Journal of Occupational Psychology (1990), 63, 245-261 Printed in Great Britain 245

© 1990 The British Psychological Society

Building organizational commitment:A multifirm study

David F. Caldwell*l^avey School of Busmen and Administration, Santa Clara Uninersity Santa Clara,

CA 9505 j , (ISA

Jennifer A. ChatmanKellogg Graduate School of Management. Northwestern University

Charles A. O'ReillySchool of Business AdminiUration. University ol California

Although much research has been conducted in the area of organizational commitment,few studies have explicitly examined how organizations facilitate commitment amongmembers. Using a sample of 29 ' respondents from 45 firms, the results of this studyshow that rigorous recruitment and selection procedures and a strong, clear organiz-ational value system arc associated with higher levels of employee commitment based oninternalization and identification. Strong organiziirional career and reward systems arerelated to higher levels of instrumental or compliance-based commitment.

While thL'construct of organizational commitment has received a great deal of attention inthe organizational psychology literature, most of this attention has been directed towardsidentifying the consequences of having committed employees. Although the results ofthese studies are not always consistent, they suggest that commitment is positivelyassociare<! with motivation and Involvement (Farreil & Rusbult, 1981; Stumpf &Hartman, 1984), expressions of positive affect and loyalty {Kanter, 1968; O'Reilly &:Caldwell, 1980), some aspects of job performance (Angle & Perry, 1981; Mowday, Porter& Diibin, 1974; Steers, 1977), and prosocial behaviour (O'Reilly & Chatman, 1986;Organ, 1988). Additionally, some studies suggest that commitment is negativelyassociatetl with potentially costly behaviours such as absenteeism (Angle ik Perry, 1981)and the likelihood ot turnover (Hom, Katerberg & Hulin, 1979; Porter, Crampon &Smith, 1976; Porter, Steers, Mowday & Boulian, 1974). Given that such positiveoutcomes apparently emerge from having committed members, it is surprising thatrelatively fewer studies have actually investigated what organizations can do to enhancesuch commitment.

* Requests tiir reprints.

246 D. F. Caldwell.J. A. Chatman and C. A. O'Reilly

Antecedents to organizational commitment

Those studies which have attempted to identify the antecedents to commitment have[inked variables such as personal or job characteristics, work experiences, organizationalstructure and size, and role-related factors to commitment (e.g. Lincoln & Kalleberg,19H5; Meyer & Allen, 1987; Morris & Sherman, 1981; Steers, 1977; Stevens, Beyer &Trice, 197H). Taken together, these studies have revealed few consistent findmgs (e.g.Mowday, Porter & Steers, 1982; Reichers, 1985; Schneider & Reichers, 1983). However,one clear implication is that early experiences in an individual's employment may have alarge impact on the subsequent development of commitment (Buchanan, 1974; Louis,1980). It is at this time that an individual may be particularly sensitive to organizationalmfUiencc and the results of that influence most consequential {Bray, Campbell & Grant,1974; Hail, 1976). What these findmgs suggest is that an individual's commitment coanorganization may be shaped by the process through which he or she enters the organization(recruitment) and by those steps the organization takes to teach him or her about theorganization's values, and how work is done (socialization). While researchers haveimplicitly suggested that organizational characteristics are linked to commitment amongmembers (e.g. Etzioni, 197'5; Kanter, 1968; Mowday et al., !9H2), almost no researchexists which explicitly examines this link.

Some consistent, albeit indirect, evidence for the linkage between organizationalpractices and commitment does exist. On the recruitment side, factors such as confir-mation of pre-entry expectations (Arnold & Feidman, 1982; Premack &L Wanous, 198^)and tole clarity (Morris & Koch, 1979) which are important at entry have been shown to be-positively related to commitment to the organization. In fact, Feidman (1977) argues thatthe (Tiosr effective types of realistic job previews ate those which, in addition to givingbalanced descriptions about che work itself, provide potential entrants with inf(jrmationabout advancement op]x)ttunicies and the general work climate within die organization.In addition, evidence shows rhat factors related to an individual's decision to accept a joboffer can influence his or her subsequent commitmenr. For example, O'Reilly & Caldwell(198 I) demtJnstrate that the volitionality and irrevocability of job choices are related toindividual commitment for over a year following job acceptance. They argue, in j^art fromdissonance theory, that certain aspects of an individual's job choice and particularexperiences within the organization can serve to 'bind' the individual to that organizationand affecc both commitment and turnover.

Specific socialization practices may aftect commitment as well. Through socializationptcicesses managers can attempt to foster better employee understanding of organizationalvalues, norms and objectives (cf. Kanter, 1968; Pascale, 1985; Van Maanen &c Schein,1979). Van Maanen & Schein (1979) have made a distinction between the content ofsocialization, or what is taught to new employees, and the process of being socialized, orhow the infotrnation is transmitted to new members. While both are important. VanMaanen & Scheiii's model highlights the process aspects. Using Van Maanen & Schcin's(1979) socialization mt)del, Jones (1986) empirically investigated the socializationprocess. HIS findings suggested that the extent tt) which socialization patterns were'institutionalized' in an organization was related to commitment. Jones (1986) demon-sttated that individuals who reported that their socialization experiences were formalized,

and followed a fixed timetable and sequence also reported greater commit-

Building organizational commitment 247

ment to the organization than did individuals reporting a less institutionalized pattern ofearly experiences,

Pascale (1985) presents a useful approach by incorporating both recruitment andsocialization issues, as well as content and process issues. He argues that companies witheffective recruitment and socialization programmes use a specific set ot techniques.Among these are: (I) careful recruitment and selection; (2) experiences designed topromote a willingness to learn about and to accept the values and practices of the neworganization; (3) career paths which are based on extensive exposure to the centralfunctional area of the business; (4) careful use of training, reward and control systems toemphasize key behaviours; and (5) the reinforcement of central values through folklore andconsistent role models. Although Pascale's observations are based on interviews ofindividuals in a small sample of companies, the pnjcesses he describes are similar to thosethat are associated with behavioural commitment (Kiesler, 1971; Salancik, 1977) andconversions to religious sects (Cialanter, 1980, 1982). The purpose of the current study isto further examine how organizational recruitment and socialization processes caninfluence individual members' commitment to their employing organizations. In doingso, we will focus on the techniques suggested by Pascale as effective in buildingorganizational commitment.

