Address correspondence to: Emily K. Asencio, University of Akron, Sociology, 247 Olin Hall, Akron, OH 44325; e-mail: firstname.lastname@example.org.
Sociological Perspectives, Vol. 54, Issue 2, pp. 163–182, ISSN 0731-1214, electronic ISSN 1533-8673. © 2011 by Pacific Sociological Association. All rights reserved. Please direct all requests for permission to photo- copy or reproduce article content through the University of California Press’s Rights and Permissions website, at http://www.ucpressjournals.com/reprintinfo.asp. DOI: 10.1525/sop.2011.54.2.163.
DOeS INCARCeRAtION ChANge the CRImINAl IDeNtIty? A SyNtheSIS
Of lAbelINg AND IDeNtIty theORy PeRSPeCtIVeS ON IDeNtIty ChANge
emIly K. ASeNCIO University of Akron
PeteR J. bURKe University of California, Riverside
AbStRACt: This study uses an identity theory framework to fill in gaps left by labeling theory in the explanation of criminal behavior resulting from one’s identity. The authors examine the reciprocal relationship between the views of significant others, peers, and guards at the jail, and self-views for changes in the criminal identity, the drug user identity, and the worker identity using an incarcerated population. The findings suggest that identity theory provides a more detailed explanation of the processes by which the views of others can influence criminal behavior than labeling theory does. Identity theory allows for the explanation of different behavioral outcomes in response to the same label, which labeling theory does not provide. Keywords: identity change; criminal identity; reflected appraisals
Empirical studies of criminal behavior from a labeling theory tradition tend to focus on whether or not stigmatizing labels representing the externally imposed views of others influence the behavior of the person labeled. Two problems emerged from these studies. First, the results of these studies provided little empirical sup- port for the theory, which led to a lack of enthusiasm for pursuing labeling theory as an explanation for crime. Second, individual variability in outcomes could not be accounted for.
Initial steps to address this second issue were made with the introduction of the reflected appraisals process as intervening between the label and behavioral outcomes (Bartusch and Matsueda 1996; Heimer and Matsueda 1994; Matsueda 1992; Matsueda and Heimer 1997). While this was helpful in moving a theoretical understanding of the process forward, the full degree of individual variability was
SOP5402_03.indd 163 5/23/11 2:34:49 PM
164 SOCIOLOGICAL PERSPECTIVES Volume 54, Number 2, 2011
left unexplained because identity processes in addition to the reflected appraisals were omitted.
The current study suggests that a better theoretical understanding can be obtained by addressing the larger, more important theoretical issues surrounding the identity processes through which a label might impact behavior. We suggest that the broader perspective of identity theory (Burke 1980; 1991; Burke and Stets 2009; Stryker 1968) incorporates all of the prior developments, including the major tenets of labeling theory and the additional processes of reflected appraisals (the perceived views of others) stemming from symbolic interactionism. It also adds an understanding of the process by which a label may or may not become part of a person’s identity, thereby allowing it to influence behavior.
Our focus will be on the way that both self-views (identities) and the perceived views of others (reflected appraisals) may change over time among residents in a correctional rehabilitation program that has the goal of curbing recidivism by alter- ing the way that offenders view themselves. These programs, known as cognitive- behavioral modification programs, are the embodiment of labeling theory. They use labels and corresponding behavior to change the identities of participants with the philosophy that if offenders no longer view themselves as a criminal, they will no longer behave as a criminal (Cullen and Gendreau 2000).
lAbelINg theORy AND SymbOlIC INteRACtIONISm
Labeling theory stems from the symbolic interactionist tradition of the “looking- glass self” (Cooley 1902), or the reflexive self (Mead 1934), which means that an individual will come to view the self in a way that reflects the views of others and thus come to act in a manner consistent with these views. Lemert (1951) explained that secondary deviance can be created through labeling when normal exploratory behavior by a child or adolescent is seen as “deviant” and the child who engages in these activities is labeled as a delinquent. According to labeling theory, this label becomes internalized and forms part of the identity of the child, who subsequently comes to act out on the basis of this identity. Once a deviant or criminal label is applied, for example, it perpetuates itself by eliciting reactions from others such as law enforcement, employers, family members, friends, and other acquaintances (Becker 1963). Labeling theory explains that an individual’s behavioral adjustment to a label represents the internalization of this label and much of the early research from the labeling theory perspective focused on how the criminal justice system perpetuated the very behavior it was designed to stop (Lemert 1951; Matza 1964; Schur 1973; Tannenbaum 1938). Empirical studies, however, failed to support this simplistic version of labeling theory, leading to a decline in popularity and appli- cation of the theory in its original formulation in recent research.
Labeling theory suggests simply that an individual will behave in accordance with a label applied to him or her by others. There is no explanation as to why some who have been labeled a criminal might not engage in criminal behavior, nor why some who have never been labeled a criminal might engage in criminal behavior even though these two kinds of cases occur very often in the criminologi- cal literature (Moffitt 1993; Sampson and Laub 1993). By ignoring theoretically the
SOP5402_03.indd 164 5/23/11 2:34:49 PM
Does Incarceration Change the Criminal Identity? 165
process by which a label may influence an individual’s identity, labeling theory overlooks the agency of the individual being labeled and thus does not explain differential outcomes for people who are labeled as criminal.
In response to criticisms of labeling theory about inattention to individual dif- ferences in behavioral outcomes despite being labeled similarly by others (Cul- len and Cullen 1978; Gove 1980), Matsueda et al. (Bartusch and Matsueda 1996; Heimer and Matsueda 1994; Matsueda 1992; Matsueda and Heimer 1997) exam- ined the effects of reflected appraisals on the behavior of juveniles who are clas- sified as delinquent. In this way, they sought to offer an answer to the criticism of why the deviant behavior of some, but not all, individuals is perpetuated by a deviant label (Braithwaite and Braithwaite 2001; Matsueda 1992; Sherman 1993). They introduced the concept of reflected appraisals as an intervening mechanism between the application of a label and behavior in conformity with the label.
They suggest that through role-taking during interaction, people read the responses of others to the behaviors in which they engage while holding a par- ticular identity. Actual appraisals (labels) by others are communicated using both language and behavior. The actions and expressions of others are perceived by the self (reflected appraisals) to provide meaningful feedback about how one’s identity is coming across in the situation and is used to guide behavior accordingly (Kinch 1963).
