ($1.50/30 minutes of work or a 7¢ piece-rate vs. $0.50/30 minutes of work or a 3¢ piece- rate). A 2 X 2 X 2 factorial analysis of variance revealed no significant main effects or interactions for task performance, an expectancy main effect for performance satisfaction, and an E X I X V interaction for willingness to perform the task again. John Michael Rauschenberger Implications of these results were discussed and related to previous research and some suggestions were offered for future research attempts to study expectancy theory's multiplicative combination rule. Approved by Thesis Committee: Dr. John Wakeley, Chairman Dr. Neal Schmitt Dr. Carl Frost AN EXPERIMENTAL TEST OF AN EXPECTANCY THEORY HYPOTHESIS: THE MULTIPLICATIVE COMBINATION OF EXPECTANCY, INSTRUMENTALITY, AND VALENCE BY John Michael Rauschenberger A THESIS Submitted to Michigan State University in partial fulfillment of the requirements for the degree of MASTER OF ARTS Department of Psychology 1975 ACKNOWLEDGMENTS My thanks go out to many: To my chairman, Dr. John Wakeley, who taught a questioning graduate student how to become a researcher, and had the patience to stay with me to see that the job was done properly. To Dr. Neal Schmitt, who should have dreaded hearing my footsteps by his office door, but rather, he was there with help and encouragement above and beyond the call of duty. To Dr. Carl Frost, who heard my call for help one day, and generously answered my plea. To Dr. Clay Hamner, for his many helpful comments and methodological suggestions. I To Bryan Coyle, who helped with the data analysis. To my undergraduate assistants, Mark Jenkins, Pat Knight, Cindy Phipps, and Jim Wilson, who had the patience to stay with me when the going got tough. To Alice Hogle and Alicia Crenshaw for typing earlier rough drafts. To my Uncle Edward C. Lapinski and his family, who not only helped me build an apparatus, but helped me realize how fortunate I am to have them as relatives. ii Especially to my parents, whose patience, understand- ing, generosity, and love I often took for granted during some difficult days with this task. iii TABLE OF CONTENTS LIST OF TABLES . . . . . . . . . . . . . . . . . . . . Vi LIST OF FIGURES . . . . . . . . . . . . . . . . . . . Vii INTRODUCTION . . . . . . . . . . . . . . . . . . . . . 1 REVIEW OF THE LITERATURE . . . . . . . . . . . . . . . 8 "E" V Compared With Basic EIV . . . . . . . . . 8 Single-Factor Component Compared With Basic EIV . . . . . . . . . . . . . . . . . . . 9 Two-Factor Component Compared With Basic EIV . . . . . . . . . . . . . . . . . . . 10 EV Weighted by I . . . . . . . . . . . . . . . 12 El Weighted by V . . . . . . . . . . . . . . . 12 EIV Compared With E + I + V . . . . . . . . . . 14 Some Criticisms . . . . . . . . . . . . . . . . 15 Experimental Studies . . . . . . . . . . . .I. I 19 VARIABLES . . . . . . . . . . . . . . . . . . . . . . 22 Independent . . . . . . . . . . . . . . . . . . 22 Expectancy (E) . . . . . . . . . . . . . 22 Instrumentality (I) . . . . . . . . . . 22 Valence (V) . . . . . . . . . °_‘ . . . 23 Dependent . . . . . . . . . . . . . . . . . . . 23 Performance (P) . . . . . . . . . . . . . 23 Satisfaction (S) . . . . . . . . . . . . . 23 Willingness (W) . . . . . . . . . . . . . 24 iv Hypotheses . . . . . . . . . . . . . . . . . METHOD . . . . . . . . . . . . . . . . . . . . . . . Subjects . . . . . . . . . . . . . . . . . . Apparatus . . . . . . . . . . . . . . . . . . Procedure . . . . . . . . . . . . . . . . . . RESULTS AND DISCUSSION . . . . . . . . . . . . . . . APPENDICES . . . . . . . . . . . . . . . . . . . . . A. DIAGRAM OF THE PEG BOARD APPARATUS AND A B. COPY OF THE PURPOSE OF THE EXPERIMENT GIVEN TO SUBJECTS IN THE 8 TREATMENT CONDITIONS . . . . . . . . . . . . . . . C. COPY OF THE CLERICAL VISUAL PERFORMANCE TEST/ANSWER KEY . . . . . . . . . . . . . D. STANINE CHART FOR THE CLERICAL VISUAL PERFORMANCE TEST . . . . . . . . . . . . E. EXPECTANCY TABLE FOR THE CLERICAL VISUAL PERFORMANCE TEST . . . . . . . . . F. POSTEXPERIMENTAL QUESTIONNAIRE . . . . . G. VISUAL DEBRIEFING HANDOUT . . . . . . . .- FOOTNOTES . . . . . . . . . . . . . . . . . . . . . BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . SAMPLE DATA CARD . . . . . . . . . . . . 24 26 26 26 27 32 44 44 45 46 48 49 50 52 53 54 LIST OF TABLES Table l. 2 X 2 X 2 Analysis of Variance . . . . . . . . . . 32 2. t-test Manipulation Checks of the Expectancy, Instrumentality, and Valence Variables . . . . . . 33 3. Performance Means and Standard Deviations for the 8 Experimental Conditions . . . . . . . . . . 35 vi Figure 1. LIST OF FIGURES Graph of the E X I X V Interaction for Performance . . . . . . . . . . . . . vii 39 INTRODUCTION For many years, industrial and organizational psy- chologists have been concerned with the problem of human motivation in the work environment. Lawler (1971) notes that a number of psychological theories of motivation have been used in an attempt to explain worker motivation, although none of them were conceived with this sole purpose in mind. While admitting that there are a great many theories of motivation, he has identified those which have had the greatest impact on current thinking in work motiva- tion, especially with respect to pay as a motivator of behavior. One theory which has received a great deal of atten- tion in the last decade is expectancy theory. Although its early beginnings were rooted in hedonism, the theory became more formalized with the work of Tolman (1932) and Lewin (1938). Basically, theirs was a cognitive approach which hypothesized that individuals tend to choose, among a number of possible outcomes, those outcomes which will minimize their losses and maximize their gains. In other words, individuals tend to evaluate the consequences of their behavior on a cognitive level, and choose those most appropriate for the situation of interest. A number of theorists (most notably Edwards, Atkinson, Rotter, Peak, 1 and Vroom) have taken these notions and "built" expectancy theories of their own. Although these theories differ somewhat in terminology, they all contain the notion that force to perform is related to a multiplicative function of expectancy and valence. With the exception of Vroom (1964), these expectancy theories were not designed with work motivation as the primary focus. Nevertheless, an early test of some of these notions in the work environment was attempted by Georgopoulos, Mahoney, and Jones (1957). Six—hundred and twenty-one subjects in a medium-sized, nonunionized, house- hold appliances company were asked for their path-goal perceptions of a number of job related goals, specifically, how instrumental high (or low) productivity was for attaining these goals. Other independent variables were level of need (for goal attainment) and level of freedom (on the job). The dependent variable of interest was a self? reported measure of productivity. Employing a Chi-squared analysis, Georgopoulos, EE.El° found significant differences in percent-responses in the predicted directions to conclude that productivity was a function of these three variables. A "modified multiple regression technique" was employed in an attempt to explain the percent of variance in the dependent measure explained by the path-goal perceptions, which was "modest" (no multiple R given). They conclude that while these variables are important, other variables, such