Participants
Mothers of incarcerated boys
In the present study, interview data on mothers of incarcerated boys are included, as youth detained under criminal law in juvenile justice institutions merely consist of male detainees, and participating parents in this study were predominantly mothers. The boys were consecutively taken into pre-trial detention in 10 (out of 11) juvenile justice institutions in the Netherlands between May 2006 and February 2008; data collection including interviews with the mothers of these adolescents took place during the same time period. In the Netherlands, criminal legislation for youths applies to persons aged 12 to 18. Pre-trial detention is enforced when a youngster is suspect of an offence, awaiting trial, and if detention is thought to be necessary for the protection of others or the adolescent itself. These adolescents were all suspects of one or more offences.
Eligible for inclusion in the present study were those mothers of whom the son remained in a juvenile justice institution and participated in the study. Dutch mothers had to be able to speak and read Dutch, Moroccan mothers had to be able to speak Moroccan-Arabic. All mothers received a letter containing the aims of the present study. Dutch mothers were contacted by telephone to make an appointment for the interview which took place at the participants' home. Moroccan mothers were sent an introductory letter in Dutch and Moroccan-Arabic and within a couple of weeks a trained Moroccan interviewer visited the parents' home to request them to participate.
Data collection took place at the participants' homes, where questionnaires were handed over to the parent. The questionnaires were filled out by the Dutch mothers. For Moroccan mothers, the questionnaires had been translated into Moroccan-Arabic and to check the accuracy of the translation we performed an independent back translation into Dutch. Since Moroccan parents did not have to be able to read Moroccan-Arabic or Dutch in order to participate in an interview, the questions were read aloud and were filled out by the interviewers. Participants were assured of the confidentiality of their spoken and written responses and data were archived anonymously. Moreover, written informed consent was obtained from the participants. Participants received compensation (a gift certificate). The research protocol was approved by the Ethical Board of the Department of Social Sciences of Leiden University and the Ministry of Justice in the Netherlands. For a more detailed description of the data collection procedure see [8].
Initially, two-hundred seventy-three parents, fathers as well as mothers, were asked to participate in the study. Eighty-four parents refused to participate and 22 parents were not found at home by the interviewers. Thus in total, 167 parents were interviewed, which is a total response rate of 61% (response rate Dutch parents 50%, and response rate Moroccan parents 70%). Of these 167 parental interviews, 129 interviews were conducted with the mother and only these were used in the present study. Sixty-six mothers were of Moroccan origin (i.e., she or the father of her son was born in Morocco) and 63 mothers were of native Dutch origin. Since incomplete participation of the parents may have caused some bias in the findings of the present study, we tested if adolescents, whose parents completed the interview, scored significantly different on self-reported internalizing and externalizing problems as measured by the Youth Self-Report (version 1991), than adolescents whose parents did not participate in the study. No differences were found on internalizing problems (F = 0.392, df = 1, p = 0.53) or externalizing problems (F = 1.106, df = 1, p = 0.29).
Moroccan immigrant parents in the general population
Data were used of interviews conducted with mothers originating from a larger study, in which a sample of 1,127 children aged 4 through 18 with at least one parent born in Morocco, were randomly selected from municipal registers of Rotterdam and The Hague. Parents and adolescents were sent an introductory letter in Dutch and Arabic describing the aims of the study and within a couple of weeks a trained Moroccan interviewer visited the respondents' homes to request them to participate. Data collection took place at the participants' homes, where questionnaires were handed over to the parent. The questions were read aloud and filled out by the interviewers. Data collection took place from April 2001 to July 2002. A total of 819 parents participated in the total study (response rate 73%). In the present study mother-reported data of male adolescent participants, aged 13 through 18 (N = 116), were used. The data collection procedure and in- and exclusion criteria are described in detail elsewhere [32].
Dutch parents in the general population
Data were used of interviews with mothers that were collected as part of a national research on children/adolescents and their parents, 'Child-rearing in the Netherlands in the 90s'. The families were selected from a larger sample of 10,000 families representative of Dutch population and were first contacted by phone. In the phone conversation the general purpose of the study was explained and it was checked if the parents had an adolescent child. From all contacted families with adolescent children, 53% agreed to participate. Data collection took place from 1993 to 1995 at the participants' homes, where questionnaires were administered individually to adolescents, mothers, and fathers. The sample consisted of 508 families with adolescents aged 12 through 18. In the present study we used mother-reported data of male adolescent participants who were in the age of 13 through 18 (N = 208). The data collection procedure and in- and exclusion criteria are described in further detail elsewhere, see [11].
