An Empirical Test of Differential Association Theory*

Post on 23-Jan-2022

7 views 0 download

Transcript of An Empirical Test of Differential Association Theory*

5

An Empirical Test of Differential Association Theory*ALBERT J. REISS, JR., AND A. LEWIS RHODES

The University of Michigan

The main empirical question for this paper is whether boys inclose friendship groups have the same specific patterns of delinquentbehavior. The delinquent behavior of boys in close friendship triadswas compared with that expected for six kinds of delinquentbehavior. Two ways of accounting for the observed distribution wereexamined, one based on a random model and the other on a deduc-tion from differential association theory. We conclude that the

probability of an individual committing a specific kind of delinquentact depends upon the commission of the act by other members ofthe triad, though this is not independent of the social class of boys.The actual delinquent behavior of boys in triads departs somewhatless from the random than the differential association hypothesis, atleast for the more serious offenses.

BASIC postulate in sociological writingA about delinquents is that delinquentbehavior is essentially group behavior.

Sociologists have shown that groups en-ter into delinquent activity in a numberof ways. Breckinridge and Abbott wereamong the first to point out that not onlyare most delinquent offenses committedin groups but that most lone offendersare influenced by companions.’ Somewhatlater, Shaw and Meyer2 and Shaw and11cKay3 estimated the extent to which

juvenile delinquency is group activity,showing that less than 20 percent are

lone offenders in juvenile court samples.Shaw and 1~’IcKay also showed that themodal size of offending groups is two andthree participants, and that not all groupdelinquency is committed by well organ-ized gangs.4 4 More recently Enyon andReckless demonstrated that companion-ship is usually present at the onset ofadmitted delinquency as well as in official-ly recorded delinquency.5 These studiesfor the United States and similar onesin other countries clearly establish thatmost delinquent behavior is committed asgroup activity.

Following Sutherland, many socio-

logists reason that delinquent behavior isgenetically a function of learning delin-quency through association with delin-

quents within intimate personal groups.6

* We wish to thank Otis Dudley and BeverlyDuncan, Lloyd E. Ohlin and Guy E. Swansonfor critical comments and suggestions.1 Sophonisba P. Breckinridge and Edith

Abbot, The Delinquent Child and the Home,New York: The Russell Sage Foundation,1917, pp. 34-35.

2 Clifford R. Shaw and Earl D. Meyer, "TheJuvenile Delinquent," in The Illinois CrimeSurvey, Illinois Association for Criminal Jus-tice, 1929, p. 662.3Clifford R. Shaw and Henry D. McKay,

"Social Factors in Juvenile Delinquency: AStudy of the Community, the Family, and theGang in Relation to Delinquent Behavior,"National Commission on Law Observance andEnforcement, Report on the Causes of Crime,Washington, D. C.: USGPO, 1931, Volume II,No. 13, Chapter VI, esp. pp. 194-199.

4 Ibid., p. 195.5 Thomas G. Enyon and Walter C. Reckless,

"Companionships at Delinquency Onset," TheBritish Journal of Criminology, 2 (October,1961), 167-68.6Albert Cohen, Alfred Lindesmith and Karl

Schuessler, The Sutherland Papers, Blooming-ton: Indiana University Press, 1956, pp. 8-11.

6

That this hypothesis is not demonstratedis troublesome to some SoCiologiStS7 anda basis for criticism by others.8 Criticismof the hypothesis rests on a logical argu-ment that empirical evidence of associa-tion in delinquent acts merely demon-strates concomitance of behavior, whereasa temporal sequence of the effects of as-sociation must be demonstrated.9

It is one thing to demonstrate thatmost delinquents associate with other

delinquents, participate with them in de-linquent activity or are members of a

group where others are delinquent andthat conforming boys and girls generallyassociate with other conformers, or belongto groups where behavior is essentiallyconforming to societal norms. It is quiteanother to demonstrate that delinquentbehavior occurs after induction into a de-

linquent group, or that delinquency oc-curs as group activity after a group isformed. Apart from the methodologicalissues raised by a causal demonstrationof group effects on individual behaviorand the nature of criteria for an adequatetest of differential association theory,there are problems of conceptualizinggroup effects and operationalizing con-cepts in differential association theory.