Matsueda used data from the National Youth Survey (NYS) to test the influ- ence of reflected appraisals of parents, teachers, and peers on delinquent behavior. His results provide empirical support for the hypothesis that reflected appraisals do affect delinquent behavior and thus establish a connection between reflected appraisals and delinquency using a cross-sectional study. His work also suggests that reflected appraisals from different sources (parents, teachers, peers) may impact behavior differently for each individual. He does not, however, examine the relationship between reflected appraisals and self-views to assess the mecha- nism through which self-views might be impacted by others’ views.
We turn now to examine the approach of identity theory, which explains why examination of the influence of reflected appraisals on self-views is necessary to provide examination of a more complete set of identity processes than just reflected appraisals.
Identity theory also grows out of symbolic interaction, especially its structural ver- sion (Burke and Stets 2009; Stryker 1968). While we recognize there are several strands of symbolic interaction approaches to identity including those of Blumer (1968), Couch, Saxton, and Katovich (1986), McCall and Simmons (1978), and Stryker (1968), we understand that these approaches share common concepts such as the idea that both society and individual “selves” are formed and maintained via social interaction (Blumer 1980; Stryker 2002). Consistent with the structural symbolic interactionist framework applied here, we use a definition of the self as consisting of the multiple identities an individual may hold at any given time (Stryker 2002).
SOP5402_03.indd 165 5/23/11 2:34:49 PM
166 SOCIOLOGICAL PERSPECTIVES Volume 54, Number 2, 2011
Within identity theory, an identity is understood as the set of meanings applied to the self in a particular social position in society (Burke 1980; 1991; Stryker 1968). These identity meanings are contained in what is called the identity standard (Burke 2004; Burke and Cast 1997; Tsushima and Burke 1999). This is true for all identities, including role identities such as professor and truck driver, as well as “deviant” identities such as criminal or drug user. From an identity theory per- spective, identities operate by comparing perceptions about how one is coming across in the situation (based on both direct as well as reflected appraisals from others) to the identity standard.
If there are any differences between the defining identity standard meanings and the perceptions of how one comes across in the situation, an error, or dis- crepancy, results that serves as a guide for behavior. By acting to reduce the error or discrepancy to zero, people act to alter the perceived self-relevant meanings in the situation so that they match the self-meanings in their identity standard. This matching verifies their identity by making the (perceived) meanings in the situation consistent with the way they define themselves (a process known as self- verification). The output behavior in the situation is thus a function of the discrep- ancy or error between perceptions and standards and is motivated by the desire for self-verification.
Identity theory is thus a perceptual control theory that states people use their behavior to control their perceptions. They act in whatever ways they can to change the meanings in the situation so that their perceptions of self-relevant meanings in the situation (including the reflected appraisals) come to match the meanings in their identity standard, thus verifying their identity.
In this way identities are active agents rather than passive receptors. They act to defend identity meanings by counteracting attributions and labels that are inconsistent with the identity standard. In order for labels to change the identity (become internalized), the normal behavior to counteract the label and maintain the identity must be prevented. We deal with this issue of change next.
We are interested in the capacity for labels as communicated through reflected appraisals to change the criminal identity for incarcerated individuals over time. As suggested earlier, under normal conditions, individuals will adjust their be- havior to reduce any discrepancy between the identity standard and meanings applied by others to the self in the situation by acting to change the way others see the self. In this way, the identity is maintained over time. However, identity theory suggests that in cases when efforts to alter perceived situational meanings do not result in reducing the discrepancy between the input and the identity stan- dard (for example, if the person has little power or few resources to alter the situ- ation), the identity standard itself will slowly change over time in the direction of the reflected appraisals (Burke 1991; Burke and Stets 2009). In this way, reflected appraisals can affect identities. Specific to this study, the context of incarceration presents a unique power differential in which inmates may be more subject to
SOP5402_03.indd 166 5/23/11 2:34:49 PM
Does Incarceration Change the Criminal Identity? 167
change the identity standard in accordance with the views of others due to institu- tional norms and restrictions on interactions.
Based upon this discussion, we present the following initial hypothesis:
Hypothesis 1: Reflected appraisals will influence self-views.
The identity protection/verification processes are also disrupted if the person is removed from his or her normal interaction partners and the mutual verification contexts (Burke and Stets 1999) that help to maintain the stability of self-views against possible influences (Ichiyama 1993). Recent observations with respect to the identities of members of terrorist groups demonstrate how changes in identities may occur as a result of being moved out of their normal interactional environ- ments (Arena and Arrigo 2000; 2004; 2005; 2006)
We contend that when persons are in the context of incarcerated criminal offenders, with few resources or power, in a controlled environment, and away from their normal interaction partners, the reflected appraisals of others may be highly influential for their identity. This is because not only are specific interactions governed and monitored by the criminal justice system, but also the environment itself limits the number of different others with whom one can interact and limits the number of identities an incarcerated individual can invoke. Stryker and Serpe (1982) suggest that when interaction options are limited the interaction that is possible becomes highly relevant in shaping one’s identity. Examples of such restrictions include having a limited number of possible identities to play out, a limited number of interaction partners, and a limit to the amount of interaction that can occur. All of these conditions exist in the context of incarceration.
Asencio (2011) finds that interaction occurring in the context of an incarceration facility influences the order in which the views of others impact identity. While prior identity research outside of the incarceration context demonstrates that the views of significant others are the most influential for identity, Asencio’s results find that within the incarceration facility the views of other inmates are the most influential for identity. Restrictions on interaction in an incarceration facility make identities such as the criminal identity highly salient as all interaction is centered on the fact that one is in jail for having been convicted of committing a crime. These restrictions may also make those with whom incarcerated participants are permitted to interact highly influential for identities as there are little or no other individuals available to provide feedback about the self such as that obtained through reflected appraisals (Asencio 2011).
At the same time, as identity theory makes clear, the self-verification process does operate as the self tries to change the (perceived) meanings around him or her to support the meanings contained in the identity standard, so that change, even when it occurs, is not entirely one sided. From the identity theory perspec- tive, therefore, we hypothesize that while reflected appraisals have an effect on the self, the self also influences reflected appraisals through attempts to control the meanings in a situation.
Hypothesis 2: Self-view and reflected appraisals mutually influence each other over time.