Defining and measuring organizational commitment

Developing a clear understanding of the antecedents to commitment has been hamperedby some ambiguity in the definition and measurement ol the construct itself. Morrow(1983), for example, notes that there are over 25 commitment-related concepts ormeasures. And as Staw (1977) has observed, it is difficult to separate the concept ofcommitment from other affective concepts such as m(jtivation, satisfaction and involve-ment. The mixed nature ol the construct is demonstrated in the OrganizationalCommitment Questionnaire, one of the most commonly used measures of commitment{VontT et al., 1974), Specifically, this instrument includes three constructs; intentions,motivations and values. At least three ot the questions on the OCQ measure a person'sintent to behave (e.g. 'I would accept almost any type of job assignment in order to keepworking tor this organization'), while other items focus on commitment as a motive (e.g.'This organization really inspires the very best in me in the way of job performance'), orcommitment as value agreement (e,g. 'I tind that my values and the organization's valuesare very similar'). This ambiguity has made it difficult to compare results of studies and todevelopappropriare tests of the antecedents to commitment (McGee& Ford, 1987; Meyer& Allen, 1984).

O'Rt'illy Si Chatman (1986) argue that commitment is best defined as the basis of anindividual's psychological attachment to the organization. This basis of aittachment isdistinct from either the antecedents of commitment or from its consequences. Drawingfrom Kelrnan (1958), they argue that commitment to an organization is predicated onthtee separate bases of attachment: compliance, identification and internalization.Compliant e refers to instrumental attachment undertaken for specific rewards; identifica-tion refers to attachment based on a desire for aftiliation with the organization; andinternaiization reters to congruence between individual and organization values. O'Reilly

248 D. F. Caldwell, J. A. ChatmanandC. A.

& Chatman (1986) further demonstrate that the consequences of commitment varyaccording to the individual's basis of attachment.

A similar argument is likely to hold ttue for the antecedents of commitment; that is,particular policies and perceptions on the part of new entrants may be associated with oneform ol commitment but not with others. For instance, organizations whose recruitmentpractices clarify the organizations' values for potential employees are more likely to selectfor and enhance internalized attachment among new recruits than are organizations whodo not screen applicants for value congruence. This is because if values are clear andsalient, candidates will have more information on which to determine if they agree with orcan comfortably conform to those values, and organizations can more easily matchprospective candidates both to rhe specific job and the organization culture {C:hatman,I9H9; O'Reilly, Chatman & Caldwell, 1989).

In addition to the ambiguities in measurement, two other factors may make it difficultto dtaw concrete conclusions about the relationship between an organization's practicesand individual commitment. These ate the reliance on individual reports of bothcommitment and other variables of interest, and the tendency to draw samples from arelatively small number of organizations. In the first instance, many .studies of commit-ment potentially sutler from a tc.sponse--tes]-)unse bias. In a typical commitment study,rcs[>ondents are asked to provide self-reports of both their level of commitment a.s well asof othct antecedent or consequent variables. Kven when data are collected longitudinally,this response bias may be problematic since commitment can be viewed either as aprospective process or a retrospective process through which the individual justifiesprevious actions (O'Reilly & Caldweli, 19H 1; Salancik, 1977). In addition, because theysample subjects from a single or few organizations, many studies are unable t(5 investigatedifferences in recruitment and socialization policies across organizations. Thus, a largesample of firms is de.sirable in order to ensure adequate variation in firm practices.Obviously one barrier to multifirm samples is their inaccessibility compared to singlefirms.

Although no multifirm studies exist which link firm recruitment and socializationpraictices to individual commitment, two general propositions can be offered. Drawingfrom Pascale (1985), one proposition is that more intensive recruitment practices shouldbe associated with higher levels of individual commitment. A seconci proposition is thatmote intensive organizational socialization practices should be associated with higherlevels ot individual commitment. More specifically, this study tested three hypotheses:

Hypothesis I: Recruitment processes which provide individuals with a realistic sense ofwhat is expected, and which provide opportunities for individuals to choose not to joinshould be relatecl to higher levels of commitment based on internalization andidentification, but not compliance (e.g. O'Reilly &c Caldwell, 1981; Premack &Wanous, 1985).

Hyf}otheii.s 2: Socialization processes which emphasize strong organizational values,through role models anci management actions, should also be associated with higherlevels of individual internaiizatiun and identification (e.g. Kanter, 1968; Ouchi,1980).

Hypothesis H: Socialization processes that rely on formal control and reward systemsshould be related tohigher levels of compliance-based organizational commitment, and

Building organizational commitment 249

lower levels of commitment based on internalization and identification (e.g. Etzioni,1961; Jones, 1986).

Method

Information about organizational sotiaiizatmn processes and individual commitment was collected from 323prolc'ssiiina! or technical employees of 47 firms. Data were collected from US firms In rwo industries. The first(siimplc 1) represented 196 emph>yees of 59 hi^h technology companies. The second (sample 2) consisted ot127 enrry level accountants employed in the western regional ofJices ol ei^ht large public accounting firms. Atotal of i J r(Spc)ndenrs was dropped from the analysis of the roral sample-because-of incomplete data or becausethey beliin^'ed to a firm from which there were fewer than three respondents. This reduced the final sample si7.cro 29 i individuals In i^ lirms.

Since we designetl this study with the intent of minimizing commun method biases which can beproiilem.itic in sur\'ey-based studies (ci. Abrahamson, 198 i, p, ^2i}, a brief overview of our general design iswarranred. We asked individuals from participating firms to respond to a survey (described more fully below)whi( h (onraint-d rhe two measurement mstrumtnts. The first mcasuremcnr instriimenr, the Recruitment itndSotkili/ation Questionnaire (Pascale, 19'S'i), was designed to query respondents about recruitment andsocialization practices which are characteristic of their organization. The second instrument, theOimmitmentQuestionnaire (OReilly & (Ihatman, 1986), was designed to get respondents to think about their ownindividual (ommitment to their employing firm. Once the surveys were completed, we aggregated theorgani/atioii-leve! responses to the Recruitment and Socialization Questionnaire for each lirm. Then wecalculated s( ores for the organizational dimensions for each individual based on the aggregate ofthe others inthe firm, hut eliminated the focal individual's own score. The effect of such a calculation is to create separateorganizational scores for each individual by averaging the factor scores ofthe other respondents in the sameorganization. Thus, the relationship between organiiiational practices and commitment is imcontaminated byconsistency or response bias. Since the consensual firm ratings require multiple raters (and one respondent isalways deleted in the aggregation), firms were only included in the analysis if we received responses from threeor more respondents

Subjects

Sample !. An initial set of ')') participants in a mid-level management training programme for managers inclecrromcs, or fully employed MBA students working for high rechnology companies, took part in the study,fiai h of rhese initial respondents was asked to recruit ai least two other individuals in their company toparru ipate in the study. A!! respondents were provided with stirveysand return envelopes. For the entire set ofsub|c<Lts, th<-average hrm tenure was f 9^ years with a fairly high variance {SD = '5.79). All but two o{ therespondents had attended college and over S") per tent possessed at least a bachelors degree. One hall oftheres|)on<ieiits had supervisory positions.