TESTING THE THEORY

Sutherland never explicitly formulatedhis hypothesis of differential association

in operational terms and Short questionswhether it lends itself to operationaliza-tion without reformulation.&dquo;O Short, how-ever, devised a test of differential associa-tion theory to show that the frequency,duration, priority and intensity of associa-tion with delinquent and anti-delinquentculture and behavior varies among delin-quent and nondelinquent groups. He de-fined intensity of asociation as a subject’sperception of the delinquency of his bestfriends and concludes that, among his

operational measures of differential asso-ciation, this measure of intensity is mostconsistently and strongly related to the

delinquency of youth.ll Short’s test restson a subject’s definition of best friendsas delinquent. The main purpose of thispaper is to make a test similar to Short’son the effect of intensity of association,using, how ever, data on the actual delin-quency known and reported by a boy andhis best friends. VVe propose to examinewhether the probability of an individualengaging in several different kinds of

delinquent acts is associated with his closefriends also having engaged in these acts.It should be apparent that a failure to

demonstrate that one’s close friends havedelinquent behavior patterns similar to

one’s own in no way contradicts or sup-ports the hypothesis that most delinquentbehavior occurs as group activity. Rather,it would simply put in doubt the judg-ment that boys who engage in a kind ofdelinquent activity are generally also inintimate association with one another.

Sutherland’s differential association hy-pothesis holds that variation in frequency,duration, priority and intensity of associa-tion with delinquent behavior patterns ac-counts for delinquent behavior. The

homophily hypothesis holds that one is

likely to select as best friends those whosevalues and behavior are similar to one’sown 12 while coalition theory argues that

7 Donald R. Cressey, "Epidemiology and In-dividual Conduct: A Case from Criminology,"Pacific Sociological Review, 3 (Fall, 1960),and James F. Short, Jr., "Differential Associa-tion as a Hypothesis: Problems of EmpiricalTesting," Social Problems, 8 (Summer, 1960),pp. 14-25.8The most pointed criticism has been made

by the Gluecks. Sheldon and Eleanor Glueck,Unraveling Juvenile Delinquency, Cambridge :Harvard University Press, 1950, pp. 146-149and 163-164. See also Marshall Clinard, "Crimi-nological Research," in Robert K. Merton,Leonard Broom and Leonard S. Cottrell, Jr.(eds.), Sociology Today: Problems and Pros-pects. New York: Basic Books, Inc., 1959,Chapter 23.9Sheldon Glueck, "Theory and Fact in

Criminology," British Journal of Delinquency, 7 (July, 1956), 92-109.

10 James F. Short, Jr., op. cit., p. 17.11 James F. Short, Jr., op. cit., p. 18.12Paul F. Lazarsfeld and Robert K. Merton,

"Friendship as Social Process: A Substantiveand Methodological Analysis," in Morroe

Berger, et. al., Freedom and Control in Modern

7

all other things being equal, constraintson members who deviate from the expect-ations of the group lead to their behavingin conformity with these standards. 13

Though not explicitly stated in any’ the-ory, both group selection and group con-straint hypotheses lead to the same con-clusion : one’s close friends should havea delinquency history similar to one’sown.

It is apparent, however, that there is

considerable variation over time in the

cliques to which an adolescent belongs,in whom he will select as his best friends,and in the kinds of delinquent activityin which he will engage. Shaw pointedout, for example, that Sidney in the courseof his delinquent career from 7 to 17

years of age was officially known to havebeen involved in delinquency with 11 dif-ferent companions, representing three

distinct groups whose activities and tradi-tions were delinquent in character, andthat he was never implicated in any of-fense with more than three delinquentsRecognizing that current best friends arenot necessarily companions from pastdelinquent association, it seems consistentwith differential association theory to

argue that, if current best friends com-

prise a salient primary group, and if pastbehavior serves as a basis for mutualcommunication and action within it

(which it need not), then boys currentlyin intense association with one anothershould show similar patterns of delin-

quency. Assuming that specific techniquesfor committing delinquent acts are com-municated in primary association, it fol-lows that all, or none, of the boys in

close friendship triads should report com-mitting a given kind of offense. Within

a triadic friendship group, there shouldbe no dyads committing a given type ofoffense, since group constraint should

produce homogeneity in behavior. It

should be clear that whether or not boysin close friendship groups show similar-ity in delinquent behavior because theyselect one another on this basis, or as aresult of association, failure to show thatdelinquency histories of boys in close

friendship groups are the same casts

doubt at least upon the specificity of anylearned delinquent behavior in intense as-sociation with others. Since Sutherlanddid not restrict his hypothesis to lowerclass delinquents, behavioral homophilyshould hold regardless of social class.15To show that one’s close friends are

also delinquent is not to show that theyhave an effect on all of one’s delinquentactivity. Shaw and McKay early showedthat stealing is more likely to be a groupoffense than are offenses against the homeand school,.16 Enyon and Reckless havegone further to show that companion-ship characterizes first participation insome kinds of offenses more than inothers. Companions were present in 100per cent of boys’ first involvement in

gang fights but only 56 per cent of thecases of first running away from home. 17While our study cannot demonstrate theprecise effect of friendship on delinquencypatterns, it investigates the extent to

which there is covariation in a boy’s delin-quent behavior and that of his friendsfor different kinds of delinquent be-havior,.