SOP5402_03.indd 167 5/23/11 2:34:49 PM
168 SOCIOLOGICAL PERSPECTIVES Volume 54, Number 2, 2011
IDeNtIty ChANge AmONg INCARCeRAteD CRImINAl OffeNDeRS
This study uses a sample of incarcerated criminal offenders who are participating in a correctional drug rehabilitation program aimed at changing how they view themselves. We focus on three identities held by the criminal offender. First is the criminal identity. An individual who views him or herself as a criminal is more likely to engage in criminal behavior as a method of maintaining and verifying this identity. In the case of incarcerated offenders, the criminal identity is likely to be salient as it is the cause of incarceration. However, opportunities for counseling and/or training while in the correctional facility (such as the drug rehabilitation program of which the participants for this study were a part) make it possible for the criminal self to diminish over time, leading to a corresponding reduction in criminal behavior, as prior work suggests (Giordano, Cernkovich, and Rudolph 2002; Maruna 2001).
Successfully working through programs of this nature may present new possibili- ties to offenders for a different type of lifestyle upon release. Such opportunities rep- resent the application of a positive label, as opposed to the criminal label, which has negative connotations. Correspondingly, others should treat those categorized as successful program participants as though they have the potential to be productive members of society in contrast to counterparts in the criminal justice systems who do not have such opportunities. Prior work examines identity change for homeless individuals (many of whom are former criminal offenders) transitioning through a “recommunalization” process. Recommunalization is an eight-stage process in which participants are housed in a group setting and resocialized to become active participants in conventional society. Qualitative studies of this approach suggest that successful navigation through programs aimed at changing participants’ self- views contributes to the possibility of identity change among “disenfranchised” groups such as criminal offenders (Arrigo and Takahashi 2006; 2007). Because of the nature of the program under study in the current research, we hypothesize that the degree to which the participants see themselves as criminal will diminish over time under the conditions of the present study as described earlier.
Hypothesis 3: The criminal identity will weaken over time.
Since the population under study consists of drug offenders who are in a cor- rectional drug rehabilitation program, we anticipate the drug user identity is also highly salient for this group. Additionally, we anticipate that since the curriculum of the correctional drug rehabilitation program is specifically focused on using cognitive-behavioral motivation techniques to change the way these individuals see themselves, the drug user identity will change over time for these partici- pants. Because of the nature of this program, we hypothesize that the degree to which the participants see themselves as drug users will diminish over time.
Hypothesis 4: The drug user identity will weaken over time.
Finally, we look at the worker identity as an example of a non-deviant iden- tity that is held by most adults and has high relevance in society. This identity
SOP5402_03.indd 168 5/23/11 2:34:49 PM
Does Incarceration Change the Criminal Identity? 169
represents one’s ties to conventional society. If an individual does not view him- or herself as a worker, then he or she is less likely to consider using conventional means of accomplishing goals. Incarcerated offenders are not likely to see them- selves as workers in the traditional sense, as they are not employed while serving jail time. However, they are required to complete chores on a daily basis and some are actually assigned jobs such as kitchen and laundry detail while incarcerated. We anticipate the worker identity should become stronger over time. As these in- dividuals begin to see themselves as less of a deviant in the sense of criminal and drug user, they may begin to see themselves as more a part of conventional soci- ety, adopting more conventional self-views such as worker as they get closer to release.
Hypothesis 5: The worker identity will strengthen over time.
DAtA AND methODS
the Research Setting
Data were collected from a sample of incarcerated offenders at a Southern Cali- fornia medium-security correctional facility. Participants for the study were par- ticipating in a court-ordered, six-month correctional substance abuse treatment program. All were incarcerated for a drug-related offense and have no history of violence or other problem behaviors during previous incarceration periods.
After passing a security clearance, one researcher was allowed to enter the prem- ises while accompanied by a member of the administrative team for the facility. The researcher was allowed to administer the survey to inmates during a routine class time. Participants were seated at desks in their normal classroom configuration and the researcher explained that the voluntary survey was to collect data about how they see themselves. It was explained that the survey was confidential, and participants were asked to sign an informed consent sheet attached to the survey. The researcher assured participants that their responses would not be shared with the jail or other employees of the criminal justice system and that their responses could not be used against them for any reason. It is possible that participants had doubts about whether their responses would be used against them, however they were not asked to discuss any illegal activities in the survey. All participants were offered an alternative activity if they chose not to participate in the survey. The alternative consisted of the normally scheduled curriculum involving indepen- dent study through a series of workbooks associated with the cognitive-behavioral modification program. Only one chose the alternative. The participants were given approximately 30 minutes to complete the survey. The survey was administered at three different time points approximately eight weeks apart (referred to as t1, t2, and t3).
The total number of respondents included in the study is 124. The group consists of 62 males and 62 females with a mean age of 29.5. The sample’s ethnic diversity is reflective of the geographic location (Southern California) and is made up of ap- proximately 52 percent whites, 36 percent Hispanics, 5 percent African-Americans,
SOP5402_03.indd 169 5/23/11 2:34:49 PM
170 SOCIOLOGICAL PERSPECTIVES Volume 54, Number 2, 2011
3 percent American Indians, and 3 percent Other based upon self-report data. Roughly 35 percent of those included in the study completed high school or passed an equivalency exam.
The program components include daily counseling, life skills training, and cognitive-behavioral modification techniques aimed at providing offenders with alternatives to drugs upon release. The participants in this program are housed together, are kept separate from the rest of the incarcerated population at this cor- rectional facility, and attend classes and counseling sessions. They also interact with one another more frequently than the offenders who are not in the program. They perform daily chores in cooperation with one another and interact as a group of students might when in the classroom setting. The program encourages them to provide moral and emotional support for one another in the quest to overcome drugs and come “clean.” Because of this, these particular inmates tend to have a greater camaraderie among them than inmates in the general jail population do. This means that they are more likely to interact regularly and have some influence on one another’s reflexive self-views.
These inmates interact with the guards at the jail on a daily basis. However, the County Sheriff’s Department consistently rotates the guards assigned to the correctional facilities so no personal relationships can be established between the guards and the inmates. This means that the interactions between offenders and guards are more likely to follow patterns set forth by expectational norms than by personal experiences. The inmates are also allowed to have visits from significant others once a week. The friends and families of these offenders may come to the facility to visit during specified visiting hours one day a week.
