Everybody Is…Drinking! Interpretation Bias in Problematic Drinkers With and Without Mild to Borderline Intellectual Disability

ABSTRACT

Background: Problematic alcohol use is characterized by disrupted associative processing of environmental clues, where problematic drinkers interpret ambiguous, alcohol-relevant clues in an alcohol-related way. The present study examined the strength of this interpretation bias in a large sample (N = 230) of light and problematic drinkers with and without mild to borderline intellectual disability (MBID, IQ 50–85). Method: All participants were asked to finish 24 open-ended, ambiguous scenarios with their first, spontaneous response. Results: Consistent with the hypothesis, problematic drinkers with and without MBID were found to have an interpretation bias toward alcohol. The difference in the strength of the bias between light and problematic drinkers was strongest for negative scenarios. Participants with MBID showed a stronger interpretation bias compared to participants without MBID, especially on the negative scenarios. Conclusion: Problematic drinkers tend to interpret ambiguous, alcohol-relevant clues in an alcohol-related way and this tendency increases with higher levels of alcohol use–related problems. These results extend our knowledge on substance use disorder and provide new lines of inquiry for the assessment and treatment of problematic alcohol use in individuals with MBID.

Keywords:

• Addiction
• alcohol
• mild intellectual disability
• interpretation bias
• implicit memory associations

Environmental stimuli often require interpretation, explanation, and evaluation. For example, when hearing the word “joint” you can think of the human anatomy, a place at which two or more things are joined, or a marijuana cigarette. Your first association with a given (ambiguous) word depends on the context and your personal experiences and memories. That is, this type of associative processing is founded on memory, which acts as an associative network (Bechara, Noel, & Crone, 2006; Strack & Deutsch, 2004; Wiers et al., 2007). The ease with which a cluster of associations is activated depends on the accessibility of that cluster or the strength between the elements and occurs fast, unintentionally and outside one’s control or awareness. Through a process of classical conditioning, experiences shape the associative network by forming clusters which can activate motivational tendencies and subsequently influence behavior by responding consistently with those tendencies (Collins & Loftus, 1975).

Although associative processing is not problematic in itself, its influence on behavior can become problematic, for example in substance-use disorders (SUD). As examining these processes might improve our understanding of SUD and its treatment options, several attempts have been made to measure automatic associations and interpretations. Typically, automatic associations and interpretations have been assessed using direct measures (e.g., questionnaires and rating scales) and indirect measures. Indirect measures are thought to tap into implicit cognitive processes, are considered to measure less-accessible memory associations than those assessed by self-report, and are thought to reduce self-presentation influences or social desirability because they do not directly mention the targeted behavior (Greenwald et al., 2002; Stacy & Wiers, 2010). Examples of indirect measures include the implicit-association task (Greenwald, McGhee, & Schwartz, 1998) and word-association tasks, in which individuals are asked to generate their first, spontaneous response when hearing an ambiguous word, sentence, or scenario. Despite the control participants can assert over their responses in word-association tasks, their responses probably relate more to implicit associations and interpretations because of the stronger focus on first, “gut” responses compared to traditional self-report measures (Ranganath Smith & Nosek, 2008). Word-association tasks have been found to be the strongest predictors of alcohol use in comparison with other indirect measures (Thush et al., 2007; Van der Vorst et al., 2013).

Using word-association tasks, an interpretation bias has been found in problematic drinkers. In other words, problematic drinkers have been found to interpret ambiguous, alcohol-relevant words, phrases or scenarios in an alcohol-related way (Ames, Sussman, Dent, & Stacy, 2005; Krank, Schoenfeld, & Frigon, 2010; Salemink & Wiers, 2014; Woud, Fitzgerald, Wiers, Rinck, & Becker, 2012; Woud et al., 2014). For example, Krank et al. (2010) found that problematic drinkers associated words such as “pitcher” or “draft” in an alcohol-related manner more often compared to light drinkers. Woud et al. (2012) found similar results using short scenarios—an approach that increases the ecological validity of the task and allows for individual differences in the underlying associative network. Using this approach, Salemink and Wiers (2014) and Woud, Becker, Rinck, and Salemink (2015a) found that problematic drinkers associate both positive (e.g., a party, being out with friends) and negative scenarios (e.g., feeling down or stressed) with alcohol use. In addition, similar correlations between the strength of the biases for positive and negative scenarios and the severity of alcohol use–related problems have been reported by these authors.