Sociological theories on delinquent sub-cultures that are consistent with Suther-land’s differential association hypothesispostulate that members of delinquent sub-cultures become highly dependent uponone another, particularly for status grati-fication. As Short points out, it followsthat members of such groups, having a

Society, New York: D. Van Nostrand Co.,Inc., esp. footnote 19.18 John W. Thibaut and Harold H. Kelley,

The Social Psychology of Groups, New York:John W. Wiley, Inc., 1959, pp. 208 and 210,and L. Festinger and J. Thibaut, "InterpersonalCommunication in Small Groups," Journal ofAbnormal and Social Psychology, XLVI

(1951), 92-100.14Clifford R. Shaw and Henry D. McKay,

op. cit., p. 221.

15 The Sutherland Papers, op. cit., p. 19,pp. 32-33 and pp. 58-59.16Clifford R. Shaw and Henry D. McKay,

op. cit., pp. 195-196.17Thomas G. Enyon and Walter C. Reckless,

op. cit., Table 3, p. 170.

8

more intense association with one another,should show greater similarity in their

patterns of delinquency than do membersof other delinquent groups.18 Cohen’s

general theory of delinquent subculturesholds that subcultural delinquent groupsshould be homogeneous in behavior for avariety of delinquent offenses againstproperty and persons. It is implicit inhis theory that middle class boys willshow less similarity and versatility intheir delinquency,.19 Miller holds that

delinquency is endemic in lower classculture. It would be consistent with his

theory to argue that delinquent behaviorof lower class boys is independent of thecommission of the act by other membersof the group. 20

THE INVESTIGATION

The investigation was designed to

gather information on the actual delin-

quent behavior of boys in close friend-

ship cliques. A sample of 378 boys wasdrawn from a base population of all whitemales between the ages of 12 and 16 who

were registered in one of 45 public, pri-vate or parochial junior or senior highschools in Davidson County, Tennesseeduring the 1957 school year. Strata weredesigned so as to select disproportionatelylower- and middle-class delinquent boys.21Each clique is a triad composed of a

boy selected in the stratified probability

sample of 378 boys and his two closestfriends. Given a large population fromwhich the sample of boys was drawn, onlya few sample cases chose the same

&dquo;closest&dquo; friend. Effects of overlappingfriendship choice or of pyramiding there-fore are negligible. Information was

gathered for 299 triads and 79 dyads. Thedyads are pairs where a boy selected onlyone &dquo;best friend.&dquo; Data are presented inthis paper only for the 299 triads. Eachstep was replicated for the 79 dyads andthe results are similar where the num-ber of dyads makes comparison possible.The index person in each triad was

classified into one of seven conformingor delinquent types.22 The career-orienteddelinquent is the most delinquent personin the classification schema. He is orient-ed toward the adult criminal world andmaintains contact with adult criminals.The largest group of delinquents are

peer-oriented and directed in their goalsand behavior. The lone delinquent is ournonconforming isolate. There are four

types of conforming boys. The con f orm-ing nonaclziever is comparable to WilliamVVhyte’s &dquo;corner boy&dquo; and the confor1n-ing achiever to his &dquo;college boy,&dquo; if socialclass attributes are disregarded.23 Thelz3,perconfortiter disregards conventionalfor strict conformity while the con f orm-ing isolate is outside the clique system.Peer-oriented delinquents, conformingnonachievers, and achievers are dividedinto white-collar and blue-collar statusbased on father’s occupation. 24The dependent variable, self-reported

delinquent behavior, was measured byasking each boy how often he had doneany of the following things, whetheralone or with others, and by inquiringabout the conditions related to it: takenlittle things worth less than $2 ? $2 to $50?

18 James F. Short, Jr., op. cit., p. 17.19Albert K. Cohen, Delinquent Boys: The

Culture of the Gang, Glencoe: The Free Press,1955, pp. 157-169.20 Walter Miller, "The Impact of a ’Total

Community’ Delinquency Control Project,"Social Problems, 10 (Fall, 1962), 169-191.21This is a more efficient sample design

inasmuch as delinquency is a relatively lowincidence phenomenon in the general popula-tion. The methodological problems encounteredin dealing with a low incidence phenomenon ina population have been discussed in DanielGlaser, "Differential Association and Crimi-nological Prediction," Social Problems, 8

(Summer, 1960), p. 7, and Albert J. Reiss, Jr.,"Unraveling Juvenile Delinquency II: An Ap-praisal of the Research Methods," AmericanJournal of Sociology, 57 (September, 1951),118-119.