Identities. The measure of how the respondent sees him- or herself with respect to a particular identity comes from items on the survey in which respondents are asked to rate themselves on a Likert-type scale on various characteristics that correspond to different identities of interest. Items were coded such that a higher score for each set of characteristics presented in the survey indicates that the identity strength for the corresponding identity under study is greater than a lower score. The characteristics on the self-view scale are the same characteristics used consistently throughout the survey instrument as indicators of these same identities. The indicators were selected using a combination of theoretical relevance and informal interviews with an avail- able sample of participants from the correctional facility to ensure that they were meaningful to the sample. Each identity was measured on a 5-point scale ranging from 1 to 5 where each end of the scale was anchored with a descriptor of the identity. For example, the “criminal” identity was anchored with 1 being a “law-abiding per- son” and 5 being an “unlawful person.” The “drug user” identity was anchored with 1 being a “non-drug user” and 5 being a “drug user,” and the “worker” identity was anchored with a 1 being a “non-worker” and a 5 being a “worker.”
Reflected Appraisals. Because reflected appraisals represent how the self thinks others see the self, a measure of reflected appraisals requires one to think about the self as an object. In this study, we measure reflected appraisals by asking
SOP5402_03.indd 170 5/23/11 2:34:49 PM
Does Incarceration Change the Criminal Identity? 171
participants to report, on a Likert-type scale like the one used to rate the self, how particular others see them. Specifically, we asked respondents how significant others (including a spouse or other important family members and/or friends), peers (others with whom they are incarcerated), and guards at the jail would rate them on the various identities under study. Each was measured on the same 5-point scale used to measure self-views.
In addition to the measures of reflected appraisals for each of the significant others, peers, and guards, we created a combined measure of reflected appraisals that averaged the reflected appraisals of both significant others and peers. These two sources were highly correlated, and the effects differ from the effects of the reflected appraisals from guards, so creating a combined measure increases the reliability of the measure of reflected appraisals over the single-item appraisals of significant others and peers separately. This was done for each of the three identities of criminal, drug user, and worker.
Because the current study uses a longitudinal design, we are able to sepa- rately assess the effect of the identity on the reflected appraisals (self-verification process) and the effect of reflected appraisals on the identity (labeling process). Because measures are examined at three points in time, we are also able to assess the influence of an identity at one time on itself at a later time (indicating internal- ization). We can also examine the influence of reflected appraisals at one time on the appraisals at a later time (the persistence effects) as shown in Figure 1.
Time. The length of time the respondent has been incarcerated for the current offense was retrieved from court records. Participants are required to complete six months in the program before release; however, the program is ongoing such that new people start each day. This variable is reported in weeks, which allows us to assess the way that self-views and reflected appraisals change over time.
To test Hypothesis 1, that reflected appraisals influence self-views (replicating earlier results), we use linear fixed-effect models (West, Welch, and Galecki 2007).
ref app 1
ref app 2
ref app 3
e2 e3 e4
fIgURe 1 Model of the Relationship between Identities and Reflected Appraisals over Time
SOP5402_03.indd 171 5/23/11 2:34:51 PM
172 SOCIOLOGICAL PERSPECTIVES Volume 54, Number 2, 2011
This procedure estimates the fixed-effect parameters for the effects of reflected appraisals on identity within persons allowing for random intercepts across per- sons. For this analysis, we also include the effects of time in program on identities in order to control for the possible spurious effects of changes over time in both the reflected appraisals and identities. We also examined the influence of gen- der, ethnicity, age, and criminal history of participants on identity and reflected appraisals and found no significant effects. The results of these additional analyses are not presented here but are available from the authors.
The effects of time in the program on the reflected appraisals used to test Hypotheses 3 (the criminal identity will weaken over time), 4 (the drug user identity will weaken over time), and 5 (the worker identity will strengthen over time) are also estimated using the procedure described earlier. The linear trends in identities over time were estimated using OLS regression using time period (t1, t2, and t3, coded as 1, 2, and 3) as the independent variable. Again, we allowed the errors to be correlated among different observations of the same respondents.
To test the second hypothesis, that self-views and reflected appraisals mutually influence on another, we use structural equation methods assuming the model shown in Figure 1. This explicitly examines the within- and between-time effects and allows us to sort out the potential mutual influence of reflected appraisals on the identity and the identity on the reflected appraisals. Although the mutual effects between identities and reflected appraisals are shown as occurring at a point in time, these effects may be viewed as cumulated effects occurring over the prior time period. We also examine an alternative model with cross-lagged effects over time.
Our first hypothesis concerns the effects of reflected appraisals on one’s identity. Resulting coefficients of the within-person, linear fixed-effects model analysis are presented in Table 1 and may be interpreted as regular standardized regression coefficients. We see in Part 1 of Table 1 that, consistent with Hypothesis 1, reflected appraisals from significant others influence identity for each of the three identi- ties. The greater the degree to which reflected appraisals from significant others indicate a criminal identity, the greater the strength reported for the self-views on the criminal identity. There are similar effects for the drug user identity and the worker identity. Also consistent with Hypothesis 1, we have similar effects of the reflected appraisals from peers (others in the jail). The higher the reflected ap- praisals from peers for a particular identity (criminal, drug user, or worker), the greater the strength of the self-view for the particular identity. Finally, we find that reflected appraisals from the guards have no effect on identity for any of the three identities tested. The guards, it would appear, are not relevant others for defining any of the three examined identities of the respondents in this study.
We also note in the table that the length of time one has been in the program in- fluences two of the identities. Consistent with Hypotheses 3 and 4, the longer one is in the program, the less strength the criminal self-view has (Hypothesis 3), and
SOP5402_03.indd 172 5/23/11 2:34:51 PM
Does Incarceration Change the Criminal Identity? 173
the longer one is in the program, the less strength the drug user self-view has (Hy- pothesis 4). However, there is no effect of time in the program on the worker iden- tity (Hypothesis 5). In Part 2 of Table 1, we see the effects of the pooled reflected appraisals strongly mirror the separate effects of significant others and peers.
Table 2 continues this part of the model, showing the effects of time in the pro- gram on the reflected appraisals. Here we see that not only did time in the pro- gram reduce the strength with which respondents saw themselves as criminal or as a drug user (Table 1), but also time in the program reduced the degree to which the reflected appraisals of both significant others and of peers indicated a criminal or drug user identity. Finally, the longer one was in the program, the more re- flected appraisals from peers indicated a strong worker identity. This result should be considered with caution since there were no effects on the reflected appraisals of significant others or the reflected appraisals of the guards and there was no ef- fect on the pooled measure.
