Assessment of identity development already in adolescence is important to study developmental paths in general and to enable valid conclusions about specific pathological risks. This is true especially in the light of the new revision of DSM-5, where “identity” has been discussed extensively to be integrated as a core criterion for impaired self-related personality functioning.
The Swiss-German-American questionnaire AIDA provides a reliable and valid assessment of the complex construct “pathology-related identity integration vs. identity diffusion” in adolescents by integrating different theoretical approaches and a reasonable structure of known subconstructs. Valid assessment tools must also meet the requirements of international usability in cross-cultural studies (e.g. as described by the International Test Commission; [35]), to model different phenotypes in different cultures and to enable valid comparisons of identity development in different societies by providing true equivalence in content of the assessment tool.
The Spanish culture-adapted translation of AIDA showed good psychometric properties in the Mexican sample, with similar patterns in results compared to the original version. We conclude that all constructs and subconstructs contained in the AIDA model to constitute “identity development” had been successfully transposed into the “Spanish-speaking culture” with good content equivalence.
However, detailed analysis revealed some problems on the item level in the Mexican sample. In the following, the results are discussed in detail, and suggestions for changes in item formulation with respect to the special need of the Mexican culture are presented. Each class of results is contrasted with the results of the original version to illustrate the special techniques and consequences of cultural test adaption.
Compared to the German items of AIDA, the Spanish items seemed to be “easier to answer in a symptomatic way”, i.e. to say “yes” coded towards identity diffusion in the Mexican school sample. While in the German sample, the mean percentage of symptomatic answers (pi) was 30% and 20 of the 58 items showed a percentage below 10% [19], in the Mexican sample the mean pi was 40% and only 3 items showed an extreme value for “difficulty to be answered with yes” with a pi below 10%. This means that the items were in general more difficult to answer with “yes” in the German version than in the Mexican version of AIDA. Thus, score differences between Mexican and German adolescents cannot be interpreted directly as different levels of identity diffusion because score equivalence cannot be assumed [33]. Therefore, population norms, e.g. T-values extracted from representative populations, have to be used for valid comparisons of samples concerning their “levels of identity development”.
All Spanish AIDA items matched the criterion for item bias and proved to be “age and gender fair” in the Mexican study. This agrees well with the results in the German study. The classical example for explaining “item fairness” is the “soccer item”: it would be unfair to judge the frequency of general physical activity by asking “How often do you play soccer” because girls usually do not like this game as much as boys do. Girls would probably say “never” more often, and therefore would be judged falsely as “physically inactive” in contrast to the boys. As all AIDA items can be regarded as gender and age fair, differences on the score level between gender and age groups can be interpreted as true developmental differences.
Not all Spanish items fully met the criterion for item total correlation (rit) in the Mexican sample. Four items (items 8, 27, 12, and 20) showed coefficients slightly below the criterion but would be generally acceptable for a translated version, especially taking into account that in all cases the item total correlations were excellent in the delinquent subsample with assumed behavioral problems. As the pooled school sample did not contain any subjects with diagnosed identity diffusion (e.g. a clinical sample with personality disorders), the data variance may not reflect the full range of true variability and relations in total, and the coefficients might be improved when “the pathological side” of the construct could better unfold its effects. Three items showed slightly weak rit in more than one category or one coefficient far below the preset criterion (items 2, 33, 49). These items should be discussed in detail to detect a possible cultural bias in translation that might be eliminated by improved wording.
Item 2 (“I feel at home in my community, here is where I belong to”) showed the weakest rit in the delinquent group, and we realized that in Mexico it might be difficult to feel ‘at home’ in a community suffering from a high crime rate. A better wording might be “I am proud of my roots and I feel like belonging to this group” to capture “Discontinuity-relations and roles” in terms of potential identity-giving and stabilizing cultural and/or ethnic roles.
Item 33 (“Just as I was as a child and as I am now, I can imagine how I could be in a few years”) was only slightly below the criterion and might be improved by a simpler wording, i.e. “As time goes by, I can imagine well how I will be in the future.”
Item 49 (“Many people are very "fake" and do not behave the way they really are”) showed a weak item total correlation with rit = .11 in the subscale-referred analysis and rit = .18 in the scale-referred analysis in the school sample, implying that this item has too little in common with the variance of the whole scale and the other items. Thus, it is not suitable for the scale “Incoherence-cognitive self-reflection” in terms of having shallow or superficial mental representations. At the same time, the item showed a high rit = .55 in the delinquent subsample. This can be interpreted as a specific concordance with behavioral problems and may constitute improved quality of the assessment if psychiatric patients are included. Additionally, we realized that calling someone “a fake” is somehow “bad language” in the Mexican society and that students might refuse to respond to such unsuitable questions. To address this, the item should be expressed more politely, e.g. “Many people behave differently from what they really are” to adequately reflect the original wording of “not understanding complexity and variety of others’ behavior”.
Thus, for all “problematic” items, issues with cultural adaption of the contents were considered, and improved formulations were suggested. The high scale reliabilities α, with .94 for the total scale Identity-Diffusion, .85 and .92 for the two primary scales Discontinuity and Incoherence, and .70 to .83 for the subscales, are expected to further improve in the next pilot test with adapted item wording.