22 Albert J. Reiss, Jr. and A. Lewis Rhodes,"The Distribution of Juvenile Delinquency inthe Social Class Structure," American Socio-logical Review, 26 (October, 1961), Chart I.23William F. Whyte. Street Corner Society,

Chicago : University of Chicago Press, 1937.24Albert J. Reiss, Jr. and A. Lewis Rhodes,

op. cit., pp. 721-722.

9

more than $50? purposely damaged ordestroyed property? taken a car withoutthe owner’s permission or knowledge?beat up somebody bad enough to be ar-rested ?25 Self-reports include virtually allcases of officially recorded delinquency.Only delinquent acts committed after

age 10 are data for this paper.Self-reported delinquent acts were

tabulated for each of the six categoriesof delinquent act for the 299 triads ar-ranged in the 10 types of conforming-delinquent, SES groups.28 This tabulationprovided information on kind of delin-

quent behavior reported by none, one,

two or all members of the triad in eachof the 10 conforming-delinquent groups.A model was then constructed to givethe expected delinquent behavior com-position of the triad, using the actualrate of delinquency for the sample of

boys for estimation purposes. The modelis based on the expansion of the bino-mial. 27 Our use of the binomial ignoresvariability in response patt2rns by friend-ship choice, e.g. + (original) - (firstbest friend) + (second best friend) and+ + - or -++ + are all treated as two

boys expected (or actual) to commit theact. This disregard of response orderseems warranted for we cannot determinewhether the original subject models hisbehavior on that of his two closest friends,or whether he chooses friends who havesimilar behavior, or whether they copyhis behavior.The observed distribution of boys in

triads reporting they committed an actof delinquency is then compared with theexpected distribution. The chi squaretest of goodness of fit is used to test thesignificance of the departure of the ob-served measure from the hypothetical oneof the binomial. 211 Occasionally, the con-ventional test of the significance of dif-ference between two proportions is usedto test whether there is any significantdifference in the number of observed andexpected triads where all members of thetriad reported committing the act.

Very briefly, this paper attempts to

shed light on three closely related ques-tions that are germane to propositionsabout the group nature of delinquencyand the empirical testing of differentialassociation theory: (1) Does the proba-bility of an individual committing kindsof delinquent acts depend upon his closefriends comitting these acts? (2) Is therevariation in dependence upon friends

committing delinquent acts among dif-ferent kinds of delinquent behavior? (3)Is the probability of committing a delin-quent act less dependent upon one’sfriends committing the act in some kindsof conforming or delinquent groups thanin others?

FINDINGS

Boys generally choose boys as closefriends whose law-abiding or delinquentbehavior is similar to their own. Table 1answers our first question in comparingreported delinquent behavior of boys intriads with that expected from the pro-portion of boys in the sample who re-

ported committing specific kinds of delin-quency. For each kind of delinquent be-havior, the probability of an individualcommitting a specific delinquent act de-pends upon the commission of tlze act byother members o f the f riendshi p triad.More of the triads in Table 1 than ex-pected from the binomial are made upof boys, all or none of whom engaged in

25 Readers will note the similarity of thesequestions with the Nye-Short delinquency scaleitems: Ivan F. Nye and James F. Short, Jr.,"Scaling Delinquent Behavior," American So-ciological Review, 22 (June, 1957), 326-331.24The authors express their appreciation to

NAL, State University of Iowa for use of theIBM 650. A special program for rapid tabula-tion of response patterns in triads was devel-oped for this study.

where x is the number of boys expected tocommit the act (0, 1, 2 or 3) ; p is the propor-tion of the subgroup reporting commission ofthe act; q is the proportion of the subgroup notreporting commission of the act.

28No test was made if the expected frequencyin any cell was less than two or less thanfive in two cells of a 2 X 3 table.

10

the same kind of delinquent act. Themore serious the offense, the greater thedifference between observed and expectedproportions of triads where all of the boyscommitted the same kind of delinquentoffense. Confidence in the finding that

the probability for a boy committing a

delinquent act is not independent of thebehavior of his close friends is increasedwith the observation that fewer of thetriads than expected have only one boyreporting he engaged in the delinquentactivity. Table 2 restates the conclusionin a way that aids the interpretation.

