ＣＯＮＴＩＮＧＥＮＣＹＦＡＣＴＯＲＳＵＰＯＮＳＴＲＵＣＴＵＲＥＡＮＤＯＲＧＡＮＩＺＡＴＩＯＮＡＬＰＥＲＦＯＲＭＡＮＣＥ：

(1)

8１

ＡＮＥＭＰＩＲＩＣＡＬＥＸＡＭＩＮＡＴＩＯＮＯＦＩＮＴＥＲＡＣＴＩＯＮＥＦＦＥＣＴＳＯＦＣＯＮＴＩＮＧＥＮＣＹＦＡＣＴＯＲＳＵＰＯＮＳＴＲＵＣＴＵＲＥＡＮＤ

ＯＲＧＡＮＩＺＡＴＩＯＮＡＬＰＥＲＦＯＲＭＡＮＣＥ：

ＡＭＵＬＴＩＰＬＥＨＩＥＲＡＲＣＨＩＣＡＬＬＥＶＥＬＳＡＰＰＲＯＡＣＨ１

KENJIYOSHIDA

HoseiUniversity

Anumberofdeterminantsoforganizationalstructurehavebeenproposedby organizationalresearchers・However，interesthasgenerallyconcentratedonthree factors：size，technology，ａｎｄenvironment（FordandSlocum，1979；Hall，1982；

Daft，1983；Pfeffer，1985)．Also，moststudieshavefocusedontherelationship betweenasinglefactorandstructure・Ｆｏｒexample，Blauandhisassociates（e９．，

Blau，1968；BlauandSchoenherr，1971）ａｎｄｔｈｅＡｓｔｏｎＧｒｏｕｐ（e９．，Pugh，Hickson，

andTurner，1968;Hickson，Pugh，andPheysey，1969）havearguedfortheimportan-

ceofsize；Woodward（1965）andPerrow（1967）haveemphasizedtechnology；and

BurnsandStalker（1961）andLaWrenceandLorsch（1967）haveconsideredenviron- menttobethemostimportantfactordeterminingstructure・Fewresearchershave consideredmultiplefactorssimultaneously．（Exceptionsareseveralstudiesoftechn- ologyandsize［e９．，ｂｙＤｅｗａｒａｎｄＨａｇｅ，1978；ＤｅｗａｒａｎｄＳｉｍｅｔ,1981；Marshand Mannari，1981；Ｃarter，1984；andSingh，1986].）Andfewindeedhaveconsidered relationsbetweenstructureandallthreeoftheprincipalcontingencyfactorsat

once．Ｔｈｉｓｒｅｓｅａｒｃｈｈａｓｄｏｎｅｓｑ

Ｈｏｗａｎｏｒｇａnizationdesingsitsstructurewhenitfacesmultiplecontingency factorswhichhaveconflictingimplicationsforstructure（Child，１９７７；Fordand

Slocum，1977；Mintzberg，１９７９；ＶａｎｄｅＶｅｎａｎｄＤｒａｚｉｎ，1985）isanobviously interestingbutstillunansweredquestion．Ｆｏｒexample，whenanorganization

faceshighlyroutinetasksandperceivesahighlyuncertainenvironment，how

formaldoesitsstructurebecoｍｅ？Contingencyfactorsmaybegenerallythought

tohaveconflictingimplicationsforstructureandperformance，ｂｕｔｓｏｆａｒｔｈｅｒｅｈａｖｅｂｅｅｎｎｏｓｙｓｔｅｍａtictheoreticalstatementsabouttheinteractioneffectsofthe

factorsAsastarttowarddescribingtheseinteractioneffects，［ofthecontingency factors（size，technology，ａｎｄenvironment）onstructureandperformance,］this

studytookasapointofdepartureMintzber９，ｓ（１９７９）argumentthateach contingencyfactoraffectsthevariouslevelsofstructuredifferentially・

Therehasbeenachronicproblemofhowbesttomeasurestructure，

technology，ａｎｄenvironment，particularlywithrespecttotheappropriateunit ofanalysis・Thisstudyusedsubunitsrathertha、anorganizationasawholeas theunitofanalysis．Manyresearchershavefoundsubunitswithinlargeand complexOrganizationｓｔｏｖａｒｙｉｎｔｅｒｍｓｏｆｔｈｅｉｒｓｔｒｕｃｔｕｒｅ（Hall，1962；Dewarand Werbel，1979；ＦｒｙａｎｄＳｌｏｃｕｍ，1984)，technology（GrimesandKlein，1973；Vande VenandDelbecq，１９７４；ComstockandScott，１９７７），andtheirenvironment

＝

(2)

8２

(LawrenceandLorsch，1967；Duncan，1972；Tung，1979)．Anotherreasonforusing subunitsastheunitofanalysisisthatseveralstudieshaveshownstronger relationshipsbetweentechnologyandstructUreatthesubunitlevelthanateither individualororganizationlevels（Hrebiniak，１９７４；ＶａｎｄｅＶｅｎ，Delbecq，and Koenig，１９７６；Fry，１９８２）．Useofthesubunitlevelforanalysishasalso beensuggestedasameansofexaminingrelationshipsbetweentechnologyand structureandbetweenenvironmentandstructurebysucｈｗｒｉｔｅｒｓａｓＶａｎｄｅＶｅｎ，

Delbecq，andKoening（1976)，Aldrich，（1979)，Schoonhoven（1981)，FryandSlocum U984)，ａｎｄＴｏｓｉａｎｄＳｌｏｃｕｍ（1984)．

Thepurposeofthisstudy，ｔｈｅｎ，wastotesttheeffectofhierarchicallevelon therelativeimportanceｏｆｓｉｚｅ，technology，andenvironmentforstructureandto testspecificcontingencyhypothesesabouttheserelations，ｔｈａｔｉｓ，howsize，

technology，andenvironmentcombinewithstructuretoaffectperformanceatthe subunitleveLAreviewoftheliteratureonrelationshipsamongthethree contingencyfactors，structure，andperformancewillserveasabasisforderiving thespecificcontingencyhypotheses．

ＬＩＴＥＲＡＴＵＲＥＲＥＶＩＥＷ

Size，Structure，andPerformance

Therehavebeenmanyempiricalstudiestoshowthatsizeisrelatedto structure（ＰｕｇｈｅｔａＬ，1969；Hickson，Pugh，andPheysey，1969；Blau，1970；Blau andSchoenherr，1971；ChildandMansfield，1972；Meyer’１９７２；Mileti，Gillespie，and Haas，1977；Mintzberg，1979；Hall，1982)．Thesestudieshavefocusedprimarilyon therelationshipbetweensizeandstructurebuthaverarelymeasuredperformance directly、Consequentlytheyhavebeenunabletoshedlightonhowsizeand structureaffectorganizationalperformanceoutcomes・Ａｔｍｏｓｔ，therefore，they provideonlyindirecttestsofacontingencytheory、Thefewempiricalstudiesthat doincludeaperformancemeasure（CumminsandKing，1973;Child，1975；Evers，

Bohlen，andWarren，1976)，however，seemtosuggestthatassubunitsizeincreases，

moreformalizedandspecializedbutlesscentralizedstructuresresultinmore effectiveperformance．

Technology，StrucCure，ａｎｄＰｅｒｆｏｒｍａｎｃｅ

Ｔｈｅｒｅｈａｖｅｂｅｅｎｍａｎｙｅｍｐｉｒｉｃａｌｓｔｕｄｉｅｓｔｏｓｈｏｗｔｈａｔｔｅｃhnologyisrelatedto structure（Bell，1967；HageandAiken，1969；GrimesandKlein，１９７３；Hrebiniak，

１９７４；ＶａｎｄｅＶｅｎａｎｄＤｅｌｂｅｃｑ，１９７４；ＶａｎｄｅＶｅｎ，Delbecq，andKoenig，１９７６；

ComstockandScott，1977)．However，theyhavelookedmostlyattherelationship betweentechnologyandstructureand，againfailedtomeasureperformance directlyandsohavebeengenerallypreventedfromexaminingtheeffeｃｔｓｏｆｔｈｅ

“fit，，betweentechnologyandstructureonperformance．Hencelthesestudies，

too，provideonlyindirecttestsofacontingencytheory-

Lately，however，therehavebeenanincreasingnumberofstudiesthatinclude

、

(3)

8３

aperformancemeasuretotestacontingencytheory（DewarandWerbel，１９７９；

Schoonhovenl981；Argote，１９８２；ＦｒｙａｎｄＳｌｏｃｕｍ，１９８４；AlexanderandRandolph，

1985)．Ｔｈｅｓｅｓｔｕｄｉｅｓｓｅｅｍｔｏｓｈｏｗｔｈａｔａｓｔｅchnologymovesfromroutinetononro- utine，subunitsemploylessformalizedandcentralizedbutmorespecializedstructu-

resinordertobeeffective．

Environment，Structure，andPerformance

Someempiricalstudieshaveexaminedtheeffectsofenvironmentandstructure

onperformance（BurnsandStalker，１９６１；LawrenceandLorsch，１９６７；Duncan，

1973；Khandwalla，1973；NegandhiandReimann，1973；Child，1975；Pennings，l975l Theresultsofthesestudiesindicatethatasperceivedenvironmentaluncertainty increases，subunitsemploylessformalizedandcentralizedbutmorespecialized

structurestobeeffective．

ＨＹＰＯＴＨｎＳＦｌＳ

Therehasbeenmuchdisagreementabouthowtotestcontingencypropositions，

thatｉｓ，howrelationshipsbetweenstrucCureandthecontingencyfactorsofsize，

technology，andenvironmentarerelatedtoperformance（Pennings，１９７５；Miller，

1981；Schoonhoven，１９８１；TosiandSlocum，１９８４；ＶａｎｄｅＶｅｎａｎｄＤｒａｚｉｎ，１９８５；

ＤｒａｚｉｎａｎｄＶａｎｄｅＶｅｎ，1985；ＦｒｙａｎｄＳｍｉｔｈ，1987)．ThisstudyusedJoyce，

Slocum，andGlinow'ｓ（1982）generalcongruencymodeloffit・Thismodelargues thatperformancewillincreasewhenstructureismatchedwithsize，technology，

andenvironment、Ｅｍｐｈａｓｉｓｉｎｔｈｅｍｏｄｅｌｉｓｏｎｔｈｅｓｉｍｉｌarityandmatchingof levelsofstructurewitheachcontingencyfactorasdeterminantsofperformance、

ThisdefinitionoffitisidenticalwiththeonemadebySchoonhoven（1981）and FryandSlocum（1984)．

HierarchicallevelsandContingencyFactors

Mintzberg（1979）proposedthateachcontingencyfactoraffectedthevarious levelsofstructuredifferentially．Ｔｈａｔｉｓ，sincepresumablythetoplevelmust functioncontinuouslyatanOrganization'ｓboundaries，itismostaffectedby

environment.ThemiddleleveLontheotherhand,producesextensivestructural

elaborationbycreatingchangesinthefavoredmechanismofcoordination，and thereforeisprimarilyaffecteｄｂｙｓｉｚｅ・Finally，thelowerlevelisaffectedmainly

bytechnology．Mintzberg（1973)，Ｋａｔｚ（1974），andPaolillo（1981）seemto supporttheMintzber,ｓ（1979）proposition，ｂｕｔAlexander（1979）andPavettand

