A Declaration of Independence: Implicit Alcohol Associations Have Independent, not Interactive, Relationships with Alcohol Consumption and AUD Risk

A Declaration of Independence: Implicit Alcohol Associations Have Independent, not Interactive,... Abstract Aims The current study aimed to test for potential interactive effects of three implicit alcohol-related associations (drinking identity, alcohol approach and alcohol excitement) in predicting concurrent and prospective alcohol consumption and risk of alcohol use disorders (AUDs) in two samples of the US undergraduate drinkers and non-drinkers. Short summary We investigated the independent and interactive effects of three implicit associations on alcohol consumption and risk of AUD in two US undergraduate student samples. We found that implicit associations had independent but not interactive effects on concurrent and subsequent alcohol consumption and risk of AUD in two independent samples. Methods Implicit drinking identity, alcohol approach and alcohol excitement associations were assessed in two US undergraduate student samples (Sample 1: N = 300, 55% female; Sample 2: N = 506, 57% female). Alcohol consumption and risk of AUD were assessed at baseline (Samples 1 and 2) and 3 months later (Sample 2). We fit zero-inflated negative binomial models to test for independent and interactive effects of the three implicit associations on alcohol consumption and risk of AUD. Results Although we found multiple, unique main effects for alcohol associations, we found minimal evidence of interactions between implicit alcohol-related associations. There was no reliable evidence of interactions in models in predicting alcohol consumption or risk of AUD, concurrently or prospectively, in either sample. Conclusions Contrary to expectations, results from both studies indicated that implicit alcohol-related associations in the US undergraduate samples generally have independent, not interactive, relationships with alcohol consumption and risk of AUD. INTRODUCTION Although more than half of the US students attending college are under the minimum legal drinking age of 21 (American College Health Association, 2012), 78% report lifetime alcohol use and 40% report having been drunk in the past 30 days (Johnston et al., 2015). The US college years are a time of both initiation of drinking for a substantial proportion of student (Fromme et al., 2008) as well as escalation of alcohol consumption (see Naimi et al., 2003). Moreover, 36% of the US college students report engaging in heavy episodic drinking (i.e. five or more drinks) at least once in the previous 2 weeks (Johnston et al., 2015). Heavy drinking during the college years prospectively predicts risk of alcohol use disorders (AUDs; O’Neill et al., 2001), further cementing college-age drinking as a major public health concern. Recent theoretical models of drinking emphasize the joint contribution of explicit and implicit (automatic/reflexive/impulsive) cognitive factors in alcohol use and misuse (see Wiers et al., 2007; Stacy and Wiers, 2010). Multiple implicit cognitive factors, including implicit associations about alcohol and drinking (i.e. implicit alcohol-related associations), have found to be robust predictors of college students’ alcohol use and misuse even after controlling for explicit measures (see Wiers et al., 2002; Reich et al., 2010; Lindgren et al., 2013, 2016a). As research on implicit alcohol-related associations has matured, attention has turned to identifying potential moderators of the relationship between implicit associations and drinking, such as cognitive capacity or mood (see Hofmann et al., 2008). One set of potential moderators that has not been considered are implicit alcohol-related associations, themselves. That is, there are multiple, different implicit alcohol-related associations and whether they might have interactive effects on alcohol consumption and risk of AUD has not, to our knowledge, been evaluated. Thus, the current study tested for possible interactive effects among well-established implicit alcohol-related associations on measures of alcohol use and misuse. Implicit Alcohol-related Associations Multiple implicit alcohol-related associations have been studied (see Rooke et al., 2008; Reich et al., 2010), with implicit drinking identity, implicit alcohol excitement and implicit alcohol approach emerging as key predictors of alcohol outcomes among the US college students (Lindgren et al., 2016a). First, implicit drinking identity—or the extent to which one associates drinking with the self—has been found to be a consistent predictor of college student alcohol consumption and risk of AUD, both cross-sectionally (Lindgren et al., 2013) and over time (see Gray et al., 2011; Lindgren et al., 2016a, 2016b). The theoretical interest in implicit drinking identity stems from social and cognitive psychology theories emphasizing the importance of the constructs that become associated with the self (e.g. Greenwald et al., 2002; Back et al., 2009), such as drinking behaviors and/or drinking social groups, positing that those associations can become unique, important drivers of drinking behavior and risk of AUD (see Dingle et al., 2015; Lindgren et al., 2017). Second, implicit alcohol excitement—or the extent to which one associates alcoholic beverages with excitement—has been linked to college alcohol use previously (Lindgren et al., 2013) and is similar to Wiers et al.’s (2002) alcohol arousal Implicit Association Test (IAT). Theoretically, since enhancement of positive mood is cited as one of the most common reasons for drinking among young adults, alcohol-excitement associations may represent an implicit expression of intention to drink alcohol for this specific reason (see Houben et al., 2010; Lindgren et al., 2013). Third, implicit alcohol-approach associations—or the extent to which one associates alcoholic beverages with approach—are assumed to reflect an individual’s appetitive inclinations for acquiring and consuming alcohol (Ostafin and Palfai, 2006). Conceptually, alcohol-approach inclinations could be viewed as analogous to the ‘wanting’ (versus ‘liking’) alcohol that occurs during compulsive alcohol use (see Berridge et al., 2009). They have been shown to be associated with alcohol consumption among adolescents and young adults (Janssen et al., 2015a; Lindgren et al., 2016a), and have been successfully targeted in interventions (Wiers et al., 2010). There is also some evidence that measures of these associations are not redundant (they are only weakly correlated with one another) and they have unique effects on alcohol use and risk of AUD in the US undergraduates (see Lindgren et al., 2016a). Collectively, they may provide a comprehensive view of key implicit associations related to drinking for this population: these associations focus on associations about alcohol as a substance (implicit alcohol approach and excitement, can be viewed as analogous to wanting and liking) and behavior linked to the self (implicit drinking identity). Simultaneously, theories about what implicit associations are, and how they affect behavior, suggest the possibility that they may also have interactive effects. Implicit associations have been conceptualized as a network of associations stored in memory, with constructs (e.g. ‘alcohol’, ‘excitement’, ‘approach’ and ‘me’) representing nodes in that network (e.g. Greenwald et al., 2002; Lindgren et al., 2017; but see De Houwer, 2014, for an alternative view). Furthermore, the activation of one (or more) set of associations could spread to other associations (i.e. spreading activation). Hence, because both implicit alcohol approach and alcohol excitement associations focus on alcohol, one might expect that activation of one would lead to activation of the other (and vice versa) and they could have an increased effect on drinking behavior. Thus, these associations could interact and have multiplicative effects on alcohol consumption. Further justification for hypothesizing interactive effects stems from social psychologies of the self (e.g. Markus and Wurf, 1987; Greenwald et al., 2002), which suggest that the self is a key, organizing/central aspect of mental constructs and guides information processing (i.e. self-relevant info). Thus, the degree to which one has strong alcohol-related associations combined with a strong drinking identity might be expected to boost alcohol-related information processing (and spreading activation) and further increase inclinations towards drinking. Study Overview The goal of the current study was to evaluate whether scores on measures of implicit alcohol-related associations (i.e. drinking identity, alcohol excitement and alcohol approach associations) have interactive effects on college student alcohol consumption and risk of AUD. We tested the following hypothesis: implicit associations would amplify one other—i.e. have multiplicative effects—and thus have an even stronger relationship on alcohol consumption and risk of AUD beyond their previously shown individual effects. We did so via secondary data analysis of two independent samples of the US undergraduates (one cross-sectional, Lindgren et al., 2013 and one longitudinal, Lindgren et al., 2016a). We expected that each of the three two-way interactions would be significant and that the pattern of the interactions would be similar and indicate multiplicative effects in the prediction of alcohol consumption/risk status and amount of alcohol consumption/risk, concurrently and prospectively. Because the undergraduate years in the US are associated with both initiation and escalation of drinking and there is evidence that implicit alcohol associations can predate the initiation of drinking (see Van Der Vorst et al., 2013; Janssen et al., 2015a), both samples included drinkers and non-drinkers. Therefore, we also expected these effects to differentiate drinkers and non-drinkers, in the same way that they differentiate greater versus smaller alcohol consumption and risk for AUD. The studies used identical measures of implicit alcohol-related associations and alcohol consumption and risk of AUD. METHODS Procedures All procedures were approved by the university’s (a large Pacific Northwest public university) Institutional Review Board. Sample 1 A randomly sampled list of 18- to 25-year-old undergraduates was obtained from the Registrar’s Office. Participants were recruited via email but completed the study in the laboratory. Written informed consent was obtained during the lab session. Up to four participants shared a room at a time. Partitions separated individuals and privacy screens were placed over the laptops to minimize risks to privacy and confidentiality. All measures were computer-based and presented in random order. Participants were compensated $30. Sample 2 A randomly sampled list of full-time students in their first or second undergraduate year, ages 18–20, was obtained from the university’s Registrar’s Office. Students were recruited via email for a 2-year online study of alcohol and cognition. Informed consent and assessments were administered online, on participants’ choice of computer. Assessments consisted of the measures of implicit alcohol associations and alcohol outcomes. Participants were compensated $25. Data for the current study come from the first (enrollment) assessment and the first follow-up after 3 months. Eighty-six percent (n = 437) of participants completed this follow-up assessment. Missing the follow-up assessment was not associated with any baseline sociodemographic characteristic (gender, age, race and ethnicity) but was significantly predicted by greater baseline risk of AUD (mean Alcohol Use Disorders Identification Test (AUDIT, Babor et al., 2001) score not missing: 4.25, mean AUDIT score missing: 6.49, t(1,502) = −3.260, P < 0.01); see Lindgren et al., 2016a, for more details on attrition. Participants Sample 1 Participants (N = 300, 55% women) were between the ages of 18 and 25 (M = 20.47, SD = 1.52). Fifty-seven percent identified as White/Caucasian, 30% as Asian, 9% as multiracial and the remaining 4% as other, or declined to answer. Sample 2 Participants (N = 506, 57% women) were first- and second-year undergraduates, between the ages of 18 and 20 (M = 18.57, SD = 0.69). Fifty-two percent identified as White/Caucasian, 31% as Asian, 11% as multiracial and the remaining 6% as other, or declined to answer. Measures and Materials Implicit alcohol associations We used three IATs (Greenwald et al., 1998) to evaluate implicit alcohol associations. IATs are computer-based reaction time tasks used to measure the strength of associations between concepts, relative to an alternative (see target constructs below). A series of words and/or pictures are presented center-screen, and participants are instructed to press the corresponding key (d and k in Sample 1; e and i in Sample 2) to classify each stimulus into one of two contrasting target or attribute categories, or pairs of categories, as quickly and accurately as possible. Scores were calculated using the D score algorithm (Greenwald et al., 2003). IAT data were screened (per Nosek et al., 2007) and IATs with ≥10% trials faster than 300 ms or ≥30% trials with errors were considered invalid. Implementation of the drinking identity IAT (Lindgren et al., 2013), the alcohol approach IAT (Ostafin and Palfai, 2006) and the alcohol excitement IAT (Lindgren et al., 2011) was previously described for Sample 1 (Lindgren et al., 2013) and Sample 2 (Lindgren et al., 2016a). Higher scores indicated stronger associations between ‘alcohol’ and ‘me’ (versus ‘not me’), between ‘alcohol’ and ‘approach’ (versus ‘avoid’), and between ‘alcohol’ and ‘excite’ (versus ‘depress’), respectively. Internal consistencies (calculated by correlating D scores from Blocks 3 and 6 with D scores from Blocks 4 and 7) were: drinking identity, 0.51 (Sample 1), 0.58, (Sample 2); alcohol approach, 0.48 (Sample 1), 0.55, (Sample 2) and alcohol excitement, 0.52 (Sample 1), 0.57, (Sample 2). Alcohol consumption The Daily Drinking Questionnaire (DDQ; Collins et al., 1985) assesses individuals’ typical alcohol consumption over the past 3 months. Participants were asked to report how many standard drinks they consumed each day of the week. Participants were provided with a list of US standard drink equivalencies. Daily consumption was summed to reflect total drinks per week. Risk of AUD The AUDIT (Babor et al., 2001) is a 10-item measure used to assess alcohol-related problems and individuals’ risk of AUDs, considering the past year. Items were summed. Higher scores indicate a greater risk of an alcohol use disorder. Cronbach’s alphas were: 0.80, Sample 1 and 0.83, Sample 2. Analytic Strategy Analyses were run separately for each sample so that we could