as work setting, effort, and social factors should be considered in future applications of path—goal studies. As mentioned earlier, Vroom (1964) was the first theorist to formalize expectancy theory notions for the work setting. He specified two combination rules to explain the interactive effects of: (l) Expectancy (E)--a behavior- outcome subjective probability that actions will lead to outcomes; (2) Instrumentality (I)--the degree to which an outcome(s) is perceived to be followed by another out- come(s); (3) Valence (V)--the extent to which the individual is satisfied with the outcome(s). (Other researchers have identified V as "importance," "value," and "attractive- ness" of outcomes; a conceptual problem which will be noted later.) The first rule states that, The valence of an outcome to a person is a monotonically increasing function of the algebraic sum of the product of the valences of all the outcomes and his conceptions of its instrumentality for the attainment of these other outcomes (Vroom, 1964, p. 17). Symbolically, Vj = 23(Vk Ijk) for all j and k outcomes. The second rule contends that, The force on a person to perform an act is a monotonically increasing function of the algebraic sum of the products of the valences of all outcomes and the strength of his expectancies that the act will be followed by the attainment of these outcomes (Vroom, 1964, p. 18). Symbolically, Fi = Z(Eij Vj) for all i acts and j outcomes. The values of the E, I, and \7 variables are hypothesized to range as follows: E from 0.0 to +1.0 (i.e., a probability) I from -1.0 to +1.0 (i.e., a correlation) V from —1.0 to +1.0 Theories and models which contain these variables have been labeled in different ways. At times they are called "expectancy" theories, "instrumentality" theories, "instrumentality-valence" theories, and "EIV" theories. The last notation will be adopted for this paper. Actually these two rules provide a conceptual basis for two different models of work motivation (Mitchell and Biglan, 1971). Rule 1, V = 2(VI), can be viewed as a job satisfaction model, while Rule 2, F = 2(EV), can be inter— preted as a job performance model. These models (as well as others which shall be mentioned shortly) have been used to predict a variety of criteria, such as leadership behavior (House, 1971; Nebeker and Mitchell, 1974), academic effort and performance (Mitchell and Nebeker, 1973), and job satisfaction (Porter and Lawler, 1968; Graen, 1969; Mitchell and Albright, 1972). Most of the research emphasis has, however, been concentrated on job effort and/or job per- formance (Georgopoulos, Mahoney, and Jones, 1957; Galbraith and Cummings, 1967; Lawler and Porter, 1967; Hackman and Porter, 1968; Lawler, 1968; Goodman, Rose, and Furcon, 1970; Schuster, Clark, and Rogers, 1971; Arvey, 1972; Arvey and Mussio, 1973; Dachler and Mobley, 1973; Jorgenson, Dunnette, and Pritchard, 1973; Lawler and Suttle, 1973; Pritchard and DeLeo, 1973; Pritchard and Sanders, 1973; Schwab and Dyer, 1973; Gavin, 1974; and Turney, 1974). Vroom's ideas have been questioned by a number of theorists (Galbraith and Cummings, 1967; Lawler and Porter, 1967; Porter and Lawler, 1968; Graen, 1969; Campbell, Dunnette, Lawler, and Weick, 1970; Lawler, 1971; and Dachler and. Mobley, 1973). These researchers have con- structed their own expectancy theories to expand on and clarify some of Vroom's original notions. Galbraith and Cummings (1967), for example, included various intrinsic and extrinsic outcome factors, as well as work role variables, ego involvement and ability measures. In like manner, Lawler and Porter (1967) also consider abilities, role perceptions, and effort, but hypothesize that effort is a function of the value of rewards and an effort-—+ rewards probability. This last factor is actually a combination of E and I, reducing the EIV model to an "E" V model (Berner, 1974). This model has been subsequently redefined by Porter and Lawler (1968) and Campbell, gt al. (1970) to its most recent definition forwarded by Lawler (1971). This highly expanded version of the model no longer includes the combined "E" V notions, but rather separates "E" into very distinct components. The main portions of this model are as follows: 6 Z IIE-—*P) x Z[(P——*O) (V)] ——+ effort -—+ performance where E -» P probability that effort will lead to performance P -—+ O = probability that performance will lead to outcomes V = valence of outcomes (Lawler, 1971, p. 108). A number of other variables (such as self esteem, personal experiences, internal and external controls, equity of outcomes, need satisfaction, problem solving approaches, and ability) enter the model at various stages. Graen (1969) offered a model which included work role attractions and work behavior probabilities as well as the notion of boundary conditions for the theory. In other words, predictions of work satisfaction and/or performance should be tempered with knowledge of organizational and job related work conditions which, according to Graen, affect the theory's predictability. Dachler and Mobley (1973) proposed a model which includes many of the revisions already mentioned plus a few more. Most notable among these additions are the concepts of performance utility, work outcome desirability and task goals. Perhaps the best explanation to account for these different models can be traced back to a couple of problems that have plagued expectancy theory in general and the Vroom (1964) definition in particular. One of these is the lack of distinction between "actions" and "outcomes" and the expectancies associated with each (Lawler and Suttle, 1973). The other problem is the attempt to identify additional variables in the model which account for the unexplained variance in performance and satisfaction measures after E, I, and V have been explained. Actually, both of these problems share a common thread, i.e., the combination of variables in the theory. (All of the models reviewed thus far have contained multiplicative functions of E, I, and V, although other combinatorial functions are possible. REVIEW OF THE LITERATURE Although there have been reviews of expectancy theory literature (Mitchell and Biglan, 1971; Heneman and Schwab, 1972; Mitchell, 1974) only one source (Berner, 1974) has reviewed these studies in terms of their combinatorial functions. This review draws heavily from his findings. "E" V Compared With Basic EIV Only one study has attempted to compare the Lawler and Porter (1967) "E" V model with the basic EIV model. Lawler and Suttle (1973) in a study of 69 managers in 6 retail stores, found no significant differences in the correlations between effort rankings by self, boss, or peers, with a number of different types of expectancy measures. E-—+ 0 measures correlated .31, .25, (p < .05) and .18 I respectively with the criteria, while (E—-+ P) 2(P-—» O) correlated .39, .29, (p < .01) and .15. The addition of V did not increase the predictability of the expectancy measures. Although some support is evidenced for both models, the "E" V approach has received less attention in the liter— ature than the basic EIV model, and, as noted earlier, others have subsequently redefined the "E" V model to include distinct E, I, and V components. 