Lau（1983）ｄｏｎｏｔｓｕｐｐｏｒｔｉｔ、Theseempiricalstudiesaretoofew，however，fora

conclusionontheeffectofhierarchicallevelontherelativeimportanceofsize，

technology,andenvironmentforstructure，Therefore,Mintzberg'ｓ（1979）proposition

wasfurthertestedinthisstudy・

Hypothesisl：Eachcontingencyfactorwillaffectthevariouslevelsof structuredifferentially．Environmentisthemostdominantfacｔｏｒａｔｔｈｅ

-1〆

(4)

8４

toplevel，sizeatthemiddlelevel，andtechnologyatthelowerlevel．

Size，Structure，andPerformance

Theempiricalstudieswehavenotedshowthatthelargerthesubunits，the moreformalizedandspecializedbutlesscentralizedstructuresarerelatedtoperfo- rmanceeffectivenessHence，

Hypothesis2： Sizewillinteractwithformalizationtoinfluence performance・Whensizeislarge，increasesinformalizationwillpositively influenceperformance・Ｗｈｅｎｓｉｚｅｉｓｓｍａｌｌ，increasesinformalizationwill negativelyinfluenceperformance・

Performance3： Sizewillinteractwithcentralizationtoinfluence

performance，Whensizeislarge，increasesincentralizationwillnegatively influenceperformance．Ｗｈｅｎｓｉｚｅｉｓｓｍａｌｌ，increasesincentralizationwnl positivelyinfluenceperformance・

Hypothesis4： Sizewillinteractwithspecializationtoinfluence performance．Whensizeislarge，increasesinspecializationwillpositively influenceperformance・Ｗｈｅｎｓｉｚｅｉｓｓｍａｌｌ，increasesinspecializationwill negativelyinfluenceperformance．

Technology，Structure，andPerformance

Giventheempiricalstudiesnotedearlier，itappearsthatastechnologymoves fromroutinetononroutine，subunitsemploylessformalizedandcentralizedbut morespecializedstructuretobeeffectivaHence，

Hypothesis5： Routinetechnologywillinteractwithformalizationto influenceperformance・ Whentechnologyisroutine，increasesin formalizationwillpositivelyinfluenceperformance・ Whentechnologyis nonroutine，increasesinformalizationwillnegativelyinfluenceperformance、

Hypothesis6： Routinetechnologywillinteractwithcentralizationto influenceperformance、Whentechnologyisroutine，increasesin centralizationwillpositivelyinfluenceperformance・Whentechnologyis nonroutine，increasesincentralizationwillnegativelyinfluenceperformance・

Hypothesis7： Routinetechnologywillinteractｗｉｔｈspecializationto influenceperformance Whentechnologyisroutine，increasesin specializationwillnegativelyinfluenceperformance・ Whentechnologyis nonroutine，increasesinspecializationwillpositivelyinfluenceperformance．

Environment，Structure，andPerformance

Theempiricalstudiesreviewedaboveseemtoshowthatasperceived environmentaluncertaintyincreases，subunitsemploylessformalizedand centralizedbutmorespecializedstructureinordertobeeffective・Hence，

Hypothesis8： Perceivedenvironmentaluncertaintywillinteractwith formalizationtoinfluenceperformance・ Whenperceivedenvironmental uncertaintyishigh，increasesinformalizationwillnegativelyinfluence

(5)

8５

performance・Whenperceivedenvironmentaluncertaintyislow，increasesin formalizationwillpositivelyinfluenceperformance・

Hypothesis9： Perceivedenvironmentaluncertaintywillinteractwith centralizationtoinfluenceperformance・ Whenperceivedenvironmental uncertaintyishigh，increasesincentralizationwillnegativelyinfluence performance．Whenperceivedenvironmentaluncertaintyislow，increases incentralizationwillpositivelyinfluenceperformance・

HypothesislO： Perceivedenvironmentaluncertaintywillinteractwith specializationtoinfluenceperformance・ Whenperceivedenvironmental uncertaintyishigh，increasesinspecializationwillpositivelyinfluence performance・Whenperceivedenvironmentaluncertaintyislow，increases inspecializationwillnegativelyinfluenceperformance．

ApplicationsoftheHypothesestoHierarchical Tablelsummarizesthehypothesized contingencyfactorsandstructure．

Levels

relationshipsbetweenthethree

ＴＡＲｍｎｌ

HypothesizedRelationshipsBetweenContingency

FactorsandStructurｅ

Forma1ization Centralization Specialization

Ｓｉｚｅ

(numberofemployees） ^＋ ^＋

(tas鰯lHIfl;f:Ls） ^＋ ^＋

Environment

(perceivedenviron‐

mentaluncertainty）

＋

Thistablepositsthatsizeandtechnologyaffectformalizationpositively；

however,environmentaffectsformalizationnegativelyTherefore，size／technology

andenvironmenthaveaconflictingimplicationforformalizationWhenasubunit islarge，faceshighlyroutinetasks，andperceivesahighlyuncertainenvironment，

ｈｏｗｆｏｒｍａｌｄｏｅｓｉｔｓｓｔｒｕｃｔｕｒｅｂｅｃｏｍｅ？

Atthelowerlevelofhierarchy，ｔｅｃｈｎｏｌｏｇｙｉｓｔａｋｅｎｔｏｂｅｔｈｅｍｏｓｔｄｏminant

factordeterminingformalization（Hypothesisl）．Thepositiveeffectsoftechnology

andsizeonformalizationexceedthenegativeeffectofenvironment・Therefore，

Hypothesis5willapplyatthelower１ｅＶeLAtthemiddlelevelofhierarchy，ｓｉｚｅｉｓａｒｇｕｅｄｔｏｂｅｔｈｅｍｏｓｔｄｏｍｉｎａｎtfactordeterminingformalizatio、（Hypothesisl)．

Thepositiveeffectsofsizeandtechnologyonformalizationexceedthenegative

(6)

8６

effectofenvironment、Therefore，Hypothesis5willapplyatthemiddleleveLAt thetoplevelofhierarchy，environmentisthemostdominantfactordetermining formalization（Hypothesisl）．Thenegativeeffectofenvironmentonformalization exceedsthepositiveeffectsoftechnologyandsize．Therefore，Hypothesis5will notapplyatthetoｐｌｅｖｅｌＴｈｅｓａｍｅｌｏｇｉｃｉｓｆｏｌｌｏｗｅｄｉｎexaminingapplications ofHypotheses2and8toeachhierarchicalleveLHence，

HypothesislL Hypotheses2and5wilｌａｐｐｌｙａｔｔｈｅｌｏｗｅｒａｎｄｍｉｄｄｌｅｌｅｖｅｌｓｏｆｈｉｅｒａｒｃｈｙｂｕｔｗｉｌｌｎｏｔａｐｐｌｙａｔｔhetopleveLHypothesis8 willapｐｌｙｏｎｌｙａｔｔｈｅｔｏｐｌｅｖｅｌａｎｄｗｉｌｌｎｏｔapplyａｔｔｈｅｌｏｗｅｒａｎｄ

ｍｉｄｄｌｅｌｅｖｅｌｓ

Ｔｈｅｆｏllowinghypothesesfollowthesamelogicastheabove

Hypothesisl2：Ｈｙｐｏｔｈｅｓｅｓ３ａｎｄ９ｗｉｌｌａｐｐｌｙａｔｔｈｅｍiddleandtop levelsofhierarchybuｔｗｉｌｌｎｏｔａｐｐｌｙａｔｔｈｅｌｏｗｅｒｌｅｖｅＬＨｙpothesis6will applyonlyattheloweｒｌｅｖｅｌａｎｄｗｉｌｌｎｏｔａｐｐｌｙａｔｔｈｅｍｉｄｄｌｅａｎｄｔｏｐ

ｌｅvels・

Hypothesisl3：Ｈｙｐｏｔｈｅｓｅｓ４ａｎｄｌＯｗｉｌｌａｐｐｌｙａｔｔｈｅmiddleandtop levelsofhierarchybｕｔｗｉｌｌｎｏｔａｐｐｌｙａｔｔｈｅｌｏｗｅｒｌｅｖｅＬＨypothesis7will applyonlyattheloｗｅｒｌｅｖｅｌａｎｄｗｉｌｌｎｏｔａｐｐｌｙａｔｔｈｅｍｉddleandtop

levels

Hypothesesll，１２，ａｎｄｌ３ａｒｅｓｕｍｍａｒｉｚｅｄｉｎＴａｂｌｅ２．

ＴＡＢＬＥ２

SummaryOfHypothesesll，１２，ａｎｄｌ３

TopLevelMiddleLevelLowerLevel

ＨＨＨＨＨＨＨＨＨｖ》ｖ）ｖ〉ｖ〉ｙｖ〉ｖ》ｖ〉ｖ》Ｄ△ｐ△ｐ△Ｄ△Ｄ＆Ｄこり△ｐＣｐＣｏＣＯＯｏＣＯＣＯｔｔｔ’しｔｔｔｔｔＵｎ⑪ｎｎｎ１ｎ１ｎｎｎ⑪ｎコｎ１ｎｅｅｅｅｅｅｅｅｅｓｓｓｓｓｓｓｓｓ ’１・１●１●１・１・１・１・１・１ｓｓｓＳｓｓｓｓｓ１２３４５６７８９０＊＊＊＊ａ

＊

＊＊＊＊＊

ａ＊showsthatthehypothesisappliesatthehierarchicalleveL

ＭＥＴＨＯＤＳ

SampleandProcedures

(7)

8７

Thestudyusedorganizationalsubunitsasitsunitofanalysis・ Survey respondentsmight，ｏｆcourse，definesubunitsindifferentways．Therefore，the definitionofsubunitwasgiveｎｉｎｔｈecoverletterofthequestionnaire： ^“Your unitconsistsofyou（asthesupervisor）andallindividualswhoreportdireｃｔｌｙｔｏ

,，

ｙｏｕ・

Sixquestionnairesweresentto223manufacturingｃｏｍｐａｎｉｅｓｔｈａｔｈａｖｅｍｏｒｅｔｈａｎｌＯＯｅｍployeesintheBuffalo，Jamestown，andNiagaraFalls，NewYork areasAdirectorofhumanresourceswhoreceivedthequestionnaireswasasked tochoosetwopeople（ｉｆpossible，ｏｎｅｉｎａ“high-performing'，ｕｎｉｔａｎｄａｎｏｔｈｅｒｉｎａ‘`low-performing'，unit，ｈｏｗｅｖｅｒｈｅｍｉｇｈｔｄｅｆｉｎｅｈｉｇｈａｎｄｌｏｗperformance）from eachofthreelevels（top，middle，ａｎｄlower)，atotalofsixpeople、Then，ｈｅｗａｓｉｎｓｔｒｕｃｔｅｄｔｏａｓｋｅａｃｈｏｆｔｈｅｍtofilloutashortquestionnaireanｄｒｅｔｕｒｎｉｔｔｏｔｈｅＵｎｉｖｅｒｓｉｔｙｉｎａｎｅｎclosedpre-addressedpostage-paidenvelope．Thenumber ofusablequestinnairesｒｅｃｅｉｖｅｄｗａｓ５６ａｔｔｈｅｔｏｐｌｅｖｅｌ，５０atthemiddlelevel，ａｎｄ