Putting these two effects of time in the program and reflected appraisals together, Table 3 summarizes the changes in strength of the mean self-view of these offenders over time for each of the identities. Both the criminal identity and the drug user identity diminish over time with significant linear trends (values range from 1, indicating a low score for the identity, to 5, indicating a high score). The mean strength of the criminal identity is 2.53 (slightly below the midpoint on the 5-point scale) at t1, 2.16 at t2, and 1.78 at t3. Overall, there is a significant lowering in the
tAble 1 Standardized Coefficients from Cross-Sectional Time-Series Fixed-Effects Regression
Showing the Effects of Reflected Appraisals and Time-in-Program on Identities Tables
Part 1: Effects of Separate Sources
Views of: Criminal Drug User Worker
Significant others 0.28** 0.49** 0.36** Peers 0.29** 0.29** 0.32** Guards ns ns ns Time-in-program –0.13* –0.17** ns
R2 0.30 0.50 0.29
Part 2: Effects of Pooled Sources
Criminal Drug User Worker
Reflected appraisals 0.48** 0.63** 0.54** Time-in-program –0.12* –0.17** ns
R2 0.28 0.49 0.36
*p ≤ .05; **p ≤ .01.
SOP5402_03.indd 173 5/23/11 2:34:51 PM
174 SOCIOLOGICAL PERSPECTIVES Volume 54, Number 2, 2011
strength of the criminal identity of .38 from t1 to t3. The mean strength of the drug user identity is 2.71 (also slightly below the midpoint of the scale) at t1, 1.83 at t2, and 1.61 at t3. Overall, there is a significant lowering in strength of the drug user identity of .58 from t1 to t3. The worker identity, however, does not show the same trends over time. Neither self-views or reflected appraisals with respect to one’s worker identity seem to change over time. The mean strength of the worker identity is 3.32 (which is above the midpoint of the scale and significantly stronger than either the criminal or drug user identities) at t1, 3.48 at t2, and 3.39 at t3, with none of the slight changes between t1 and t3 being statistically significant. The initial self-view for the worker identity seems simply to be maintained at whatever level it started.
The above analyses replicate prior conceptualizations and treat identity simply as a passive outcome of reflected appraisals (and time in program). This analysis does not recognize that the self may act to change others’ view by changing the mean- ings in the situation as identity theory asserts. In the following analyses, we use a structural equation approach that allows us to test the more complicated identity
tAble 2 Standardized Regression Coefficients Showing the Effects of Time-in-Program
on Reflected Appraisals
Criminal Reflected Appraisals By:
Significant Others Peers Guards Pooled Measure
Time-in-program –0.19** –0.21** ns –.20**
Drug User Reflected Appraisals By:
Significant Others Peers Guards Pooled Measure
Time-in-program –0.25** –0.23** ns –.28**
Worker Reflected Appraisals By:
Significant Others Peers Guards Pooled Measure
Time-in-program ns 0.17* ns ns
*p ≤ .05; **p ≤ .01.
tAble 3 Mean Self-Ratings on Each Identity Over Time
Time Point Criminal Drug User Worker
t1 2.53 2.71 3.32 t2 2.16 1.83 3.48 t3 1.78 1.61 3.39
Linear trend –0.38** –0.58** ns
**p ≤ .01.
SOP5402_03.indd 174 5/23/11 2:34:52 PM
Does Incarceration Change the Criminal Identity? 175
processes suggested by identity theory. Here, we examine the extent to which each identity is a function of reflected appraisals and its own independent structure over time, as well as the extent to which each identity is actively defended through the self-verification process. The model for each of the identities includes both self- views and the combined reflected appraisals measure and is illustrated in Figure 1.
This model is identified because cross-lagged effects (e.g., from self-view at Time 1 to reflected appraisals at Time 2) are not included. The reciprocal effects (labeled “c” and “d”) in the figure reflect the cumulative effects of each of the variables on the other over the eight-week period. An alternative specification that includes the cross-lagged effects but leaves out the reciprocal effects to achieve identification was also estimated. Overall, this model did not have a good fit with the data.
One note about the estimates of these models: In each case, the reciprocal effects at Time 2 are not significantly different from the reciprocal effects at Time 3 and are therefore constrained to be equal in the estimation process. Similarly the per- sistence effects from Time 1 to Time 2 are not significantly different from the effects from Time 2 to Time 3. Therefore, these are constrained to be equal in the estima- tions. These constraints are suggested because of the lack of theoretical reasons to expect differences over time or at different points in time.
The results of this analysis are presented in Table 4. We consider each of the iden- tities in turn. First, with respect to the criminal identity, we see that the criminal identity is influenced by reflected appraisals (Path D, beta = .41), but the reverse is not true (Path C). In this context, the strength of the criminal identity is determined by (perceptions of) others’ appraisals (as well as other effects summarized in the error or residual term). At the same time, the criminal identity does have its own independent structure and persists over time, as do the reflected appraisals. These persistence effects (Paths A and B, .43 and .70 in the table) are not particularly strong. This indicates that there are changes over time of both reflected appraisals and the criminal identity, but they also show some stability over time. This model fits very well as seen by the small chi-square/df values at the bottom of the table,
tAble 4 Standardized Coefficients from Structural Equation Model Showing Mutual Effects
of Identities and Reflected Appraisals over Time
Criminal Drug User Worker
Path in Model
Identity Time t
Ref App Time t
Identity Time t
Ref App Time t
Identity Time t
Ref App Time t
A Identity, time t-1 .43** — .00 a — .46** — B Ref App, time t-1 — .70** — .43** — .00 a
C Identity, time t — .00 a — .00 a — .38** D Ref App, time t .41** — .81** — .00 a — Chi-square/df 14.9/16 17.1/17 23.0/17 Alternative cross-lagged model Chi-Square/df 16.0/16 58.3/17 36.7/17
a Constrained to zero. **p ≤ .01.
SOP5402_03.indd 175 5/23/11 2:34:52 PM
176 SOCIOLOGICAL PERSPECTIVES Volume 54, Number 2, 2011
which indicates little difference between the model and the data: No other paths or restrictions are necessary to achieve a good fit. The poor fit of the alternative model with cross-lagged effects is also given at the bottom of the table.