Sample 2 The second set ol respondents were graduates of university accounting programmes employedby eight of ihe largest public accounting hrms When surveyed, al! had been members of their employingorganiziition for approximately one year. All were employed in the auditing function. Respondents averaged24. ^ years in age, and 47 per a.nt were male. All po.ssessed a bachelors degree and approximately 2^ per cent.ilso had .111 MBA degree.

An advanragc of selecting these two samples was that we could assess a large- number of firms within eachindustry (toi example, rhe eight accounting firms conducted S'5 per cent ofthe public auditing business--Fmerson, 1987) and rhus varianc'c- in recruitment and socialization practices across the lirms could beraptured However, it is apparent that substantial diffcrenas existed between the rwo samples, and evenwithin s,un|>lc' I, Ehc- variation in [enure was quire- large. In order to rule our tin- possibility that rhesediffereniL-s atfc-cted subsequent results, we controlled lor industry, tenure, firm size and the number ofrespondents Irom each fitm in regression equations. 'l"hese conrrol variables are de^ribed below.

l)rgaiuzani,ii recrunment and wcialization practues. Respondents completed a 16-item scale developed by Pascale(198^). This scale is ilesigned lo measure the degree to which companies engage in actions ot have pohcies

250 D, F. Caldwell. J. A. Chatrmn andC. A. O'ReillysimiUr to those* ot companies that have a reputation for successtully recruiting atid socializing newcomers.Respondents indicated on a hvc-point scale the extent to which each item was true of their company.

'!"hc 16 items were- Initially analysed using a principal components analysis with a varimax rotation. After jninspecnon of the item statistics and rotittL'd .solution, hve items which loaded on more than two hictors ')rwhich liaii very l(iw variance were dtopped Irom futthet analysis. The I I remaining items were tean.dysetj,again usnig J principal components analysis with a vurimax rotation, arid the results are shown in Table I,

Three clear factors emerged. The first fat. tor, defined by four items, is based on the existence ofa common setof organizational values. We labelled this factor 'vaiues' and it corresponds with the notion ofa strong, clear,visible organizational value system manifested through role models and management actions (suggested in rhesecond hypothesis). We labelled the second factor, 'rewards' and ir is defined by three items with loadingsgreiiier than ,')5. The "rewards' fauor reflects the existence of cleat rewards and career paths available toal! newentrants. This factor is used rn lesc the rhird hypothesis which posriilares that formal reward and controlsysteiivs should be positively related ro higher levels ot instrumental or lompliaiice-based commitment. Thethird factor is defined by tour items, all reflecting the existence ofa rigorous recruitment and selection processas proposed in the first hypotliesis. We labelled this factor 'selection'.

To ensure independence atnoiig ilie three organizational practice dimensions ar the individual level ofanaiysi.s. t"actor scores were computed and used in subsequenr analyses. It should be pomtc-d out, however, thatalthough the factors are completely independent at the individual level used in the factor iiniilysis, subsequentaggregation of individual scores ro chanKteri/<' the organizational level may induce some small correlationamong the dimensions This is not unreasonable given that organizations with strong value systems may alsohave recruitment and reward systems designed to maintam these,

Tlie use of multiple mdividual estimates of socialization practices to form an overall score is also argued to bea reascinable procedure since noE all entrants to a given organization will be exposed to precisely the same set otexperiences. Thus, rhe use ot multiple perceptual assessments may be as accurate an estimate of the true scoreas a more objet rive index, suih as the number of interviews rhe personnel department recommenils, but whichmay luit be unitormly applied m .lil instances.

Two indicators support the view iliar the composite hrm pnicnce scores reliably represented anddisiuiguished firm pracn..es. ['irst, we found that tirm raters haci relatively uinsisteiit perceptions of firmpractites as indicated by the mean loict-item correlations lor each Ijcror (vaiues = .7.!; rewards = .'16,selection = .'iO). And second, il recruitment and socialization practices .\n- distinct charac terisncs otorgaiiizatiims, small wirhin-firm variances and large between-firm variances should result. In order to test rheextent to which recruitment and socializatioti practices were distinct from one firm to the next, an analysis otvariance (ANOVA) was conducted using firms as the independent variable and the three organization practicefarti>rs (values, rewards, selection) as dependent variables. This analysis was done separately for each ot the twoindustries (high technology and accounting).* Findings indicated thar between-firm variance (within eachindustry) was significantly greater than within-tirm variance for all three of the iirm practice variables(vahies,,,,,,/,,/•• = 2 l . ( )4 , / '< .01 , K ' ^ ,S4; rewards/,,,,,;,:/'^ 29,'Xt,/)< .() 1;«-'^ ,88; selection;,,,,,,,:/' ^*i ,99, / ' ' • .01 . R- --- ,S9; valuers.,,,,,, /' ^ ' i 7 . ' i 8 , / ' • ' . ( ) I, /^' ^ I). 78, rewards,,,,,,,. /' ^ -^^^.^S,/; <" .01,R' - .9^; selecnoiv,,,: /• = 447,OS. p -- ,01, R = ,97).

Com»/itm>it. (.ommirment was measured using O'Reilly & Chatman's (19X6) scale, A principal comjionentsanalysis with varimax rotation of the 12 items yielded two unambiguous factors in coiurast to the threeidentified by O'Reilly & Chatman. Table .1 shows the results ot this analysis.