evaluate whether results were consistent. Both samples included drinkers and non-drinkers, and preliminary analyses showed a large number of zero scores on alcohol outcomes (Sample 1: 21.7% and Sample 2: 50.3% at enrollment, 47.4% at follow-up). We elected to retain non-drinkers in analyses because of preliminary evidence that implicit associations predate initiation of drinking and this is a period when initiation and escalation of drinking occurs. Further, their distributions were positively skewed. Therefore, data were modeled using a zero-inflated negative binomial (ZINB) distribution for alcohol consumption and risk of AUD, consisting of a count portion and an inflated portion. This approach addresses both the large number of zeros and the positively skewed distributions. In ZINB, two simultaneous regressions are conducted. The first—the inflated portion—predicts the probability of being one of the large number of zeros the distribution, which one can conceptualize of predicting the probability of being a ‘true’ non-drinker (or not at risk of AUD) versus the probability of being a drinker (or at risk of AUD). The second—the count portion–predicts the remainder of the distribution (i.e. the full range of alcohol consumption and risk of AUD). (For a detailed and informative overview of count distributions, see Atkins et al., 2013). We ran a series of stepwise zero-inflated negative binomial regression models predicting concurrent alcohol consumption and risk of AUD (Samples 1 and 2) and prospective alcohol consumption and risk of AUD (Sample 2), investigating the interactive effect of the three implicit associations. In Step 1, we entered the main effects of the three implicit associations. In Step 2, we entered the three possible two-way interactions between the three associations. All analyses controlled for gender and race. We evaluated the significance of interactive effects and significance of changes in model fit (based on two indexes comparing the model with main effects to the model with interactions included: loglikelihood chi-square and sample size-adjusted Bayesian Information Criterion, sBIC; Desmarais and Harden (2013)). All analyses were performed in Mplus 7.2. Missing data was handled using Full Information Maximum Likelihood under assumption of missing at random (Enders, 2011). Finally, for longitudinal models using Sample 2, we investigated prospective changes in alcohol consumption and risk of AUD by predicting Time 2 alcohol consumption and risk of AUD, while controlling for Time 1 alcohol consumption and risk of AUD, respectively, along with all previously included covariates. RESULTS Table 1 shows the means and correlations between implicit associations and alcohol outcome variables for each sample, as well as descriptive details on sociodemographic and alcohol consumption details of each sample. Table 2 shows the regression results for the stepwise ZINB models with DDQ and AUDIT scores as outcome, for both samples. While we note main effects, our main focus is on the presence or absence of interaction as those tests are the primary aim of the current study. Table 1. Descriptive statistics and correlations for variables for Sample 1 and Sample 2 Measure 1 2 3 4 5 6 7 8 9 10 M SD 1. Drinking identity IAT – 0.13* 0.21*** 0.32*** 0.34*** −0.02 −0.23*** 0.09 −0.03 −0.08 0.03 0.39 2. Alcohol approach IAT 0.22*** – 0.44*** 0.15* 0.23*** −0.03 −0.02 0.08 −0.03 −0.08 –0.15 0.35 3. Alcohol excitement IAT 0.13* 0.39*** – 0.24*** 0.26*** −0.03 −0.07 0.06 0.00 −0.09 –0.01 0.41 4. Alcohol consumption 0.38*** 0.20** 0.16** – 0.76*** −0.12* −0.20** 0.25** −0.23** −0.06 8.22 9.96 5. Risk of AUD 0.44*** 0.28*** 0.20*** 0.79*** – −0.05 0.15** 0.23** −0.22** −0.03 6.39 5.53 6. Age 0.05 −0.03 −0.02 0.19*** 0.22*** – – 0.07 −0.06 −0.02 20.47 1.52 7. Gender −0.13** −0.11* −0.12* 0.00 0.01 0.06 – −0.08 0.01 0.11 55% F – 8. Race: White 0.10* 0.02 0.11* 0.15** 0.18** −0.06 −0.03 – – – 57% – 9. Race: Asian-American −0.09 −0.07 −0.17** −0.18 −0.21 −0.04 0.00 – – – 30% – 10. Race: Othera −0.03 0.05 0.06 0.02 0.02 0.12** 0.03 – – – 14% – M −0.08 −0.21 −0.10 5.10 4.37 18.57 57% F 52% 31% 17% – – SD 0.45 0.40 0.46 8.54 5.07 0.69 – – – – – – Measure 1 2 3 4 5 6 7 8 9 10 M SD 1. Drinking identity IAT – 0.13* 0.21*** 0.32*** 0.34*** −0.02 −0.23*** 0.09 −0.03 −0.08 0.03 0.39 2. Alcohol approach IAT 0.22*** – 0.44*** 0.15* 0.23*** −0.03 −0.02 0.08 −0.03 −0.08 –0.15 0.35 3. Alcohol excitement IAT 0.13* 0.39*** – 0.24*** 0.26*** −0.03 −0.07 0.06 0.00 −0.09 –0.01 0.41 4. Alcohol consumption 0.38*** 0.20** 0.16** – 0.76*** −0.12* −0.20** 0.25** −0.23** −0.06 8.22 9.96 5. Risk of AUD 0.44*** 0.28*** 0.20*** 0.79*** – −0.05 0.15** 0.23** −0.22** −0.03 6.39 5.53 6. Age 0.05 −0.03 −0.02 0.19*** 0.22*** – – 0.07 −0.06 −0.02 20.47 1.52 7. Gender −0.13** −0.11* −0.12* 0.00 0.01 0.06 – −0.08 0.01 0.11 55% F – 8. Race: White 0.10* 0.02 0.11* 0.15** 0.18** −0.06 −0.03 – – – 57% – 9. Race: Asian-American −0.09 −0.07 −0.17** −0.18 −0.21 −0.04 0.00 – – – 30% – 10. Race: Othera −0.03 0.05 0.06 0.02 0.02 0.12** 0.03 – – – 14% – M −0.08 −0.21 −0.10 5.10 4.37 18.57 57% F 52% 31% 17% – – SD 0.45 0.40 0.46 8.54 5.07 0.69 – – – – – – Note. Correlations for Sample 1 (N = 300) are presented above the diagonal; correlations for Sample 2 (N = 506) are presented below the diagonal. Sample 2 drinking outcomes were assessed at baseline. N’s for each correlation vary slightly due to missing data. Means and standard deviations for Study 1 are presented in vertical columns, and those for Study 2 are presented in horizontal rows. IAT = Implicit Association Test; higher scores = stronger associations between the concepts in the IAT’s name. Alcohol consumption = number of drinks consumed in a typical week (assessed in the Daily Drinking Questionnaire). Risk of AUD = score on the Alcohol Use Disorders Identification Test; higher scores = greater risk of an alcohol use disorder. Gender is coded 0 = male; 1 = female. aIn Sample 1, the composition of this ‘other’ group was 1% African-American, 0.7% Native Hawaiian, 1.0 American Indian/Alaska Native and 9.3% multiracial, with 0.7% declining to answer. In Sample 2, this composition was 1.4% African-American, 0.8% American-Indian/Alaska Native, 11.3% multiracial and 3.5% unknown/declined to answer. *P < 0.05, **P < 0.01 and ***P < 0.001. Table 1. Descriptive statistics and correlations for variables for Sample 1 and Sample 2 Measure 1 2 3 4 5 6 7 8 9 10 M SD 1. Drinking identity IAT – 0.13* 0.21*** 0.32*** 0.34*** −0.02 −0.23*** 0.09 −0.03 −0.08 0.03 0.39 2. Alcohol approach IAT 0.22*** – 0.44*** 0.15* 0.23*** −0.03 −0.02 0.08 −0.03 −0.08 –0.15 0.35 3. Alcohol excitement IAT 0.13* 0.39*** – 0.24*** 0.26*** −0.03 −0.07 0.06 0.00 −0.09 –0.01 0.41 4. Alcohol consumption 0.38*** 0.20** 0.16** – 0.76*** −0.12* −0.20** 0.25** −0.23** −0.06 8.22 9.96 5. Risk of AUD 0.44*** 0.28*** 0.20*** 0.79*** – −0.05 0.15** 0.23** −0.22** −0.03 6.39 5.53 6. Age 0.05 −0.03 −0.02 0.19*** 0.22*** – – 0.07 −0.06 −0.02 20.47 1.52 7. Gender −0.13** −0.11* −0.12* 0.00 0.01 0.06 – −0.08 0.01 0.11 55% F – 8. Race: White 0.10* 0.02 0.11* 0.15** 0.18** −0.06 −0.03 – – – 57% – 9. Race: Asian-American −0.09 −0.07 −0.17** −0.18 −0.21 −0.04 0.00 – – – 30% – 10. Race: Othera −0.03 0.05 0.06 0.02 0.02 0.12** 0.03 – – – 14% – M −0.08 −0.21 −0.10 5.10 4.37 18.57 57% F 52% 31% 17% – – SD 0.45 0.40 0.46 8.54 5.07 0.69 – – – – – – Measure 1 2 3 4 5 6 7 8 9 10 M SD 