Turning to the drug user identity, we see again that the strength of this identity is influenced by the reflected appraisals (Path D, beta = .81). Again, the reverse is not true (Path C). Indeed, the drug user identity, while strongly influenced by the reflected appraisals, seems to have no independent structure in itself. The persis- tence of the drug user identity is not different from zero (Path A). Apparently, the drug user identity is fully recreated at each point in time by the current reflected appraisals. The strength of this identity simply mirrors what the respondent feels others think of him or her with no internalization or commitment to the drug user identity. The reflected appraisals, on the other hand, do have a structure of their own with a persistence coefficient of .43 (Path B). That is, what the respondent feels others think does persist over time. Again, the model fits the data very well as indicated by the chi-square value. The poor fit of the alternative model with cross-lagged effects is also given at the bottom of the table.
Finally, with respect to the worker identity, we see a third pattern. Here, the worker identity influences the reflected appraisals (Path C, beta = .38), but the worker identity is not influenced by the reflected appraisals (Path D). This appears to show the verification process at work to counteract any changes in the reflected appraisals to make them different from the worker identity standard. The worker identity has its own independent structure with a significant persistence (Path A, beta = .46), but the reflected appraisals have no persistence (Path B) and are thus created anew by self-views at each point in time. These results provide a better understanding of the lack of change in the worker identity over time, as shown in Table 3. The active self posited in identity theory that is revealed in the results of Table 4 is working to maintain itself over time. It is also working to change mean- ings in the environment in order to verify itself, rather than change to conform to meanings provided by others. The poor fit of the alternative model with the cross- lagged effects is again given at the bottom of the table.
Overall, these results show only partial support for the mutual effects sug- gested in Hypothesis 2. For the criminal and drug user identities, the effects go from reflected appraisals to self-views, consistent with our earlier analysis and suggestions in the literature. For the worker identity, however, the effects go from the identity meanings to reflected appraisals as the participants seek to maintain self-verification by creating situations in which the views of others are consistent with self-views. While Cast, Stets, and Burke (1999) found mutual influence for the spousal identity between self-views and reflected appraisals over time (though the relative strength of each was a function of power differentials), we find effects for the identities we studied to be in one direction or the other, but not both.
We began by reviewing problems that have been raised in the literature about labeling theory concerning the question of why there is individual variability in the response to labels. It has been suggested that there is some intervening process
SOP5402_03.indd 176 5/23/11 2:34:52 PM
Does Incarceration Change the Criminal Identity? 177
between the application of a label and the internalization of the label and subse- quent behavior in accord with the label. For example, Matsueda et al. (Bartusch and Matsueda 1996; Heimer and Matsueda 1994; Matsueda 1992; Matsueda and Heimer 1997) went back to symbolic interaction theory and suggested that the reflected appraisal process lay between the application of the label and behavior that accounted for this variability.
We then introduced some ideas from identity theory (Burke and Stets 2009), which suggests that reflected appraisals are only part of the full set of identity processes that might intervene between application of a label and behavior with respect to the label. In identity theory, the self-view plays a much more active role that generally rejects reflected appraisals that differ from the identity stan- dard working to make them more like the standard. In identity theory, reflected appraisals that differ from the identity standard become internalized only if the person lacks the power and resources to alter them and to verify the identity or if the person desires to change his or her identity. Thus, reflected appraisals can influence identity (consistent with labeling theory) and the identity can influ- ence reflected appraisals (consistent with the verification process discussed in identity theory). Our findings show both processes at work and demonstrate the potential value of identity theory in explaining differential outcomes for people who are labeled.
We examined three identities among incarcerated criminal offenders: the crimi- nal identity, the drug user identity, and the worker identity. Our results suggest that the internalization of reflected appraisals is dependent upon the identity at issue and the source of the reflected appraisal. We showed that the strength of the criminal identity and the drug user identity (both deviant identities) were influenced by the reflected appraisals of significant others and peers (though not the guards), but the verification process in which these identities influence the reflected appraisals was absent. The worker identity showed the effects of the verification process but was not influenced by the reflected appraisals that were measured. The initial strength of the worker identity is greater than that of the criminal or the drug user identity, which may explain why this identity appears to be less influenced by reflected appraisals. The criminal identity and worker identity demonstrated some stability over time, indicating these identities had been internalized. However, the drug user identity was completely built on the reflected appraisals of others but evidenced no internalization or stability over time. Thus, for each of the identities, we saw a different pattern to the identity processes that built or maintained the identity.
Before discussing these results further, we should remind the reader that while we have embedded our understanding in the overall labeling process, we do not have a measure of the label itself, nor do we have a measure of behavior resulting from any of these identities. We do know the nature of the rehabilitative program and the way in which it is structured. We know that the main components of the program consist of counseling and education aimed at accomplishing socio- cognitive changes in participants’ views of themselves in an attempt to accom- plish behavioral changes. Specifically, the program aims to assist participants in going from non-conventional views of themselves (criminal, drug user) to more
SOP5402_03.indd 177 5/23/11 2:34:52 PM
178 SOCIOLOGICAL PERSPECTIVES Volume 54, Number 2, 2011
conventional views of themselves (non-criminal, non–drug user). This provides us with some sense of the labels being applied, but all we know for sure is what the respondents perceived (reflected appraisals). The outcome behavior is not mea- sured, but numerous studies of the impact of identities on behavior show that people generally behave in accord with the meanings in their identity standards (Burke and Stets 2009). Thus, we are only looking at the potential for identity processes to mediate the application of a label and subsequent behavioral outcomes that help explain individual variability in the effect of labeling on behavior.
In the present study, we find that the criminal identity behaves as Matsueda et al. suggest. It is a function of the reflected appraisals that are internalized. In the context of the correctional rehabilitation program in which this study took place, it appears that the verification processes are at a minimum for the criminal identity. It is not defended but absorbed in the sense that discrepancies between the reflected appraisals for the criminal identity are integrated into the self-view rather than an adjustment to the identity feedback loop being made in an effort to compensate for the discrepancy. While this pattern of simply taking in the reflected appraisals has not been found in prior research, being influenced by reflected appraisals has been found to occur when persons are lower in power (Cast, Stets, and Burke 1999) or when persons are desiring to change their identities (Ichiyama 1993; Kiecolt 1994), the latter of which seems appropriate in the present case.
Like the criminal identity, the drug user identity is also a function of the reflected appraisals of peers and significant others. Unlike the criminal identity, however, it apparently is not internalized and has no persistence over time. It is as if the drug user identity is personally meaningless, and in this sense perhaps it is not an identity at all.