8 Single-Factor Component Compared With Basic EIV Four studies have compared the predictability of individual E, I, and V components with the full EIV model. In this case, Lawler and Suttle (1973) found E to correlate .37, .28, (p < .01) and .05 with self, boss, and peer rankings respectively, while I correlated .29, (p < .05), .19, and .22 (p < .05). No correlations were computed for V. The full EIV model correlated .39, .28, (p < .01) and .16 with these criteria. Schwab and Dyer (1973) in a study of 124 workers in a large manufacturing facility of consumer goods in the midwest found E to correlate with performance .38 (p < .01), I, .13, and V, .17 (p < .05). The EIV model correlated .39 (p < .01) with this criterion. Graen (1969) employed 169 female workers for two number rounding tasks under three working conditions. In an achievement condition (prompting climate), subjects were led to believe rewards were inducements to effective per- formance. In the money condition (reciprocating climate), rewards were viewed as contingent upon effective performance. A control condition, in which rewards were neither contingent upon nor inducements to effective performance was also established. For rounding task 1, E correlated .33 (p < .01), -.23 (p < .05), and .07 with posttreatment gain in performance scores in the three conditions respectively, while the full E XIV model correlated .28 (p < .05), -.15, and .04 with 10 the criterion. Similar results were found for the second task. Pritchard and Sanders (1973) asked 146 postal employees in a 30 hour training program to rate various job factors in E, I, and V terms via Likert scaled measures. Also, self-reported effort, supervisory rated effort, and supervisory performance measures were obtained as criteria measures. They found E to correlate .13, .01, and .00 with these criteria respectively, I correlated .22, -.02, and .02, while V correlated .54, .22, and .24. The full E(IV) correlated .47, .16 and .17. Pritchard and Sanders conclude that while the full EIV model is fairly good at predicting self-reported effort, V is the best single predictor. The results of these studies offer mixed findings. The general trend appears to be that, in any particular study, the single factor components predicted the criteria of interest as effectively as the full EIV model, however, these single factor components varied from study to study. Two-Factor Component Compared With Basic EIV Four studies have investigated the predictability of the full EIV model with various two-factor combinations of E, I, and V. As noted earlier, Lawler and Suttle (1973) found E ZIV to correlate .39 (p < .01), .28 (p < .01) and .16 with self, boss, and peer ranked effort respectively. XIV products correlated .31 (p < .05), .17, and .20 (p < .05) 11 with these criteria, while EZI correlated .39 (p < .05), .29 (p < .01) and .15. In like manner, Pritchard and Sanders (1973) found IV to correlate .50, .16, and .17 with self-reported effort, supervisory effort, and supervisory performance respectively, while VE correlated .52, .21, and .23 with these criteria. As noted earlier, the full EIV model correlated .47, .16, and .17. Dachler and Mobley (1973) calculated actual perfor- mance scores at five levels for each subject at two plants. Plant 1 consisted of 181 predominantly female sewing workers, while plant 2 subjects were 412 male production workers. They found that adding E to the ZIV scores improved pre- dictability in plant 1 (r = .04 to r = .30, p < .01) while no difference was evident in plant 2 (r = .12, p < .05 to r = .1l,p < .05). Mitchell and Albright (1973), in a study of 51 naval aviation officers, compared I, V, and two E scores for each subject with self and superior rated effort and performance. The first E score (E1) was obtained by having the subject select from among five statements the one which best described how hard he would have to work to perform effectively, while the second E score (E2) was obtained by having the officers select from among five statements the one which best described how his performance would increase if he raised his effort level significantly. 12 They found that EIV scores correlated .22 and .50 (p < .01) with superior and self-rated effort and .29 (p < .05) and .36 (p < .01) with superior and self-reported perfor- mance. E XIV correlated .26 (p < .05), .64 (p < .01), l and .31 (p < .05) and .19 with these criteria respectively, while EZZIV correlated .01, .25 (p < .05), .16, and .08. From this evidence, Dachler and Mobley conclude that weighting IV by E does not improve the predictability of effort or performance scores. As with the single-factor studies, the two-factor approach seems to offer mixed results. Once again, the full EIV model appears, in general, to be as informative as the various two-factor combinations in predicting criterion variables of interest. The notion of weighting variables in the model has received further attention by a few investigators. EV Weighted by I A couple of studies have considered the importance of weighting EV by I. Pritchard and Sanders (1973) and Arvey and Mussio (1973) found no support for the notion that I contributed to the predictability of job effort and job performance. EI Weighted byAV Some support, however, has been found for weighting EI by V. Lawler and Porter (1967), employing the previously 13 mentioned "E" V model, obtained expectancy-reward perceptions and importance scores for seven job related variables (pay, promotion, prestige, security, autonomy, friendship, use of skills and abilities) from 154 managers in five organi- zations (ranging from a manufacturing firm to a YMCA group). Criteria measures were superiors, peers, and self-ratings of job effort and performance. They found the "E" measure to correlate with job performance .17, .21, and .25 with the three criteria measures respectively, while job effort correlated .22, .25, and .32 (p < .05). After weighting "E" by the importance scores (V), the correlations with performance were .18, .21, and .38 (p < .01) respectively, while job effort correlations were .27 (p < .10), .30 (p < .05) and .44 (p < .01). Hackman and Porter (1968) obtained similar support in their study of 82 female telephone service representa— tives. These subjects rated a number of different job related behaviors in "E" and V terms via a 7-point Likert scale. Criteria were job involvement and effort ratings, an Employee Appraisal Form, error rates, sales data, and a composite criterion. They found a median correlation of .11 for the "E" scores. When weighted by V, the median correlation was .27. Gavin (1974) failed to demonstrate this relationship. One hundred and ninety—two male and.ll75 female managerial candidates at a large insurance company rated 21 rewards 14 in "E" and V terms on 9-point Likert scales. Nine dimensions of superior ratings served as criteria. Gavin found Z"E" correlations with these criteria to be .28 (p < .01) for females and .25 (p < .01) for males. Weighting Z"E" by V resulted in correlations of .22 (p < .01) for females and .28 (p < .01) for males. The correlation between "E" and Z"E" V was .91, which Gavin sites as evidence for the lack of additional predictability of the V weighted relationship. Although a few instances of rather mild success are reported, the bulk of the evidence indicates that these weighting strategies have not been an effective way to reduce unexplained criterion variance. EIV Compared With E + I + V Some researchers have attempted to directly test the multiplicative combination rule with correlational techniques. Two of these studies (Galbraith and Cummings, 1967; and Schwab and Dyer, 1973) attempted to use multiple correlation techniques to assess the additive and multiplicative relation- ships of the E, I, and V variables. However, as Berner (1974) points out, multiple correlation techniques are "incongruent with the expectancy theory model" (p. 15). Hackman and Porter (1968) found a median correlation of .27 with their Z"E" V model, while the additive (2"E" + V) model median correlation was .17 (employing the same criteria as noted earlier in this review). 