1. Drinking identity IAT – 0.13* 0.21*** 0.32*** 0.34*** −0.02 −0.23*** 0.09 −0.03 −0.08 0.03 0.39 2. Alcohol approach IAT 0.22*** – 0.44*** 0.15* 0.23*** −0.03 −0.02 0.08 −0.03 −0.08 –0.15 0.35 3. Alcohol excitement IAT 0.13* 0.39*** – 0.24*** 0.26*** −0.03 −0.07 0.06 0.00 −0.09 –0.01 0.41 4. Alcohol consumption 0.38*** 0.20** 0.16** – 0.76*** −0.12* −0.20** 0.25** −0.23** −0.06 8.22 9.96 5. Risk of AUD 0.44*** 0.28*** 0.20*** 0.79*** – −0.05 0.15** 0.23** −0.22** −0.03 6.39 5.53 6. Age 0.05 −0.03 −0.02 0.19*** 0.22*** – – 0.07 −0.06 −0.02 20.47 1.52 7. Gender −0.13** −0.11* −0.12* 0.00 0.01 0.06 – −0.08 0.01 0.11 55% F – 8. Race: White 0.10* 0.02 0.11* 0.15** 0.18** −0.06 −0.03 – – – 57% – 9. Race: Asian-American −0.09 −0.07 −0.17** −0.18 −0.21 −0.04 0.00 – – – 30% – 10. Race: Othera −0.03 0.05 0.06 0.02 0.02 0.12** 0.03 – – – 14% – M −0.08 −0.21 −0.10 5.10 4.37 18.57 57% F 52% 31% 17% – – SD 0.45 0.40 0.46 8.54 5.07 0.69 – – – – – – Note. Correlations for Sample 1 (N = 300) are presented above the diagonal; correlations for Sample 2 (N = 506) are presented below the diagonal. Sample 2 drinking outcomes were assessed at baseline. N’s for each correlation vary slightly due to missing data. Means and standard deviations for Study 1 are presented in vertical columns, and those for Study 2 are presented in horizontal rows. IAT = Implicit Association Test; higher scores = stronger associations between the concepts in the IAT’s name. Alcohol consumption = number of drinks consumed in a typical week (assessed in the Daily Drinking Questionnaire). Risk of AUD = score on the Alcohol Use Disorders Identification Test; higher scores = greater risk of an alcohol use disorder. Gender is coded 0 = male; 1 = female. aIn Sample 1, the composition of this ‘other’ group was 1% African-American, 0.7% Native Hawaiian, 1.0 American Indian/Alaska Native and 9.3% multiracial, with 0.7% declining to answer. In Sample 2, this composition was 1.4% African-American, 0.8% American-Indian/Alaska Native, 11.3% multiracial and 3.5% unknown/declined to answer. *P < 0.05, **P < 0.01 and ***P < 0.001. Table 2. Results from concurrent zero-inflated negative binomial regression Sample 1 Sample 2 Count B e^B [95% CI] P Zero-inflated P Count B e^B [95% CI] P Zero-inflated P OR [95% CI] OR [95% CI] Risk of AUD  Main effects   Intercept 1.70 5.45 [4.54–6.54] <0.001 0.41 [0.25–0.66] <0.001 2.03 7.61 [6.50–8.90] <0.001 0.06 [0.02–0.15] <0.001   Excite 0.13 1.13 [1.00–1.28] <0.05 0.92 [0.64–1.31] 0.64 0.07 1.08 [0.97–1.19] 0.15 0.55 [0.32–0.94] <0.05   Identity 0.33 1.40 [1.26–1.55] <0.001 0.56 [0.41–0.78] <0.001 0.21 1.23 [1.12–1.35] <0.001 0.91 [0.55–1.52] 0.72   Approach 0.05 1.06 [0.94–1.19] 0.38 0.56 [0.39–0.81] <0.001 0.12 1.12 [1.01–1.25] <0.05 0.78 [0.42–1.45] 0.43   Gender (Female) 0.04 1.04 [0.84–1.27] 0.74 0.47 [0.26–0.84] <0.01 −0.12 0.89 [0.74–1.08] 0.23 1.11 [0.39–3.17] 0.85  Race (Ref: White)   Asian −0.28 0.76 [0.59–0.97] <0.05 2.03 [1.10–3.73] <0.05 −0.30 0.74 [0.59–0.93] <0.01 3.31 [1.18–9.25] <0.05   Other −0.04 0.96 [0.73–1.25] 0.75 0.71 [0.26–1.92] 0.50 −0.09 0.92 [0.67–1.24] 0.57 0.70 [0.05–9.26] 0.79  Interactions   Excite*Identity −0.10 0.91 [0.61–1.34] 0.63 0.91 [0.80–1.03] 0.14 0.03 1.03 [0.93–1.14] 0.60 1.19 [0.72–1.98] 0.49   App*Identity 0.17 1.19 [0.76–1.86] 0.45 1.02 [0.90–1.16] 0.76 −0.06 0.94 [0.86–1.03] 0.17 0.79 [0.46–1.37] 0.41   App*Excite −0.62 0.54 [0.35–0.84] <0.01 0.95 [0.86–1.06] 0.36 −0.09 0.91 [0.83–1.00] 0.06 1.13 [0.67–1.90] 0.66 Alcohol consumption  Main effects   Intercept 2.13 8.40 [6.57–10.73] <0.001 1.23 [0.80–1.89] 0.35 2.42 11.22 [9.24–13.63] <0.001 0.10 [0.05–0.23] <0.001   Excite 0.10 1.10 [0.95–1.28] 0.21 0.82 [0.62–1.07] 0.15 0.06 1.06 [0.93–1.21] 0.40 0.44 [0.27–0.70] <0.001   Identity 0.38 1.46 [1.26–1.70] <0.001 0.49 [0.38–0.65] <0.001 0.22 1.25 [1.10–1.41] <0.001 0.74 [0.50–1.12] 0.16   Approach −0.05 0.95 [0.82–1.11] 0.52 0.55 [0.42–0.72] <0.001 0.15 1.17 [1.02–1.33] <0.05 1.12 [0.71–1.76] 0.62   Gender (Female) 0.00 1.00 [0.75–1.32] 0.98 0.54 [0.32–0.89] <0.05 −0.34 0.71 [0.56–0.90] <0.01 0.80 [0.35–1.81] 0.59  Race (Ref: White)   Asian −0.28 0.76 [0.54–1.05] 0.09 1.41 [0.82–2.45] 0.22 −0.40 0.67 [0.50–0.90] <0.01 4.32 [1.78–10.48] <0.001   Other −0.11 0.90 [0.63–1.28] 0.55 0.82 [0.40–1.69] 0.59 0.06 1.06 [0.71–1.59] 0.78 3.01 [0.98–9.28] 0.06  Interactions   Excite*Identity −0.18 0.84 [0.70–1.00] 0.05 1.04 [0.77–1.41] 0.80 0.03 1.03 [0.90–1.18] 0.67 0.43 [0.26–0.72] <0.001   App*Identity 0.01 1.01 [0.85–1.19] 0.95 0.97 [0.71–1.32] 0.84 −0.10 0.91 [0.81–1.02] 0.11 0.74 [0.46–1.20] 0.23   App*Excite −0.06 0.94 [0.83–1.07] 0.36 0.80 [0.59–1.07] 0.13 −0.06 0.94 [0.83–1.07] 0.36 1.37 [0.81–2.31] 0.25 Sample 1 Sample 2 Count B e^B [95% CI] P Zero-inflated P Count B e^B [95% CI] P Zero-inflated P OR [95% CI] OR [95% CI] Risk of AUD  Main effects   Intercept 1.70 5.45 [4.54–6.54] <0.001 0.41 [0.25–0.66] <0.001 2.03 7.61 [6.50–8.90] <0.001 0.06 [0.02–0.15] <0.001   Excite 0.13 1.13 [1.00–1.28] <0.05 0.92 [0.64–1.31] 0.64 0.07 1.08 [0.97–1.19] 0.15 0.55 [0.32–0.94] <0.05   Identity 0.33 1.40 [1.26–1.55] <0.001 0.56 [0.41–0.78] <0.001 0.21 1.23 [1.12–1.35] <0.001 0.91 [0.55–1.52] 0.72   Approach 0.05 1.06 [0.94–1.19] 0.38 0.56 [0.39–0.81] <0.001 0.12 1.12 [1.01–1.25] <0.05 0.78 [0.42–1.45] 0.43   Gender (Female) 0.04 1.04 [0.84–1.27] 0.74 0.47 [0.26–0.84] <0.01 −0.12 0.89 [0.74–1.08] 0.23 1.11 [0.39–3.17] 0.85  Race (Ref: White)   Asian −0.28 0.76 [0.59–0.97] <0.05 2.03 [1.10–3.73] <0.05 −0.30 0.74 [0.59–0.93] <0.01 3.31 [1.18–9.25] <0.05   Other −0.04 0.96 [0.73–1.25] 0.75 0.71 [0.26–1.92] 0.50 −0.09 0.92 [0.67–1.24] 0.57 0.70 [0.05–9.26] 0.79  Interactions   Excite*Identity −0.10 0.91 [0.61–1.34] 0.63 0.91 [0.80–1.03] 0.14 0.03 1.03 [0.93–1.14] 0.60 1.19 [0.72–1.98] 0.49   App*Identity 0.17 1.19 [0.76–1.86] 0.45 1.02 [0.90–1.16] 0.76 −0.06 0.94 [0.86–1.03] 0.17 0.79 [0.46–1.37] 0.41   App*Excite −0.62 0.54 [0.35–0.84] <0.01 0.95 [0.86–1.06] 0.36 −0.09 0.91 [0.83–1.00] 0.06 1.13 [0.67–1.90] 0.66 Alcohol consumption  Main effects   Intercept 2.13 8.40 [6.57–10.73] <0.001 1.23 [0.80–1.89] 0.35 2.42 11.22 [9.24–13.63] <0.001 0.10 [0.05–0.23] <0.001   Excite 0.10 1.10 [0.95–1.28] 0.21 0.82 [0.62–1.07] 0.15 0.06 1.06 [0.93–1.21] 0.40 0.44 [0.27–0.70] <0.001   Identity 0.38 1.46 [1.26–1.70] <0.001 0.49 [0.38–0.65] <0.001 0.22 1.25 [1.10–1.41] <0.001 0.74 [0.50–1.12] 0.16   Approach −0.05 0.95 [0.82–1.11] 0.52 0.55 [0.42–0.72] <0.001 0.15 1.17 [1.02–1.33] <0.05 1.12 [0.71–1.76] 0.62   Gender (Female) 0.00 1.00 [0.75–1.32] 0.98 0.54 [0.32–0.89] <0.05 −0.34 0.71 [0.56–0.90] <0.01 0.80 [0.35–1.81] 0.59  Race (Ref: White)   Asian −0.28 0.76 [0.54–1.05] 0.09 1.41 [0.82–2.45] 0.22 −0.40 0.67 [0.50–0.90] <0.01 4.32 [1.78–10.48] <0.001   Other −0.11 0.90 [0.63–1.28] 0.55 0.82 [0.40–1.69] 0.59 0.06 1.06 [0.71–1.59] 0.78 3.01 [0.98–9.28] 0.06  Interactions   Excite*Identity −0.18 0.84 [0.70–1.00] 0.05 1.04 [0.77–1.41] 0.80 0.03 1.03 [0.90–1.18] 0.67 0.43 [0.26–0.72] <0.001   App*Identity 0.01 1.01 [0.85–1.19] 0.95 0.97 [0.71–1.32] 0.84 −0.10 0.91 [0.81–1.02] 0.11 0.74 [0.46–1.20] 0.23   App*Excite −0.06 0.94 [0.83–1.07] 0.36 0.80 [0.59–1.07] 0.13 −0.06 0.94 [0.83–1.07] 0.36 1.37 [0.81–2.31] 0.25 Note. Risk of AUD is assessed as the sum of items on the AUDIT scale. Alcohol Consumption is assessed as the sum of drinks per day on the Daily Drinking Questionnaire. Both outcomes are assumed zero-inflated negative binomial