The sample consisted of 32 participants, of whom 19 were pathological gamblers (PG group) who were married or who had a life partner, and 13 were spouses/partners of pathological gamblers (GS group). Initially, the sample included an additional three subjects; however, these subjects fulfilled the criteria for both groups (PG and GS) and were thus eliminated. The PG group had an average length of time as part of a marriage/consensual union that was equal to 12.00 years (SD = 12.67), whereas the GS group had an average of 17.08 years (SD = 12.67). The difference between the groups was not statistically significant (p > .05, Mann-Whitney test). When the number of years in a marriage/consensual union was divided into 10-year bands, both groups were more likely to have participants in the 1–10 years of marriage/consensual union band (PG: N = 11, 57.89%; GS: N = 6, 46.15%). It is important to note that all of the participants were married or had a marital life partner but that couples were not present; that is, the PG and GS groups were independent samples. The PG group was primarily composed of male participants (N = 16, 84.21%) with an average age of 40.39 years (SD = 10.70). The majority of participants in this group had an upper secondary education (N = 6, 31.57%) or a bachelor’s degree (N = 5, 26.32%) (United Nations Educational Scientific and Cultural Organization (UNESCO) Institute for Statistics, 2012), lived in urban areas (N = 19, 100%) (Statistics Portugal (INE), 2009), and had average socio-economic status (N = 14, 73.68%) (Simões, 1994). The GS group was composed of nine women (69.23%) and four men (30.77%) with an average age of 43.46 years (SD = 10.41). The majority of participants in this group had completed a lower secondary education (N = 4, 30.77%) or had a bachelor’s degree (N = 3, 23.08%) (UNESCO Institute for Statistics, 2012), lived in urban areas (N = 13, 100%) (Statistics Portugal (INE), 2009), and had average (N = 3, 23.08%) or low socio-economic status (N = 3, 23.08%) (Simões, 1994). The two groups had no statistically significant differences with regard to socio-demographic variables (p > .05), except for differences in sex (p = .004, Fisher’s exact test). These differences were primarily because more females were in GS group than would be expected if these variables were independent (residual 2.0).

Specialized entities in gambling (e.g., Gamblers Anonymous and Families Anonymous) were asked to advertise for participants in the current study among their members. The entities were provided with copies of the protocol, and the person in charge was asked to distribute these copies to anyone who was interested in participating. This advertising resulted in the recruitment of seven PG and eight GS subjects. In addition, online dissemination of the research protocol (via social networks and email) was utilized using the virtual equivalent of the snowball method of recruitment (Goodman, 1961). This online dissemination resulted in the recruitment of 12 PG and 5 GS subjects. In the PG group, there were no statistically significant differences between the two collection methods for any of the study variables (p > .05). In the GS group, there was one statistically significant difference with regard to Family Difficulties (SCORE-15), which had a lower score for the online participants (U(17) = 0.29, p = .008, r = .07). The following criteria were implemented for inclusion/exclusion in the current study: (1) being older than 18 years old; (2) having a diagnosis of pathological gambling; and (3) being married to or having a life partner who is a pathological gambler.

The subjects’ participation request was framed by a set of preconditions (i.e., study objectives, respect for confidentiality and anonymity, voluntary participation, and provision of the contact details for the free specialized clinical support service at the authors’ host institution), which were described in the cover page of the protocol. Participants did not sign an informed consent form to protect the voluntary, anonymous and confidential nature of the information being collected (American Psychological Association, 2010). The Foundation for Science and Technology (FCT), which is an external agency, both approved of the current study and sponsored the project.

Questionnaire examining socio-demographic characteristics. To characterize the sample, a questionnaire examining socio-demographics (e.g., gender, marital status, nationality, residence, age, education and occupation) was administered that took into consideration the need to preserve the anonymity of the participants. It included questions that permitted the operationalization of the third criterion for inclusion/exclusion in the sample.

Systemic Clinical Outcome and Routine Evaluation – 15 (SCORE-15) (Stratton, Bland, Janes, & Lask, 2010; Portuguese version by Vilaça, Silva, & Relvas, 2014).

SCORE-15 is a self-report instrument that assesses family functioning. It consists of 15 items measuring three dimensions, which are Family Strengths, Family Communication and Family Difficulties, as well as five additional questions that relate to the family routine, the nature and impact of family problems and possible therapeutic needs. The participants evaluate whether each item describes their family according to a five-point Likert scale, in which 1 represents “describes us very well” and 5 represents “describes us very poorly”. A higher score corresponds to more problematic family functioning. SCORE-15 has good internal consistency both overall and dimensionally (.82 < α < .85).