In this study, we expand on these findings by including problematic drinkers with mild to borderline intellectual disability (MBID; IQ 50–85; American Psychiatric Association, 2013). They are at risk for developing problematic alcohol use and alcohol use disorders (Burgard, Donohue, Azrin, & Teichner, 2000; McGillicuddy, 2006) and often experience more severe negative consequences from alcohol use compared to individuals without MBID (Slayter, 2008). Although this group has gained attention over the past years, the current knowledge on SU(D) in individuals with MBID is scarce and there is a need for valid screening and assessment tools and effective treatment interventions (Kerr, Lawrence, Darbyshire, Middleton, & Fitzsimmons, 2013; Van Duijvenbode et al., 2015). From a scientific point of view, studying the interpretation bias would thus extend our knowledge on the role of this bias in SUD in individuals with MBID and shed light on the role of cognitive functioning on the interpretation bias. From a clinical point of view, studying the interpretation bias would provide new ways for the assessment and treatment of problematic alcohol use in individuals with MBID. For example, as word-association tasks provide indirect measures of high-risk situations for alcohol use or relapse (Woud et al., 2012), these tasks could provide possibilities for tailoring treatment to the needs and characteristics of the individual and provide implications for treatment and relapse prevention, for example by focusing treatment more specifically on personal high-risk situations that are associated with alcohol use and directly changing the alcohol associations in an interpretation retraining procedure (Woud et al., 2012). Preliminary evidence shows that the alcohol-related interpretation bias can indeed be trained in such interpretation retraining procedures (Woud, Hutschemaekers, Rinck, & Becker, 2015b).

The aim of our study was twofold. First, we aimed to compare the strength of the interpretation bias between light and problematic drinkers. In line with earlier results on the interpretation bias in problematic drinkers (for an overview, see Stacy & Wiers, 2010), our first hypothesis was that problematic drinkers with and without MBID would show an interpretation bias toward alcohol and that this bias would be significantly stronger in problematic drinkers compared to light drinkers. Second, we wanted to explore the dynamics of the interpretation bias in more detail by calculating a bias score for positive and negative scenarios separately. In line with Salemink and Wiers (2014) and Woud et al. (2015a), we expected that problematic drinkers would show an interpretation bias on both the positive and the negative scenarios, and that both bias scores would be equally strongly correlated with the severity of alcohol use–related problems. As this is the first study on the interpretation bias in problematic drinkers with MBID, we did not formulate any a priori hypotheses regarding the role of intellectual functioning on the strength or manifestation of the interpretation bias but ran exploratory analyses instead.

Method

Participants

This sample is a combination of the sample of two of our other studies (Van Duijvenbode, Didden, Korzilius, & Engels, 2016b; Van Duijvenbode, Didden, Korzilius, & Engels, resubmitted). Participants were recruited from organizations within ID care (n = 109, 47%) or addiction medicine (n = 30, 13.0%), via advertisements on social media, the Radboud University, and word of mouth (n = 91, 39.6%). Inclusion criteria were an age of 18 years or older, an IQ of minimally 50, and stable functioning (i.e., no active psychotic or manic state, as assessed by the treatment team). All participants were required to have had access to and/or consumed alcohol in the last 1.5 months. Those with a history of problematic alcohol use but currently abstaining for longer than 1.5 months were excluded from participating.

We included 230 participants (139 men, 60.4%) with a mean age of 32.3 years (SD = 12.45, range = 18–61 years). Highest completed form of education ranged between none (4.3%, n = 10) to university (13.0%, n = 30). Most participants finished special education (24.3%, n = 56) or secondary school (24.3%, n = 56). The majority of participants originated from The Netherlands (92.2%, n = 212). The other participants originated from Surinam/The Antilles (2.1%, n = 5), Morocco/Turkey (1.6%, n = 4) or other Western and non-Western countries (3.9%, n = 9). All participants spoke Dutch fluently and had no trouble understanding the instructions. Almost half of the participants (46.1%, n = 106) were diagnosed with one or more psychiatric disorders, of which substance use disorder (28.3%, n = 65), autism spectrum disorder (11.7%, n = 27), and attention deficit hyperactivity disorder (7.0%, n = 16) were diagnosed most often. Seventy-five participants (32.6%) used psychotropic medication, including antipsychotics and antidepressants.