distributed. For the prediction of the zero-inflated portion, the coefficients reflect the change in likelihood of being an excess zero. In other words, scores under 1 reduce the likelihood of having lifetime risk of problems (top rows) or the likelihood of being a lifetime drinker (bottom rows). Excite: Implicit Alcohol-Excitement Associations. Identity: Implicit Alcohol-Identity Associations. App: Implicit Alcohol-Approach Associations. Table 2. Results from concurrent zero-inflated negative binomial regression Sample 1 Sample 2 Count B e^B [95% CI] P Zero-inflated P Count B e^B [95% CI] P Zero-inflated P OR [95% CI] OR [95% CI] Risk of AUD  Main effects   Intercept 1.70 5.45 [4.54–6.54] <0.001 0.41 [0.25–0.66] <0.001 2.03 7.61 [6.50–8.90] <0.001 0.06 [0.02–0.15] <0.001   Excite 0.13 1.13 [1.00–1.28] <0.05 0.92 [0.64–1.31] 0.64 0.07 1.08 [0.97–1.19] 0.15 0.55 [0.32–0.94] <0.05   Identity 0.33 1.40 [1.26–1.55] <0.001 0.56 [0.41–0.78] <0.001 0.21 1.23 [1.12–1.35] <0.001 0.91 [0.55–1.52] 0.72   Approach 0.05 1.06 [0.94–1.19] 0.38 0.56 [0.39–0.81] <0.001 0.12 1.12 [1.01–1.25] <0.05 0.78 [0.42–1.45] 0.43   Gender (Female) 0.04 1.04 [0.84–1.27] 0.74 0.47 [0.26–0.84] <0.01 −0.12 0.89 [0.74–1.08] 0.23 1.11 [0.39–3.17] 0.85  Race (Ref: White)   Asian −0.28 0.76 [0.59–0.97] <0.05 2.03 [1.10–3.73] <0.05 −0.30 0.74 [0.59–0.93] <0.01 3.31 [1.18–9.25] <0.05   Other −0.04 0.96 [0.73–1.25] 0.75 0.71 [0.26–1.92] 0.50 −0.09 0.92 [0.67–1.24] 0.57 0.70 [0.05–9.26] 0.79  Interactions   Excite*Identity −0.10 0.91 [0.61–1.34] 0.63 0.91 [0.80–1.03] 0.14 0.03 1.03 [0.93–1.14] 0.60 1.19 [0.72–1.98] 0.49   App*Identity 0.17 1.19 [0.76–1.86] 0.45 1.02 [0.90–1.16] 0.76 −0.06 0.94 [0.86–1.03] 0.17 0.79 [0.46–1.37] 0.41   App*Excite −0.62 0.54 [0.35–0.84] <0.01 0.95 [0.86–1.06] 0.36 −0.09 0.91 [0.83–1.00] 0.06 1.13 [0.67–1.90] 0.66 Alcohol consumption  Main effects   Intercept 2.13 8.40 [6.57–10.73] <0.001 1.23 [0.80–1.89] 0.35 2.42 11.22 [9.24–13.63] <0.001 0.10 [0.05–0.23] <0.001   Excite 0.10 1.10 [0.95–1.28] 0.21 0.82 [0.62–1.07] 0.15 0.06 1.06 [0.93–1.21] 0.40 0.44 [0.27–0.70] <0.001   Identity 0.38 1.46 [1.26–1.70] <0.001 0.49 [0.38–0.65] <0.001 0.22 1.25 [1.10–1.41] <0.001 0.74 [0.50–1.12] 0.16   Approach −0.05 0.95 [0.82–1.11] 0.52 0.55 [0.42–0.72] <0.001 0.15 1.17 [1.02–1.33] <0.05 1.12 [0.71–1.76] 0.62   Gender (Female) 0.00 1.00 [0.75–1.32] 0.98 0.54 [0.32–0.89] <0.05 −0.34 0.71 [0.56–0.90] <0.01 0.80 [0.35–1.81] 0.59  Race (Ref: White)   Asian −0.28 0.76 [0.54–1.05] 0.09 1.41 [0.82–2.45] 0.22 −0.40 0.67 [0.50–0.90] <0.01 4.32 [1.78–10.48] <0.001   Other −0.11 0.90 [0.63–1.28] 0.55 0.82 [0.40–1.69] 0.59 0.06 1.06 [0.71–1.59] 0.78 3.01 [0.98–9.28] 0.06  Interactions   Excite*Identity −0.18 0.84 [0.70–1.00] 0.05 1.04 [0.77–1.41] 0.80 0.03 1.03 [0.90–1.18] 0.67 0.43 [0.26–0.72] <0.001   App*Identity 0.01 1.01 [0.85–1.19] 0.95 0.97 [0.71–1.32] 0.84 −0.10 0.91 [0.81–1.02] 0.11 0.74 [0.46–1.20] 0.23   App*Excite −0.06 0.94 [0.83–1.07] 0.36 0.80 [0.59–1.07] 0.13 −0.06 0.94 [0.83–1.07] 0.36 1.37 [0.81–2.31] 0.25 Sample 1 Sample 2 Count B e^B [95% CI] P Zero-inflated P Count B e^B [95% CI] P Zero-inflated P OR [95% CI] OR [95% CI] Risk of AUD  Main effects   Intercept 1.70 5.45 [4.54–6.54] <0.001 0.41 [0.25–0.66] <0.001 2.03 7.61 [6.50–8.90] <0.001 0.06 [0.02–0.15] <0.001   Excite 0.13 1.13 [1.00–1.28] <0.05 0.92 [0.64–1.31] 0.64 0.07 1.08 [0.97–1.19] 0.15 0.55 [0.32–0.94] <0.05   Identity 0.33 1.40 [1.26–1.55] <0.001 0.56 [0.41–0.78] <0.001 0.21 1.23 [1.12–1.35] <0.001 0.91 [0.55–1.52] 0.72   Approach 0.05 1.06 [0.94–1.19] 0.38 0.56 [0.39–0.81] <0.001 0.12 1.12 [1.01–1.25] <0.05 0.78 [0.42–1.45] 0.43   Gender (Female) 0.04 1.04 [0.84–1.27] 0.74 0.47 [0.26–0.84] <0.01 −0.12 0.89 [0.74–1.08] 0.23 1.11 [0.39–3.17] 0.85  Race (Ref: White)   Asian −0.28 0.76 [0.59–0.97] <0.05 2.03 [1.10–3.73] <0.05 −0.30 0.74 [0.59–0.93] <0.01 3.31 [1.18–9.25] <0.05   Other −0.04 0.96 [0.73–1.25] 0.75 0.71 [0.26–1.92] 0.50 −0.09 0.92 [0.67–1.24] 0.57 0.70 [0.05–9.26] 0.79  Interactions   Excite*Identity −0.10 0.91 [0.61–1.34] 0.63 0.91 [0.80–1.03] 0.14 0.03 1.03 [0.93–1.14] 0.60 1.19 [0.72–1.98] 0.49   App*Identity 0.17 1.19 [0.76–1.86] 0.45 1.02 [0.90–1.16] 0.76 −0.06 0.94 [0.86–1.03] 0.17 0.79 [0.46–1.37] 0.41   App*Excite −0.62 0.54 [0.35–0.84] <0.01 0.95 [0.86–1.06] 0.36 −0.09 0.91 [0.83–1.00] 0.06 1.13 [0.67–1.90] 0.66 Alcohol consumption  Main effects   Intercept 2.13 8.40 [6.57–10.73] <0.001 1.23 [0.80–1.89] 0.35 2.42 11.22 [9.24–13.63] <0.001 0.10 [0.05–0.23] <0.001   Excite 0.10 1.10 [0.95–1.28] 0.21 0.82 [0.62–1.07] 0.15 0.06 1.06 [0.93–1.21] 0.40 0.44 [0.27–0.70] <0.001   Identity 0.38 1.46 [1.26–1.70] <0.001 0.49 [0.38–0.65] <0.001 0.22 1.25 [1.10–1.41] <0.001 0.74 [0.50–1.12] 0.16   Approach −0.05 0.95 [0.82–1.11] 0.52 0.55 [0.42–0.72] <0.001 0.15 1.17 [1.02–1.33] <0.05 1.12 [0.71–1.76] 0.62   Gender (Female) 0.00 1.00 [0.75–1.32] 0.98 0.54 [0.32–0.89] <0.05 −0.34 0.71 [0.56–0.90] <0.01 0.80 [0.35–1.81] 0.59  Race (Ref: White)   Asian −0.28 0.76 [0.54–1.05] 0.09 1.41 [0.82–2.45] 0.22 −0.40 0.67 [0.50–0.90] <0.01 4.32 [1.78–10.48] <0.001   Other −0.11 0.90 [0.63–1.28] 0.55 0.82 [0.40–1.69] 0.59 0.06 1.06 [0.71–1.59] 0.78 3.01 [0.98–9.28] 0.06  Interactions   Excite*Identity −0.18 0.84 [0.70–1.00] 0.05 1.04 [0.77–1.41] 0.80 0.03 1.03 [0.90–1.18] 0.67 0.43 [0.26–0.72] <0.001   App*Identity 0.01 1.01 [0.85–1.19] 0.95 0.97 [0.71–1.32] 0.84 −0.10 0.91 [0.81–1.02] 0.11 0.74 [0.46–1.20] 0.23   App*Excite −0.06 0.94 [0.83–1.07] 0.36 0.80 [0.59–1.07] 0.13 −0.06 0.94 [0.83–1.07] 0.36 1.37 [0.81–2.31] 0.25 Note. Risk of AUD is assessed as the sum of items on the AUDIT scale. Alcohol Consumption is assessed as the sum of drinks per day on the Daily Drinking Questionnaire. Both outcomes are assumed zero-inflated negative binomial distributed. For the prediction of the zero-inflated portion, the coefficients reflect the change in likelihood of being an excess zero. In other words, scores under 1 reduce the likelihood of having lifetime risk of problems (top rows) or the likelihood of being a lifetime drinker (bottom rows). Excite: Implicit Alcohol-Excitement Associations. Identity: Implicit Alcohol-Identity Associations. App: Implicit Alcohol-Approach Associations. Similar to reports in previous work (Lindgren et al., 2013, 2016a), results from the main effects only model (Step 1) indicated that implicit drinking identity was significantly and positively associated with the full range of alcohol consumption and risk of AUD across both samples (as observed in the count portion of the ZINB), and greater drinking identity was significantly associated with a decreased likelihood of being an abstainer (from alcohol consumption) and having never been at risk of AUD in Sample 2 (as observed in the inflated portion of the ZINB). The findings for implicit approach and excitement associations were mixed, both in terms of outcomes and for which Sample they offered significant predictive validity (see Table 2 for details). Significant prediction by drinking identity did not extend to prospective prediction of alcohol consumption and risk of AUD after controlling for baseline alcohol consumption (Notably, differences appear between main effects reported here and those reported in Lindgren and colleagues (2016a) because of differences in the analytic approach. Specifically, in the results from Lindgren and colleagues, data from multiple time points were used in a model that also