５６ａｔｔｈｅｌｏｗｅｒｌｅｖｅｌ．

Measures

Formalization、 Mostresearchershavedefinedformalizationintermsof

writtendocuments（Hall，1982）．Formalizationwasmeasuredherebylnkson，

Pugh，andHickson，ｓ（1970）formalizationoftherole-definitionscale、Ｔｈｉｓｈａｓｂｅｅｎｃｏｍｍｏｎｌｙｕｓｅｄｉｎｔｈｅｌｉterature（e９．，Child，1973；Sathe，1974；Ford，１９８１；

MillerandDroge，１９８６；Singh，１９８６)．Cronbach，ｓａｌｐｈａｆｏｒｔｈｅｓｃａｌｅｍｔｈｉｓｓｔｕｄｙｗａｓ、471．

centralizationCentralizationisthedegreetowhichpowerisdistributed withinasubunit（Hall，1982)．ThiswasmeasuredbyAikenandHage，ｓ（1968）

scaleofthehierarchyofauthority，ｂｅｃａｕｓｅｏｆｉｔｓｈｉｇｈｌｅｖｅｌｓｏｆｂｏｔｈｖａlidityand reliability（Dewar，Whetten，andBoje，1980)．Gronbach，ｓａｌｐｈａｆｏｒｔｈｉｓｓｃａｌｅ

ｗａｓ、853．

SpecializationSpecializationisthenumberofjobtitles，whichisanindicator ofdivisionoflabor（BlauandSchoenherr，1971；ＤｅｗａｒａｎｄＨａｇｅ，1978；ＶａｎｄｅＶｅｎａｎｄＦｅｒｒｙ，1980)．Weaskedsupervisorshowmanydifferentjobtitleswere representedamongthepeoplethevsupervised．Thenspecializationoftheunit wascalculatedbyaddingone（forthesupervisor）ｔｏｔｈｉｓｎｕｍｂｅｒ・

ＳｉｚｅＡｔｏｔａｌｎｕｍｂｅｒｏｆｅｍｐｌｏｙｅｅｓｗａｓusedtomeasuresizeinthisstudy・

Thisisthemostcommonlyusedmeasureintheliterature（Kimberly，1976)，ａｎｄis applicabletoallorganizations．Weaskedsupervisorshowmanypeoplethey superviseddirectly．Thensizeoftheunitwascalculatedｂｙａｄｄｉｎｇｏｎｅ（for thesupervisor）ｔｏｔｈｉｓｎｕｍｂｅｒ、

Technology．Thereareanumberofdefinitionsoftechnology（Gerwin，

１９８１；Fry，１９８２）．Becauseofremarkablyconsistentfindingsforthe routine-nonroutinetechnology（Fry，1982)，thisstudydefinedtechnologyinterms oftheroutinetononroutineconception（Perrow，1967,1970)．Routinenesswas measuredbyHageandAiken，ｓ（1969）scale、Itsquestionnaireitemspertainedｔｏ

(8)

8８

ｔｈｅｎumberofexceptionsusingPerrow'ｓ（1967,1970）measureoftechnologyand showedhighreliabilityandvalidity（Dewar，Whetten，andBOje，1980；Whitney，

Daft，andCooper,1983)．Cronbach,ｓａｌｐｈａｆｏｒｔｈｉｓｓｃａｌｅｗａｓ８３８・

EnvironmentpDuncan，ｓ（１９７２）perceivedenvironmentaluncertaintyscale representsthemostadvancedoperationalizationoftheuncertaintyconstructto date（Sathe，１９７４）andhasreceivedwidespreadattention

However，this

instrumentrequirestherespondenttoidentifyfactorsassociatedwithagiven decision-makingsituation，whichisdifficultandtime-consumingSathe（1974）

modifiedDuncan'ｓ（1972）instrumenttoproduceasimpleroneforquestionnaire

administration、Internalreliability（Kuder-Richardson8）forthisscalewas804

for531subjectsfromtwenty-twodepartmentsofaninsurancecompanyand、834 for60subjectsfromsevenbranchesofacommercialbank（Sathe，1974).Therefore，

thisstudyusedSathe'ｓ（1974）perceiveduncertaintyscale、Cronbach，salphafor

thisscalewas、728．

Performance．AperceivedunitperformanceindexｄｅｖｅｌｏｐｅｄｂｙＶａｎｄｅＶｅｎ andFerry（1980）wasusedtomeasuretheperformanceofasubunitinthisstudy．

Itmeasuredthedegreetowhichtheunithasachieveditsperformancetargetsand alsotherelativestanding（rating）ｏｆｔｈｅｕｎｉｔｉｎｃｏｍｐａｒｉｓｏｎｔｏｏｔｈｅｒｕｎits・The eightquestionnaireitemsusedforthismeasureseemedtobeapplicabletoall subunits・Thefirstquestionaskedsupervisorswhatpercentagesofperformance targetswereattainedoverallbytheirunitsduringthelastyear・Theyindicated theirresponseonascalethatvariedfrom“１”（notargetswereset）through

“２'，（０－２０％）ｔｏ“８，，（ｍｏｒｅｔｈａｎ１００％)．Theremainingsevenquestionsasked supervisorstoratehowtheirunitｄｉｄｏｎｅａｃｈｏｆｔｈｅｆａｃｔｏｒｓｄｕｒｉｎｇｔｈｅpastyear inrelationtoothercomparableorganizationalunitswithinoroutsideoftheirorg-

anizationCronbach'ｓａｌｐｈａｆｏｒｔｈｉｓｓｃａｌｅｗａｓ、819.

Analyses

TotestHypothesisl，aseriesofmultipleregressionanalyseswasconducted forstructureonthethreecontingencyfactors（size，technology，ａｎｄenvironment）

ateachhierarchicalleveL

Hypotheｓｅｓ２ｔｏｌ３ａｓｓｅｒｔｅｄｔｈａｔｔｈｅｔｈｒｅｅｃｏntingencyfactorswouldinteract

withstructuretoinfluenceperformance・Ｔｈｅｙｓａｙｔｈｅ“form'’oftherelationship betweenstructureandperformanceisconditionaluponthethreecontingency factors（Arnold，1982)．Inordertodeterminewhethertheformofrelationship betweentheindependentvariable（Ｘ，inthisstudy，structure）andthedependent variable（Ｙ，performance）isconditionaluponthemoderatorvariable（Ｚ，size，

technology，ｏｒenvironment），wetestedthesignificanceofthesemipartial

coefficientassociatedwiththeXZproductterm（ie.，ｓｒ２ｙ(x２．x,z)）inhierarchical

multipleregressionanalysis（Arnold，1982)．Ｔｈｅｒｅgressionequationanalyzedwas

-へ

Ｙ＝Ａ＋ＢｉＸ＋Ｂ２Ｚ＋Ｂ３ＸＺ・

Thesignificanceofthepartialcoefficientassociatedwiththeproducttermwas

testedby

(9)

8９

Ｂ３ｗｉｔｈｎ－４ｄｆ，

t＝

SＥＢ３

ｗｈｅｒｅＳＥＢ３＝standarderrorofestimateB3

n＝ｎｕｍｂｅｒｏｆｃａｓｅｓ

ｌｆｔｈｅｓｅｍｉｐａｒｔｉａｌｃoefficientB3issignificantfortheR2incrementassociated withXZinthehierarchicalmultipleregression，itisinterpretedassupportforthe hypothesis．

ＲＥＳＵＬＴＳ

EffectsofHierarchicalLevelsonVariables

Table3presentsthemeansandstandarddeviationsofthecontingency factors，structure，andperformanceateachhierarchicalleveL

ＴＡＢＬＥ３

MeansandStandardDeviationsforContingencyFactors，Structure，

andPerformanceatEachHierarchicalLevel ＴｏｐＬｅｖｅｌ

－￣

Ｍｅａｎｓｓ.．．

MiddleLevel

￣￣

Ｍｅａｎｓｓ.．．

ＬｏｗｅｒＬｅｖｅｌＭｅａｎｓｓ.．．

ContingencyFactors

Size

Technology

Environment Structure

Formalization Centralization

Specialization

Performance

Targetsachieveda

Relativeperformance

12.964 1.893 2.189

17.635 .583 .617

10.065 2.208 2.166

11.822 .630 .525

16.309 2.436 2.114

16.099 .623 .547

5.296 1.486 6.667

1.574 .494 2.819

5.000 1.718 5.130

1.400 .610 2.315

5.071 2.143 5.436

1.412 .676 3.196

6.184 3.756

1.131 .586

5.929 3.653

1.135 .567

6.040 3.500

1.124 .604

ａＡ“Notargetswereset'，responsewascodedas“missingvalue''・

Ｔｈｅｎｕｍｂｅｒｏｆｔｈｉｓｒｅｓｐｏｎｓｅｗａｓ６ａｔｔhetoplevel，５ａｔｔｈｅｍｉｄｄｌｅｌｅｖｅｌ，

ａｎｄ４ａｔｔｈｅｌｏｗｅｒｌｅｖｅＬ

One-wayanalysesofvariancewereperformedtoexamineeffectsof hierarchicallevelsonthecontingencyfactors，structure，ａｎｄperformance，Results oftheseanalysesindicatedthatthehierarchicallevelshadasingnificanteffecton

technology（Ｆ＝11.11,ｐ＜001)，centralization（Ｆ＝17.39,ｐ＜，001),ａｎｄspecial‐

(10)

9０

ization（Ｆ＝４１３，Ｐ＜０５）ｂｕｔｔｈａｔｔｈｅｙｄｉｄｎｏｔａｆｆｅｃｔａｎｙｏｔｈｅrvariablesat thesignificancelevｅｌｏｆＯ５、

T-testswereappliedtomakemorespecificcomparisonsonthethreevariables

betweenthethreehierarchicallevels・ Resultsindicatedthattherewere

significantdifferencesfortechnologybothbetweentopandmiddlelevels（ｐ＜０１）

andbetweentopandlowerlevels（ｐ＜001)．Ｔｈａｔｉｓ，asonemovedfromtop throughmiddletolowerlevels，subunitshadmoreroutinetechnology，although therewasnosignificantdifferenceoftechnologybetweenmiddleandlowerlevels Thisresultisｔｈｅｓａｍｅａｓｗｈａｔｏｔｈｅｒｓ（Hall，1962；HageandAiken，1969）have

found

Resultsoft-testsalsoindicatedthatcentralizationincreasedsignificantlyfrom

toptomiddlelevels（ｐ＜、０５）ａｎｄｆｒｏｍｍｉｄｄｌｅｔｏｌｏｗｅｒｌｅｖｅｌｓ（ｐ＜、01)．Ａｓ

ｏｎｅｍｏｖｅｄｄｏｗｎｔｈｅｈｉｅｒａｒｃｈｙ，subunitsbecamemorecentralizedTherewere

significantdifferencesofspecializationbothbetweentopandmiddlelevels (ｐ＜０１）andbetweentopandlowerlevels（ｐ＜､05）buttherewasnosignificant

differencebetweenmiddleandlowerlevelｓＴｈａｔｉｓ，ａｓｏｎｅｍｏｖｅｄｆｒｏｍｔｏｐｔｏｍｉｄｄｌｅｏｒｌｏｗerlevels，subunitsbecamelessspecialized・