TABLE 1

Observed (f ), Expected (fe)1 and Sum of Expected (f~:)2 for Conforming-DelinquentSubgroups. Number of White Male Triads Classified by Number of Males in EachTriad VVho Reported Delinquent Behavior One or More Times for Six Kinds of

Delinquent Behavior

1 Expected frequencies are calculated for the binomial using the proportion of boys in the samplewho reported committing each kind of delinquency.$ Expected frequencies were calculated for the binomial for each of 10 conforming-delinquent

subgroups. The sum of these expected values is reported here.’ X2 computed for actual (f ) with expected (fe) values only.It Offense committed two or more times.* Cell frequencies combined for computation of y2.

11

For each kind of act, significantly moreof the original sociometric subjects whoreported the offense, than of those whodid not, have friends who also committedthe act.

Nonetheless, Tables 1 and 2 make ap-parent considerable variation in delin-

quent behavior homophily of close friend-ship triads. Of the triads in Table 1 whereat least one boy reported committing autotheft or assault, three-fifths have only oneboy reporting he committed the act. Byway of contrast, but one-fourth of thetriads where at least one boy committedan act of vandalism and one in five for

petty larceny are made up of boys whereonly one reported committing the act.

Original sociometric subjects in Table 2are more likely to choose boys as friends

TABLE 2

Per Cent of Triads with Number ofFriends Committing Delinquent Act byOriginal Subject’s Delinquent Behavior,for Six Kinds of Delinquent Behavior

who also committed acts of vandalismor theft under two dollars than they areto have chosen boys as friends who alsocommitted acts of auto larceny or assault,when they report having done these

things. We must conclude in answer toour second question that although, in theaggregate, commission of a kind of delin-quent act is not independent of the com-mission of the act by other members of aclose friendship triad, the correlationvaries with kind of delinquent behaviorand is far from perfect for any kind.We know from previous studies that

roughly four-fifths of all boys arrestedfor delinquency had associates in the of-fense for which they were arrested, andthat at least that high a proportion ofdelinquent boys have as close friends

boys who have committed some kind ofdelinquent act. We must conclude then,that close friendship choices are more

closely correlated with delinquency per sethan with specialization or engagement inall specific kinds of delinquency.

Attention has been called to the am-

biguity in formulation of Sutherland’sdifferential association theory renderingdifhcult both operationalization of the

theory and deductions from it. An alto-gether literal deduction from Sutherland’stheory, though he never made it, is thateither all or none of the boys in a closefriendship triad should report committingthe same kind of offense. It is immedi-ately apparent from inspection of Tables 1and 2 that there is a substantial numberof triads where only one or two mem-bers of the triad committed the same kindof delinquent act, thereby calling into

question any postulate about the homo-geneity of law-violative behavior in triadsthrough differential association. Let us as-sume, however, as does a variant of coali-tion theory, that when two members ofa triad engage in a given kind of be-havior, the third member is under strongpressure to do likewise.29 We would ex-

pect, then, that there should be relativelyfew, if any, close friendship triads with29See footnote 13.

12

only two members engaging in delinquentbehavior. Expressing the triads where allmembers commit the same kind of delin-

quent act as a per cent of all triads wheretwo or three members commit the act,the following distribution results; auto

theft (65 per cent) ; theft over $50 (64per cent) ; theft $2-$50 (80 per cent) ; as-sault (83 per cent) ; vandalism (71 percent) ; theft under $2 (82 per cent). Thedistribution supports the contention thatthere is pressure toward uniformity ofbehavior in these triads. In two-thirds ormore of the triads, for each kind of delin-quent offense, all members report theycommitted the act, i.e., if more than onedid it, it was probably three. There re-mained, nonetheless, a substantial minor-ity of triads in which only two mem-bers committed the same kind of delin-quency. The more serious offenses areleast likely to show triadic uniformity.Thus far two main ways of accounting

for the observed distribution of delinquentassociates in close friendship triads havebeen introduced. We first examinedwhether the sample of triads was a sampledrawn from a binomial based on the rateof a specified kind of delinquency, andwe concluded that the departure of theobserved distribution from the binomialexceeded that ordinarily encountered inrandom sampling. The probability of anindividual committing a specific kind ofdelinquent act depends upon the com-mission of the act by other members ofthe friendship triad. B&dquo;’1 e then examinedwhether the sample of triads conformedto predictions from Sutherland’s differ-ential association theory or coalition

theory. Inasmuch as there was a sub-stantial number of triads with only oneor two members reporting they engagedin a specific kind of delinquent behavior,we are led to question the postulate thatdifferential association is a necessary andsufficient condition explaining delin-quency. Table 3 summarizes these com-

parisons and is a convenient way of rais-ing the further question whether the ob-served distribution of triads departs more

from the random distribution than theexpected one based on the differential as-sociation hypothesis. Although no test ofstatistical significance is employed, it