included a time covariate.). Interactive Effects In Step 2, we added the three two-way interaction terms to the prediction of both the count- and inflated-portions of the models (see Tables 2 and 3). Table 3. Results from prospective zero-inflated negative binomial regression Sample 2 Count B e^B [95% CI] P Zero-inflated P OR [95% CI] Risk of AUD after 3 months  Main effects   Intercept 1.07 2.90 [2.41–3.49] <0.001 1.89 [0.86–4.16] 0.12   T1 alcohol problems 0.11 1.12 [1.10–1.14] <0.001 0.04 [0.01–0.21] <0.001   Excite 0.06 1.06 [0.97–1.17] 0.22 1.15 [0.73–1.83] 0.55   Identity 0.01 1.01 [0.92–1.09] 0.89 1.23 [0.76–2.00] 0.39   Approach −0.03 0.98 [0.89–1.07] 0.60 0.70 [0.42–1.17] 0.17   Gender (female) −0.12 0.88 [0.75–1.05] 0.15 1.26 [0.50–3.15] 0.62  Race (Ref: White)   Asian −0.03 0.97 [0.80–1.17] 0.74 1.78 [0.70–4.56] 0.23   Other −0.24 0.79 [0.62–0.99] <0.05 0.87 [0.20–3.78] 0.86  Interactions   Excite*Identity −0.08 0.92 [0.84–1.01] 0.09 0.68 [0.32 – 1.47] 0.33   App*Identity −0.02 0.98 [0.90–1.07] 0.70 1.82 [0.88–3.76] 0.11   App*Excite −0.05 0.95 [0.88–1.03] 0.21 1.04 [0.65–1.67] 0.86 Alcohol consumption after 3 months  Main effects   Intercept 1.55 4.70 [3.77–5.85] <0.001 2.14 [1.16–3.96] <0.05   T1 alcohol consumption 0.05 1.05 [1.04–1.06] <0.001 0.19 [0.07–0.54] <0.001   Excite 0.10 1.11 [0.98–1.24] 0.09 0.73 [0.49–1.08] 0.11   Identity 0.01 1.01 [0.91–1.13] 0.83 0.77 [0.52–1.14] 0.19   Approach 0.05 1.05 [0.92–1.19] 0.45 0.94 [0.63–1.41] 0.78   Gender (female) −0.08 0.92 [0.74–1.14] 0.45 1.30 [0.64–2.64] 0.47  Race (Ref: White)   Asian −0.33 0.72 [0.56–0.93] <0.01 1.63 [0.73–3.61] 0.23   Other −0.08 0.92 [0.68–1.24] 0.59 1.81 [0.59–5.58] 0.30  Interactions   Excite*Identity −0.09 0.91 [0.79–1.05] 0.20 0.87 [0.56–1.36] 0.54   App*Identity 0.04 1.04 [0.91–1.19] 0.59 0.96 [0.61–1.52] 0.87   App*Excite −0.05 0.96 [0.86–1.06] 0.40 1.03 [0.71–1.49] 0.87 Sample 2 Count B e^B [95% CI] P Zero-inflated P OR [95% CI] Risk of AUD after 3 months  Main effects   Intercept 1.07 2.90 [2.41–3.49] <0.001 1.89 [0.86–4.16] 0.12   T1 alcohol problems 0.11 1.12 [1.10–1.14] <0.001 0.04 [0.01–0.21] <0.001   Excite 0.06 1.06 [0.97–1.17] 0.22 1.15 [0.73–1.83] 0.55   Identity 0.01 1.01 [0.92–1.09] 0.89 1.23 [0.76–2.00] 0.39   Approach −0.03 0.98 [0.89–1.07] 0.60 0.70 [0.42–1.17] 0.17   Gender (female) −0.12 0.88 [0.75–1.05] 0.15 1.26 [0.50–3.15] 0.62  Race (Ref: White)   Asian −0.03 0.97 [0.80–1.17] 0.74 1.78 [0.70–4.56] 0.23   Other −0.24 0.79 [0.62–0.99] <0.05 0.87 [0.20–3.78] 0.86  Interactions   Excite*Identity −0.08 0.92 [0.84–1.01] 0.09 0.68 [0.32 – 1.47] 0.33   App*Identity −0.02 0.98 [0.90–1.07] 0.70 1.82 [0.88–3.76] 0.11   App*Excite −0.05 0.95 [0.88–1.03] 0.21 1.04 [0.65–1.67] 0.86 Alcohol consumption after 3 months  Main effects   Intercept 1.55 4.70 [3.77–5.85] <0.001 2.14 [1.16–3.96] <0.05   T1 alcohol consumption 0.05 1.05 [1.04–1.06] <0.001 0.19 [0.07–0.54] <0.001   Excite 0.10 1.11 [0.98–1.24] 0.09 0.73 [0.49–1.08] 0.11   Identity 0.01 1.01 [0.91–1.13] 0.83 0.77 [0.52–1.14] 0.19   Approach 0.05 1.05 [0.92–1.19] 0.45 0.94 [0.63–1.41] 0.78   Gender (female) −0.08 0.92 [0.74–1.14] 0.45 1.30 [0.64–2.64] 0.47  Race (Ref: White)   Asian −0.33 0.72 [0.56–0.93] <0.01 1.63 [0.73–3.61] 0.23   Other −0.08 0.92 [0.68–1.24] 0.59 1.81 [0.59–5.58] 0.30  Interactions   Excite*Identity −0.09 0.91 [0.79–1.05] 0.20 0.87 [0.56–1.36] 0.54   App*Identity 0.04 1.04 [0.91–1.19] 0.59 0.96 [0.61–1.52] 0.87   App*Excite −0.05 0.96 [0.86–1.06] 0.40 1.03 [0.71–1.49] 0.87 Note. Risk of AUD is assessed as the sum of items on the AUDIT scale. Alcohol consumption is assessed as the sum of drinks per day on the Daily Drinking Questionnaire. Both outcomes are assumed zero-inflated negative binomial distributed. Excite: Implicit Alcohol-Excitement Associations. Identity: Implicit Alcohol-Identity Associations. App: Implicit Alcohol-Approach Associations. Table 3. Results from prospective zero-inflated negative binomial regression Sample 2 Count B e^B [95% CI] P Zero-inflated P OR [95% CI] Risk of AUD after 3 months  Main effects   Intercept 1.07 2.90 [2.41–3.49] <0.001 1.89 [0.86–4.16] 0.12   T1 alcohol problems 0.11 1.12 [1.10–1.14] <0.001 0.04 [0.01–0.21] <0.001   Excite 0.06 1.06 [0.97–1.17] 0.22 1.15 [0.73–1.83] 0.55   Identity 0.01 1.01 [0.92–1.09] 0.89 1.23 [0.76–2.00] 0.39   Approach −0.03 0.98 [0.89–1.07] 0.60 0.70 [0.42–1.17] 0.17   Gender (female) −0.12 0.88 [0.75–1.05] 0.15 1.26 [0.50–3.15] 0.62  Race (Ref: White)   Asian −0.03 0.97 [0.80–1.17] 0.74 1.78 [0.70–4.56] 0.23   Other −0.24 0.79 [0.62–0.99] <0.05 0.87 [0.20–3.78] 0.86  Interactions   Excite*Identity −0.08 0.92 [0.84–1.01] 0.09 0.68 [0.32 – 1.47] 0.33   App*Identity −0.02 0.98 [0.90–1.07] 0.70 1.82 [0.88–3.76] 0.11   App*Excite −0.05 0.95 [0.88–1.03] 0.21 1.04 [0.65–1.67] 0.86 Alcohol consumption after 3 months  Main effects   Intercept 1.55 4.70 [3.77–5.85] <0.001 2.14 [1.16–3.96] <0.05   T1 alcohol consumption 0.05 1.05 [1.04–1.06] <0.001 0.19 [0.07–0.54] <0.001   Excite 0.10 1.11 [0.98–1.24] 0.09 0.73 [0.49–1.08] 0.11   Identity 0.01 1.01 [0.91–1.13] 0.83 0.77 [0.52–1.14] 0.19   Approach 0.05 1.05 [0.92–1.19] 0.45 0.94 [0.63–1.41] 0.78   Gender (female) −0.08 0.92 [0.74–1.14] 0.45 1.30 [0.64–2.64] 0.47  Race (Ref: White)   Asian −0.33 0.72 [0.56–0.93] <0.01 1.63 [0.73–3.61] 0.23   Other −0.08 0.92 [0.68–1.24] 0.59 1.81 [0.59–5.58] 0.30  Interactions   Excite*Identity −0.09 0.91 [0.79–1.05] 0.20 0.87 [0.56–1.36] 0.54   App*Identity 0.04 1.04 [0.91–1.19] 0.59 0.96 [0.61–1.52] 0.87   App*Excite −0.05 0.96 [0.86–1.06] 0.40 1.03 [0.71–1.49] 0.87 Sample 2 Count B e^B [95% CI] P Zero-inflated P OR [95% CI] Risk of AUD after 3 months  Main effects   Intercept 1.07 2.90 [2.41–3.49] <0.001 1.89 [0.86–4.16] 0.12   T1 alcohol problems 0.11 1.12 [1.10–1.14] <0.001 0.04 [0.01–0.21] <0.001   Excite 0.06 1.06 [0.97–1.17] 0.22 1.15 [0.73–1.83] 0.55   Identity 0.01 1.01 [0.92–1.09] 0.89 1.23 [0.76–2.00] 0.39   Approach −0.03 0.98 [0.89–1.07] 0.60 0.70 [0.42–1.17] 0.17   Gender (female) −0.12 0.88 [0.75–1.05] 0.15 1.26 [0.50–3.15] 0.62  Race (Ref: White)   Asian −0.03 0.97 [0.80–1.17] 0.74 1.78 [0.70–4.56] 0.23   Other −0.24 0.79 [0.62–0.99] <0.05 0.87 [0.20–3.78] 0.86  Interactions   Excite*Identity −0.08 0.92 [0.84–1.01] 0.09 0.68 [0.32 – 1.47] 0.33   App*Identity −0.02 0.98 [0.90–1.07] 0.70 1.82 [0.88–3.76] 0.11   App*Excite −0.05 0.95 [0.88–1.03] 0.21 1.04 [0.65–1.67] 0.86 Alcohol consumption after 3 months  Main effects   Intercept 1.55 4.70 [3.77–5.85] <0.001 2.14 [1.16–3.96] <0.05   T1 alcohol consumption 0.05 1.05 [1.04–1.06] <0.001 0.19 [0.07–0.54] <0.001   Excite 0.10 1.11 [0.98–1.24] 0.09 0.73 [0.49–1.08] 0.11   Identity 0.01 1.01 [0.91–1.13] 0.83 0.77 [0.52–1.14] 0.19   Approach 0.05 1.05 [0.92–1.19] 0.45 0.94 [0.63–1.41] 0.78   Gender (female) −0.08 0.92 [0.74–1.14] 0.45 1.30 [0.64–2.64] 0.47  Race (Ref: White)   Asian −0.33 0.72 [0.56–0.93] <0.01 1.63 [0.73–3.61] 0.23   Other −0.08 0.92 [0.68–1.24] 0.59 1.81 [0.59–5.58] 0.30  Interactions   Excite*Identity −0.09 0.91 [0.79–1.05] 0.20 0.87 [0.56–1.36] 0.54   App*Identity 0.04 1.04 [0.91–1.19] 0.59 0.96 [0.61–1.52] 0.87   App*Excite −0.05 0.96 [0.86–1.06] 0.40 1.03 [0.71–1.49] 0.87 Note. Risk of AUD is assessed as the sum of items on the AUDIT scale. Alcohol consumption is assessed as the sum of drinks per day on the Daily Drinking Questionnaire. Both outcomes are assumed zero-inflated negative binomial distributed. Excite: Implicit