The hrst factor contains eight items with loadings of greater than ,60, These items include those originallylabelled by O'Reilly & Chatman (I9S6) as representing internalization arid identification. Since rhese itemsrepresent commitment to the organization based on shared values, we labelled this factor normarivecommitment. The second factor, defined by four items with loadings greater than ,^0, is quite similar rowhiir OKeilly & Chatman describe as compliance. This factot was labelled'instrumental commirment'since It

• li should Itc'noted that in ordc t to lit consist(.-nc with cur treatment ofthc two indcismes in cithtTanalyses (e,g.ciiii[r(illmgt,,r mdustrv m our regressitin analyses) we report the between-firm diffrrencc-b tot fach of the iwo industries separately Inlii.t we kiund idenric;»l te.suits (suhstaniially ^-reaier beiween-fitm varwme than within-firm variance) when we ranANOVAs on rhe entice data sei. We <.lso found (hat wlien using imlustry as the mdepenclent variabi.' the reward;, andselection factors differed significantly across ihc industries, bur the values tacror did not differ significandy across ihe rwo

Building organizational commitment 251

Table I. Varimax factor loadings for recruitment/socialization dimensions (N = 323)

Varimax factor loadings

Item Values Rewards Selection

1. Recruiters receive at least one week oi —.13 .46 .57intensive training.

2. Recruitment forms identify several key .12 .31 .66traits deemed crucial to the firm's success,traits are defined in concrete terms anctinterviewer records specific evidence ofeach trait.

3. Recruits are subjected to at least four in- .22 —.14 .69depth interviews.

4. {^impany actively facilitates deselection .23 .31 .64during the recruiting process by revealingminuses as well as pluses.

5. All professional employees in a particular .00 .7.5 .05discipline begin in entry level [X)sitionsregardless of prior experience or advanceddegrees.

6. Reward systems and promotion criteria .18 .56 .23require mastery of a core discipline as aprecondition of advancement.

7. The career path for profc'ssional employees .24 .74 —.02ill relatively consistent over the first six totc-n years with the company.

S. Virtually all professional employees can .70 .26 .15identify and articulate the firm's sharedvalues (i.e. the purpose or mission thatt!c's the firm to society, the customer orIts employees).

9. There are very few instances when actions ,6y .12 .06ol management appear to violate thefirm's espoused values.

10. Hm(>]oyees frequently make persona] .70 - . 1 8 .18sacrifices for the firm out of commitmentto the firm's shared values.

11. When confronted with trade-offs between .62 .24 .15systems measuring short-term results anddoing what's besr for the company in thelong term, the firm usually decides intavour ofthe long-term.

Percentage of variance explamed 29 13 10

252 D. F. Caldwell,]. A. Chatman and C A. O'Reilly

Table 2. Varimax factor loadings for commitment dimensions {N — 323)

Item

Whar this organization stands for is important tu

Varimax factor loadings

Normativet-ommirment

.72

Instrumentalcommitment

2. I talk up this organization to my friends as a great .HIorganization to work for,

3. If the values of this organization were different, I .6?would not be as attached to this otganization,

4. How hard I work for the organization is directly .37linked to how much I am rewarded.

5. In order for me to get rewarded around here, it is .06necessary to express the right attitude.

6. Since joining this orgatiization, my personal values .72and those of the organization have become moresimilar.

7. My private views about this organization are —.44different from those 1 express publicly.

H. The reason I prefer this organization to others is .82because of what it stands for, that is, its values.

y. My attachment to this organization is primarily .83based on the similarity of my values and thoserepresented by the organi/.ati<i(i.

10. Unless I'm rewarded for it in some way, I see no —.32reason to exptncl extra effciri on bfhait of thisorganization.

11. 1 am proud to tell others that I am part of this .80organization.

12. I feel a sense of'ownership' for this organization .68rather than being just an employee.

.02

.57

.68

50

.01

.66

- . 1 0

- . 2 8

Percentage of variance explained 42

Building organizational commitment 253

describes commitmc-nr bast-d on involvement cxch.ingc-d for specific rewards. Using the results described in

Table 2, two factor scores were coni[iu[ed for each respondent.

Control variables

in order to examine the hypothesized associiitions between or^anizarional recruitment and socializationprat tices ,IIKI commitment, numerous potentially confounding variables need tn be ruled out. Therefore, tourcontrol variables were used in the analysis. I'lrst, since tenure in the orj,'anuation has been shown to be relatedrouimrniiment (Hall, Schneider & Ny^rer., I9:'O; Sheldon, 1971). each respondent reported the length oftime he or she had worked in the hrm. Second, since recruitment and sociali^^atiiin practices may vary betweenlarije ;md small hrms (e.,i:. budj^et allotted) and since an individual's commitment may be S(Mnewhat• lepcndcnr (,n the si/e of ilie hrm (Steers, 19^7), respondents reported the number of employees in theiriir^ani/aiion (median —- 1500). Similarly, since recruitmint and socialization practices may vary acrossmdij.'.trii.s, a dummy variable was used tu index rhe difference between high technology and accounting firmsil) -•• hi,i;h tech, I ~ acc<Kmtmg) Finally, since perceptual measLires of recruirmeiit and .socialization wereused, aod since the lunnber of respondent.s per firm varied from \ t o 2 i , the actual number oOespondents foreach linn w.is entered as a control variable to <apture variations attributable to sample size (Al = 6.7 1,

SD = ^ -'.:^i.

Calculation of recruitment and socialization scores

Since multiple respondents described the socialization pra<tices for each firm, the responses had to beaggregated before relating the otgaiiii'ation variables to individual commitment. The most straightforwardmethod for talculating rhe scores ol a firm would be to average the responses of individual members on thethree ree rent merit and socialization factors. However, this woeild introduce some response'—response bias intoihe analysis To eliminate such bias, scores for the organizational dimensions were calculated for eachindividual based on the aggregate of the otiiers in the firm, but eliminating the fcKal individual's own score.This was done u.sing rhe following formula:

where x, represents the organization score assigned to the individual, x^ represents the set of scores of allrespondents within the organization, and N represents the number of respondents from the organization.Separate scores were calculated for each of the three organizational factors.