15 Pritchard and Sanders (1973) found E(VI) to correlate .47, .16, and .17 with their self-reported effort, super- visory effort, and supervisory performance criteria respec- tively, while the additive E + (V + I) correlated .36, .07, and .09 with these criteria. Both studies conclude that modest support is demon- strated for the multiplicative relationship, which is probably a liberal conclusion at best. Some Criticisms One deficiency shared by these correlational studies (all those reviewed thus far) is that the significance testing they employed considered only zero-order differences. No attempt was made to assess the significance of the differences of the obtained correlations. Therefore, even though the correlations for the multiplicative combina- tion of E, I, and V might be higher than an additive combina— tion (and both significantly different than zero), one cannot be confident of the conclusion that the multiplicative model is supported unless the differences between the correlation for it and the additive model are significantly different. The observed difference could be due to chance. DeLeo and Pritchard (1974) have noted further method- ological problems. They attempted to assess the reliability and construct validity of the E, I, and V variables. Two measures of V (attractiveness and importance) and two measures of I (correlation and probability) were evaluated 16 via a questionnaire by 171 Air Force trainees on two occasions. Two measures of B were also obtained. Subjects evaluated the various E, I, and V components of task related behaviors in their two 16 week technical training courses. Each subject completed two forms of the questionnaire with one form being completed on one day and the second form on the following day. They found the median test—retest reliability for V (importance) to be .60, V (attractiveness) .61, I (correlation) .56, and I (probability) .47. No test-retest reliability coefficient was computed for E, but a Cronbach Alpha for the E items was .64. From these results, they conclude that the lack of predictability of the EIV model is easier to understand when these rather low reliabilities are observed. They go on to clarify this issue by noting an example in which the reliability of an EIV model's predictions is .64, and "Since the square of the reliability (.41) is the maximum validity, one should not be surprised that correlations in the literature between the model and effort/performance criteria are so low" (p. 147). This, however, is a rather serious misconception since maximum obtainable validity is ng£_the square of reliability (.41) but rather the square root of reliability (in this example, .80). With a maximum validity of .80 (as Opposed to .41), one wonders about the implications of their interpretation. Mitchell (1974) offers more global criticisms. He believes the manner in which the E, I, and V variables have 17 been operationally defined could depart from the theoretical interpretation. V (as was noted earlier) has been defined as value or attraction of rewards. Most authors have operationalized it as importance of rewards or outcomes. The extent to which subjects perceive differences in these terms, argues Mitchell, could affect their responses to the V variable. Furthermore, I has been treated differently. Although Vroom (1964) defines it in correlational terms (i.e., values range from —l.0 to +1.0), most authors have treated it as a subjective probability (i.e., values range from 0.0 to +1.0). The last of Mitchell's major criticisms is concerned with the method of analysis employed in most studies. Since expectancy theory is a theory of behavioral choice, it is ipsative in nature and calls for a within- subjects paradigm. With the exception of Dachler and Mobley (1973), all studies have employed across-subjects analyses. Perhaps the most resounding criticism of expectancy theory research to date has been forwarded by Schmidt (1973). His basic argument is that researchers have been treating interval data as though it were ratio. Most expectancy theory studies have measured the E, I, and V variables with Likert—type scales--producing interval data at best. However, E, I, and V scale scores have been multiplied (in keeping with the theoretical structure of the model) to end up with a final "score" which is then 18 correlated with a criterion score in an attempt to predict effort and/or performance. As Schmidt points out, multi- plication of scales is not a meaningful Operation with interval data. To demonstrate his point, Schmidt generated two sets of correlations (within the range of empirically observed correlations) and performed a number of different X + b and ax + b transformations (a and b = a positive or negative constant) on them. While no differences were observed with the X + b transformations, the ax + b trans- formations affected the ratio of the standard deviations of the scales and scores and hence, the correlation co— efficients. He observed differences of 152 correlation points (+.76 to —.76) with the ax + b transformations. Schmidt offered three suggestions for future expectancy theory research. First, one could employ the Thurstone and Jones (1957) approach of constructing ratio scales of preference. Second, is the conjoint measurement approach of Krantz and Tversky (1971) and Krantz, Luce, Suppes, and Tversky (1968). For the third option he notes that "Laboratory studies in which levels of expectancy, instru- mentality, and valence are directly manipulated are an obvious possibility" (p. 250). Actually, Hackman and Porter (1968) were the first to note the problem of nonratio data. However, they argued that one could opt for Comrey's (1951) "practical validity criteria," which contends that if scores relate to criterion 19 variables of interest, then the scores can have practical interpretability. Schmidt (1973) contends that while this is true, this procedure is inappropriate if one is interested in testing the theoretical meaningfulness of the constructs and their multiplicative combinations. In other words, adequate tests of the model cannot be made under situation- specific conditions. Experimental Studies Following Schmidt's (1973) suggestions, Berner (1974) conducted an experimental study employing conjoint measure— ment. Fourteen of 125 middle managers from the Detroit area returned a mailed copy of the Work Motivation Questionnaire containing narrations of various E, I, and V conditions. Only two of the 14 subjects satisfied the conjoint measurement criteria to support the multiplicative (EIV) model, while five others met the requirements for an alternative dual distributive (E + IV) model, and one questionnaire was unusable. A multidimensional scaling and hierarchical clustering analysis yielded similar results. Two more studies attempted to investigate