Quality of Life (QOL) (Olson & Barnes, 1982; Portuguese version by Cunha & Relvas, 2015). This instrument assesses the perception of the quality of one’s family life. It is a self-report questionnaire that consists of 20 items representing the following four dimensions: (1) Family, Friends, and Health, (2) Time, (3) Media and Community, and (4) Financial Well-Being. The participants respond to the items according to a five-point Likert scale, in which 1 represents “dissatisfied” and 5 represents “extremely satisfied”. A higher score corresponds to a better quality of life. This instrument has reasonable/good internal consistency regarding the total scale and all of its dimensions (.72 < α < .89).

Dyadic Adjustment Scale (DAS) (Spanier, 1976; Portuguese version by Lourenço, 2006). The DAS assesses dyadic adjustment through 32 items that are grouped into the following four dimensions: Dyadic Consensus, Dyadic Satisfaction, Affectional Expression, and Dyadic Cohesion. The responses on a Likert scale ranged from 5 to 6 points for most items, whereas two items (items 29 and 30) had dichotomous response scales. Higher scores for the dimensional and total results correspond with better dyadic adjustment. Overall, this instrument has good internal consistency (α = .93), and the subscales vary between .65 and .88.

Marital Satisfaction Assessment Scale (Escala de Avaliação da Satisfação em Áreas Conjugais [EASAVIC]) (Narciso & Costa, 1996). The Marital Satisfaction Assessment Scale is a self-report instrument consisting of 44 items that are distributed across two large dimensions, which consist of the following areas: (1) Love (i.e., Feelings and Expression of Feelings, Sexuality, Emotional Intimacy, Continuity of the Relationship, and Physical and Psychological Traits) and (2) Functioning (i.e., Functions, Free-time, Autonomy/Privacy, Communication and Conflicts, and Relationships outside of the Family). The participants evaluate their levels of satisfaction regarding various areas of marital life according to a Likert scale, in which 1 represents “not at all satisfied” and 6 represents “completely satisfied”. The internal consistency of this scale is very good (Love: α =.97 and Functioning: α =.90).

Differentiation of Self Inventory – Revised (DSI-R) (Skowron & Schmitt, 2003; Major, Rodríguez-González, Miranda, Rousselot, & Relvas, 2014) The DSI-R is a self-report inventory consisting of 46 items that assess the differentiation of the self in adults using a Likert scale, in which 1 represents “not true for me” and 6 represents “very true for me”. This instrument has the following four dimensions: Emotional Reactivity, Emotional Cut-Off, I-Position and Fusion with Others. A higher score corresponds to a greater differentiation of the self. The DSI-R has good internal consistency for the overall scale (α = .86) and the subscales vary between .51 and .86.

Congruence Scale (CS) (Lee, 2002; Portuguese version by Cunha, Silva, Vilaça, Gonçalves, & Relvas, in press1). The CS is a self-report questionnaire for adults that consists of 16 items assessing congruence through the relationship with oneself, with others and with life. The 16 items are divided into the following two subscales: Spiritual/Universal and Intra/Interpersonal. Congruence is the central construct of the Congruence Couples Therapy for Pathological Gambling (CCT), which is an intervention model that has been adapted for couples with a pathological gambler element (Lee, 2009). The internal consistency of this scale and its subscales varies between reasonable and very good (.75 < α < .93).

Psychopathological Symptom Inventory (BSI) (Derogatis & Spencer, 1982; Portuguese version by Canavarro, 1999). This instrument is a self-report inventory consisting of 53 items with responses provided on a Likert scale ranging from “never” (0) to “very often” (4). The items encompass the following nine dimensions: Somatization, Obsessions-Compulsions, Interpersonal Sensitivity, Depression, Anxiety, Hostility, Phobic Anxiety, Paranoid Ideation, and Psychoticism. This instrument also provides information regarding the following three global indices: the General Symptom Index (GSI), the Positive Symptom Total (PST) and the Positive Symptom Index (PSI), which summarily rates emotional disorders. The PSI provides the cut-off point after which individuals are (probabilistically) considered emotionally disturbed (PSI ≥ 1.70). This instrument has internal consistency ratings that range from “reasonable” to “good” for all of its subscales and global indices (.62 < α < .80).

South Oaks Gambling Screen (SOGS) (Lesieur & Blume, 1987; Portuguese version by Lopes, 2009). Previous literature states that the SOGS is the most widely used tool for assessing pathological gambling (Shaffer, 1997). Composed of 20 items that are based on the DSM-III, this instrument evaluates the impact of gambling on various areas in a gambler’s life, including family, social, professional, financial and emotional aspects. A gambler is considered pathological when he or she scores 5 or more points out of 20. Higher scores indicate more severe gambling problems. The SOGS also provides data (via informational items that are not included in the calculation of the overall score) regarding the (1) type and frequency of the gambling behavior, (2) amounts involved in the betting process and (3) presence of family and friends with problems related to gambling. In the current study, this instrument had good internal consistency (α = .89), which allowed for the operationalization of the second and third criteria for exclusion/inclusion in the sample (e.g., diagnosis of pathological gambling, being the spouse or partner of a pathological gambler, respectively). Both the PG and GS subjects completed this questionnaire with regard to their own gambling behavior. Therefore, it was possible to eliminate the GS subjects from the sample who had scored within the pathological gambling range, combining conditions to integrate the two groups.