Results

Several patterns among the bivariate correlations, presented in Table 3, are worth noting.First, the only control variable related to commitment is industry. Accountants tend to bemore likely to report being instrumentally committed (r — .22., p < .01). This may bedue to the fact that one tnajor requirement for becoming a certified public accountant isthat a specific number of hours of public auditing be completed (Emerson, 1987). Thus,many of the accountants in this sample may have joined the firm in order to fulfil thisrequirement rather than to spend their careers in that particular firm. Longer tenurerespondents are more likely to be in larger firms {r = .31, p < -01) and high technologyfirms (r = .39, p < .01). This refiects the nature of the participants drawn fromaccounting firms, all of whom were comparatively new staff auditors. This can be seenmost clearly with the dummy variable which shows strong associations with both thenumber of respondents (r — .64, p < .01) and an emphasis on clearly structured rewardand career systems (r — .74, p < .01). The former finding reflects the larger number ofrespondents per firm obtained with the accounting firm sample, while the latter refiectsthe more formal promotion ladders in accounting firms (e.g. Stevens, 1981). A final

D. F. CaldwelLj. A. Cbatman and C. A. O'Rei/ly

Table 3. Correlations among variables (N = 291)

Variables i 2 3 4 "j 6 7 S

Control variablei1. Tenure —2. Firm size ,3 1**i. Industry

(0 - Hi-tech,I = Acag.) - , 3 9 " - . 26** —

4. Respondentsper firm - , ! 2 * - . 1 4 * .64**

Socialization5. Values - . 0 6 - . 0 5 - . 0 3 .14* —6. Rewards -.20** .07 .74** .44** - . 0 3 —7. Selection - . 0 7 . IS** .29** .30** . 17** .24**

Commitment

S, Normative .01 .01 .03 - . 0 1 .21** - . 0 6 .15**9- Insrrumenral - . 0 9 .02 .22** .10 - . 1 1 .14* .01 .00

• p <. O.O'i; ** p < 0 .01.

pattern of interest is the significant positive correlations between the number ofrespondents per firm and the organizational variables. Several explanations are possiblebut the most relevant one is that the number of respondents per firm is a potentialconfounding variable and should be controlled f(jr when testing the hypotheses.

There are also relationships among the organizational practice variables. The presenceof a set oi consistent, well-exphcated selection pohcies is pi^sitively related to anarticulated set oi shared values (r — 0.17,;? < 0.01) and to a well-defined set of rewardsand career paths (r = 0.24, {> < 0.01). Although these factors are orthogonal at thelndivitlual level of analysis, aggregation to the organi/^atiijnal level reveals these associ-ations; that is, while no correlations are present in terms of individual responses, a smallrcliitionshi]-) exists at the tirm level.

The general question motivating this study focuses on the relationship betweenintensive recruitment and socializatioti practices and the nature of the individual'scommitment to the organization. While Table 3 offers some general support for thesepropositions, more specific tests that examine these associations independently from otherpotential confounding factors such as individual tenure, industry and organizational sizeare needed. To test out three hypotheses we report two sets ot regressions, one for each typeot commitment, both without control variables (equations 1 and 3) and with controlvariables (equations 2 and 4) in Table 4.

The results in Table 4 are supportive of the general association, predicted iu hypothesesI and 2, between firm practices and normative commitment. Equations 1 and 2 show thatrecruitment procedures (hypothesis 1) and socialization practices (hypothesis 2) which aremote intensive, and specifically those which emphasize strong organizational valuesystems, are positively related to higher levels of commitment based on internalizationand identification. When the control variables are included in the equaticjn, the presenceoi a uniiorm formal reward system is negatively related to normative commitment.

Building organizational commitment 255

Table 4. Regression results; The effects of firm socialization practices on individual

commitment" •

Independentvariables

1. Values2, Rewards3. Seleeticyn4. Individual tenure5. Size of firm6. Industry

(0 == Hi-tech;i ---- Acetg.)

7. Respondentsper firm

Adjusted /-;"'

/-' ratio

Normativecommitment(no controls)

I

. IK**- . 0 9

—-

06

3. 2870.^6**

Normativecommitment

(controls)2

.20**- . 2 1 *

.14*,0K.04

- . 1 3

.067, 2763,60**

Instrumentalcommitment(no controls)

3

- . 1 0.14*.00—-

—

' • - •

-03^, 287^.06*

Instrumentalcommitment

(controls)

4

- . 0 9- . 1 1- . 0 7

- .02.16*.38**

- .03

.0^7. 2763.H**

• p • W); " • [' ••-- , 1 ) 1 .

IVH/I Rsiines represent stanelacdi«-d regression euethcitnts

providing partial support for hypothesis 3. In addition, the dummy variable tor industryshows that respondents Irom aicoimting firms have hi>;hc-r levels of normative commit-ment, j--erhaps reflecting the fac t that new accountants were more likely to have recentlyparncipatfd in orientation and training programmes.

The findings with respect co mstrumental or compliance-based commitment are alsointeresring. Hypothesis .i i.s icsted more fully in equations \ and 4. The equation withoutcontrol variables (3) shows that forniai reward systems are related to instrumentalcomrnirnu-nt m the positive direction which was [predicted in hypothesis ^. When thetontrt)! variables are entered this association becomes non-significant, but instrumentalcommiuncnt is then positively predicted by the size ofthe firtn and the industry dummy.Responeients from larger firms atid accounting firms report higher levels of instrumentalcommitment. Given the- increased tormalization typically found in larger organizations(c-.i . Hugh, Hickson, I-imnings &; Turner, 1968) and the rule-based nature of the publicaccounnni; profession (t-.g. Stevt-ns, 1981) these results are reasonable. It is alsoinrcresiiiis; to note that ncithiT rlie length of the individual's tenure with tlie orgaruzationiKtr till number of respondents sampled from tlie hrm is ever significant, suggesting thatthese variables do noi affect the lindings.

It should be noted that while the technique ot removing the focal individual from theanalysis solves the problem ot cornmon-method bias, it may simultaneously reduce themagnitude of the predicted associarions, and it thus constitutes a conservative test ot thehypotheses. To illustrate this pomt, consider the example of a respondent who describesorganizational values as strong and who also scores high on normative commitment as we

256 D. F. Caldwell,]. A. Chatfnan and C. A. O'Reilly

predict in hypothesis 2. Since this person's score on the independent variable {e.g.organizational values) is excluded when we calculate the index of organizational values forthat firm, the resulting composite score on the independent variable will be lowcted. If weassume that the same effect occurs in the opposite direction for respondents who score lowon normative commitment, the resulting correlation beween the firm estimate and thefocal individual's normative commitment score will be smaller than if the focal individ-ual's score had been included in the firm's composite socialization score. Therefore, ourresults may mderestimate the magnitude of the relationship between socialization practicesand mdividual commitment. Clearly, the extent to which our results systematicallyunderestimate this relationship depends, at least ia part, on the consistency with whichthe firm characteristics are rated. Thus, the higher the internal consistency among raters(excluding any one rater), the less likely it is that removing the focal individual from thecalculation will alter the ratings of the firm characteristics. In this study, as we reported inthe Method section, the raters were highly consistent about firm values, and moderatelyconsistent about rewards and selecti()n practices.