Measures
Mother-son relationship
To assess maternal parenting practices, two subscales of the Nijmegen Rearing Questionnaire [33] were used: Affection Expression and Discipline. Mothers were asked to indicate on a 6-point scale (1 = highly disagree to 6 = highly agree) whether they agreed with the items. Affection Expression consists of nine items which measure the extent to which the mother shows positive affection towards the child (e.g., 'I often tell my child that I love him/her'). Discipline consists of five items concerning different types of punishment used by the mother (e.g. 'Most of the time, when my child does something he/she is not allowed, I slap him/her', 'I punish my child by sending him/her to his/her room'). Maternal monitoring was measured by means of a six-item instrument on a 4-point scale (1 = nothing to 4 = everything). Mothers were asked to indicate how much they know about, for example, their child's friends, how their child spends free time or how their child spends money [34]. To assess the amount of conflicts between adolescents and their mothers, the Parent-Adolescent Conflict List [35] was used. Mothers were asked to indicate on a 5-point scale (1 = never to 5 = very often) how often they quarrel with their son/daughter about 15 issues (e.g. 'academic achievement', 'curfew', 'home chores', 'son's/daughter's friends', etc.). Reliabilities of the Moroccan-Arabic translations of the scales were comparable to the reliabilities of the Dutch versions. The alphas of the Affection Expression scale were .84 for the Dutch version and .87 for the Moroccan-Arabic version. The alphas of the Discipline scale were .79 for the Dutch version and .86 for the Moroccan-Arabic version. The alphas of the Maternal monitoring scale were .83 for the Dutch version and .91 for the Moroccan-Arabic version.
In order to determine underlying dimensions of the Parent-Adolescent Conflict List, factor analysis was used. Exploratory factor analysis (i.e., Principal Component Analysis) revealed three factors (eigenvalues >1.0) and Varimax rotation (with Kaiser Normalization) showed three distinct factors. Factor 1 represented mother-child conflicts about issues outside the home, whereas Factor 2 represented mother-child conflicts about in-home issues, and Factor 3 represented one item (conflicts about son's girlfriend). The first two factors indicated a 46% explanation of the variance across all 15 items. As Factor 3 consisted of one item only, this item was not further used in the analyses. One item (conflicts about father's/mother's new partner) had low factor loadings on all extracted factors, this item was not retained. Thus, two underlying dimensions of the Parent-Adolescent Conflict List were used as two subscales: Conflicts about issues outside the home and Conflicts about in-home issues. Reliabilities of the two subscales were comparable for Dutch and Moroccan-Arabic versions; the alphas of the Conflicts about issues outside the home subscale were .82 for the Dutch version and .87 for the Moroccan-Arabic version, the alphas of the Conflicts about in-home issues subscale were .70 for the Dutch version and .80 for the Moroccan-Arabic version.
Educational level
Parental educational level was scored on a 4-point scale: 0 = elementary school or uncompleted elementary school, 1 = lower level of secondary or vocational education, 2 = medium level of secondary or vocational education and 3 = higher level of vocational education or university. The highest educational level of the father or the mother was used to score educational level of the family. For statistical analyses, the scores were classified into 'low educational level' (0-1), 'moderate educational level' (2) and 'high educational level' (3).
Statistical analyses
In order to identify different mother-son relationship types within the total sample of native Dutch and Moroccan parents of incarcerated and non-incarcerated adolescents, Latent Class Analysis (LCA) was used. LCA is a statistical method which is used to identify a set of mutually exclusive latent classes that account for the distribution of cases that occur within a cross tabulation of observed variables [36]. In other words, the purpose of a LCA is to find the smallest number of classes of individuals with similar patterns of, in this case mother-son relationships, which can explain associations of a set of variables. The parameters in a LCA model are class specific symptom profiles (which give the probabilities of a set of items for a particular class) and latent class probabilities (which estimate the likelihood for individuals to belong to each of the classes). Individuals are classified to the group with their highest class probability. The number of latent classes is determined by testing the goodness of fit of models with N latent classes using the Vuong-Lo-Mendell-Rubin likelihood ratio test and goodness of indices such as the Akaike Information Criterion, the Bayesian Information Criterion and Entropy. Latent Class Analyses were conducted in the software package Mplus version 5. To test the representation of native Dutch and Moroccan mothers of incarcerated and non-incarcerated adolescents in each of the classes, Chi-square tests were used. In order to identify the relationship of incarceration of the child and ethnicity on each of the mother-son relationship types, logistic regression analyses were conducted. Interaction effects between incarceration and ethnicity on mother-son relationships were tested using logistic regression analyses.