Statistical analyses were conducted on the total scores and all of the dimensions for each instrument using nonparametric statistics; however, the dimensional results should be viewed and interpreted as merely indicative because of the small sample size. This issue is particularly relevant for the instruments with more dimensions (i.e., EASAVIC and BSI). Cohen’s (1988, 1992) reference values were used to classify the effect sizes (r and V: 0.1 – small effect, .3 – medium effect, and .5 – large effect; η_p²: .01 – small effect, .06 – moderate effect, and .14 – large effect).

The results will be presented from the macro level to the micro level of the ISMPG, i.e., from the family context to the individual context. First, the quality of family life will be discussed (QOL) (Table 1). Both the PG and GS subjects had total and dimensional scores that were similar to the general population, except with regard to the Financial Well-Being and Family, Friends and Health dimensions. For these dimensions, the GS group had lower scores than the general population (if the range that resulted from the difference between the mean and standard deviation is considered) (Cunha & Relvas, 2015). Comparing across the two groups, a Mann-Whitney test indicated that the GS group had higher levels of dissatisfaction than the PG group. This moderate difference was marginally significant for the total score and the Media and Community dimension.

With regard to family functioning (SCORE-15) (Table 1), both the PG and GS groups reported having higher levels of difficulty with family functioning than the general population (Pereira, 2011), which was reflected in their total scores and in the Family Difficulties and Family Communication dimensions (considering the mean and standard deviation). A Mann-Whitney test showed that the groups differed from each other with regard to Family Difficulties, with the GS group reporting higher scores than the PG groups (i.e., suggesting greater difficulty; marginally significant differences and medium effect).

Given the cut-off point that has been proposed in previous literature (Spanier, 1976), the GS group showed poor levels of marital adjustment (DAS) (Table 1). According to a Mann-Whitney test, the differences between the two groups were statistically significant for the total scores, Dyadic Satisfaction and Dyadic Consensus (medium effect), with the GS group having lower scores than the PG group.

Marital satisfaction (the total score; EASAVIC) (Table 1) for the PG and GS groups was lower than for the general population (Pires, 2008), if the range that resulted from the difference between the mean and standard deviation is considered. A Mann-Whitney test indicated that the differences between the groups (in the total scores) were statistically significant (medium effect), with the GS group having a lower score than the PG group. Dimensionally, this trend was maintained, with lower satisfaction levels for the GS group than the PG group (statistically or marginally significant differences; medium effect).

On the individual level, both the PG and GS groups had lower scores on the total congruence scale and for the Intra/Interpersonal dimension (i.e., CS) than the general population (regarding the mean and standard deviation) (Table 2). For the Spiritual/Universal dimension, the PG group had lower scores than the general population, whereas the GS group had similar scores to the general population (regarding the mean and standard deviation) (Cunha, Silva, & Relvas, 2014). A Mann-Whitney test revealed that the differences between the two groups for the total scores and the Spiritual/Universal dimension were marginally significant (medium effect), with lower levels of congruence for the PG group than the GS group.

Regarding the differentiation of the self (DSI-R) (Table 2), both the PG and GS groups scored in accordance with the reference values for the general population (Major et al., 2014). Moreover, a Mann-Whitney test did not show any statistically significant differences between the two groups, which was also evident with regard to the dimensions.

Regarding the psychopathological symptoms (BSI) (Table 2), according to the PSI cut-off for emotionally disturbed populations (> 1.7) (Canavarro, 1999), the GS group’s scores showed emotional disturbance (PSI = 1.82). The PG group showed borderline levels of emotional disturbance (PSI = 1.69); however, rounding the values up revealed that members of this group had a high likelihood of becoming emotionally disturbed. A Mann-Whitney test revealed that the two groups did not differ significantly with regard to the global indices (i.e., GSI and PST) or the dimensions of the BSI.