As in the German validation sample, the AIDA scores differed with about medium effect size between boys and girls in the Mexican school sample. However, in contrast to the findings in Germany, the Mexican girls showed systematically lower scores than the boys in the AIDA, implying healthier development, i.e. better identity integration. Therefore, differentiated norms for boys and girls should be extracted based on a representative Mexican population sample.
In contrast to the German subjects, Mexican subjects showed systematic differences between the two age groups (12-14 years and 15-19 years) with small to medium effect sizes. Therefore, it can be assumed that in Mexico distinct developmental stages related to age can be found. In line with the general theory of developmental identity, the younger adolescents displayed higher levels of “identity diffusion” without reaching pathological levels. This is viewed as a sign of an expected identity crisis at this age. Given this, differentiated age-specific population norms should be extracted in Mexico.
The socioeconomic background seemed to have no remarkable impact on the adolescents’ identity development in Mexico. Thus, students from different schools can be pooled for statistical analyses without affecting the results.
The EFA on the item level resulted in a very similar factorial structure as the one found in the Swiss-German validation sample. In the Mexican sample, 15 extracted factors explained 66.2% of the total variance with the first component alone explaining already 25.4%, while in the German sample, 15 extracted components explained 62.6% of the variance (first component 24.3%). This clearly documents favorable equivalence and effective culture-specific test adaption for assessing this complex construct in Mexico, as the translated version showed comparable patterns of results in a similar statistical analysis.
The confirmed “i-factor” is in line with the expected overall congruence on phenotype level, as all modeled contents (i.e. items) had been constructed to reflect current pathology-related identity development. The AIDA model combines distinct aspects of healthy identity concerning sources and/or consequences for its development and uses these theory-based distinct units in terms of scales and subscales experimentally to facilitate communication and research. However, empirical confirmation of the assumed structure is needed. As soon as different AIDA versions with convincing basic psychometric properties for item characteristics and scale reliability have been tested in normative samples, the instrument will be tested on the item and subscale level in a multi-sample study with a cross-cultural focus. Optimally, data obtained in various countries and continents should be suitable for pooling to analyze the underlying structure.
Because AIDA is a pathology-oriented inventory, the central quality standard lies in the diagnostic or predictive validity, i.e. the potential to differentiate healthy from impaired development. We evaluated criterion validity of the Spanish AIDA in Mexico by comparing the scores of the school sample to a “clinical sample” recruited from the juvenile justice system sample. The delinquent boys were a highly heterogeneous group with respect to their behavioral problems and comorbidities. The most probable behavioral problem in this subsample, i.e. “externalizing disorder”, is not assumed to be directly associated with severe identity diffusion in terms of “having no inner continuity and subjective self-sameness (Discontinuity)” or “having no consistently defined inner self-picture and autonomy (Incoherence)”. However, we clearly expected relevant consequences for identity development and detectable differences compared to the normal students as identity diffusion can be seen as a basis for several types of psychopathology, and the prevalence of mental disorders has proven to be high in incarcerated adolescents [22–24].
To create a more homogeneous contrast group, we used the Mexican pilot test version of the borderline screening inventory Ab-DIB with Canadian cut-offs. The subgroup of delinquent boys with signs of borderline pathology in this test (N = 14) was compared to the whole school sample on the one hand, and to a school subsample matched for age and gender on the other hand. In both analyses, the delinquent boys showed higher frequencies of identity pathology than the adolescents in the school sample, pointing towards satisfactory criterion validity of AIDA in Mexico. The Discontinuity scores differed with high effect sizes of d = 1.21 and 1.17, and the Incoherence scores with about medium effect sizes d = 0.62 and 0.46 standard deviations between the groups. However, the discriminative power of AIDA in this study was lower than in the original study that contained a true clinical sample of diagnosed PD patients. The strikingly different impact of the AIDA subscales on differentiating between the school and conflict samples implies that it is appropriate to treat the subscales as distinct units in an experimental fashion. Especially the subscales representing the psychosocial function “mental representation”, split into the domains “emotional” (part of Discontinuity; 1.3) and “cognitive” (part of Incoherence; 2.3), showed a different pattern compared to the other subscales. While subscale 1.3 showed a weaker discriminative potential than the other subscales of Discontinuity, 2.3 showed a stronger potential than the other subscales of Incoherence to differ between the school group and the “clinical” group. This may be due to the special characteristics of the delinquent sample with probable current behavioral problems like aggression and externalizing disorders that may be related to specific deficits leading to a special AIDA profile of this group.
Limitations
A limitation of the study is the lack of psychiatric disorder assessment in the school sample. Based on epidemiological studies, we assumed that up to 15-20% of this representative sample of adolescents may exhibit minor to major signs of mental problems. However, without enrichment with a clinical subsample displaying extremer levels of identity diffusion, the heterogeneity of the sample for evaluating the basic psychometric properties of the Spanish AIDA is not optimal. Similarly, a true clinical sample of patients with defined diagnoses would be more informative for comparing their AIDA scores with those of the school sample in order to evaluate criterion validity. To extract population norms for Mexico, a representative sample with a higher participation rate is needed with adequate sample size in the different targeted groups for gender and age.