A power analysis (with G*Power Version 3.1.92) showed that with the number of participants in the sample and the statistical tests used a power of .97 was achieved at a medium effect size of .3 and α of .05.

Materials

Substance Use

The Substance Use and Misuse in Intellectual Disability Questionnaire (SumID-Q; VanDerNagel, Kiewik, Van Dijk, De Jong, & Didden, 2011) was used to assess participants’ alcohol use. This is an interview method adapted to the needs of those with MBID, for example by avoiding lengthy phrases and difficult wording (VanDerNagel, Kemna, & Didden, 2013). Participants reported their general frequency and quantity of alcohol use, which was converted into standard units of 10 g of alcohol (International Center for Alcohol Polities, 2010).

The severity of alcohol use–related problems was measured using the Alcohol Use Disorders Identification Test (AUDIT; Babor, Higgins-Biddle, Saunders, & Monteiro, 2001; Dutch translation: Schippers & Broekman, 2010). This questionnaire consists of ten questions about the frequency, quantity, and consequences of alcohol use which are answered on a 5-point rating scale ranging from “never” (0 points) to “almost every day” (4 points). Total scores range between 0 and 40, with higher scores reflecting more severe alcohol use–related problems. A score of 8 or more is indicative of hazardous alcohol use (Babor et al., 2001) and was used in this study to classify participants as either light drinkers (score < 8) or problematic drinkers (score ≥ 8). The internal consistency of the AUDIT in the current study was good (Cronbach’s alpha = .88, mean inter-item correlation = .42).

Interpretation Bias

The interpretation bias was measured with the open-ended, ambiguous scenarios developed by Woud et al. (2012). This task consists of 24 situations (8 positive, 6 negative, and 10 neutral) that each allow for different interpretations (see Table 1). The positive and negative scenarios are based on the Inventory of Drinking Situations (Annis, 1982) and tap into situations associated with alcohol use. The scenarios were adapted to ensure feasibility in an older (clinical) sample and for individuals with MBID.

Table 1. Examples of the Positive, Negative, and Neutral Scenarios Used (Woud et al., 2012) and Possible Answers Given by Participants.

	Scenario	Possible answers
Positive scenario	Movie night	New film (alcohol unrelated)
	Movie night at your friend’s house. “One more?” one of your friends asks. You cannot resist temptation and reach for a…	Your glass (ambiguous)
		Beer (alcohol related)
Negativescenario	Bad day	Chocolate (alcohol unrelated)
	It is a horrible day and nothing works. You want to lose this bad feeling and treat yourself. You get a strong craving for …	A drink (ambiguous)
		Alcohol (alcohol related)
Neutral scenario	Poker	Good (alcohol unrelated)
	You play poker with your friends every other week. Everything is ready and the cards have been dealt. This time, your cards are very…	Bad (alcohol unrelated)
		Difficult to win with (alcohol unrelated)

Each scenario consisted of a title and three lines. The last sentence ended abruptly and participants were asked to finish each scenario with their first, spontaneous response. All participants were ensured there were no correct or incorrect answers and they should respond with whatever came up in their mind first. There was no time limit for the administration of the task. Taking into account that participants with MBID often have reading and writing difficulties, all scenarios were read out loud. The answers provided by the participants were written down verbatim by the researcher. To control for carry-over effects, we used three different booklets with a different order of scenarios. The order of the booklets was counterbalanced across participants.¹

All answers provided by participants were coded individually by the first author and a trained student in a conservative and liberal way (Frigon & Krank, 2009; Woud et al., 2012). Only the conservative approach will be reported in this article (cf. Woud et al., 2014). All responses were coded as binary variables in this approach: “1” for alcohol-related answers and “0” for ambiguous or alcohol-unrelated answers. Consensus scores agreed upon by both raters were used to calculate mean bias scores (i.e., total score, positive-scenario score, negative-scenario score) for each participant. Total bias scores ranged from 0–24, while the bias scores for positive and negative scenarios had a maximum of 8 and 6, respectively. Alcohol-related answers on neutral scenarios were only included in the total bias score, but not calculated separately. Interrater reliabilities for the total bias score (Cohen’s kappa = .99, p < .001; percentage of agreement: 99.8%) and the bias scores for positive scenarios (Cohen’s kappa = .98, p < .001; percentage of agreement: 99.3%) and for negative scenarios (Cohen’s kappa = .99, p < .001; percentage of agreement: 99.9%) separately were excellent.