the inter— active prOperties of the EIV model in experimental settings. Arvey (1972) measured 180 male college undergraduates on an arithmetic task under three levels of the Campbell, gt al. (1970) EXpectancy I (E) and two levels of Expectancy II (I). Number of correct responses on the task served as the criterion measure. Employing an analysis of variance 20 design, he found no main effect for E or I (F = 2.84, p < .06, F = .30, p < .60). The E X I interaction was also not significant (F = 2.06, p < .13). A posttest valence measure (degree to which subjects wanted extra credit experimental points) was acquired to test for an E X I X V interaction and none was found (no F reported). Pritchard and DeLeo (1973) tested the I X V inter— action in their study of 60 mostly college students on a cataloging task presented as a real job. Two levels of I (piece-rate and salary) and two levels of V (amount of money paid) were investigated. After controlling for differences in ability, a significant I and V main effect was found, but V was in the opposite direction (low V performance was greater than high V performance). The I X V interaction was not significant (p < .08). The results of these experimental studies do not offer impressive support for the multiplicative combination rule hypothesized in expectancy theory models of work motivation. However, with the exception of Berner's (1974) conjoint measurement analysis, these experimental studies have failed to manipulate all three of the theory's variables in their experimental settings. The purpose ofthe present research was to follow Schmidt's (1973) suggested use of laboratory studies for testing the multiplicative combination of E, I, and V employing a 2 X 2 X 2 factorial (analysis of variance) 21 design. Although Pritchard and DeLeo (1973) and Arvey (1972) noted the advantages of the analysis of variance paradigm, they manipulated only two of the three main variables of the theory. This study manipulated all three of the theory's variables (E, I, and V). The task in this study required subjects to sort packs of data cards onto a peg board, and this measure constituted one of the dependent variables (performance). Other dependent measures included task performance satisfaction and the willingness of the subject to perform the task again.l Specific verbal and operational definitions of the independent and dependent variables in this study are outlined below. VARIABLES Independent Expectancy (E) Verbal Definition: The extent to which a subject believes that effort on his part will lead to successful task performance. (Values range from 0.0 to +1.0). Operational Definition: Two levels (+ = high, - = low) of E were manipulated. All subjects were required to take a fabricated Clerical Visual Performance Test. Regardless of actual test performance, subjects were informed (depending on treatment condition) that their performance was greatly above their peers (E+) or greatly below their peers (E-). Instrumentality (I) Verbal Definition: The extent to which a subject believes that effective performance on his part will lead to various outcomes. (Values range from 0.0 to +1.0). Operational Definition: Two levels (+ = high, . + . . - = low) of I were manipulated. I was a modified (no base pay) piece-rate pay condition (i.e., outcomes were performance dependent). I- was a salary condition (i.e., outcomes were independent of performance). 22 23 Valence (V) Verbal Definition: The extent to which a subject considers the outcomes of performance desirable or undesirable. (Values range from .1.0 to +1.0). Operational Definition: Two levels (+ = high, - = low) of V were manipulated. V+ was a high pay condition ($1.50/30 minutes or 7¢ per work unit completed). V— was a low pay condition ($0.50/30 minutes or 3¢ per work unit completed). Dependent Performance (P) Verbal Definition: The number of tasks the subject completed. Operational Definition: The number of data cards the subject placed on the peg board (see Appendix A) in a 30 minute work period. Since only one of the possible twelve positions on the peg board would correspond with any given data card, this task provided a unidimensional measure of performance. The advantage of this measure is that perfor- mance scores are easier to interpret since they vary as a function of the quantity of performance, rather than some combination of quantity and quality.2 Satisfaction (S) Verbal Definition: The extent to which a subject was satisfied with his performance on the peg board task. 24 Operational Definition: The subject's response to a 5-point Likert scaled item on the postexperimental question- naire, anchored very satisfied to very dissatisfied. (See Appendix F, question 6). Willingness (W) Verbal Definition: The extent to which a subject was willing to perform the peg board task again. Operational Definition: The subject's response to a 5-point Likert scaled item on the postexperimental questionnaire, anchored very willing to very unwilling. (See Appendix F, question 7.) Since Pritchard and DeLeo (1973) have noted that direct tests of expectancy theory's multiplicative combina- tion rule would require accurate estimates of the values of these E, I, and V variables, but that when one manipulates the levels of E, I, and V in the experimental setting the presence of significant interactions with main effects would ". . . support the presence of a multiplicative relationship," (p. 265), the hypotheses for this study were as follows: Hypotheses H : There will be significant interaction effects among the expectancy, instrumentality, and valence variables with the presence of main effects. 25 The performance, satisfaction, and willingness of subjects in the high (+) expectancy, instrumentality, and valence conditions will be significantly greater than that of subjects in the low (—) conditions. METHOD Subjects This study employed 120 college-aged male volunteers from undergraduate social science, psychology, and manage- ment classes at a large midwestern university. In addition to cash payment, some subjects also received extra credit points. Apparatus Each experimenter was equipped with a test file which included: copies of the Clerical Visual Performance Test (CVPT), a CVPT answer key, a CVPT stanine chart, a CVPT expectancy table, and a sheet describing the purpose of the study. Other equipment included a stopwatch, timer, 17" x 20" peg board, 3" x 5" data cards (35 packs of 24 each), a sample data card, a receipt pad, pencils, and rubber bands. The experiment was conducted in three 10' X 15' rooms with a desk and two chairs, illuminated by eight 4' long flourescent ceiling lights. Also used in the study were subject sign-up sheets, a postexperimental questionnaire, and a debriefing visual aid (see Appendices). 