Thus，wecanrecognizeeffectsofhierarchicallevelsontechnology，

centralization，andspecialization，ｂｕｔｔｈｉｓｓｅｅｍｓｌｅｓｓａｐparentfortheothervaria-

bles．

andPerformanceCorrelations Pearsoncorrelationcoefficients

perfｏｒｍａｎｃｅａｔｔｏｐ，middle，

WithinContingencyFactors，Structure，

Tableｓ４，５，ａｎd6presentthe contingencyfactors，structure，and hierarchicallevels，respectively-

ａｍｏｎｇｔｈｅ

ａｎｄｌｏｗｅｒ

ＴＡＲＴＪＥ４

P…．､雛麓鰡翻諾鵠晋;侭競鄙F…

６７

５４

３２

Variables １

ContingencyFactors

lSize

2、Technology２５５簾

3．Environment、191、l77 Structure

4・Formalization-.112-.225-.204

5．Centralization、110、320*．、229＊-.188

6．Specialization551…．255＊、207、201-.027

Performance

7．TargetsachievedIG、116、0００－０７９．０１２－０６５０５２

８．Relativeperformance-O64-.105-.325鉢－．350..-.017、149、451…

(11)

9１

aCorrelationsexcludedcaseswithmissingvaluesonananalysis-by‐

analysisbasis

bBecauseofmissingｄａｔａ，Ｎｓｒａｎｇｅｄｆｒｏｍ４５ｔｏ５６ｆｏｒｔｈｅａｎａｌｙｓｅｓｌｎ

ｔｈｉｓｔａｂｌｅ

ｃＡ“Notargetswereset'，responsewascodedas“missingvalue"．

＊Ｐ＜,０５

＊＊ｐ＜０１

＊＊＊ｐ＜,001

ＴＡＲｍＦ】５

P…｡造脇謡iWh鶴Ｈ１:H1:h概鯛:ｉＷ,lFa…

Variables １２３４５６７

ContingencyFactors

lSize

2・Technology、211

3．Environment－０４７ Structure

4・Formalization．010 5．Centralization．168 6．Specializationl70

Performance

7・Targetsachievedc－．１００

８．Relativeperformance-.071

一.045

.051-.226 .495…．217 .019-.064

-.053

-062-.018

-.076 -.226

-.393...219 -.362...319。

-.195 -.363..

-.112 .129.442*。

aCorrelationsexcludedcaseswithmissingvaluesonananalysis-by-

analysisbasis、

bBecauseofmissingdata，Ｎｓｒａｎｇｅｄｆｒｏｍ３９ｔｏ５０ｆｏｒｔｈｅａｎａｌｙｓｅｓ

ｉｎｔｈｉｓｔａｂｌｅ

ｃＡ“Notargetswereset'，responsewascodedas“missingvalue''．

＊Ｐ＜０５

＊＊ｐ＜０１

＊＊＊ｐ＜００１

(12)

9２

ＴＡＢＬＥ６

P.…１s:lUHflli:j珊嚇麓､鰹F:鶴iWJF…

Variables １２３４５６７

ContingencyFactors

LSize

2・Technology、093

3．EnvironmentO31 Structure

4・Formalization、127 5．Centralization244.

6．Specialization、361.．

Performance

7．Targetsachievedc

^l65

8・Relativeperformance-.040

.168

-.201 .532…

、090

》４５調肥羽０←●００

-.099

-.012１５０

６６０８２１ ●● 亡巾戸坑叩）〈．ｈ〉凹釧一一 _-.045

-.052

８９３９００ ●● ｜’ -.025

-.097.701…

aCorrelationsexcludedcaseswithmissingvaluesonananalysis-

by-analysisbasis

bBecauseofmissingdata，Ｎｓｒａｎｇｅｄｆｒｏｍ４９ｔｏ５６ｆｏｒｔｈｅａｎａｌｙｓｅｓ

ｉｎｔｈｉｓｔａｂｌｅ

ｃＡ“Notargetswereset，，responsewascodedas“missingvalue，'．

＊ｐ＜０５

＊＊ｐ＜０１

＊＊＊ｐ＜００１

Thesetablesshowthatcorrelationｓａｍｏｎｇｓｉｚｅ，technology，andenvironment weregenerallylowandnotsignificantattheO51evel，withtheexceptionofa

significantpositiverelationshipbetweensizeandtechnologyatthetoplevel (ｒ＝、255,ｐ＜０５)．ThisfindingSuggeststhatthethreecontingencyfactors

weresubstantiallyindependentofoneanother・

Fromthesetables，ｉｔｉｓｓｅｅｎｔｈａｔｔｈｅｒｅｗｅｒｅｎｏｓｉｇｎｉｆｉcantintercorrelations amongformaliZation，cenCralization，andspecializationatanyofthethree hierarchicallevels．Thisiscontrarytothegeneralpatternofpreviousstudies

(Mintzberg，1979；Gerwin，1981；Fry，1982）．２

Tableｓ４，５，ａｎｄ６ａｌｓｏｓｈｏｗｔｈａｔｔｗｏｐｅｒｆｏｒｍａｎｃｅvariables，targets achievedandrelativeperformance，weresingnificantlyintercorrelatedataUthree hierarchicallevels、Assubunitsachievedhigherpercentagesoftheirperformance targets，theyratedhigheronthesevenfactorsofperformanＣｅ．

ContingencyFactorsandStructureCorrelations

SizeandStructureCorrelationsTables4，５，ａｎｄ６ｓｈｏｗｔｈａｔｔｈｅｒｅｗａｓｎｏ

気'

(13)

9３

relationshipbetweensizeandformalizationatanyhierarchicalleveLThe correlationbetweensizeandcentralizationwassignificantonlｙａｔｔｈｅｌｏｗｅｒｌｅｖｅｌ (ｒ＝２４４，ｐ＜、05)．Assizeofsubunitsbecamelarger，subunitsweremore centralizedatthelowerleveLThisiscontrarytowhatothers（ＰｕｇｈｅｔａＬ，1969；

Hickson,PughandPheysey,1969;ChildandMansfield,1972;Child,1975;Evers

etaL，1976）havefoundSizewasexpectedandfoundtobepositivelyrelated tospecializationatbothtop（ｒ＝、551,ｐ＜、001）ａｎｄｌｏｗｅｒｌｅｖｅｌｓ（ｒ＝８６１，

ｐ＜０１)．Assizeofsubunitsbecamelarger，subunitsweremorespecializedat

bｏｔｈｔｏｐａｎｄｌｏｗｅｒｌｅｖｅｌｓＴｈｉｓｉｓｔｈｅｓameresultasothers（Ｐｕｇｈｅｔａ1.,1969；

Hickson，Pugh，andPheysey，1969；Blau，1970；BlauandShoenherr，1971；Childand Mansfield，1972；Child，1975；Mileti，Gillespie，andHaas，1977）havefound

TechnologyandStructureCorrelations、FromTables4，５，ａｎｄ６，itisseen thattechnologywasnotsignificantlyrelatedtoformalizationatanyofthethree hierarchicallevels．However，technologywaspositivelyrelatedtocentralizationat

eachlevel（top：ｒ＝、320,ｐ＜01；middle：ｒ＝、495,ｐ＜001；ｌｏｗｅｒＩｒ＝、532, ｐ＜’００１）．Astechnologybecamemoreroutine，subunitsemployedmore

centralizedstructure．ＴｈｉｓｉｓｔｈｅｓａｍｅｒｅｓｕｌｔａｓｒｅｐｏｒｔｅｄｂｙＨａｇｅａｎｄＡｉｋｅｎ (1969)，Hrebiniak（1974)，ａｎｄＤｅｗａｒａｎｄＷｅｒｂｅｌ（1979）Thecorrelationbetween technologyandspecializationwassignificanｔｏｎｌｙａｔｔｈｅｔｏｐｌｅｖｅｌ（ｒ＝．２５５，

ｐ＜０５)．Assubunitshadmoreroutinetechnology，theyweremorespecialized

atthetopleveLMostothers（ComstockandScott，1977；ＦｒｙａｎｄＳｌｏｃｕｍ，1984）

havefoundthetechnology-specializationcorrelationstrongeratｌｏｗｅｒthanat higherlevels

EnvironmentandStructureCorrelationsTables4，５，and6showthat environment（perceiveduncertainty）didnotcorrelatewitheitherformalizationor specializationatanyhierarchicalleveLEnvironmentwassignificantlyrelatedto

centralizationatbothtop（ｒ＝、229,ｐ＜､05）ａｎｄｌｏｗｅｒｌｅｖｅｌｓ（ｒ＝、395,ｐ＜０１)．

Assubunitsperceivedmoreenvironmentaluncertainty，theybecamemore centralizedatbotｈｔｏｐａｎｄｌｏｗｅｒｌｅｖｅｌｓＴｈｉｓｉｓａｃｏｎｔｒadictoryresulttowhat others（NegandhiandReimann，1973；Pennings，1975）ｈａｖｅfound．

EffectsofContingencyFactorsonStructurebyLevel（HypothesisD

Aseriesofmultipleregressionanalyseswasconductedtoassesstherelative importanceofthecontingencyfactorsonstructure・ Table７presentsthe

full-equationbetaweights（standardizedregressioncoefficients)，incrementsinR2，

ａｎｄＦｖａｌｕｅｓｆｏｒｉｎｃｒｅｍｅｎｔｓｉｎＲ２ｆｏｒformalization，centralization，and specializationatthetoplevel．

(14)

9４

ＴＡＲＴ』F】７

ResultsofMultipleRegressionAnalysesofStructureatTopLevela

Centralization

Formalization Specialization ^■

lncre-

ments Beta

inR2FWeights

Incre-

ments Beta

inR2FWeights

IｎｃｒｅｍｅｎｔｓｉｎＲ２Ｆ Beta

VariablesWeights

Size

Technology

Environment R2

AdjustedR2

OverallF

１２５５３９５３５５０４０１１００９ ●ＣＯ■●● 002

.016 .023

、123 ,846 1.207

５０３６２４２１６８３００２１００６ ●●●白●●１

001 .041 .025

､032 2.283

1.409

、525*＊＊

､013 .117 .317 .273 7.270*．＊

､220 .000 .013

15.133…

０１０

．９０８

aListwiSedeletionofmissingdatareducedtheNsto53for

formalization，５５forcentralization，and51forspecialization．

＊＊＊ｐ＜､001

Theseresultsshowthatsize，technology，andenvironmentcontributedlittleto explainingvarianceineitherformalizationorcentralization．Neithertechnology norenvironmentaddedasignificantincrementtothepredictionofspecialization；

sizewastheonlyvariabletoaccountforsignificantvarianceinpredictingspecializ- ation､Sizeexplained3q4（３１．７－０．０－１．３）percentofthevarianceinspecialization・

Thus，theresultsdidnotsupportHypothesisL

Table8presentsbetaweights，incrementsinR2，ａｎｄＦｖａｌｕｅｆｏｒｉｎｃｒｅｍｅｎｔｓｉｎＲ２ｆｏｒｆormalization，centralization，ａｎｄspecializationatthemiddlelevel

(15)

9５

ＴＡＢＬＥ８

ResultsofMultipleRegressionAnalysesofStructureatMiddleLevela

Formalization Centralization Specialization

Iｎｃｒｅｍｅｎｔｓ

ｉｎＲ２

Iｎｃｒｅ－

ｍｅｎｔｓｉｎＲ２Ｆ

Incre‐

ｍｅｎｔｓｉｎＲ２Ｆ Beta

Variables Weights

Beta

FWeights

Beta

Weights

､000 .001 .039

002 .058 1.729

．080 .466..