IQ

IQ was measured using the most recent scores on the Dutch version of the Wechsler Adults Intelligence Scale third edition (WAIS-III-NL; Uterwijk, 2000) in the participants’ files. If IQ was unknown, we administered a short version of the WAIS-III (47.8%, n = 110) to overcome time constraints and possible problems with participant fatigue, agitation, and frustration (Van Duijvenbode, Didden, Van den Hazel, & Engels, 2016a). The WAIS-III short form is based on the Wechsler Abbreviated Scale of Intelligence (WASI; Wechsler, 1999) and can be administered in approximately 30 minutes. It consists of four sub-tests (vocabulary, similarities, block design, and matrix reasoning) and provides a valid and reliable estimate of full-scale IQ (Van Duijvenbode et al., 2016a). Full-scale IQ was used to identify participants with MBID (IQ < 85) and without MBID (IQ ≥ 85).

Procedure

All participants were provided with written information about the study before signing informed consent forms. The session started with questions regarding general demographic information and, if necessary, the WAIS-III short form to estimate full-scale IQ. The scenario task was then used to measure the interpretation bias, after which the SumID-Q was administered to assess general patterns of alcohol use and the severity of alcohol use– related problems. As this study is part of a larger PhD project on the neuropsychology of SUD in individuals with MBID, all participants also completed other computerized tasks to measure cognitive biases (i.e., visual dot probe task, approach avoidance task; Van Duijvenbode et al., 2016b) and executive control (i.e., Corsi block tapping task, Go/No-go task; Van Duijvenbode et al., resubmitted). These tasks were administered after completion of the word-association task. In between tasks, participants were allowed to take a break whenever necessary. Finally, participants were thanked for their time and received a gift worth €5 (US $6.50, GBP £3.70) for their participation. The study was approved by the Ethical Committee of the Faculty of Social Sciences, Radboud University, Nijmegen, the Netherlands (ECG2012-1301-003).

Results

Group Characteristics

Participants were divided into four groups based on severity of alcohol use–related problems and IQ: light drinking participants without MBID (n = 57), problematic drinking participants without MBID (n = 55), light drinking participants with MBID (n = 50), and problematic drinking participants with MBID (n = 68). There were no missing data on any of the key variables. As expected, a one-way between groups ANOVA with post-hoc Tukey HSD comparisons showed that participants with MBID had a significantly lower (estimated) full-scale, verbal, and performance IQ compared to participants without MBID. Similarly, problematic drinkers had a significantly higher AUDIT score and weekly alcohol consumption compared to light drinkers (see Table 2). Groups also differed in gender ratio (χ²(3) = 42.56, p < .001), with larger proportions of light drinking (n= 60, 26.1%) compared to problematic drinking gen (n = 31, 13.5%).

Table 2. Participant Characteristics per Group (N = 230): Light Drinkers With (Below) Average IQ (n = 57), Problematic Drinkers With (Below) Average IQ (n = 55), Light Drinkers With Mild to Borderline Intellectual Disability (MBID; n = 50), and Problematic Drinkers with MBID (n = 68).

	(Below)Average IQ		MBID
	Light drinkers	Problematic drinkers	Light drinkers	Problematic drinkers
	M (SD)	M (SD)	M (SD)	M (SD)	F (3, 226)	p	ηp^2
Age	29.07 (12.06)	31.47 (12.00)	33.22 (11.90)	34.82 (13.12)	2.42	.067	.03
Est. full-scale IQ	105.42 (11.09)	103.49 (11.36)	70.52 (8.53)	68.73 (8.60)	229.54	< .001	.76
Est. verbal IQ	102.21 (11.10)	102.02 (12.91)	71.00 (10.19)	69.78 (9.93)	149.62	< .001	.68
Est.performance IQ	107.56 (12.62)	105.46 (12.77)	73.09 (10.63)	88.77 (21.07)	171.64	< .001	.71
AUDIT score	4.23 (2.13)	16.05 (6.62)	3.24 (2.21)	17.34 (6.81)	123.05	< .001	.62
Weekly alcohol consumption	3.35 (4.10)	48.09 (74.65)	1.94 (3.16)	61.99 (97.82)	13.35	< .001	.15

Note. AUDIT = Alcohol Use Disorders Identification Test (Babor et al., 2001); η_p² = partial eta squared.