Alcohol-Excitement Associations. Identity: Implicit Alcohol-Identity Associations. App: Implicit Alcohol-Approach Associations. Concurrent results in Sample 1 Interactions did not significantly predict risk of AUD in the count- or inflated-portion of the model. Models featuring the interactions did not significantly improve model fit (Δ–2LL (6): 3.44, P = 0.75; sBIC main: 1529.8, sBIC interactions: 1537.5). Interactions also did not significantly predict alcohol consumption in count- or inflated-portions and model fit was not improved by their addition (Δ–2LL (6): 3.29, P = 0.77; sBIC main: 1606.2, sBIC interactions: 1614.2). Concurrent results in Sample 2 The interaction between implicit approach and excitement associations significantly predicted the count of AUD risk indications in Sample 2 (e^b = 0.54, 95% confidence interval (CI): [0.35–0.84], P < 0.01). No other interactions were significant. Furthermore, models predicting risk of AUD featuring the interactions did not significantly improve model fit (Δ–2LL (6): 6.85, P = 0.34; sBIC main: 1951.4, sBIC interactions: 1954.7). Interactions also did not significantly predict alcohol consumption in the count- or inflated-portions, and model fit was not improved by their addition (Δ–2LL (6): 4.40, P = 0.62; sBIC main: 1752.9, sBIC interactions: 1761.1). Prospective results in Sample 2 Interactions did not significantly predict risk of AUD in the count- or inflated-portion of the model. Models featuring the interactions did not significantly improve model fit (Δ–2LL (6): 5.38, P = 0.50; sBIC main: 1449.1, sBIC interactions: 1454.6). Interaction also did not significantly predict alcohol consumption in count- or inflated-portions, and model fit was not improved by their addition (Δ–2LL (6): 1.70, P = 0.95; sBIC main: 1325.5, sBIC interactions: 1338.3) (We exploratively tested three-way interactions between all three implicit predictors, and found that none significantly predicted any of the concurrent or prospective outcomes in either sample.). DISCUSSION The purpose of the current work was to examine the potential interactive effects of implicit drinking identity, excitement and approach associations in predicting the US undergraduates’ alcohol consumption and risk of AUD. Contrary to predictions, results indicated that interactive effects did not contribute significantly to the prediction of drinking outcomes in any of the models. We found almost no evidence that measures of implicit alcohol associations had interactive effects in predicting alcohol consumption and risk of AUD cross-sectionally or prospectively. We offer three possible interpretations for the lack of interactions in our measures of implicit associations. First, it may be that the notion of spreading activation in associative networks did not influence behavior in our samples to the degree that it might influence behavior in populations exhibiting riskier, unregulated drinking behavior. Since our samples included a substantial number of young adults who appeared not to have a had lot of experience with alcohol use, implicit associations may not have been formed sufficiently to meaningfully interact with each other. Second, implicit measures, including the IAT, are not process-pure (e.g. in O’Connor et al., 2012). That is, implicit measures are used to infer underlying processes (Nosek et al., 2011), and they may not be direct representations of the process intended to be represented (i.e. actual implicit associations) and responses to those measures likely calls for at least some use of controlled/reflective processes (see De Houwer, 2014). For instance, higher scores on an alcohol-approach IAT may reflect a strong approach tendency or may reflect a failure of executive functions aimed at suppressing automatic approach tendencies. Alternatively, they may reflect individual differences in the motivation to suppress such automatic tendencies, since implicit measures do not presuppose that individuals are necessarily unaware of their associations. Third, it has been suggested that implicit cognitions are not necessarily associative (De Houwer, 2014). In other words, what we called associations (e.g. alcohol approach) may, in fact, be a propositional statement (e.g. ‘I want to approach alcohol’) expressed in a way that is less subject to suppression by executive functions. Note that a propositional model does not necessarily negate the utility of implicit measures, but rather suggests that underlying associative theory is flawed. Thus, the lack of observed interactions could be viewed as evidence in support of assertions that associative cognitive models are flawed. Despite not finding support for the interaction hypothesis, investigation of interactive effects of implicit associations remains an important endeavor. From a theoretical vantage, theories and research on implicit associations tends not to disambiguate the function of different associations in terms of their unique impact on behavior (particularly, ‘when’, ‘where’, ‘for whom’, and ‘why’ they influence drinking behavior). This represents a core weakness to theories explicating implicit associations. Although we did not find interactions in the current study, the fact that no prior study had sought to investigate them highlights that the current thinking on different alcohol associations tends to be less nuanced. By investigating whether implicit associations interact, we seek to start a conversation on further disambiguating their functioning. This study has several strengths worth noting. First, this study evaluated these models using two independent samples, thereby increasing confidence about the lack of observed interactive effects. Moreover, the samples were somewhat different in age range (Sample 1 included students from all years of college between the ages of 18–25, whereas Sample 2 only included students in the first or second year of college between the ages of 18–20), which suggest that the null findings may apply broadly to the college years. Whether these findings will generalize to non-college populations is, of course, unknown. Second, this work examined these associations as concurrent and prospective predictors of consumption and risk of AUDs and found little evidence of interactive effects cross-sectionally or over time. This is important due to the consistent lack of findings across both time points and samples. Furthermore, due to the current ‘replication crisis’ in science (see Open Science Collaboration, 2015), we felt that it was important to evaluate the extent to which similar findings were observed across two different samples. This is particularly important as the current work represents secondary data analyses and used data from studies that were not originally designed to address these questions. While this study advances our understanding of the independence of implicit processes, findings should be considered in light of several limitations. First, it is important to note that both these samples are college student samples, which may limit generalizability to other populations, potentially excluding those at greatest risk of AUD. Second, participants in Sample 2 completed the assessment battery outside the laboratory setting. Although multiple steps were taken with the intent of minimizing participant’s distraction and maximizing the quality of responses, i.e. asking participants to complete the assessment at a time