In addition, while the amount ot variance accounted for appears to be relatively small,interpretations of the results should reflect the techniques used in this study. Specifically,since each tocal individual's ratings ate removed from the calculation of organizatujnalrecruitment and socialization practices, the relationships demonstrated in the regressionequations represent a conservative estimate—one which is unaffected by common-methodbias—-ol how organization level phenomena affect individuals' attitudes and behavii)urs.In fact, studies which do not take steps to remove common-method variance may generateresults which overestiinaie the amount of variance accounted for by the rclationshi|T betweenfirm practices and individual commitment (cf. Abrahamson, 19H3).

This point can be demonstrated empirically by comparing our findings with common-method bias removed (Table 4), with the results which wouid have emerged if we had notremoved common-method bias. We calculated equations 1 -1 (from Table 4) using theidentical set of variables; however, the three organization characteristics variables did nothave the focal individual's ratings removed. Results of this 'biased' analysis revealed thatwhile the pattern ot significance among the standardized coefficients was virtuallyidentical to the 'unbiased' models m Table 4, the overall equations m the 'biased' modelsacc(]unted lor substantially more variance in normative-based commitment (equation 1:F = 5').29, /' < .01, R' - ,^4; equation 2: /• = 24.66, p < .01, R^ = .37) andslightly more variance in instrumental-based commitment (equation y. I' ~ 9.633,/; <• .()],R^- = .08; equation 4: f" - 4 .58, / ' < .1)1, R' - .08). In sum, one outcomeof this research has been to demonstrate the importance of removing common-method biasin studies which rely on self-report data in order to arrive at a truer estimate ofrelationships between independent and dependent variables.

Discussion

Overall, the tindings ot this study provide support for the hypotheses proposed. Theresults show a significant positive relationship between strong organizational recruitmentand socialization practices, and individual commitment. When firms have well-developedrecruitment and orientation procedures and well-defined organizational value systems,respondents manifest higher levels of normative commitment to the organization. This

Building organizational commitment 257

finding is consistent with much earlier theorizing, but it has seldom been empiricallydemonstrated (e.g. Jones, 1986; Pascale, 1985; Van Maanen & Schein, 1979). Further,these findings exist alter controlling for possible response-response bias, and for a set ofpotential alternative explanatory variables such as individual tenure, industry, thenumber of firm members rating organizational recruitment and socialization practices,and organizational size. In addition to these effects, well-articulated reward systems arepositively related to instrumental based commitment, also suggested in earlier theoreticalwork (e.g. Becker, 1960; Etzioni, 1975; Gould, 1979).

These findings are useful increments to our understanding of the relationship betweenorganizational practices and individual commitment in several ways. First, as suggestedby previous research, the concept of organizational commitment may reflect multipleunderlying factors. O'Reilly & C:hatman (1986) argue that commitment can develop fromthree separate sources of attachment. Using a larger and a more representative sample thantheir original research, our results showed that two of the three factors identified byO'Reilly & Chatman (1986) collapsed to form a single dimension. Despite this difference,the implications of this study are similar to theirs. Findings from both studies support thenotion that commitment is a multifaceted construct, and that without clarity in thespecific aspects of commitment being studied results may be somewhat ambiguous.

In addition, the findings in Table 4 suggest that, at a general level, both extrinsic andintrmsic factors can operate to influence commitment in a way that is consistent with asufficiency ofjustification hypothesis (Caldwell, O'Reilly & Morris, 1983). For example,the positive effects of selection and strong organizational value systems on normativecommitment may stem from processes of incremental commitment and strong socialconstructions. These early organizational experiences may lead to a lack of extrinsiciusrification for behaviour, and thus may be manifested in attachments based onitientihcation and internalization (e.g. O'Reilly & Caldwell, 1981; Salancik & Pfeffer,1978) At the same time, a clear formal reward system may undermine normativecommitment, as suggested by the negative coefficient for the reward factor in Table 4,while simultaneously promoting instrumental based commitment. The potential contra-dictory effects are reminiscent of earlier research which alludes to the tension betweencosmopolitan and local orientations of professionals in bureaucratic organizations (e.g.Flango &. Brumbaugh, 1974; Gouldner, 1957).

The present study is alsti useful m highlighting areas in which additional investigationsare needed. First, since organizational practices such as reward systems and recruitmentprocedures vary across firms, there is a need for multi-organizational samples. Althoughthis study used a sample oi 45 firms and an analytic approach to help control forresponse response bias, clearly, a more focused and objective examination of specificptactices IS needed. For instance, studies which measure more objective aspects ofrecruitment and orientation practices, such as the number of hours an intlividual spends inselection procedures and indoctrination programmes, and the types oi activities, couid beimportant (e.g. Chatman, 1989; Feldman, 19^6; Louis, 1980).

The significance o( the results of this study may seem rather small in terms ofexplanatory power. However, in addition to the specific statistical explanation provided inthe Results section, a number of compelling conceptual explanations can be offered aswell. Fitst, our measurement of recruitment and socialization practices is somewhatimpressionistic -that is, such practices are less proximal and more reliant on member

l^- P- Caldwell.]. A. Chatman and C. A. O'Reillyperceptions than on 'obiective' accounts of recruitment and socialization. Perceptuafiissessments may in fact be more accurate since it may be easier for people to report globalimpressions rather than specific experiences (e.g. Louis, 1980; Van Maanen & Schein1979). Second, while there have been other explanations offered fot individual commit-ment (e.g. Steers, 1977), no other studies have used class level variables which are basedon group perceptions. In this sense, it is somewhat remarkable that any variance at theindividual level was explained. The literature on strong culture organizations may alsooftet important insights. Researchers and practitioners have argued that strong cultureorganizations, which arc usually defined as having strong recruitment and socializationpractices (e.g. Davis, 1984; Deal & Kennedy, 1982), are also more likely to have a highlycommitted workforce (e.g. Peters & Waterman, 1982). If the strength of an organiz-ation's culture exists on a continuum (e.g. Jackson, 1966; O'Reilly, 1983), the presentsample of firms would clearly fall in the middle of this continuum. In a sense, then, thisstudy offers a conservative test of the hypotheses. If we had examined strong culture firms,where recruitment and socialization practices were heavily emphasized, the amount of"variance m individual attitudes and behaviours explained by such organization leveljiractices would be higher. Thus, a second focus for future research is on studies whidiexamine the impact of strong culture firms on individual attitudes and bcliaviours,