TABLE 3

Comparison of Reported Behavior in 299Delinquent Triads (f) with RandomExpectation (fe) and Number Expectedunder Differential Association Effect onBehavior of Original Member of Socio-metric Triad (fd)2 for Six Kinds of

Delinquent Behavior

’ The proportion of boys in the sample whoreported each kind of delinquency is used toset up the binomial of triads.’ The expected values for differential associa-

tion are the marginal frequencies of originalsubjects committing and not committing a spe-cific kind of delinquent act.* Less than one case.

13

seems clear that the observed distributionis closer to the binomial than to the ex-

pected distribution based on the differ-ential association hypothesis.The reciprocation of sociometric

choices, the delinquency orientation andbehavior of boys chosen as close friends,and the content and seriousness of a boy’sdelinquent offenses were the main criteriain classifying a boy into a particularlyconforming or delinquent type in our

study. The type and content of the delin-quent offenses of his close friends werenot used as criteria in classifying a boyinto a particular group. Although classifi-cation of a boy and his friends into a con-forming or delinquent subtype then isnot independent of classification by typeof delinquent behavior, there still can beconsiderable variation in the delinquentbehavior among the members of a triadwithin any kind of delinquent group.Given the possibility of variation in typeand content of delinquent offense withina triad, we compared the triads in each

subtype of conforming or delinquentgroup to see whether boys in each subtypechose as close friends boys who committedacts of delinquency similar to their own.Table 4 compares the reported behaviorof boys in each subtype of conforming ordelinquent triad with the behavior ex-

pected from the proportion of boys ineach type who reported committing eachkind of act. Such comparisons should

permit us to learn whether membershipin a specific kind of conforming or delin-quent group has any effect upon one’sdelinquent behavior independent of therate of delinquency within that type ofgroup.

Inspection of Table 4 shows that thereis little significant variation between ob-served and expected values for any ofthe conforming-delinquent groups. Theanswer to our third question then is thatselection of close friends who commit aspecific kind of act within a given typeof conforming-delinquent group is largelya function of the rate of that kind of delin-

quency within each group. The more boys

there are committing any kind of offensein a type of group, the more likely oneis to have groups in which all memberscommit that kind of offense. Put in an-other way, our classification of boys intoconforming and delinquent types of groupsaccounts in large part for the tendencyfor boys to choose as close friends boyswho commit delinquent acts similar to

their own. This can be seen by turningagain to Table 1 where we observe thatthe sum of the expected values for the

conforming-delinquent subgroups is re-

markably like that observed for all triads,particularly for the serious offenses ofauto theft and theft over $50. These twotypes of offenses are more clearly con-centrated in the career- and peer-orienteddelinquent types, of course. The similar-ity between the sum of the expected valuesfor subgroups and the actual behavior

reported within triads is less marked forthe less serious offenses, offenses whichoccur quite frequently in most conform-ing and delinquent groups.

These observations (a summary ofwhich is aided by comparing the fand fz values in Table 1) suggest that amodel of random selection accounts forsubcultural or career-oriented delinquentsassociating most frequently with boyswho commit delinquent acts similar to

their own, given our classification of theminto that type of delinquent group. Thisfinding, of course, should not obscure thefact that the classification system doesdiscriminate among types of conformingand delinquent boys. Career-oriented de-linquents are easily distinguished from allother types by the fact that for every kindof offense, at least two-thirds of thetriads are made up of boys who com-mitted the same kind of offense. Thereare significantly more career-orientedtriads in which all members engaged in

every kind of offense than in any other

type of delinquent group except that pecr-oriented white-collar delinquents have

significantly more triads in which allmembers committed theft under two dol-lars.

14

buoG4

G4 GM_O

2tl3a.~ü5l#/l~.&dquo; 4-.,t- cf-4~

:3

.cO&dquo;u C

Li Aflc J5 ho

C

S’E~gwc

08u

o~ 0.a£#z~HE.c u’5t3~t:.1N C

+ 1’~

y$ 1# C= V) CJ~ &dquo;’C :3

1-4 .2 g-P§vflni-O~-~’~ c%k£~ !<

fisiu &dquo;&dquo;,ESJ=z~,m

-0

&dquo;’0&dquo;&dquo;VOV<J ft!12K ~w a1w-C C

Cd U:3- 0&dquo;

+~Eu....’T-~34J i-I

t0IT.aC

w

0mbz

+

O

J5

0

ejSo.