The words that the participants used to describe their families (i.e., “What words best describe your family?”) were categorized into the following three groups: (1) positive (e.g., union, love, respect, support, and harmony), (2) negative (e.g., secretive, distant, jealous, and complex), and (3) neutral (e.g., normal and family). For the PG group, 12 (63.15%) participants used positive words, three (15.79%) used negative words, and three (15.79%) used neutral words (with one missing value). For the GS group, five (38.46%) participants used neutral words to describe their families, five (38.46%) used negative words and three (23.07%) used positive words. To calculate the statistical significance of these differences, the neutral and negative words were combined to satisfy the minimum expected cell frequency assumption for the chi-square test, which indicated that the differences between the PG and GS groups were statistically significant [χ²(1, N = 31) = 4.13, p = .042]. Given that this was a residual analysis, these difference were primarily because of the following: (1) there were more positive and fewer negative/neutral words listed by the PG group than would be expected if these variables were independent variables (residual 2.4 and -2.4), and (2) there were fewer positive and more negative/neutral words listed by the GS group than would be expected if these variables were independent (residual -2.4 and 2.4). The magnitude of the differences was medium (V = .43). The question “What is the main problem/difficulty that your family is currently facing?” triggered a wide range of answers, which were grouped into the following four categories: financial problems (e.g., financial difficulties and economic crisis), gambling, oneself (e.g., “It is me”), and various other responses—specifically, diverse problems, such as health, death in the family, birth of a child, excessive domestic responsibilities, and marriage. For the PG group, nine (47.37%) participants responded that there were financial problems, five (26.32%) mentioned various problems, three (15.79%) considered themselves as the primary family problem, and 2 (10.53%) mentioned gambling. For the GS group, the category “various” was the most representative (N = 6, 46.15%), followed by financial problems (N = 5, 38.46%) and gambling (N = 2, 15.38%). A Fisher’s exact test revealed that the differences between the groups were not statistically significant (p = .408). Another question was the following: “In your opinion, what treatment would be most useful to help with the gambling problems: individual therapy, couples therapy, family therapy or none?” The majority of PG group subjects chose the individual therapy option (N = 12, 63.16%), five (26.32%) chose family therapy and one (5.26%) chose the “none” option (with one missing value). For the GS group, five (38.46%) chose individual therapy, four (30.77%) chose family therapy, and three (23.08%) chose couples therapy (with one missing value). The differences between the groups were not statistically significant according to the results of a Fisher’s exact test (p = .093).

A two-way between-groups analysis of variance (two-way ANOVA) (Table 3) was conducted to examine the influences of sex and group (i.e., PG and GS) on the dependent variables (i.e., the total scores). An adjusted alpha level (p ≤ .10) was used to compensate for the small sample size (Stevens, 2009). The interaction effect (sex*group) did not reach statistical significance for any of the independent variables. The main effect of sex was statistically significant for the QOL, EC, DSI-R and BSI, with higher scores for women on the QOL, EC and DSI-R and higher scores for men on the BSI. The effect sizes were moderate to large (.11 < η_p² < .21). There were statistically significant main effects of group for the QOL, SCORE-15, DAS and EASAVIC, with the GS group having higher scores (i.e., greater difficulty) on the SCORE-15 and lower scores (i.e., greater levels of dissatisfaction) on the QOL, DAS and EASAVIC than the PG group. The effect sizes were moderate to large (.11 < η_p² < .27).

The influence of the socio-demographic variables on the instruments’ total scores (i.e., the dependent variables) within the groups was also examined (with a Mann-Whitney test, chi-square test or Fisher’s exact test). The socio-demographic variables (i.e., sex, age, education, and socioeconomic status) were not statistically different across any of the dependent variables (i.e., the total scores) for the PG group. For the GS group, a Mann-Whitney test revealed that women had stronger results (i.e., statistically significant) with regard to quality of life (i.e., the total score) [U(11) = 4, p = .030, r = -.60] (large effect) than men. The remaining socio-demographic variables (i.e., age, education, and socioeconomic status) were not statistically different across any of the dependent variables for the GS group.

This results of the current study show that PGs view their family and marital functioning as less marked by difficulties than GSs, and this difference is more evident with regard to marital issues. However, in terms of individual functioning, both the PGs and GSs showed psychopathological symptoms with characteristic values ??similar to emotionally disturbed populations. It should be noted that the PGs had more problems related to congruence than the GSs.

This study has the following two major limitations: (1) there was marital independence between the two subsamples, i.e., the gamblers and gamblers’ spouses belonged to different marital dyads; and (2) a small sample size that limits the implications of the current results (e.g., issues with generalization, a need to use nonparametric statistics and restrictions related to basic analyses). However, this study contributes to the current literature by reducing the empirical gap evident in studies examining the marital lives of PGs. Furthermore, the current results question a number of “truths” that are taken for granted in previous literature (e.g., PGs’ perception of marital problems). Couples therapists who work with these dyads should be aware of the challenges that are inherent in divergent perspectives (cf., Cunha et al., in press2).

Future research should focus on developing qualitative or mixed method studies to analyze the narratives of PGs and GSs regarding their marital lives to shed light on the nuances within their perspectives.