A related criticism of this study might also be that the questions on each of the twosurveys used here can be seen as similar in content (e.g. their focus on values) and form.However, our design allowed us to separate ratings of the firm from individualcommitment responses. In essence, the two instruments were designed to ca[-)ture twodifferent levels of analysis: organization level practices, and individual level commitment.Given the difficulties of comparing individual level constructs to organization levelpractices (Rousseau, 1985), a contribution of the approach taken here is that it offers a wayof crossing levels of analysis. Future studies which rely on self-report data for bothindependent and dependent variables should be sure to assess the extent to which theconsistent y within firms is high (as would be the case in strong culture firms). When thereIS liigh consistency among firm raters about important organizational characteristics,removing common-method bias, as described here, will generate more accurate resultsthan wouid be generated if such bias were not taken into account.

f'irially, while organizational commitment is one important outcome of recruitmentand socialization experiences, there are other important outcomes such as motivation,withdrawal and performance, which should be examined. Such outcome-foe used researchoffers another approach to clarifying the nature of organizational commitment.

Overall, much that is written suggests that the commitment of individuals will berelated to their organizational experiences. The results of this study indicate that thisproposition is true and further suggest that these experiences may differentially affect theform that such individual attachment to the organization takes.

References

Abrahamson. M. (1983). Social Research Methodi. Englewood CWfh, NJ: Prenrice-Hall,Angle, H. &Per ry , J . (198!). An t-mpirical assessment of organisational commitmenr and org

e t i c c r i v c n e s s . AdmintUralive Science Quarterly. 2 6 , 1 - l ' i .

Building organizational commitment 259

Arne)ld, H. & Feidman, D. (1982). A mulnvariate analysis oi the determmants ofthe job turnover Journal oj

Applied Psychology. 87, ^5()-36O.Becker, H. S. (I960). Notes on the concept of commitment. American Journal oj Sociology, 66, 32-40.Bray, D. W., Campbell, R. J . & Grant, D. I.. (1974). The Formative Years in Business. New York:

WyBuchanaii, B. (1974), Building organizational commitment: The socialization of managers in work

organizations. Administrative Science Quarterly, !9, 533-546.CaMwel!, D. F., O'ReilJy, C, A. ikMorm.,}. \i. (1983). Kt-spo/ise to an organjzational reward: A iieJd tcsrof

the siiffiuency ot justification hypothe^h. Journal oj Personality and Social Psychology, 44, 506-514.Chatman, | . (1989). Matching people and organizations: Selection and socialization in public accounting

firms. Academy of Management Best Paper Proceedings. Washington, DC: Academy of Management.Davis, S (1984). Managing Corporate Culture. Cambrtd,i;c, MA: BalHnger.Deal, 'V. 6. Kennedy, A, 11982), Corporate Cultures. Rea<lirig, MA: AddiSon-WesSey,limctsoii, ), (1987). Careers in publii. aceouniin^ in 1986-87 A comprehensive comparison of the big eight

firms. 'I htf Bij; Eight Revieu\ St'attle, WA,Eizioni, A, (1975). A Comparative Anal)M\ oj Complex Organizations. 2nd ed. New York: Free Press.Fiirrcll, D. &. Rusbult, C. (198!), Exchange vatiables as predictors e f job satisfaction, job commitment, and

turnover. The impact e)f rewards, ceists, alternatives, and investments. Organizational Behavior and HumanPerfirniancK. 28, 7«-95.

R'ldman, D. (1976). A eontinfjency thenry nl socialuation. Administrative Science Quarterly, 21, 4-^3^52,Feidman, D. (1977), The role ot initiation activities in socialization. Human Relations. 30, 977-990.FLuigf., V & Brumbaugh, B. (1974). The eiimen-sionaliry ot the cosmopolitan-local construct. Administrative

Science Quarterly, 19, 19H-2iO.Galanter, M. (1980). Psychological induction into the large groups: Findings from a modern religious sect.

American Journal oj Psychiatry. 1 3 7 1 5 7 4 - - - I 5 7 9 .

C,al:inie-r. M. (1982). (^harisinarii teli^'ious sects and psychiatry: An ovetview. American Journal of Psychiatry,

GouM, S. ( 1979). An equity-c-xchange model of otgani/ational involvement. Academy of Management Revien\4, ')^-62,

GouMiK'f, A. (19") 7). f.bsmopolitan?. and locals: Toward an analysisof latent social roles. Administrative ScienceQuarterly, 2, 281-^06.

Hall, 1.). '1'. (1976). Careers in Organizatiom. Santa Monica, CA: Gooeiyeat.Hall, I), r . , Schneider, B. & Nygten, H. J. (1970). Personal factors in organizational identification.

Administrative Silence Quarterly. 1 $ , 1 7 6 - 1 8 9 .

Hom, P., Katerberg, R. &L Hulm, (.. (1979). (Comparative examination of three approaches to the predictionot turnover, journal oj Appied Psychology. 64, 280-290.

Jyik.s(>j), ). (1966) A cunecptua} and nii'.isurement model ior norms and roles, Pacijic Sociolognal Rti'itw. 9,^^ 17.

Jones, ( j . R. (1986). Socialization tactics, selt-etficacy, and newcomers' adjustments to organizations.Auidemy oj Management Journal, 29, 262—279.

Kanter, K. M. (1968). Commitment and Community. Cambridge, MA: Harvard University Press.Keiinan, H. (,, (1958). Compliance, identification, and internalization: Three pnxesses of attitude change.