Interpretation Bias

To test our first hypothesis that problematic drinkers would show an interpretation bias toward alcohol, we first calculated Pearson product-moment correlation coefficients. There was a strong, positive correlation between the severity of alcohol use–related problems (AUDIT score) and total bias score (r = .63, p < .001). Second, we conducted one-sample t-tests to compare the mean total bias scores to zero, meaning no bias. The total bias scores of both light (M = 2.08, SD = 1.80; 95% confidence interval (CI): 1.74–2.43) and problematic drinkers (M = 4.55, SD = 2.72; 95% CI: 4.07–5.04) differed significantly from zero, t(106) =11.96, p <.001; t(122) = 18.50, p < .001, respectively. In addition, an independent samples t-test was conducted to compare the strength of the interpretation bias between light and problematic drinkers. There was a significant difference in total bias scores between light drinkers and problematic drinkers, t(228) = 7.97, p < .001, Cohen’s d = 0.94.

Positive and Negative Scenarios

To test our second hypothesis that the strength of the interpretation bias between light and problematic drinkers would be similar for positive and negative scenarios, we first calculated Pearson product-moment correlation coefficients. There were medium to strong correlations between the severity of alcohol use–related problems (AUDIT score) and the positive (r = .42, p < .001) and negative bias scores (r = .63, p < .001). A Fisher r-to-z-transformation indicated that the bias score for negative scenarios correlated significantly more strongly with the severity of alcohol use–related problems than the bias score for positive scenarios (z difference = −3.13, p < .001). These results were complemented with an independent samples t-test to compare the strength of the bias scores for light and problematic drinkers. Results showed that problematic drinkers scored significantly higher than light drinkers on the positive scenarios (M = 2.39, SD = 1.37; 95% CI: 2.15–2.63; M = 1.39, SD = 1.19; 95% CI: 1.16–1.62, respectively; t(228) = 5.87, p < .001, Cohen’s d = 0.78) and the negative scenarios (M = 2.10, SD = 1.82; 95% CI: 1.77–2.42; M = 0.65, SD =0.94; 95% CI: 0.47–0.83, respectively; t(228) = 7.40, p < .001, Cohen’s d =0.98). Last, a linear regression analysis was conducted to assess the ability of the bias scores for positive and negative scenarios to predict the severity of alcohol use–related problems (AUDIT score). The full model was statistically significant (F(2, 227) = 81.52, p < .001) and explained 64.7% of the variance. Both the bias score for positive scenarios (β SE = .35, p = .006) and the bias score for negative scenarios (β SE = .29, p < .001) significantly predicted the severity of alcohol use–related problems.

Role of IQ

Last, to investigate the role of IQ in the interpretation bias, we first calculated Pearson product-moment correlation coefficients. Estimated full-scale IQ correlated negatively with the total bias score (r = −.22, p < .001) and the bias score for negative situations (r = −.28, p < .001), but not with the bias score for positive scenarios (r = −.07, p = .291).

These results were supplemented with linear regression analyses to assess the predictive validity of severity of alcohol use–related problems and estimated full-scale IQ for total, positive, and negative bias scores. When predicting the total bias score, the full model was statistically significant (F(2, 219) = 76.95, p < .001) and explained 64.2% of the variance. Both AUDIT score (β = .20, SE = .02, p < .001) and estimated full scale IQ (β = −.02, SE = .01, p = .017) predicted the total bias score significantly. Similar results were found when predicting the bias score for negative scenarios. This model was also statistically significant (F(2, 219) =82.40, p < .001) and explained 65.5% of the variance. AUDIT score (β = .12, SE = .01, p < .001) and estimated full-scale IQ (β = −.02, SE = .01, p < .001) both predicted the strength of the bias score for negative scenarios significantly. Last, when predicting the bias score for positive scenarios, the full model was statistically significant (F(2, 219) = 25.58, p < .001) and explained 43.5% of the variance. Unlike the other two models, AUDIT score was the only significant predictor (β = .08, SE = .01, p < .001). Estimated full-scale IQ did not significantly predict the strength of the bias score for positive scenarios (β = .00, SE = .01, p < .925).