The worker identity is internalized and has persistence over time. It is also defended in the sense that people act to change reflected appraisals to match their worker identity standard, but there is no effect of the reflected appraisals from significant others and peers on the worker identity. Its source, therefore, is outside the context of the correctional facility. We also saw that there is no change in the worker identity over time while the respondents are in the correctional facility, and that could be a problem. If the respondents do not increasingly see themselves as workers, they may encounter problems upon release. It would seem that as the criminal identity decreases, something like the worker identity ought to increase to takes its place.
The one-way effects between reflected appraisals and identities for the three identities studied thus represent something not seen in identity theory studies in the past where the mutual influence of identities and reflected appraisals were always evident (e.g., Cast, Stets, and Burke 1999). The extent to which these results are an effect of the content (deviant identities) or context (incarcerated individuals) is something that needs to be investigated in the future.
We noted that the different sources of reflected appraisals (significant others, peers, and guards) have different effects on the identities of the respondents. We suggest this is due in part to the nature of the relationships with these particular others given the context of incarceration. We see that the reflected appraisals of both peers and significant others are highly relevant for all three identities. Since
SOP5402_03.indd 178 5/23/11 2:34:52 PM
Does Incarceration Change the Criminal Identity? 179
the guards at the jail are not likely to interact with participants on a regular basis due to the rotation schedule, it is likely that participants do not consider what the guards think of them to be relevant to how they see themselves. These results are consistent with the idea that others who are not close to the self have less influence on the self-view, which identity theory also asserts (Burke and Stets 2009). It is also consistent with the logic of the criminal justice system for rotating the guards in the first place, to avoid the establishment of any personal relationships between the guards and the offenders. Labeling theory should also take into account that the source of the label plays a big role in its effectiveness.
This type of study offers an additional assessment tool for evaluation of cor- rectional rehabilitation programs and may assist in yielding more clear results for these studies. Since the goal of these programs is cognitive-behavioral modifica- tion, measuring whether cognitions about the self (identities) are actually modi- fied by such programs is an effective way to evaluate success. Future evaluation studies of cognitive-behavioral modification programs should move toward using a control group of incarcerated offenders not in the program to help in assessing the impact of program curriculum on the self-views of participants, rather than focusing strictly on behavioral outcomes as a way of measuring success.
We acknowledge several limitations with this study that have implications for future research. Our sample is limited only to individuals who are in a correctional rehabilitation program. While these participants provide a great deal of informa- tion about the identity processes of others in similar programs, future research (specifically on the drug user identity) should include comparison of the drug user identity between individuals who are in a correctional rehabilitation program and individuals who are in a voluntary rehabilitation program. This comparison will allow researchers to see if reflected appraisals have as much weight for those who have identified themselves as drug users prior to seeking treatment as they did for those who were designated drug users by others in this study.
Our data set does not allow for follow-up of participants after they have been released from incarceration. Future studies focused on the criminal identity from an identity theory perspective should include follow-ups with those who are re- leased after successful sentence completion to assess the criminal self-view at mul- tiple time points after reintegration into society, as well as its impact on criminal behavior patterns.
Our study does not allow for comparison with other groups. Future studies of this nature should include comparison groups in order to assess the differences in the criminal identity between an incarcerated population in a correctional rehabili- tation program and an incarcerated population not in a correctional rehabilitation program. It is possible that the criminal identity may operate differently for these two populations. Adding such a comparison group would also allow for assess- ment of the influence of the program curriculum itself on changes in identities, as would collecting data from participants as they enter the program in order to establish a baseline identity by which to assess changes. An additional comparison
SOP5402_03.indd 179 5/23/11 2:34:52 PM
180 SOCIOLOGICAL PERSPECTIVES Volume 54, Number 2, 2011
of the criminal identity for those who reoffend versus those who do not reoffend might offer some insight into the issue of recidivism and perhaps provide some policy implications for reducing the rate of reoffending.
Overall, our results suggest that the application of identity theory is useful in understanding how labels applied by others may or may not result in criminal identities. Identity theory allows for consideration of the role of the active self when internalizing the views of others and provides a better understanding of the connection between labels and behavior. Since the self varies from individual to individual and opportunities for identity verification vary from situation to situa- tion, it is conceivable that labels result in differential behavior among those labeled in the same way because the self and the situation play themselves out differently for each person.
The findings presented here demonstrate the potential role for identity theory to explain why not all of those who are labeled as criminal will behave accordingly. Identity theory, with its understanding of an active self, provides an empirically testable approach for understanding when and how a label may become internal- ized as part of an identity. We should no longer ignore the active role of the self in the maintenance and change of the criminal identity under conditions of labeling.
Acknowledgments: The authors would like to acknowledge the Social Psychol- ogy Seminar (2009) at the University of California, Riverside, for review of a previ- ous version of this article.
Arena, Michael P. and Bruce A Arrigo. 2000. “White Supremacist Behavior: Toward an In- tegrated Social Psychological Model.” Deviant Behavior: An Interdisciplinary Journal 21(1):213–41.
——— and Bruce A. Arrigo. 2004. “Identity and the Terrorist Threat: An Interpretive and Explanatory Model.” International Criminal Justice Review 14(1):124–63.
——— and Bruce A. Arrigo. 2005. “Social Psychology, Terrorism, and Identity: A Prelimi- nary Re-examination of Theory, Culture, Self, and Society.” Behavioral Sciences and the Law 14(23):485–506.
——— and Bruce A. Arrigo. 2006. The Terrorist Identity: Explaining the Terrorist Threat. New York: NYU Press.
——— and Yoshiko Takahashi. 2006. “Recommunalization of the Disenfranchised.” Theo- retical Criminology 10(3):307–36.
——— and Yoshiko Takahashi. 2007. “Theorizing Community Reentry for Male Incarcer- ates and Confined Mothers: Lessons Learned from Housing the Homeless.” Journal of Offender Rehabilitation 46(1):133–62.
Asencio, Emily K. 2011. “Familiarity, Legitimacy, or Frequency: Which Matters Most for the Criminal Identity?” Sociological Inquiry 81(1):34–52.
Bartusch, Dawn Jeglum and Ross L. Matsueda. 1996. “Gender, Reflected Appraisals, and Labeling: A Cross-Group Test of an Interactionist Theory of Delinquency.” Social Forces 75(1):145–76.
SOP5402_03.indd 180 5/23/11 2:34:52 PM
Does Incarceration Change the Criminal Identity? 181
Becker, Howard S. 1963. The Outsiders:Studies in the Sociology of Deviance. New York: The Free Press.