Journal <i( Conjhct Resolution. 2, 5 I -6(1.Kiesk-r, ( . A, (Hd ) ( 197 I). The PsychtJogy oj Commitment: Experiments Linking Behavior to Reltef. New York:

At;ideniit PressLincoln, J R. & Kallcbt-rg, A. L. (1985). Work organization and workforce commitment: A study of plants

and e'lnployees in the LI.S. and japan, American Sociological Reveiew. 50, 738-760.L)tiis, M. R. (I98U). Stirpriit a/id stviSf making: WJwr newcomers experience' in entering utiiamilmr

e)ri;iiriii'ati()nal sctrinjjs. Administrativi Science Quarterly. 25, 226-25 1.Me-(ic-e, ((. & 1-ord, R, (1987). Two (or more) dimensions ol organizational commirmetic: Reexaminarion of

rhe .ittcctive anil eonnnuame ciimmitmeii! seales. Journal oj Applied Psychology. 72, 6.^8- 642.Meyer, (. P. 6L Allen, N. J. ((984). Te.-sting the 'side-bec theory' of organizational commitment, journal oj'

A p p l i e d Piycholoi^y. 6 9 , T ^ 2 -^IH

Meyer, ). F. & Allen, N. J (1987). A longitudinal analysis of the early development and consequences ofotijani/arional lommitme'nt Canadian Journal oj Behavioral Scieme. 19, 199- 2 15.

260 D. F. Caldwell.J. A. Cbatman and C. A. O'Reilly

Morns, J. H. & Koch, J. L. (1979). Impacts of role perceptions on organizational commitment, jobinvolvement and psychosomatic illness amon^ chref vocational groupings, journal of Vocational Behavior15, 88-101.

Morns, J. H. & Sherman, J. D, (1981). Genc-ralizabiliry of an organizational commitment model. Academy ofManaj!_ementJournal, 24, 5 l 2 - ' i 2 6 ,

Morrow. P, C;. (1985). Concept redutidancy in organizational research: The case of work commitmentAcademy of Management Review. 8, 4H(^'y(H.).

Mowday, K. T.. Porter, L. W. & Dubin, R. (1974). Unit performance, situational factors, and employeeatntiides in spatially separated work units. Ovf^anizatiomil Behavior and Human Perfom/anci^. 12, 2 ^ - 2 4 8 ,

Mowday, R. T., Porter, L. W. & Steers, R. (1982). Organizational Linkages The Psyiholo^y af'commitnienl.Ahstnteenm. and Turnover. New York: Academic Press.

OReilly, C. A, (i9H9). Corporations, cults and organizational culture: motivation and social control inorganizations. California Manajiemen! Review. 3 1 , 9 - 2 ^ .

O'Redly, C. A. & C:uiawett, D. P. (!98(t). Job chmcc: The impact ol intrinsic factors on subsequentsatislacrion and commnmenr. Journal of Applied Psychology. 65, 559-56^.

O'Reilly, C, A. &Caldwell, D. F. (1981). The commitment and job tenure of new employees: Some evidenceot postdc-cisionai justification. Adminnrrative Science Quarterly. 26, ^97-616.

O'Reilly, C A. & Chatman, J, A. (19S6). Organizarional commitment and psychnioj^ical attachment: Theeftetts o( compliance, identificatioTi and incernalixation un |irosocial behavior, /ournal of Applied Pfyfholoii}

71, 492-19V.OReilly, C. A., Cliatmaa, J, A. & Oildwell, D. F. (199(1). Peopit!, jobs and ogamzathma! culture: A ^-sort

approach co assessing lit. Working paper. University of California, Berkeley.Organ, 1>. W. (1988), Organizational Citrzemhip Behavior. Lexington. MA: Lexington Books.Ouchi, W. (1980). Markets, bureaucracies, and clans, Adminntrative Science Quarterly. 25, 129-141.Pascale, R. (I9H5). The paradox c)( 'corjiorate CLitturt-': Reconciling ourselves to socialization. California

Management Revieu.'. 27. 26- ''tO.Peters. T. & Waterman, R. (1982). In Sairch of Excellence: LeMomfrm America's Best-run Cotupanm. New York:

Harper & Row.Porter, L. W., Crampon, W . J . Ac Smith, F, j . (1976). Organizational commitment and managerial turnover:

A longitudinal srudy. Organizational Behavior and Human Performance, 1 5, 86-98.Porter, L. W., Steers, R., Mowday, R. & Boulian, P. (1974). Organizational commitment, job satisfaction,

and turnover among psychiatric ivchnkmns. Journal of Applied Psychology. 59. 603-609.Premack, S. L. & Wanous, J. P, (198'i). A mt-ta-analysis of realistic job preview experiments, journal of

Applied Psychology. 70, 706-719.Pugh, D., Hickson, D,, Hinntngs, C. & Turner, C (!968). Tht context of

A d m i n n t r a t i v e Silence Quarterly, l 4 , 9 1 - 1 1 4 .Reichers, A. H. (1985). A review of reconceptiialization of organizational commitmenr. Academy of

Management Review, 10, •165-^^76.Rousseau, D. (1985). Issues of level in organizational research: Multi-level and cross-level perspectives. In

B. Sraw & L. C^ummings (Eds), Research in Organizatiomal Behavior, vol. 7. Greenwich, CT: JAI Press.Siilancik, G. (1977). Commitment and the control of organizational behavior and belief. In B. Staw &

G. Salantik (Hcis), Heu: Oirectiom tn Orgamzatitinal Behavior, pp. t -54 . Chicago: St. Claic.Salancik, t i & Pfeffer, J. (1978). Asocial informarion processing approach ro job attitudes and task design.

Administrative Science Quarterly. 2 3 , 2 2 4 — 2 5 i ,Schneicier, B. & Reichers, A, ( i98i) . On the etiology of climates. Personnel Psychology. 36, 19-39.Sheldon, M. (197 I). Investments and involvements as mechanisms producing commitment ro the organiz-

arion. Adminntnitive Science Quarterly, 16. 143—150,Staw, B. (1977). Two sicfcs of commitment. Paper presented ar the annual meeting of the Academy ot

Management, Orlando., Florida.Sreers, R. M. (1977). Anreccdetits and consequences ot organizational commitment. Adminiitratii'e Science

Quarterly. 22, 46 56,Stevens, j . M., Beyer, j . M. & Trice, H. M. (I97S). Assessing personal, role and organizational predictors of

managerial commitment. Auideiny of Management journal. 21, 3^1) .S96.Stevens, M, (198!). The Big liight. New York: McGraw-Hill.

Btiiiding organizational commitment 261

Stumpf, S. & Hartman, K. (1984). Individual exploration to or^anizationa! commicmenr or withdrawal.Academy oj Management Journal. 27. H)8—^2').

Van Maaneti, J. & Sthtin, I;. H. (1979). Towards a theory of iirganizational sociali/arion. In B. Staw (Ed.),He^mrch in Organizational Behavior, vol. 1. pp. 209-2fii.

Reieitvd IH May I9H9: revised ftrsion received 2') January 1990