Last, a 2 × 2 between-groups ANOVA was conducted. Participants were divided into four groups according to their estimated full-scale IQ and severity of alcohol use–related problems based on the AUDIT score (i.e., light and problematic drinking participants with and without MBID). The p-level of the interaction effect between IQ and severity of alcohol use–related problems did not reach statistical significance in any of the analyses (see Table 3). There were significant main effects of severity of alcohol use–related problems, both when analyzing total bias scores and when analyzing bias scores for positive and negative scenarios separately. Problematic drinkers were shown to have stronger bias scores (total: M = 4.55, SD = 2.73; positive scenarios: M = 2.39, SD = 1.37; negative scenarios: M = 2.10, SD = 1.82) than light drinkers (total: M = 2.08, SD = 2.73; positive scenarios: M = 1.39, SD = 1.19; negative scenarios: M = 0.65, SD = 0.94). In addition, there were significant main effects for full-scale IQ on the total bias score and the bias score for negative scenarios, with higher total scores and bias scores on negative scenarios for participants with MBID compared to participants without MBID. When controlling for estimated verbal IQ in a partial correlation analysis and an ANCOVA, all these results remained.

Table 3. Descriptives and Two-Way Between-Groups ANOVA Results per Participant Group (N = 230): Light Drinkers Without Mild to Borderline Intellectual Disability IQ (n = 57), Problematic Drinkers Without MBID (n = 55), Light Drinkers With MBID (n = 50), and Problematic Drinkers With MBID (n = 68).

	Without MBID			With MBID
	Light drinkers	Problematic drinkers	Total group	Light drinkers	Problematic drinkers	Total group
	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)	M (SD)	F (1, 226)	p	ηp^2
Total score Alcohol use IQ Alcohol use x IQ	1.65 (1.37)	4.33 (2.47)	2.96 (2.39)	2.58 (2.10)	4.74 (2.93)	3.82 (2.81)	61.224.700.72	< .001.031.398	.21 .02 .00
Positive scenarios Alcohol use IQ Alcohol use x IQ	1.23 (1.05)	2.55 (1.33)	1.88 (1.36)	1.58 (1.31)	2.26 (1.39)	1.97 (1.39)	34.660.043.46	< .001.834.064	.13.00 .02
Negative scenarios Alcohol use IQ Alcohol use x IQ	0.39 (0.65)	1.76 (1.72)	1.06 (1.46)	0.96 (1.12)	2.37 (1.86)	1.77 (1.73)	52.179.330.01	< .001.003.938	.19.04 .00

Note. η_p² = partial eta squared.

Discussion

Research has repeatedly shown that problematic drinkers tend to interpret ambiguous, alcohol-relevant words, sentences, or scenarios in an alcohol-related way, consistent with an interpretation bias. In the current study, we expanded on these findings by including problematic drinkers with mild to borderline intellectual disability (MBID).

In line with previous research (Ames et al., 2005; Krank et al., 2010; Salemink & Wiers, 2014; Woud et al., 2012, 2014), we found problematic drinkers with and without MBID to have an interpretation bias toward alcohol. In addition, this bias was significantly stronger in problematic drinkers compared to light drinkers and correlated positively with the severity of alcohol use–related problems. Hence, the strength of the interpretation bias increases with higher levels of alcohol use–related problems. When exploring the dynamics of the interpretation bias in more detail, we found the differences in strength of the bias between light and problematic drinkers to be strongest for the negative scenarios. The bias score for negative scenarios made a substantially bigger contribution to the prediction of the severity of alcohol use–related problems than the bias score for positive scenarios. Although speculative, these results appear in line with literature on drinking motives, which has shown that problematic drinkers—both with MBID (Didden, Embregts, Van der Toorn, & Laarhoven, 2009; Taggart, McLaughlin, Quinn, & McFarlane, 2007) and without MBID (Kuntsche et al., 2014; Mezquita et al., 2011)—frequently drink alcohol to cope with stress and other negative emotions. But note that enhancement motives have also frequently been found to be related to problematic drinking (Cadigan, Martens, & Hermans, 2015). Coping drinkers might have formed an association between experiencing unpleasant emotions and tension reduction by repeatedly drinking alcohol in response to negative scenarios or situations. These situations can then trigger the activation of alcohol-related associations. Indeed, Salemink and Wiers (2014) and Woud et al. (2015a) showed that social/enhancement drinking motives predict the strength of the interpretation bias for positive situations, whereas coping drinking motives predicted the strength of bias score for negative situations.