26 27 Procedure Subject sign-up sheets were placed in various class— rooms around campus. Each sheet contained 15 total time slots, three each at the hours of 3, 4, 5, 7, and 8 p.m. per calendar weekday, plus information related to the experiment: name of experiment (Test Validation) and experimenters, faculty member in charge, place, date, amount of credit available (usually 2 credit points), restrictions (males only), and a sentence informing the subjects that they would receive a cash payment for a portion of their participation. As each subject arrived at the designated room, he was sent to one of three rooms where he was greeted by an experimenter. After taking a seat Opposite the experimenter at the desk, the subject was given a copy of the Purpose (see Appendix B) which he read silently as the experimenter read aloud. After asking for and answering any questions, the experimenter then gave the subject a copy of the CVPT (see Appendix C), read the directions aloud and worked through an example problem with the subject. The subject was told that this was a speeded test and that both errors and time-Of-completion would be considered for the final score, upon which he was urged to work as quickly and as carefully as possible. Time was measured with a stopwatch. Utilizing an answer key for the test (see Appendix C), the experimenter then corrected the test in sight of the subject 28 and recorded the number of errors and time-Of-completion on the back of the test in the spaces provided. The experi- menter then gave the test a final score which, depending upon the eXpectancy condition being manipulated, corresponded to a particular stanine score (3 or 4 for E-, 7 or 8 for E+). The subject was shown a CVPT stanine chart (see Appendix D) to match his score with the stanine and was also shown a CVPT expectancy chart (see Appendix E) to match this stanine with his predicted chances in 100 of being classified as a successful performer of routine eyewhand tasks. This procedure completed the initial E manipulation. The experimenter then introduced the subject to the task (see also Pritchard and Curts, 1973), placing the peg board (see Appendix A) in front of him on the desk. Utilizing a sample card, the experimenter demonstrated how to properly place the cards on the board, and told him that there was only one correct position per card but that there could be more than one card at any position. In order to familiarize the subject with the task and strengthen the expectancy manipulation with a behaviorally based experience, the experimenter gave the subject a "practice" pack of cards and informed the subject that people who were predicted to be successful performers (as he either was or was not, depending upon the E manipula— tion) could complete this pack in 60 seconds or less. The E- pack contained 20 cards to insure that the subject 29 would not complete it, or the experimenter would stOp the watch early, while the E+ pack contained only 8 cards to insure that the subject would be able to complete it, or the experimenter would not stop the watch at the end of 60 seconds. The experimenter then informed the subject that this result was consistent with what the test predicted. The subject was told that he would be working at this task for 30 minutes and that this was the portion of the experiment for which he would receive a cash payment. He was then given a pack of 24 randomly shuffled cards (two cards for each of the 12 positions). The top card on each pack was numbered (from 1-35) and fastened to the pack with a rubber band. The subject was instructed to remove the rubber band and numbered card and place the cards on the peg board. Furthermore, he was informed that he could not skip any cards and that all cards would fit somewhere on the board. When he completed the first. pack, the experimenter would hand him the second, third, etc., until the 30 minutes had expired. The experimenter then introduced the appropriate I and V manipulations for the condition being tested. A 30 minute timer was placed near the subject and he was told that when the timer reached zero a bell would ring, at which time he was to stop placing cards on the board. When the 30 minute period ended, the experimenter counted the cards and gave the subject a receipt for the amount of money due and told him 30 to return to the room he had originally come from for payment. When the subject returned to the original room, a debriefer would take his receipt and pay him for the face amount, after which the subject was asked to complete a short questionnaire (see Appendix F). With the aid of a visual debriefing handout (see Appendix G), the subject was then verbally debriefed as to the nature of the study. Depending upon the condition tested, the subjects' pay was adjusted so that no subject received less than $1.50. The subject was requested not to discuss the experiment with others and given a date and place where he could inquire about the final results. Total time required of subjects was usually between 40-50 minutes. The order of the eight experimental conditions was random, and the first 15 subjects constituted the first condition, the second 15 constituted the second condition, etc., until the eight conditions were exhausted. The order of these conditions was as follows: E I V l. - + - 2. + + - 3. + - + 4. + + + 5. — - + 6. - + + 7. - - — 8. + - - 31 The non—random assignment of subjects to conditions was employed in an attempt to reduce experimenter error in manipulating the E, I, and V variables. This procedure would allow for more comparable treatment among subjects in any given condition, thereby reducing error (within cell) variance in the eight conditions. The study was completed in eight weeks. RESULTS AND DISCUSSION The hypotheses for this study predicted main effects for the three independent variables E, I, and V, and interactions for all dependent variables. As can be seen in Table 1, only two of the possible 21 F ratios attained statistical significance and neither of these with P as the dependent variable. Therefore, Hl was not supported. Table 1 2 X 2 X 2 Analysis of Variance F Source df P S W E l 0.11 l2.77** 1.01 I 1 1.30 1.42 2.80 V l 0.95 3.39 1.50 El 1 0.01 2.64 0.61 EV l 0.07 0.10 0.31 IV 1 1.59 2.64 1.01 EIV 1 2.18 0.01 6.58* ERROR 112 (17532.94) (1.71) (0.67) TOTAL 119 Note: *(pI<.05), **(p<<.01) P = performance S = satisfaction W = willingness 32 33 With regard to H2, the E main effect for satisfaction was in the predicted direction (X - = 2.75, §.+ = 2.20). E E Due to the lack of significance for the other main effects, support was not achieved for the other predictions of H2. The results of the manipulation checks from the post- experimental questionnaire demonstrate that the subjects did perceive differences in the + and - conditions. Table 2 t-test Manipulation Checks of the Expectancy, Instrumentality, and Valence Variables Question No. Variable Condition x t 1. Increasing effort E + 2.52 +0.231 - 2.47 2. Capability with E + 1.73 ** -6.649 * peers - 2.60 3. Pay-performance I + 1.77 ~10.580*** contingency - 3.93 4. Desirability V + 2.58 _ * of pay — 2.75 1'377 5. Satisfaction V + 2.60 -1 949** with pay - 2.83 ' Note: ***(p < .01), **(p < .05), *(p < .10) 118 df for each test From Table 2 one notices that although E+ and E- subjects could not be differentiated with question 1, all of the remaining t values were statistically significant. 