.254 .294 .242 5.682．＊

６７筋Ⅱ別０

●■● Size

Technology

Environment R2

AdjustedR2

OverallF

-.008 .037 -.199 .042 -,027 .607

、352 12.019..

3.745

173 -.023 -.047

.031 -.039 .443

､028 .000 .002

1．１７７

．０２１

．０９３

aListwisedeletionofmissingdatareducedtheNsto46for

formalizatiｏｎ，４５forcentralization，and45forspecialization．

＊＊ｐ＜０１

Theresultsindicatethatneithersizenorenvironmentacountedforsignificant varianceinformalizatiｏｎ，centralization，andspecialization Technologywasa significantpredictorofcentralization，butnotofformalizationorspecialization・

Technologyexplained223percentofthevarianceincentralization．Theseresults couldbeexpectedfromtheverylowcorrelationsbetweenthecontingencyfactors andstructurepresentedinTable5，withtheexcePtionoftherelationshipbetween technologyandcentralization Thus，theseresultsalsoofferednosupportfor Hypothesisl、

Table9presentsbetaweights，incrementsinR2，ａｎｄＦｖａｌｕｅｆｏｒｉｎｃｒｅｍｅｎｔｓｉｎＲ２ｆｏｒｆormaliZation，centralization，ａｎｄspecializationatthelowerlevel．

(16)

9６

ＴＡＢＩＥ９

ResultsofMultipleRegressionAnalysesofStructureatLowerLevela

Formaliz2tion Centralization Specialization

IｎｃｒｅｍｅｎｔｓｉｎＲ２

lncre-

ments Beta

inR2FWeights

Iｎｃｒｅ－

ｍｅｎｔｓｉｎＲ２Ｆ Beta

VariablesWeights

Beta

FWeights

Size

Technology

Environment R2

AdjustedR2

OverallF

、143 -.223

.142 .076 .022 1.397

020 .048 .020

1.123 2.666 1.089

、191 .461…

、308..

.408 .373 11.716..。

036 .205 .092

3.123、357..

17.677…．066 ７．９４４．．－．０５７

．１３７

．０８６ 2.697

６４３２００１００ ●●● 7.456..

.249 .190

aListwisedeletionofmissingdataTeducedtheNsto55for

formalization，５５forcentralization，and55forspecialization．

＊＊ｐ＜０１

＊＊＊ｐ＜００１

Theseresultsindicatethatsize，technology，andenvironmentdidnotcontribute toexplainingvarianceinformalization Butbothtechnologyandenvironment accountedforsignificantvarianceincentralization，althoughtheformerwasa muchbetterpredictorthanthelatter．Bothpredictorstakentogetherexplained 372percentofthevarianceincentralization、 Sizewastheonlyvariableto contributeasignificantincrementinpredictingspecialization・SizeexplainedlaO percentofthevarianceinspecializationThus，theseresultssupportedHypothesiｓｌｉｎｔｅｒｍｓofcentralization，butnotformalizationandspecialization．

ContingencyHypotheses（Hypotheses2tol3）

ＴｏｔｅｓｔＨｙｐｏｔｈｅｓｅｓ２ｔｏｌ３，moderatedregressionanalyseswereusedWe conducted54separateanalyses．Eachanalysisincludedastructurevariable，a contingencyfactor，andtheirinteractiontermasindependentvariables，anda performancevariableasadependentvariable・

EffectsofSizeontheRelationshipBetweenStructureandPerformance．The resultsshowthatalthoughformalizationexplainedasignificantportionof varianceinrelativeperformanceatbothtopandmiddlelevels，andcentralization addedasignificantincrementtothepredictionofrelativeperformanceatthe middlelevel，theinteractiontermsofsizeandstructuredidnotcontribute significantincrementsinpredictingperformanceatanyofthethreehierarchical levelsThatis，sizedidnotsignificantlymoderatetherelationshipbetween structureandperformanceThus，theseresultsofferednosupportforHypothesis

(17)

9７

'1concerninglowerandmiddlelevels,Hypothesisl2concerningmiddleandtop

levels，orHypothesisl3concerningmiddleandtoplevels．However，theydid supportHypothesisllconcerningthetoplevel，Hypothesisl2concerningthelower

level，andHypothesisl3alsoconcerningthelowerleveL

EffectsofTechnologyontheRelationshipBetweenStructureandPerformance Theresultsindicatethattheinteractiontermsoftechnologyandstructureadded

nosignificantincrementtothepredictionsofeithertargetsachievedorrelative performance，ｅｘｃｅｐｔｉｎｔｗｏｃａｓｅｓ・ＯｎｅｃａｓｅｗａｓｆｏｕｎｄａｔｔｈｅｍｉｄｄｌｅｌｅｖｅLThe

interactionoftechnologyandformalizationwassignificantlyrelatedtotargets achieved・TheincrementinR2associatedwiththisinteractionwas、164,ｌｅａｄｉｎｇｔｏａｆｉｎａｌｖａｌｕｅｆｏｒＲ２ｏｆ２０２

Inordertohelpinterpretthissignificantinteractioneffect，usingthe

proceduresuggestedbyCohenandCohen（1983)，threerepresentativeregression linesareplottedinFigurel；ｏｎｅｆｏｒａ“low”（nonroutine）technologyscore

（１standarddeviationbelowthemean)，ｏｎｅｆｏｒａｎ“average'，technologyscore (atthemean)，ａｎｄｏｎｅｆｏｒａ“high”（routine）technologyscore（１standard deviationabovethemean)．Thethreeregressionequationswere：

８９７３０５５４３

’’一一一一

〈Ⅱ〈Ⅲ〈Ⅵ

ｆｏｒ“low” （Technology＝1.578）

ｆｏｒ“average'’（Technology＝2.208）

ｆｏｒ“highj，（Technology＝2.838）

Ｙ＝TargetsAchieved

X＝Formalization Ｘ＋7.910

Ｘ＋5.505 Ｘ＋3.100 where

ＦＩＧＵＲＥ１

SignificantlnteractionEffectBetweenTechnologyand FormalizationonTargetsAchievedatMiddleLevel

８７６５４３２１

Targets

Achieved

1２３４５６７ Formalization

(18)

9８

ThisfigureprovidessupportforHypothesisllwhenperformancewas measuredbytargetsachieved・Increasesinformalizationpositivelyinfluenced performancewhentechnologywasroutineandnegativelyinfluencedperformance

whentechnologywasnonroutineatthemiddlehierarchicalleveL

AnothersignificantinteractioneffｅｃｔｗａｓｆｏｕｎｄａｔｔｈｅｌｏｗｅｒｌｅｖｅＬＴｈｅｉnteractionoftechnologyandcentralizationaddedasignificantincrementtothe

predictionofrelativeperformance TheincrementinR2associatedwiththis

interactionwasO74,ｌｅａｄｉｎｇｔｏａｆｉｎａｌｖａｌｕｅｆｏｒＲ２ｏｆ、1０７．Figure2illustrates thissignificantinteractioneffect，usingCohenandCohen'ｓ（1983）procedure．

ＦＩＧＵＲＥ２

SignificantlnteractionEffectBetweenTechnologyand

CentralizationonRelativePerformanceatLowerLevel

５４３２

Relative Performamce

１

1２３４ Centralization

Thethreeregressionequationswere：

２６９３１９２６９４３２＋＋＋ＸＸＸ

４８７９８１２０１一一一一一一一一

〈Ⅲ〈Ⅲ〈Ⅲ

ｆｏｒ“low'，（Technology＝1.813）

ｆｏｒ‘`average，，（Technology＝2.436）

ｆｏｒ“high'，（Technology＝3.059）

Ｙ＝RelativePerformance X＝Centralization

where

ThisfigureprovidessupportforHypothesisl2whenperformancewas measuredbyrelativeperformacne、Whentechnologywasroutine，increasesin centralizationpositivelyinfluencedperformance．Whentechnologywasnonroutine，

increasesincentralizationnegativelyinfluencedperformance

Thus，theresultsdidnotsupportHypothesisllforeitherperformance

variaｂｌｅａｔｔｈｅｌｏｗｅｒｌｅｖｅｌｏｒｒｅｌａｔｉｖｅｐｅrformanceatthemiddlelevel；nordidthey

supportHypothesisl2fortargetsachievedatthelowerleveLHypothesisl3was

noｔｓｕｐｐｏｒｔｅｄａｔｔｈｅｌｏｗｅｒｌｅｖｅＬＴｈｅｄａtadid，however，supportHypothesisll

(19)

9９

concerningthetoplevelonbothperformancevariablesand，atthemiddlelevel，on targetsachieved．TheyalsosupportedHypothesisl2concerningthemiddleand

thetoplevelsonbothperformancemeasuresandthelowerlevelonrelative

performanceiandHypothesisl3heldformiddleandtoplevels

EffectsofEnvironmentontheRelationshipBetweenStructureandPerformance・

Theresultsindicatethａｔｎｏｎｅｏｆｔｈｅｉｎｔｅｒａｃｔｉｏｎｔｅｒｍｓｏｆｅnvironmentand

structureexplainedsignificantvarianceineitherofthetwoperformancevariables atanyofthethreehierarchicallevels、Environmentwasfoundtoaddsignificant

incrementstothepredictionsoftargetsachievedatthemiddlelevelandrelativｅ

ｐｅｒｆｏｒｍａｎｃｅａｔｂｏｔｈｔｏｐａｎｄｌｏｗｅｒｌevels．Formalizationexplainedasignificant portionofvariancemrelativeperformanceatbothtopandmiddlelevelsandso didcentralizationatthemiddleleveLTheseresultscould，ｏｆcourse，beexpected fromthesignificantcorrelationspresentedinTables4，５，ａｎｄ６．