~!80

’A«

H

Vu

c 4-flNn

C b9+N yr

’E o

S.’5s.sox

ux’p «’5 ~

t- -04i1

L40

~ Ot11v Ld

uIn°lX a.4j= 0

~_’. a O °° L.~ 3N

a c

2 ~~.~

a

w. CH uyu o41)geoC =K *5

4o

- Hm£0>~.0uw4d 0u

@Gu.0W ir0on~ &...4u~^ E-~ .* w

15

VVe have shown that for each kind ofdelinquent behavior reported in this study,the probability of an individual com-

mitting a specific act of delinquency- is, dependent upon the commission of the

act by other members of the triad. Exceptfor Sutherland’s original formulation ofdifferential association theory, most con-temporary sociological theories emphasizea qualitative difference between middleand lower class delinquency. In Table 5we ask whether the finding that a boy’s

delinquent behavior depends upon hisclose friends engaging in it is independentof the social class status of the boys. It

clearly is not for all types of offenses.Among blue-collar boys, the probabilityof a boy engaging in any specific kindof delinquency depends upon his closefriends engaging in it but among white-collar boys this is true only for theft in-volving amounts of less than $50 or forvandalism. Apparently when middle classdelinquent boys engage in serious delin-

TABLE 5

Observed (f) and Expected (f.)’ Number of White Male Triads Classified bynumber of Male in Each Triad who Reported Delinquent Behavior for Six Kindsof Delinquent Behavior, Controlling on Social Class of Original Subject in

Each Triad

* Expected frequency is less than one case.’ The proportion of boys in each social class subgroup who reported committing each kind ofdelinquency is used to set up the binomial of triads for each social class subgroup.

16

TABLE 6

Per Cent of Triads Where All Boys Admit Delinquent Acts in Which All Boys AlsoIndicate Commission of Act with Someone Else, by Kind of Act and Type of Triad

quent behavior it is relatively independentof their close friendship choices.Our interviews with subjects were

structured so as to avoid mention byname of close friends in delinquency. Todo so would violate peer norms about

&dquo;squealers&dquo; and at times jeopardize rap-port. Many respondents neverthelessvolunteered names of their co-participantsfor delinquent acts in which they had as-sociates and these were usually personsmentioned as &dquo;closest friends.&dquo; Each re-

spondent was explicitly asked for each

reported delinquent offense whether hewas (always, usually, sometimes or never)alone (or with one or more persons)when committing it. Table 6 presents thisinformation only for those triads in whichall members reported committing a speci-fic kind of delinquent act. The objectiveis to investigate whether unanimous

reporting of engaging in a kind of delin-quent behavior in a triad means they en-gaged in the behavior as group activity.Evidently this is not always the case. Itis apparent that, for close friendshipgroups where all members committed the

same kind of act, participants are mostlikely to report vandalism as group activ-ity and least likely to report theftunder $2 as group activity. This findingis consistent with that of Enyon andReckless on the percentage of cases inwhich companions were present at firstoccurrence of admitted delinquency, it

being higher for acts of vandalism (91per cent) than for taking things under $2(69 per cent).80

If attention is directed to variation in

group involvement in delinquency amongour conforming-delinquent types of triads,there is substantial evidence that onlycareer-oriented delinquents report groupinvolvement for all types of offense otherthan theft under $2. The career-orienteddelinquent is apparently most likely to

commit his offenses with accomplices.

SUMMARY AND CONCLUSIONS

The main question for this paper waswhether boys in close friendship groups30Thomas G. Enyon and Walter C. Reckless,

op. cit., Table 4.

17

have the same specific patterns of delin-quent behavior. The reported delinquentbehavior of boys in close friendship triadswas compared with that expected for sixkinds of delinquent behavior. Two dif-ferent ways of accounting for the observeddistribution were examined.The first compares the observed delin-

quent behavior of boys in triads with abinomial based on the rate for each kindof delinquency in the population. We con-cluded that the probability for an in-dividual committing a specific kind of de-linquent act depends upon the commis-sion of the act by other members of thetriad. This dependence upon close friendsengaging in delinquent activity is not in-dependent of the social class status of

boys for all kinds of offenses, however.Among blue-collar boys, the probabilityof a boy engaging in any of the six kindsof delinquency depends upon his closefriends engaging in it but among white-collar boys this is true only for the lessserious offenses.The second comparison asks whether

the behavior of boys in triads departsfrom predictions from Sutherland’s dif-ferential association theory or coalition

theory that there be uniform conformityin conforming groups and uniformity ofspecific kinds of delinquent behavior in