34 Support for the multiplicative combination rule hypothesized in expectancy theory would have occurred if main effects for the three independent variables were found with the presence of any interaction effects among these variables. Support for the alternative additive model would have occurred if the main effects were found without the presence of the interactions. The results of the analysis demonstrated support for neither of these possibilities. Perhaps the most obvious reason for the lack of statistical support from the analysis of variance for performance can be evidenced in Table 1 by noting the large error (within cell) variance. With large variation in the dependent measures within the cells (see also Table 3), great differences in the treatment means would have been necessary for the analysis to have rendered significant main effects and interactions. Similar conclusions can be made for satisfaction and willingness. Another of the many reasons for the lack of support for the hypotheses in this study could rest with the analysis itself. In the case of the analysis of variance as used in this study, the notions of normality, homogeneity of variance, and independence are noteworthy. A graph of the performance scores revealed that the normality assumption was tenable. An Fm test (Winer, 1962) on these scores ax demonstrated that there was no violation of the homogeneity 35 Table 3 Performance Means and Standard Deviations for the 8 Experimental Conditions E" 13+ v' v+ v“ v+ 497.67 475.13 - 473.47 509.87 (137.57) (145.80) 486'40 I (100.62) (103.77) 491'67 517.20 505.00 + 570.13 474.07 (120.09) (158.24) 511'10 I (131.37) (149.88) 522'10 507.43 490.07 498.75 521.80 491.97 506.88 Note: Standard deviations are in parentheses Marginals are also included of variance assumption, (F = 2.47, p < .01). Since max there was no attempt to match subjects on any antecedent variables which would produce correlated experimental conditions, a violation of the independence assumption does not seem likely. Again, similar conclusions are drawn for satisfaction and willingness. One other source of concern rests with the effects of the E, I, and V manipulations. As noted in Table 2, the manipulations were, for the most part, statistically significant, however, a closer examination of the means raises questions as to whether these manipulations had behavioral significance. For example, one might question 36 the behavioral effect of valence, as measured by question 5, where the means for the + and - conditions are 2.60 and 2.83 respectively. Appearances here would indicate that both the + and - groups were, for the most part, "neither satisfied nor dissatisfied" with the amount of money they received. Therefore, one could wonder whether the motivational levels of the V+ and V- conditions were great enough to cause significant differences in the dependent variables. Similarly, in recalling Mitchell's (1974) criticism that I is postulated to be a subjective correlation (values ranging from -1.0 to +1.0) but that most researchers Operationalize it as a subjective probabi- lity (values ranging from 0.0 to +1.0), one might argue that this study failed to properly Operationalize this variable, since both piece-rates and salaries are positive instrumenta- lities. However, for the purpose of the study (a test of the multiplicative combination rule), the only requirement necessary was that subjects perceive differences (high and low) in this variable, regardless of where these differences existed along the continuum. Coupled with the~prob1em of the manipulation effects is concern for the individual difference variables which might have affected subject performance on the dependent measures. Comments from subjects in the debriefing situation revealed, for example, that sometimes the $1.50 pay was perceived as too low, while for others, $0.50 was appreciated. Some subjects commented that money, in any amount, was of 37 no interest to them. Other subjects were mainly concerned with discovering "how many cards I could do," (a problem of intrinsic rewards noted by Turner, 1974), and some were mostly interested in how well their performance matched that of others in the experiment (an equity condition). The expectancy manipulation was not conducted without personal interpretation either. Some subjects may have perceived E- at the p = .50 set, and could have been more motivated than E+ subjects. Comments from some E- subjects revealed that they felt the test (CVPT) was probably correct, but they wanted to see if they could do better than predicted, while some E+ subjects were surprised at their test performance and felt that the test must have been wrong in predicting their ability to successfully perform the task. The instrumentality condition seemed to be fairly straight- forward to most subjects. Whether the proportion of people with these types of perceptions were randomly dispersed in the eight treatment conditions is certainly open to question, and if they were not, this lack of randomization would also help to explain the nonsupportive results of this study. Furthermore, the manner in which the data were collected could have been a problem. As noted earlier, the eight experimental conditions were randomly chosen for order of presentation, but subjects were not randomly assigned Raconditions. Again, this strategy was chosen in hopes of reducing experimenter error in manipulating the 38 E, I, and V variables in each condition. Nevertheless, the possibility exists that subjects who participated in the beginning of the study may have been more (or less) intrin- sically motivated, anxious, interested, etc. than subjects participating in the later stages (conditions) of the study. Similarly, even though subjects were asked in the debriefing session not to communicate the nature of the study to anyone, the possibility exists that some subjects in the later conditions could have been alerted to the nature of the study sometime during the eight weeks in which it was conducted. Therefore, the possibility exists that these extraneous variables were confounded with the manipulation of the independent variables. Despite the lack of statistical significance, it is somewhat interesting to note the degree of correspondence between this study and that of Pritchard and DeLeo (1973). In both cases, the I X V interactions have V- subjects outperforming V+ subjects in the I+ condition. As can be seen from Figure 1, this is also true in the I X V inter- actions for the E+ and E- subjects. Perhaps the lack of statistical significance is not surprising when one considers the results of the other studies in the literature. Very often, these studies generate a great number of correlations between the inde- pendent variables and criteria of interest, only to find relatively few Of them significant. Even where significant, 39 OOCOEHOMHOm Mom cofluomuwucH > x H x m Ocu mo nacho I> onv omv omv com cam omm omm ovm omm com Ohm 0mm BDUEWJOJIGd .H musmam onv omw omv com oam omm omm ovm 0mm omm Ohm 0mm BOUEUIJOJISd 40 these correlations are often only in the .20 to .30 range (Berner, 1974). With correlations of this magnitude, a great deal of criterion variance is left unexplained, and the effects of the independent variables on the criterion measures are probably not tremendous. The problems evidenced in this study point to the fact that adequate tests of expectancy theory's multiplica- tive combination rule are difficult to construct. Never- theless, these problems do help in understanding the needs of research designs which attempt to assess the impact of this rule. Perhaps most apparent is the need to care- fully measure or manipulate the levels of expectancy, instrumentality, and valence. This might best be accom- plished by studyhxyintact work groups, where levels Of E, I, and V might be more fixed and stable than they are in contrived work situations. By employing the Thurstone developed scales of ratio preference (recommended by Schmidt, 1973), one