Thus，theresultsfailedtosupportHypothesisllconcerningthetoplevel，

Hypothesisl2concerningmiddleandtoplevels，andHypothesisl3concerning middleandtoplevels．However，theydidsupportHypothesisllconcerninglower andmiddlelevels，Hypothesisl2concerningthelowerlevel，andHypothesisl3

concerningtheloｗｅｒｌｅｖｅＬ

ＡｓｕｍｍａｒｙｏｆｔｈｅｒｅｓｕｌｔｓｆｒｏｍｔｅｓｔｓｏｆＨｙｐｏｔｈｅｓｅｓ２ｔｏｌ３ｉｓｐｒｏｖidedin

TablelO．

ＴＡＢＴＥ１０

SummaryofResultsFromTestsofHypoｔｈｅｓｅｓ２ｔｏｌ３

TopLevel TargetsRelative AchievedPerformance

MiddleLevel TargetsRelative AchievedPerformance

LowerLevel TargetsRelative AchievedPerformance ＨＨＨＨＨＨＨＨＨ２３４５６７８９ｎａ

＊＊

＊＊＊＊

＊＊＊＊＊＊＊＊＊＊＊＊＊

＊

＊＊＊＊＊＊

ａ

＊showsthattheresultsprovidedsupportforthehypothesis．

(20)

1００

ＤＩＳＣＵＳＳＩＯＮＡＮＤＣＯＮＣＬＵＳＩＯＮＳ

SummaryandDiscussionofResults

Hypothesislassertedthatcontingencyfactorwouldaffectthevariouslevels oforganizationalstructuredifferentially，ｉｅ.，thatenvironmentwouldbethe

dominantfactoratthetoplevel，sizeatthemiddlelevel，andtechnologyatthe lowerleveLThｅｒｅｓｕｌｔｓｏｆthestudyprovidednosupportforHypothesislexce-

ptinonecase（technology／centralizationatthelowerlevel)．Asmentionedearlier，

therehasbeenmixedsupportforMintzberg'ｓ（1979）ideasabouttheeffectsof hierarchicallevelontherelativeimportanｃｅｏｆｓｉｚｅ，technology，andenvironment forstructure；and，likethestudiesbyAlexander（1979）ａｎｄｂｙＰａｖｅｔｔａｎｄＬａｕ (1983)，thepresentstudyfailedtosupportMintzberg，ｓａｒｇｕｍｅｎｔｓｔｈａｔｗｅｒｅｔｈｅｂａｓｉｓｆｏｒＨypothesisLHierarchicallevelseeｍｓｔｏｈａｖｅｎｏｓｙｓｔｅｍａｔｉｃｅｆｆｅｃｔｏｎ

ｔｈerelativeimportanceofthethreecontingencyfactorsforstructure（atleast

asmeasuredhere）．Thus，Mintzberg，ｓ（1979）propositionmayneedtobe

reconsidered

Hypotheses2tol3assertedthatthethreecontingencyfactorswouldinteract

withstructuretoinfluenceperformance．ＲｅｓｕｌｔｓｓｈｏｗｎｉｎＴａｂｌｅｌＯｅｘｈｉｂｉｔａ

ｄiscerniblepatternofresults．Foreverycasewhereitwaspredictedthat contingencyfactorswouldnotinteractwithstructure，nosignificantinteractions

wereobserved（i､e､，２４ｏｕｔｏｆ２４)．Inall8caseswheresimilarpredictionswere madeforenvironment，nosignificantinteractionswereobserved；andinalllO

caseswheretechnologywasexpectednottointeractwithstructure，nosignificant

interactionswereobserved

Meanwhile，ｉｎｏｎｌｙ２ｏｆ３０ｃａｓｅｓｉｎｗｈｉｃｈｃｏｎｔｉｎｇencyfactorswerepredicted tointeractwithstructure，weresignificantinteractionsobservedTwosignificant interactionsoutof30predictionsyieldsa“hitrate，’ｏｆ６．７％，whichisslightly betterthanchance（5％)．However，oneofthesignificantinteractionwasbetween

technologyandformalizationandtheformalizationmeasureusedherewasnota veryreliableone、Therefore，thetwosignificantinteractions，althoughpredicted，

shouldbeviewedskeptically．Hence，theresultsleavesomedoubtthat“fit，，

betweencontingencyfactorsandstructureisanimportantpredictorofperformance

（e９．，NadlerandTushman，1980；ＶａｎｄｅＶｅｎａｎｄＤｒａｚｉｎ，1985)．Ｓｕｃｈａｆｉｔｍａｙｓｉｍｐｌｙｎｏｔｂｅｅｎｏｕｇｈｔｏensurehighperformance，ｏｒｖｉｃｅｖｅｒｓａ．

lmplicationsforTheory

ThisstudyhastwoprincipaltheoreticalimplicationsFirst，althoughfew researchershaveconsideredallthreecontingencyfactorssimultaneously，ｔｈｉｓｓｔｕｄｙｄｉｄｓｏ、 Someexpectedandsomeunexpectedeffectsofthecontingency

factorsuponstructurewerefound，ｂｕｔａｌｌｉｎall，theresultssuggestthatthethree

contingencyfactorsaresubstantiallyindependentofoneanotherintheireffects atallhierarchicallevelSArguably，therefore，anytheoreticalframeworkforthe

studyofsubunitstructureinorganizationsmustbebroadenoughtoinclude

(21)

1０１

systematicallyallthreecontingencyfactorsasdeterminantsofstructure・

Second，althoughsomeresearchers（Child，1977；FordandSlocum，1977；Mintz- berg,1979；ＶａｎｄｅＶｅｎａｎｄＤｒａｚｉｎ,1985）havesuggestedthatthethreecontingency factorssometimeshaveconflictingeffectsuponstructure，therehavesofarbeen nosystematictheoreticalargumentsabouttheinteractioneffectsofthethree contingencyfactors．Thisstudyadducedonebasisforexploringthembyusing Mintzber9,ｓ（1979）argumentthateachcontingencyfactorwouldaffectthelevels ofstructuredifferentially・Unfortunately，ｔｈｅｒｅsultsdidnotsupportMintzberg'ｓ

(1979）proposition，althoughtheydidprovidepartialsupportforthehypotheses

concerningthecontingencyfactors，structure，andperformance・Itisnecessary，

therefore，todevelopamoresatisfactorytheoryofthemteractioneffectsofthe contingencyfactorsuponstructureandperformance．

Practicallmplications

Theresultsofthisstudyhaveseveralimplicationsformanagers．First，

althoughmanyresearchers（e９．，NadlerandTushman，１９８０；ＶａｎｄｅＶｅｎａｎｄＤｒａｚｉｎ，1985）havesuggestedthat“fit'，betweencontingencyfactorsandstructure isanimportantpredictorofperformance，ｔｈｅｒｅｓｕｌｔｓｏｆｔｈｉｓｓｔｕｄｙｄｏｎｏｔｐｒｏｖide supportforthisargument・Severalunexpecteddirecteffectsofcontingencyfacto- rsandstructureuponperformancewerefound，however・Perceivedenvironmental uncertaintyhadanegativeeffecton“relativeperformance”atallthreehierarchical levels，ａｎｄｏｎ“targetsachieved”atthemiddleleveLFormalizationhadapositive effecton“relativeperformance,，ａｔｂｏｔｈｔｏｐａｎｄｍｉddlelevels；andcentralization wasnegativelyrelatedto“relativeperformance'，atthemiddleleveLThus，it wouldappearthat，regardlessofthehierarchicallevel，theperformanceoforganiz- ationsmaybeimprovedbyreducinguncertaintyfortheirworkingmembers・Sev- eralresearchers（Thompson，1967；OsbornandHunt，1974；DowneyandSlocum，

1982）havemadethesamesuggestion,i・ａ，thathighlevelsofperceiveduncertainty

aredetrimentaltoperformaｎｃｅ

Ｔｈｉｓｓｔｕｄｙａｌｓｏｆｏｕｎｄｔｈａｔａｓｏｎｅmoveddowntheorganizationalhierarchy，

subunitshadmoreroutinetechnology，ｗｅｒｅｍｏｒｅｃｅntralized，andwereless specializedHence，thestudyprovidessupportforargumentsthatsubunitswithin largeandcomplexorganizationsarelikelytovarybothinstructure（Hall，1962；

DewarandWerbel，1979；ＦｒｙａｎｄＳｌｏｃｕｍ，1984）andintechnology（Grimesand Klein，1973；ＶａｎｄｅＶｅｎａｎｄＤｅｌｂｅｃｑ,1974；ComstockandScott，1977)．Forboth practicalandmethodologicalpurposes，itisimportant，therefore，torecognizethe likelihoodofstructuralandtechnologicalvariationsamongthesubunitsatdifferent hierarchicallevelsoforganizations．Simplyput，organizationslikelycomprise subunitswithdifferentstructuresandtechnologies・

Finally，ａｓｗｅｈａｖｅｎｏｔｅｄ，ｔｈｅｒｅsultsofthisstudysuggestthatthethree contingencyfactorsaresubstantiallyindependentofoneanotheratallthree

hierarchicallevels、 Therefore，ａｇａｉｎｆｏｒｂｏｔｈpracticalandmethodological purposes，ｏｎｅｍａｙｆｉｎｄｉｔｕｓｅｆｕｌｔｏｐｌａｎｆｏｒｓｅｐarateeffectsonstructurebyeach

(22)

102

ofthecontingencyfactorsateachhierarchicalleveL LimitationsoftheStudy

Therearesomeimportantlimitationsthatshouldberecognizedbefore generalizingfromtheresultsofthisstudy．First，thestudyassumedthatthe responseｓｏｆａｎｙｓｉｎｇｌｅｐｅｒｓｏｎｉｎａｕｎｉｔｗｅｒｅｒepresentativeoftheunit．This couldbequestioned（cfSeidler，１９７４；DewarandWerbel，1979)．Alternative samplingstrategiesemployinganumberofpersonsfromeachunitarecommonly impracticable，however，becauseitisextremelydifficulttofindlargesamplesof organizationswillingtoparticipateinsuchresearch、Forthisresearch，priority wasgiventoaheterogeneouｓｓａｍｐｌｅｏｆｏｒganizationsandsubunitsratherthan ofindividualrespondents，sinceitwasorganizationsthatwerｅａｔｉｓｓｕｅｉｎｔｈｅｓｔｕｄ－

ｙ，Manyorganizationalresearchers（e９．，LawrenceandLorsh，1967；ＰｕｇｈｅｔａＬ，

1969；ChildandMansfield，1972）haveusedthesameprocedure・

Second，anoverallresponserateofquestionnaireswasl2.6％，whichisrather low．Thosewhorespondedtothequestionnairewereessentiallyself-selected Therefore,thestudymaysufferfromaserious(althoughunknown）biasassociated withnonresponse・

AnothermatterworthmentionisthatourmeasurementsofformalizatioｎａｎｄｏｆｅｎｖｉｒｏｎｍｅｎｔｗｅｒｅｌｅｓｓｔｈａｎｉdeaLThetwomeasuｒｅｓｗｅｒｅｎｏｔａｓｒｅｌｉａｂｌｅａｓ