Although we did not hypothesize a specific role of full-scale IQ on the strength of the interpretation bias, we found the total bias score and the bias score for negative scenarios of light and problematic drinkers to be higher among individuals with MBID compared to individuals without MBID. Similarly, the strength of the bias score correlated negatively with estimated full-scale IQ. These results remained when controlling for estimated verbal IQ, suggesting that vocabulary and verbal reasoning do not play a role in the assessment of the interpretation bias. One explanation for the stronger interpretation bias in individuals with MBID is that participants were aware they were participating in a research study on substance use, which could have activated the cluster of associations related to this topic. As individuals with MBID are often more vulnerable to probing questions and are more likely to please others, this could have resulted in them responding in accordance with the research goals more often than individuals without MBID (Finlay & Lyons, 2001, 2002).

From a clinical perspective, these results imply that the scenario-based approach provides a new method to assess the severity of alcohol use–related problems. For example, this method might provide a less biased way of assessing high-risk situations for alcohol use or relapse compared to typical questionnaires (Ranganath et al., 2008; Woud et al., 2012) and could therefore be incorporated into prevention programs. Measuring the interpretation bias also provides new ways to treat substance-use disorder (SUD). That is, treatment interventions could focus on directly changing the interpretation bias by training problematic drinkers to interpret ambiguous alcohol-related cues in a neutral manner (cf. Kelly, Masterman, & Marlatt, 2005). Such treatment procedures have been shown to be promising in the field of anxiety (Amir & Taylor, 2012; Salemink, Van den Hout, & Kindt, 2009), and the feasibility has recently been generalized to the treatment of SUD (Woud et al., 2015b). It should be noted, however, that cognitive bias modification procedures have also been criticized, both in the field of anxiety and depression (Hallion & Ruscio, 2011; Mogoase, David, & Koster, 2014) as well as in the field of SUD (Christiansen, Schoenmakers, & Field, 2015; Field, Marhe, & Franken, 2013).

This study has several limitations that lead to suggestions for future research. First, participants were aware they were participating in a study on alcohol use. This could have biased their response, for example by censoring their responses in line or in contrast with the research goals. Although the awareness score did not correlate with the interpretation bias score in a study by Woud et al. (2012), future research needs to take this into account as this could explain the stronger interpretation bias found in individuals with MBID. Second, a cross-sectional approach was used, which does not allow us to study the causality in the relationship between the interpretation bias and drinking behaviour. In future studies, a prospective design should be adopted to investigate the predictive validity of the interpretation bias on alcohol consumption (and vice versa). This would increase our knowledge of the role the interpretation bias plays in the development and maintenance of SUD. Third, we specifically studied the interpretation bias in problematic drinkers. To be able to draw conclusions about the role of the interpretation bias in the etiology of SUD in individuals with and without MBID in general, future research should focus on generalizing these results to other substances and studying individual differences in the strength of the interpretation bias in more detail. For example, the role of executive control (e.g., inhibitory control, working memory; Burton, Pedersen, & McCarthy, 2012; Peeters et al., 2012; Van Hemel-Ruiter, Wiers, Brook, & De Jong, 2016), craving (Field, Munafo, & Franken, 2009), poly-substance use (Marks, Pike, Stoops, & Rush, 2015) and co-morbid psychiatric disorders (Sinclair, Nausheen, Garner, & Baldwin, 2010) have been studied in relation to other cognitive biases in problematic drinkers, but have not yet been explored in relation to the interpretation bias and could therefore be examined in future studies.

To conclude, our results show an interpretation bias toward alcohol in problematic drinkers both with and without MBID. Problematic drinkers with and without MBID tend to interpret ambiguous, alcohol-relevant scenarios in an alcohol-related way and this tendency increases with higher levels of alcohol use–related problems. The differences between light and problematic drinkers are largest in negative scenarios. Participants with MBID showed a stronger interpretation bias compared to participants without MBID, especially on the negative scenarios. These results add to the knowledge base on the underlying mechanisms of SUD and provide new lines of inquiry for the identification, assessment, and treatment of SUD in clinical samples.

Acknowledgments

Support from several Dutch organisations within ID care (Arduin, Aveleijn, Ipse de Bruggen, Leekerweide, Pluryn, Trajectum) and addiction medicine (Omnizorg, Tactus) is gratefully acknowledged.

Notes

¹ We controlled for booklet number and gender in all analyses, but they had no effect. Therefore, only the results without booklet number and gender as controlling variables are reported.