might be able to more accurately assess the impact each of these independent variables has on performance, satisfaction, and effort. Secondly, as Mitchell (1974) suggests, designs which employ within-subjects analyses might be preferable to across-subject analyses because they would not only help control for the impact of individual difference variables on the dependent measures, but would also be more in line with the ipsative theoretical postulates of most expectancy theory models. This might be accomplished 41 by longitudinal studies in which the performance, satisfac— tion, and effort of individual subjects could be measured at different time levels under each of a number Of different combinations of E, I, and V situations in actual field settings or in the laboratory. Finally, as noted by Graen (1969), more consideration could be given to the boundary conditions of the theory. If these boundary conditions do affect perceived effort, satisfaction, or actual job per— formance, as Graen suggests, then they could have an effect on the combinatorial properties of the E, I, and V variables. For example, it might be found that under autocratic organi- zational settings, an additive EIV theory would best predict dependent measures, whereas a multiplicative EIV theory might better explain them in a COOperative organizational setting, or vice-versa. This type of research would, undoubtedly, require (anal further extensions of the EIV models now in existence, and thus more theory building. In a broader perspective, however, perhaps the problems evidenced in this and other studies testing the multiplicative combination rule hypothesized in expectancy theories of work motivation point to some misgivings concerning the efficacy of expectancy theory's ability to explain work motivation. In the first place, the theory was postulated by theorists as a general theory of human motivation and adapted, by Vroom (1964) and others, to "fit" the workplace. While this fact in itself does not render 42 the theory useless, it can and perhaps should raise questions in the minds of researchers about its applicability in the work environment. Secondly, there appears to be no real sense of agreement among researchers and theorists as to what actually constitutes expectancy theory. Although the three main variables (E, I, and V) are frequently mentioned, they oftentimes are merged into different types of constructs such as Expectancy I, Expectancy II, "E" V, etc. The problem would be severe enough if these terms were merely synonyms, but rather they are offered by different theorists as distinct constructs within the theory. Couple with this the ever increasing number of additional variables which enter the different models Offered by theorists such as Lawler (1971) and Graen (1969) with his boundary conditions, and one begins to wonder about the degree of correspondence between the theory's original notion of "maximizing gains and minimizing losses" and its supposed ability to predict "force to perform" in the work environment. Finally, there is no real agreement among researchers and theorists concerning what expectancy theory is supposed to predict. Some models, such as Vroom's (1964), are designed to predict a number of job related criteria such as effort, performance, and satisfaction, while other models predict only effort or performance or satisfaction. So in addition to no clear distinction among the independent variables in the theory, 43 one also finds that there is no consensus on what constitutes the dependent variables. Therefore, since there is so little conceptual and theoretical agreement concerning what expectancy theory is and what it should predict in the work setting, one cannot be overly surprised or disappointed by the extreme lack of empirical evidence to justify its multiplicative combination rule, and perhaps to some extent, its existence as a theory of work motivation. APPENDICES APPENDIX A DIAGRAM OF THE PEG BOARD APPARATUS AND A SAMPLE DATA CARD APPENDIX A DIAGRAM OF THE PEG BOARD APPARATUS AND A SAMPLE DATA CARD O O O O O O O O O O O O O I. O O O 0 O O o O O o O O 6743 SEX: O 1 2 IECE: O l 2 AGE: 0 1. 2 3 44 APPENDIX B COPY OF THE PURPOSE OF THE EXPERIMENT GIVEN TO SUBJECTS IN THE 8 TREATMENT CONDITIONS APPENDIX B PURPOSE The experiment we have asked you to participate in is designed to study some of the factors involved in eye- hand coordination in routine tasks. One specific goal of this study is to extend our knowledge of the Clerical Visual Performance Test, which is designed to measure the recognition of numbers. This test has proven to be a valid predictor of routine eye-hand tasks. We are interested in determining its predictive validity for a male college student population. In a few moments you will be working on a routine eye-hand task. This is the portion of the experiment for which you will receive a cash payment. 45 APPENDIX C COPY OF THE CLERICAL VISUAL PERFORMANCE TEST/ANSWER KEY APPENDIX C Budget No. 13-1475 Form B CLERICAL VISUAL PERFORMANCE TEST Quarter Sample Direction: When the three pairs of numbers are exactly the same make a check mark on the line between them. EXAMPLE 213 213 312 312 312 132 132 132 213 213 123 123 I/ l) 132 132 2) 213 213 3) 123 123 4) 312 312 5) 132 132 213 213 321 312 231 231 213 213 312 312 312 312 231 231 312 312 132 132 231 231 J. ._ .K. .1. _‘/- 6) 321 321 7) 123 123 8) 231 231 9) 312 312 10) 132 132 132 123 213 213 213 213 231 231 321 321 213 213 312 321 312 321 312 321 132 123 11) 312 312 12) 132 132 13) 321 321 14) 312 312 15) 231 231 213 213 213 213 132 123 123 123 231 231 321 321 321 321 213 213 213 231 123 132 __/_ .1. _... .._ _. 16) 123 123 17) 213 213 18) 213 231 19) 132 132 20) 132 132 312 312 231 231 231 231 321 321 312 312 321 321 312 312 321 321 123 123 213 231 .1. .1. ._ __‘/_ __ 21) 213 213 22) 132 132 23) 312 312 24) 132 132 25) 321 321 213 213 123 123 213 213 321 321 321 321 231 231 321 312 123 123 132 132 321 321 / / / I/ 46 -—.— — 26) 31) 36) 41) 46) 51) 56) 231 231 213 213 231 231 1. 312 312 123 123 312 312 1. 312 312 123 132 312 312 213 231 231 231 231 231 312 312 213 213 213 213 1. 312 312 132 132 123 132 213 213 231 231 123 123 / 27) 32) 37) 42) 47) 52) 57) 132 132 231 132 321 213 123 123 312 312 132 213 132 132 312 312 213 231 312 312 213 213 213 213 .11. 132 132 213 213 321 321 l/ -——- 213 213 132 123 321 321 312 312 321 312 123 123 Test Time: Errors: 28) 33) 38) 43) 48) 53) 58) 47 321 321 132 132 312 312 .11. 312 312 312 312 213 213 .11. 123 123 312 312 312 312 .1: 132 132 123 123 321 312 123 123 213 213 321 321 / 132 132 123 123 132 132 / 231 231 213 213 231 231 / 29) 34) 39) 44) 49) 54) 59) 132 132 213 213 213 231 123 123 312 312 213 213 / 213 213 231 231 231 231 / 312 312 213 213 321 321 / 132 132 132 123 213 213 312 312 213 213 312 312 / 132 132 312 321 213 213 30) 35) 40) 45) 50) 55) 60) 312 312 213 213 132 132 _1L_ 132 132 123 123 132 132 / 132 132 123 123 231 231 I/ 231 231 132 132 213 213 / 132 132 213 213 312 312 / 132 123 123 123 123 123 213 213 .132 132 213 213 / APPENDIX D STANINE CHART FOR THE CLERICAL VISUAL PERFORMANCE TEST APPENDIX D CLERICAL VISUAL PERFORMANCE TEST E9912 119 120-150 151-170 171-175 176-180 181-190 191-210 211-220 220+ 48 STANINE APPENDIX E EXPECTANCY TABLE FOR THE CLERICAL VISUAL PERFORMANCE TEST hm OOH om om on om om ow om om mocmfiuomumm Hammmmoozm mo OOH cw mmoamnu Emma mUZ¢ZmOhmmm AdDmH> A