ｃｏｕｌｄｂｅｗished（althoughCronbach'salphaforenvironmentwas728whichis aboveNunnally，ｓ［1978］“satisfactory”level)．Therefore，theresultsregarding thesemeasuresshouldbemterpretedcautiously・

AlsowecollecteddataonlyfrommanUfacturingfirms・Hence，thegeneraliza- bilityoftheresｕｌｔｓｏｆｔｈｅｓｔｕｄｙｔｏｓｅｒｖｉｃｅｏｒｏｔｈｅｒkindsoforganizationsis uncertai､．And，sincedatawerecollectedatonepoiｎｔｉｎｔｉｍｅ，ｔｈａｔｉｓ，cross- sectionally，causalinferencesofcoursecannotｂｅｄｒａｗｎ．

lmplicationsforFutureResearch

Theresultsofthisstudysuggestseveralareasforfutureresearch．First，

whiletherewereseveralsignificanteffectsofthethreecontingencyfactorson structureateachhierarchicalleveLingeneral,thecontingencyfactorscouldnot explainalargeportionofthevarianceinstructure、Therearestudiestoindicate thatthepersonalityofanorganization'ｓｃｈiefexecutiveofficermaybean importantdeterminantofstructureatleastinsmallandyoungorganizationｓ（e､g､，

MillerandDroge，1986)．Althoughourdataweregatheredfromrelativelylarge organizationstｈａｔｈａｖｅｍｏｒｅｔｈａｎｌＯＯｅｍｐｌｏｙｅｅｓａｎｄsomayhavebeenexpected toprecludeaninfluenceofCEOpersonalityuponstructure，weemployedno directmeasureofit．FutureresearchmightprofitablyincludemeasuresofCEO personalityasaprospectivedeterminantofstructureFurthermore，ｉｔｍａｙｂｅｕｓｅｆｕｌｔｏｅｘａｍｉｎｅｔｈｅｃｏｎditionsunderwhichtherelationshipsbetweenstructure andpersonalityorcontingencyfactorsholdForexample，MillerandDroge (1986）foundthattherelationshipsbetweenchiefexecutiveneedforachieveｍｅｎｔ

(23)

103

andstructurewereinfluentialmainｌｙｉｎｓｍａｌｌ（100emplyeesorless）ａｎｄyoung (２０yearsorless）firms．MillerandDroge（1986：554）alsosuggestedthatthe

impactofpersonalityonstructuremightbegreａｔｅｒｆｏｒｆｉｒｍｓｉｎｗｈｉｃｈｔｈｅＣＥＯ

ｏｗｎｓａsignificantpercentageofthefirmａｎｄｓｍａｌｌｅｒｉｎ“subsidiaryorganizations，

tradition-boundindustries，orfirmsbeholdentobanks，governmentalassistance，or

otherexternalconstituents.'’

Theresultsreportedheresupportedveryfewofthehypothesizedinteraction effectsofthecontingencyfactorsandstructureuponperformance．However，

severalunexpecteddirecteffectswerefoundTherefore，futureresearchshould examinemorecloselybothdirectandinteractioneffectsofcontingencyfactors

andstructureuponperformance・

Finally,wewereinterestedinexaminingwhetherthe“form''○ftherelationship betweenstructureandperformancewasconditionaluponthecontingencyfactors・

Therefore，thestudyusedhierarchicalmultipleregressionanalysistotestfor interactioneffectsAsArnold（1982）hassuggested，however，thereisanother aspectofamoderatedrelationship，ｎａｍｅｌｙｔｈｅ“degree，，（ｏｒstrength）ofthe

relationshipThisisindicatedbythecorrelationcoefficientbetweenthetwo variables，andthecomparisonofthemagnitudeofthecorrelationcoefficientstest fordifferentdegreesofrelationship、 Itremainspossiblethatthethree

contingencyfactorswillmoderatethｅｄｅgreeoftherelationshipbetweenstructure andperformance，ｅｖｅｎｉｆｔｈｅｙｄｏｎｏｔｍｏｄｅｒａｔｅｔｈｅｆｏｒｍ・Itwould，therefore，

beusefulinafuturestudytoexaminethedegreeoftherelationshipbetween structureandperformanceforsubgroupsdefined，forinstance，bydichotomizing

measuresofthecontingencyfactorsattheirmedians．

FＯＯＴＮＯＴＥＳ

１．ＩｗｏｕｌｄｌｉｋｅｔｏｔｈａｎｋＲａｙｍｏｎｄＨｕｎｔｆｏｒｈｉｓｈｅｌｐｆｕｌｃｏｍｍｅｎｔｓｏｎａｎｅａｒｌｉｅｒ

draftofthisarticle、 ThisresearchwassupportedbytheGraduateStudent

AssociationDiamondResearchFundandDepartmentofOrganizationand

HumanResources，SchoolofManagement，ＳＵＮＹａｔＢｕｆｆａｌｏ．

２．Analysisof specialization nonsignificant

thecorrelations aggregatedover

values．

amongformalization，centralization，and allorganizationallevelsyieldｅｄｓｍａｌｌａｎｄ

ＲＥＦＥＲＥＮＣＥＳ

Aiken，ＭａｎｄＨａｇｅ，Ｊ、１９６８．“Organizationalinterdependenceandintra-

organizationalstructure.'，AmericanSociologicalReview，３３：９１２－９３０．

Aldrich，ＨＥ、１９７９．OrganizationsandEnvironments．EnglewoodCliffs，Ｎｊ.：

Prentice-HalL

(24)

104

Alexander，』.Ｗ・ａｎｄRandolph，ＷＡ、１９８５．‘`Thefit structureasapredictorofperformanceinnursing ManagementJournal，２８：844-859.

betweentechnologyand

subunits.'，Ａｃａｄｅｍｙｏｆ

Alexander，ＬＤ、1979．“Theeffectlevelinthehierarchyandfunctionalareahave ontheextentMintzberg'srolesarerequiredbythemanageriａｌｊｏｂｓ”

AcademyofManagementProceedings，186-189.

Argote，Ｌ、1982．“Inputuncertaintyandorganizationalcoordinationinhospital emergencyunits.，，AdministrativeScienceQuarterly,２７：420-434.

Arnold，Ｈ､Ｊ、1982．“Moderatorvariables：Aclarificationofconceptual，analytic，

andpsychometricissues.'’OrganizationalBehaviorandHumanPerformance，

２９：143-174.

Bell，Ｇ、1967．“Determinantofspanofcontrol.'，AmericanJournalofSociology，

７３：100-109.

Blau，Ｐ.Ｍ、１９６８．“Thehierarchyofauthorityinorganizations.，’American JournalofSociology，７３：453-467.

Blau，Ｐ.Ｍ、1970．“Aformaltheoryofdifferentiationinorganizations.，，American SociologicalReview，３５：２０１－２１８．

andSchoenherr，Ｐ.Ａ、1971．TheStructureofOrganizations．New

BasicBooks．

Blau，ｎＭ．

Ｙｏｒｋ

Burns，Ｔ・andStalker，Ｇ・Ｍ１９６１．TheManagementofInnovation．London：

Tavistock．

Carter，ＮＭ、1984．“Computerizationofapredominatetechnology：Itsinfluence onthestructureofnewspaperorganization.'， AcademyofManagement Journal，２７：247-270.

Child，Ｊ、１９７３．“Predictingandunderstandingorganizationstructure.”

AdministrativeScienceQuarterly，１８：168-185.

Child，Ｊ、1975．“Managerial performance-PartⅡ Studies，１２：１２－２７．

andorganizationalfactorsassociatedwithcompany Acontingencyanalysis.，，JournalofManagement

Child，Ｊ、1977．Organization：AGuidetoProblemsandPractice・London：Harper

(25)

105

＆ＲＯＷ．

Child，JandMansfield，Ｒ、1972．“Technology，size，andorganizationstructure.'，

Sociology，６：369-393.

Cohen，JandCohen，Ｐ、1983．AppliedMultipleRegression／CorrelationAnalysis fortheBehavioralSciences（２ｎｄｅｄ.)．Hillsdale，Ｎｊ.：LawrenceErlbaum

Associates．

Comstock，Ｄ・andScott，Ｗ，１９７７．“Technologyandthestructureofsubunits.,，

AdministrativeScienceQuarterly，２２：197-202.

Cummins，ＲＣ・ａｎｄKing，ＤＣ、１９７３．‘`Theinteractionofgroupsizeandtask structureinanindusrtrialOrganization.'，PersonnelPsychology，２６：８７－９４．

Daft，ＲＬ、1983．OrganizationTheoryandDesign．Ｓt、Paul：WestPublishing．

Dewar，Ｒ・andHage，Ｊ１９７８．“Size，technology，complexityandstructural differentiation：Towardatheoreticalsynthesis.'’ AdministrativeScience Quarterly,２３：111-136.

1981．“Alevelspecificpredictionofspanofcontrol ofsize，technology，andspecialization.'，Ａｃａｄｅｍｙｏｆ

２４：５－２４．

Dewar，Ｒ・ａｎｄＳｉｍｅｔ，ＤＰ・

examiningtheeffects ManagementJournal，

Werbel，Ｊ１９７９．“Universalisticandcontingencypredictionsof satisfactionandconflict.'，AdministrativeScienceQuarterly，２４：

Dewar，Ｒ、and employee

426-448.

Dewar，Ｒ､，Whetton，Ｄ，andBoje，Ｄ、1980．“Anexaminationof andvalidityoftheAikenandHagescalesofcentralization，

andtaskroutineness.'，AdministrativeScienceQuarterly,２５：

thereliability formalization，

120-128.

Downey，Ｈ、andSlocum，Ｊ１９８２．“Managerialuncertaintyandperformance.'，

SocialScienceQuarterly,６３：195-207.

Drazin，Ｒ・ａｎｄＶａｎｄｅＶｅｎ，Ａ､Ｍ1985．“Alternativeformsoffitincontingency theory.，，AdministrativeScienceQuarterly,３０：514-539.

Duncan，Ｒ、１９７２．“Characteristicsoforganizationalenvironmentsandperceived environmentaluncertainty.'，AdministrativeScienceQuarterly，１７：313-327.