Blumer, Herbert. 1968. Symbolic Interactionism. Englewood Cliffs, NJ: Prentice-Hall. ———. 1980. “Mead and Blumer: The Convergent Methodological Perspectives of Social Be-
haviorism and Symbolic Interactionism.” American Sociological Review 45(1):409–19. Braithwaite, John and Valerie Braithwaite. 2001. “Revising the Theory of Reintegrative
Shaming.” Pp. 39–57 in Shame Management Through Reintegrative Shaming, edited by N. H. E. Ahmed, J. Braithwaite, and V. Braithwaite. Cambridge, UK: Cambridge Uni- versity Press.
Burke, Peter J. 1980. “The Self: Measurement Requirements from an Interactionist Perspec- tive.” Social Psychology Quarterly 43(1):18–29.
———. 1991. “Identity Processes and Social Stress.” American Sociological Review 56(6): 836–49.
———. 2004. “Identities and Social Structure: The 2003 Cooley-Mead Award Address.” Social Psychology Quarterly 67(1):5–15.
——— and Alicia D. Cast. 1997. “Stability and Change in the Gender Identities of Newly Married Couples.” Social Psychology Quarterly 60(4):277–90.
——— and Jan E. Stets. 1999. “Trust and Commitment through Self-Verification.” Social Psychology Quarterly 62(4):347–66.
——— and Jan E. Stets. 2009. Identity Theory. New York: Oxford University Press. Cast, Alicia D., Jan E. Stets, and Peter J. Burke. 1999. “Does the Self Conform to the Views of
Others?” Social Psychology Quarterly 62(1):68–82. Cooley, Charles H. 1902. Human Nature and the Social Order. New York: Scribner’s. Couch, Carl J., Stan L. Saxton, and Michael A. Katovich. 1986. Studies in Symbolic Interaction:
The Iowa School. Greenwich, CT: JAI Press. Cullen, Francis T. and John B. Cullen. 1978. “Labeling Theory and the Empty Castle Phe-
nomenon.” Western Sociological Review 9(1):28-38. ——— and Paul Gendreau. 2000. “Assessing Correctional Rehabilitation: Policy, Practice,
and Prospects.” Pp. 109–75 in Criminal Justice 2000: Volume 3—Policies, Practices, and Decisions of the Criminal Justice System, edited by J. Horney. Washington, DC: U.S. De- partment of Justice, National Institute of Justice.
Giordano, Peggy C., Stephen A. Cernkovich, and Jennifer L. Rudolph. 2002. “Gender, Crime, and Desistance: Toward a Theory of Cognitive Transformation.” American Journal of Sociology 1(7):990–1064.
Gove, Walter. 1980. The Labeling of Deviance. Beverly Hills, CA: Sage. Heimer, Karen and Ross L. Matsueda. 1994. “Role-Taking, Role Commitment, and De-
linquency: A Theory of Differential Social Control.” American Sociological Review 59(2):365–90.
Ichiyama, Michael A. 1993. “The Reflected Appraisal Process in Small-Group Interaction.” Social Psychology Quarterly 56(2):87–99.
Kiecolt, K. Jill. 1994. “Stress and the Decision to Change Oneself: A Theoretical Model.” Social Psychology Quarterly 57(1):49–63.
Kinch, John W. 1963. “A Formalized Theory of the Self-Concept.” American Journal of Sociol- ogy 68(2):481–86.
Lemert, Edwin M. 1951. Social Pathologoy: A Systemic Approach to the Theory of Sociopathic Behavior. New York: McGraw-Hill.
Maruna, Shadd. 2001. Making Good: How Ex-convicts Reform and Rebuild Their Lives. Wash- ington, DC: American Psychological Association.
Matsueda, Ross L. 1992. “Reflected Appraisals, Parental Labeling, and Delinquency Speci- fying a Symbolic Interactionist Theory.” American Journal of Sociology 97(6):1577–611.
SOP5402_03.indd 181 5/23/11 2:34:52 PM
182 SOCIOLOGICAL PERSPECTIVES Volume 54, Number 2, 2011
——— and Karen Heimer. 1997. “A Symbolic Interactionist Theory of Role-Transitions, Role-Commitments, and Delinquency.” Pp. 163–213 in Developmental Theories of Crime and Delinquency. Advances in Criminological Theory, Vol. 7, edited by T. Thornberry. New Brunswick, NJ: Transaction Publishers.
Matza, David. 1964. Becoming Deviant. Englewood Cliffs, NJ: Prentice-Hall. McCall, George J. and J. L. Simmons. 1978. Identities and Interactions. New York: Free Press. Mead, George Herbert. 1934. Mind, Self, and Society. Chicago: University of Chicago Press. Moffitt, Terri E. 1993. “Adolescence-Limited and Life-Course-Persistent Antisocial Behav-
ior: A Developmental Taxonomy.” Psychological Review 100(6):674–701. Sampson, Robert J. and John H. Laub. 1993. Crime in the Making: Pathways and Turning Points
through Life. Cambridge, MA: Harvard University Press. Schur, Edwin. 1973. Radical Non-Intervention: Rethinking the Delinquency Problem. Englewood
Cliffs, NJ: Prentice-Hall. Sherman, Lawrence W. 1993. “Defiance, Deterrence, and Irrelevance: A Theory of the Crimi-
nal Sanction.” Journal of Research in Crime and Delinquency 30(4):445–73. Stryker, Sheldon. 1968. “Identity Salience and Role Performance: The Relevance of Sym-
bolic Interaction Theory for Family Research.” Journal of Marraige and the Family 30(3):558–64.
———. 2002. Symbolic Interactionism: A Social Structural Version. Caldwell, NJ: The Black- burn Press.
——— and Richard Serpe. 1982. “Commitment, Identity Salience, and Role Behavior: A Theory and Research Example.” Pp. 192–216 in Personality, Roles, and Social Behavior, edited by W. I. A. E. Knowles. New York: Springer-Verlag.
Tannenbaum, Frank. 1938. Crime and the Community. New York: Columbia University Press. Tsushima, Teresa and Peter J. Burke. 1999. “Levels, Agency, and Control in the Parent Iden-
tity.” Social Psychology Quarterly 62(2):173–89. West, Brady T., Kathleen B. Welch, and Andrzej T. Galecki. 2007. Linear Mixed Models. Boca
Raton, FL: Chapmen & Hall.
SOP5402_03.indd 182 5/23/11 2:34:52 PM
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.