delinquent groups. We concluded there isconsiderable departure from this explan-atory model even when only those groupsare considered where at least two boysengaged in the same kind of delinquency.The observed distribution of delinquencyin close friendship triads departs some-what less from the random than the dif-ferential association model, at least forthe more serious offenses.There is in fact considerable variation

in the delinquent behavior homophily offriendship triads. The degree to whichcommission of a kind of delinquent actdepends upon its commission by othermembers of the triad varies considerablyby type of delinquency. Vandalism andpetty larceny, the more common offenses,,are commonly committed by two or three

members of the triad while a majority ofthe triads where at least one member com-mitted auto theft or assault are made upof only one member committing the of-fense. Behavioral homophily in triadsdoes not mean that boys always or usual-ly commit these offenses together, sincethere is evidence that theft under $2 isleast likely to involve group activity. Twothings seem evident from these findings,that delinquent behavior homophily inclose friendship triads does not neces-

sarily involve association in the commis-sion of offenses and that some offensesare more clearly group activity thanothers.Our classification of boys into confonn-

ing and delinquent subgroups accounts inlarge part for the selection of close friendswho commit delinquent acts similar toone’s own. BVhile career-oriented delin-

quent boys generally have the highest pro-portion of triads where boys commit thesame kind of delinquent act, they alsohave the highest overall rate of delin-

quency for each kind of act. The main

problem is to account for the higher rateof delinquency among these boys. Certain-ly the effect of one’s close friends ondelinquency does not appear to be a suf-ficient reason to account for this higherrate since a substantial minority of career-oriented delinquent boys are in close

friendship triads where at least one otherboy does not commit the same kind ofoffense and the convergence of boys whocommit the same kind of delinquent actin close friendship triads is not greaterthan that one would expect from the rateof delinquency among these boys.

This study cannot be construed as a

test of the genetic formulation of the dif-ferential association hypothesis. To theextent that the findings of this study arevalid and our logical inferences correctlydrawn, however, they may be disappoint-ing to proponents of differential associa-tion theory. The association of boys withthe samc kind of delinquent behavior inclose friendship triads while somewhat

greater than cl~ance is well below what

18

one would expect from the learning hy- <

pothesis in differential association theoryand the results are not independent of

social class. Close friendship choices aremore closely correlated with delinquencypeer se than with participation in specificpatterns of delinquency presumably learn-ed from others.The results also cannot be interpreted

as clearly supporting one of the majortheories of delinquent behavior over thatof another, though some postulates in

these theories seem supported over

others. The main sociological theories ofsubcultural delinquency such as those ofCohen and Walter Miller postulating dif-ferences between lower and middle class

gang behavior find some support in this

study. Delinquency among middle class

boys, particularly for the more serious

delinquent offenses, is independent of

friendship choices while among lower

class boys the probability of committingany kind of delinquent activity is relatedto the delinquent activity of one’s closefriends. The fact that selection of closefriends who commit specific kinds of delin-quency within each type of conforming-delinquent group is largely a function ofthe rate of delinquency within each grouplends support to Walter Miller’s conten-tion that delinquency is endemic in lowerclass culture. Nonetheless, if Miller is

correct, convergence of delinquent pat-terns of behavior in friendship groupsshould not exceed chance since he arguesthat the pressures toward deviance comefrom outside the immediate peer group.The fact that the probability of a lowerclass boy’s committing any specific kind

of delinquency is dependent upon the com-mission of the act by other members ofthe group therefore is at odds withMiller’s formulation. The model of dif-ferential association seems even a lesspowerful one in accounting for our ob-served patterns of behavior in close

friendship triads than does Miller’s for-mulation, however.The fact that a substantial proportion

of career-oriented delinquent boys doshow a marked similarity in delinquentactivity, particularly for the more seriousoffenses, is consistent with Cohen’s for-mulation emphasizing the versatility of

delinquency among subcultural delin-

quents. That some of these groups maybe specialized in specific kinds of delin-quent activity was not investigated.

This study perhaps only serves to em-phasize the difficulty in testing inferencesfrom differential association theory. It

perhaps is unnecessary to repeat what isalready well stated, that we need to

operationalize the hypothesis in such a

way as to test the relationship of associa-tion with delinquent others through time.Of considerable importance, however, infuture research would be an investiga-tion of the &dquo;deviant&dquo; cases which do notconform to expectations of the differentialassociation model. How can one accountfor the fact that all members of delinquentgroups do not conform to the same pat-terns of delinquency? Why are some

members of close friendship groups delin-quent and not others? What are the pat-terns of recruitment to peer groups and

how stable is peer group structure ?