ＣＯＮＴＩＮＧＥＮＣＹＦＡＣＴＯＲＳＵＰＯＮＳＴＲＵＣＴＵＲＥＡＮＤ ＯＲＧＡＮＩＺＡＴＩＯＮＡＬＰＥＲＦＯＲＭＡＮＣＥ：

Daft，1983；Pfeffer，1985)．Also，moststudieshavefocusedontherelationship betweenasinglefactorandstructure・Ｆｏｒexample，Blauandhisassociates（e９．，

ceofsize；Woodward（1965）andPerrow（1967）haveemphasizedtechnology；and

Slocum，1977；Mintzberg，１９７９；ＶａｎｄｅＶｅｎａｎｄＤｒａｚｉｎ，1985）isanobviously interestingbutstillunansweredquestion．Ｆｏｒexample，whenanorganization

formaldoesitsstructurebecoｍｅ？Contingencyfactorsmaybegenerallythought

factorsAsastarttowarddescribingtheseinteractioneffects，［ofthecontingency factors（size，technology，ａｎｄenvironment）onstructureandperformance,］this

Schoonhovenl981；Argote，１９８２；ＦｒｙａｎｄＳｌｏｃｕｍ，１９８４；AlexanderandRandolph，

onperformance（BurnsandStalker，１９６１；LawrenceandLorsch，１９６７；Duncan，

1973；Khandwalla，1973；NegandhiandReimann，1973；Child，1975；Pennings，l975l Theresultsofthesestudiesindicatethatasperceivedenvironmentaluncertainty increases，subunitsemploylessformalizedandcentralizedbutmorespecialized

Therehasbeenmuchdisagreementabouthowtotestcontingencypropositions，

thatｉｓ，howrelationshipsbetweenstrucCureandthecontingencyfactorsofsize，

technology，andenvironmentarerelatedtoperformance（Pennings，１９７５；Miller，

1981；Schoonhoven，１９８１；TosiandSlocum，１９８４；ＶａｎｄｅＶｅｎａｎｄＤｒａｚｉｎ，１９８５；

ＤｒａｚｉｎａｎｄＶａｎｄｅＶｅｎ，1985；ＦｒｙａｎｄＳｍｉｔｈ，1987)．ThisstudyusedJoyce，

Slocum，andGlinow'ｓ（1982）generalcongruencymodeloffit・Thismodelargues thatperformancewillincreasewhenstructureismatchedwithsize，technology，

andenvironment、Ｅｍｐｈａｓｉｓｉｎｔｈｅｍｏｄｅｌｉｓｏｎｔｈｅｓｉｍｉｌarityandmatchingof levelsofstructurewitheachcontingencyfactorasdeterminantsofperformance、

Mintzberg（1979）proposedthateachcontingencyfactoraffectedthevarious levelsofstructuredifferentially．Ｔｈａｔｉｓ，sincepresumablythetoplevelmust functioncontinuouslyatanOrganization'ｓboundaries，itismostaffectedby

elaborationbycreatingchangesinthefavoredmechanismofcoordination，and thereforeisprimarilyaffecteｄｂｙｓｉｚｅ・Finally，thelowerlevelisaffectedmainly

Lau（1983）ｄｏｎｏｔｓｕｐｐｏｒｔｉｔ、Theseempiricalstudiesaretoofew，however，fora

technology,andenvironmentforstructure，Therefore,Mintzberg'ｓ（1979）proposition

Thistablepositsthatsizeandtechnologyaffectformalizationpositively；

however,environmentaffectsformalizationnegativelyTherefore，size／technology

factordeterminingformalization（Hypothesisl）．Thepositiveeffectsoftechnology

Thepositiveeffectsofsizeandtechnologyonformalizationexceedthenegative

ａ＊showsthatthehypothesisappliesatthehierarchicalleveL

ｔｈｅｎumberofexceptionsusingPerrow'ｓ（1967,1970）measureoftechnologyand showedhighreliabilityandvalidity（Dewar，Whetten，andBOje，1980；Whitney，

EnvironmentpDuncan，ｓ（１９７２）perceivedenvironmentaluncertaintyscale representsthemostadvancedoperationalizationoftheuncertaintyconstructto date（Sathe，１９７４）andhasreceivedwidespreadattention

instrumentrequirestherespondenttoidentifyfactorsassociatedwithagiven decision-makingsituation，whichisdifficultandtime-consumingSathe（1974）

modifiedDuncan'ｓ（1972）instrumenttoproduceasimpleroneforquestionnaire

for531subjectsfromtwenty-twodepartmentsofaninsurancecompanyand、834 for60subjectsfromsevenbranchesofacommercialbank（Sathe，1974).Therefore，

thisstudyusedSathe'ｓ（1974）perceiveduncertaintyscale、Cronbach，salphafor

Performance．AperceivedunitperformanceindexｄｅｖｅｌｏｐｅｄｂｙＶａｎｄｅＶｅｎ andFerry（1980）wasusedtomeasuretheperformanceofasubunitinthisstudy．

TotestHypothesisl，aseriesofmultipleregressionanalyseswasconducted forstructureonthethreecontingencyfactors（size，technology，ａｎｄenvironment）

technology，ｏｒenvironment），wetestedthesignificanceofthesemipartial

coefficientassociatedwiththeXZproductterm（ie.，ｓｒ２ｙ(x２．x,z)）inhierarchical

multipleregressionanalysis（Arnold，1982)．Ｔｈｅｒｅgressionequationanalyzedwas

Thesignificanceofthepartialcoefficientassociatedwiththeproducttermwas

SＥＢ３

Targetsachieveda

ａＡ“Notargetswereset'，responsewascodedas“missingvalue''・

technology（Ｆ＝11.11,ｐ＜001)，centralization（Ｆ＝17.39,ｐ＜，001),ａｎｄspecial‐

significantdifferencesfortechnologybothbetweentopandmiddlelevels（ｐ＜０１）

toptomiddlelevels（ｐ＜、０５）ａｎｄｆｒｏｍｍｉｄｄｌｅｔｏｌｏｗｅｒｌｅｖｅｌｓ（ｐ＜、01)．Ａｓ

significantdifferencesofspecializationbothbetweentopandmiddlelevels (ｐ＜０１）andbetweentopandlowerlevels（ｐ＜､05）buttherewasnosignificant

7．TargetsachievedIG、116、0００－０７９．０１２－０６５０５２

aCorrelationsexcludedcaseswithmissingvaluesonananalysis-by‐

bBecauseofmissingｄａｔａ，Ｎｓｒａｎｇｅｄｆｒｏｍ４５ｔｏ５６ｆｏｒｔｈｅａｎａｌｙｓｅｓｌｎ

ｃＡ“Notargetswereset'，responsewascodedas“missingvalue"．

7・Targetsachievedc－．１００

aCorrelationsexcludedcaseswithmissingvaluesonananalysis-by-

bBecauseofmissingdata，Ｎｓｒａｎｇｅｄｆｒｏｍ３９ｔｏ５０ｆｏｒｔｈｅａｎａｌｙｓｅｓ

ｃＡ“Notargetswereset'，responsewascodedas“missingvalue''．

7．Targetsachievedc

aCorrelationsexcludedcaseswithmissingvaluesonananalysis-

bBecauseofmissingdata，Ｎｓｒａｎｇｅｄｆｒｏｍ４９ｔｏ５６ｆｏｒｔｈｅａｎａｌｙｓｅｓ

ｃＡ“Notargetswereset，，responsewascodedas“missingvalue，'．

significantpositiverelationshipbetweensizeandtechnologyatthetoplevel (ｒ＝、255,ｐ＜０５)．ThisfindingSuggeststhatthethreecontingencyfactors

etaL，1976）havefoundSizewasexpectedandfoundtobepositivelyrelated tospecializationatbothtop（ｒ＝、551,ｐ＜、001）ａｎｄｌｏｗｅｒｌｅｖｅｌｓ（ｒ＝８６１，

ｐ＜０１)．Assizeofsubunitsbecamelarger，subunitsweremorespecializedat

eachlevel（top：ｒ＝、320,ｐ＜01；middle：ｒ＝、495,ｐ＜001；ｌｏｗｅｒＩｒ＝、532, ｐ＜’００１）．Astechnologybecamemoreroutine，subunitsemployedmore

ｐ＜０５)．Assubunitshadmoreroutinetechnology，theyweremorespecialized

centralizationatbothtop（ｒ＝、229,ｐ＜､05）ａｎｄｌｏｗｅｒｌｅｖｅｌｓ（ｒ＝、395,ｐ＜０１)．

full-equationbetaweights（standardizedregressioncoefficients)，incrementsinR2，

ResultsofMultipleRegressionAnalysesofStructureatTopLevela

aListwiSedeletionofmissingdatareducedtheNsto53for

ResultsofMultipleRegressionAnalysesofStructureatMiddleLevela

aListwisedeletionofmissingdatareducedtheNsto46for

ResultsofMultipleRegressionAnalysesofStructureatLowerLevela

aListwisedeletionofmissingdataTeducedtheNsto55for

levels，orHypothesisl3concerningmiddleandtoplevels．However，theydid supportHypothesisllconcerningthetoplevel，Hypothesisl2concerningthelower

nosignificantincrementtothepredictionsofeithertargetsachievedorrelative performance，ｅｘｃｅｐｔｉｎｔｗｏｃａｓｅｓ・ＯｎｅｃａｓｅｗａｓｆｏｕｎｄａｔｔｈｅｍｉｄｄｌｅｌｅｖｅLThe

proceduresuggestedbyCohenandCohen（1983)，threerepresentativeregression linesareplottedinFigurel；ｏｎｅｆｏｒａ“low”（nonroutine）technologyscore

〈Ⅱ〈Ⅲ〈Ⅵ

ThisfigureprovidessupportforHypothesisllwhenperformancewas measuredbytargetsachieved・Increasesinformalizationpositivelyinfluenced performancewhentechnologywasroutineandnegativelyinfluencedperformance

AnothersignificantinteractioneffｅｃｔｗａｓｆｏｕｎｄａｔｔｈｅｌｏｗｅｒｌｅｖｅＬＴｈｅ ｉnteractionoftechnologyandcentralizationaddedasignificantincrementtothe

interactionwasO74,ｌｅａｄｉｎｇｔｏａｆｉｎａｌｖａｌｕｅｆｏｒＲ２ｏｆ、1０７．Figure2illustrates thissignificantinteractioneffect，usingCohenandCohen'ｓ（1983）procedure．

SignificantlnteractionEffectBetweenTechnologyand

〈Ⅲ〈Ⅲ〈Ⅲ

ThisfigureprovidessupportforHypothesisl2whenperformancewas measuredbyrelativeperformacne、Whentechnologywasroutine，increasesin centralizationpositivelyinfluencedperformance．Whentechnologywasnonroutine，

Thus，theresultsdidnotsupportHypothesisllforeitherperformance

ＣＯＮＴＩＮＧＥＮＣＹＦＡＣＴＯＲＳＵＰＯＮＳＴＲＵＣＴＵＲＥＡＮＤＯＲＧＡＮＩＺＡＴＩＯＮＡＬＰＥＲＦＯＲＭＡＮＣＥ：

AnothersignificantinteractioneffｅｃｔｗａｓｆｏｕｎｄａｔｔｈｅｌｏｗｅｒｌｅｖｅＬＴｈｅｉnteractionoftechnologyandcentralizationaddedasignificantincrementtothe

（e９．，NadlerandTushman，1980；ＶａｎｄｅＶｅｎａｎｄＤｒａｚｉｎ，1985)．Ｓｕｃｈａｆｉｔｍａｙｓｉｍｐｌｙｎｏｔｂｅｅｎｏｕｇｈｔｏensurehighperformance，ｏｒｖｉｃｅｖｅｒｓａ．