and place when they could give it their complete concentration, requiring participants to respond to periodic check questions, it is possible that participants may have been distracted. While we cannot rule out that possibility, we note that the sample means and SDs for the IATs—arguably the measures most sensitive to distraction—for both samples were roughly equivalent, and that other studies have demonstrated the predictive validity of implicit measures in settings outside the laboratory (see Houben and Wiers, 2008; Janssen et al., 2015a, 2015b). While it might have been informative to investigate the interaction of these measures of implicit associations with measures of executive functions as well as each other, this investigation was outside the scope of the current study. Furthermore, IAT measures of implicit associations most likely, at best, represent associations held in the moment, which may not represent associations as they occur in naturalistic contexts of young adult alcohol use. It is important to emphasize that there are distinct levels of analysis underlying our assessment approach and our theoretical expectations. Specifically, we can only infer the process (alcohol-related associations) by way of the procedure (IAT). Thus, in the sense of De Houwer’s (2014) work, our procedures may be elucidating propositionally held concepts that may not match the process that was intended to be measured. Finally, other important measures of implicit associations exist that have previously been related to alcohol use (e.g. alcohol and valence, see Wiers et al., 2002), but were not included in the current samples. Ecological momentary assessment studies could assess naturally occurring associations in one’s actual drinking context, providing a promising next step. CONCLUSIONS Results from two independent US-based college samples suggests that when evaluating implicit measures of drinking identity, alcohol approach and alcohol excitement, there is little support for interactive effects among these variables. Furthermore, the current findings are consistent with existing literature suggesting that implicit drinking identity is the most consistent predictor of alcohol-related outcomes when considered among other implicit alcohol variables. Thus, future work must aim to disambiguate the functioning of different implicit associations further in order more thoroughly establish their unique role in the development of addictive behaviors. FUNDING This work was supported by the National Institute on Alcohol Abuse and Alcoholism at the National Institutes of Health (R01AA021763, R01AA024732 and R00AA017669 to K.P.L. and T32AA007459 to Peter Monti). NIAAA had no role in the study design, collection, analysis or interpretation of the data, writing the manuscript or the decision to submit the paper for publication. Conflict of Interest Statement None declared. REFERENCES American College Health Association . ( 2012 ) American College Health Association-National College Health Assessment II: Undergraduate reference group executive summary Spring 2012 . Hanover, MD : American College Health Association . Atkins DC , Baldwin SA , Zheng C , et al. . ( 2013 ) A tutorial on count regression and zero-altered count models for longitudinal substance use data . Psychol Addict Behav 27 : 166 – 77 . Google Scholar CrossRef Search ADS PubMed Babor TF , Higgins-Biddle JC , Saunders JB , et al. . ( 2001 ) The Alcohol Use Disorders Identification Test (AUDIT): Guidelines for Use in Primary Care , 2nd ed . Geneva, Switzerland : World Health Organization, Department of Mental Health and Substance Dependence . Back MD , Schmukle SC , Egloff B . ( 2009 ) Predicting actual behavior from the explicit and implicit self-concept of personality . J Pers Soc Psychol 97 : 533 – 48 . Google Scholar CrossRef Search ADS PubMed Berridge KC , Robinson TE , Aldridge JW . ( 2009 ) Dissecting components of reward: ‘liking’, ‘wanting’, and learning . Curr Opin Pharmacol 9 : 65 – 73 . Google Scholar CrossRef Search ADS PubMed Collins RL , Parks GA , Martlatt GA . ( 1985 ) Social determinants of alcohol consumption: the effects of social interaction and model status on the self-administration of alcohol . J Consult Clin Psych 53 : 189 – 200 . Google Scholar CrossRef Search ADS De Houwer J . ( 2014 ) A propositional model of implicit evaluation . Soc Personal Psychol Compass 8 : 342 – 53 . Google Scholar CrossRef Search ADS Desmarais BA , Harden JJ . ( 2013 ) Testing for zero inflation in count models: bias correction for the Vuong test . Stata J 13 : 810 – 35 . Dingle GA , Cruwys T , Frings D . ( 2015 ) Social identities as pathways into and out of addiction . Front Psychol 6 : 1795 . Google Scholar CrossRef Search ADS PubMed Enders CK . ( 2011 ) Analyzing longitudinal data with missing values . Rehabilitation Psychology 56 : 267 – 88 . Google Scholar CrossRef Search ADS PubMed Fromme K , Corbin WR , Kruse MI . ( 2008 ) Behavioral risks during the transition from high school to college . Dev Psychol 44 : 1497 – 1504 . Google Scholar CrossRef Search ADS PubMed Gray HM , Laplante DA , Bannon BL , et al. . ( 2011 ) Development and validation of the Alcohol Identity Implicit Associations Test (AI-IAT) . Addict Behav 36 : 919 – 26 . Google Scholar CrossRef Search ADS PubMed Greenwald AG , Banaji MR , Rudman LA , et al. . ( 2002 ) A unified theory of implicit attitudes, stereotypes, self-esteem, and self-concept . Psychol Rev 109 : 3 – 25 . Google Scholar CrossRef Search ADS PubMed Greenwald AG , McGhee DE , Schwartz JK . ( 1998 ) Measuring individual differences in implicit cognition: the implicit association test . J Pers Soc Psychol 74 : 1464 – 80 . Google Scholar CrossRef Search ADS PubMed Greenwald AG , Nosek BA , Banaji MR . ( 2003 ) Understanding and using the implicit association test: I. An improved scoring algorithm . J Pers Soc Psychol 85 : 197 – 216 . Google Scholar CrossRef Search ADS PubMed Hofmann W , Friese M , Wiers R . ( 2008 ) Impulsive versus reflective influences on health behavior: a theoretical framework and empirical review . Health Psychol Rev 2 : 111 – 37 . Google Scholar CrossRef Search ADS Houben K , Wiers RW. ( 2008 ) Measuring implicit alcohol associations via the Internet: Validation of Web-based implicit association tests . Behav Res Methods 40 : 1134 – 43 . Google Scholar CrossRef Search ADS PubMed Houben K , Nosek BA , Wiers RW . ( 2010 ) Seeing the forest through the trees: a comparison of different IAT variants measuring implicit alcohol associations . Drug Alcohol Depend 106 : 204 – 11 . Google Scholar CrossRef Search ADS PubMed Janssen T , Larsen H , Vollebergh WA , et al. . ( 2015 a) Longitudinal relations between cognitive bias and adolescent alcohol use . Addict Behav 44 : 51 – 7 . Google Scholar CrossRef Search ADS PubMed Janssen T , Wood MD , Larsen H , et al. . ( 2015 b) Investigating the joint development of approach bias and adolescent alcohol use . Alcohol Clin Exp Res 39 : 2447 – 54 . Google Scholar CrossRef Search ADS PubMed Johnston LD , O’Malley PM , Bachman JG , et al. . ( 2015 ) Monitoring the Future National Survey Results on Drug Use, 1975-2014: Volume 2, College Students and Adults Ages 19-55 . Ann Arbor : The University of Michigan . Lindgren KP , Hendershot CS , Neighbors C , et al. . ( 2011 ) Implicit alcohol motives predict unique variance in drinking in Asian American college students . Motiv Emotion 35 : 435 – 43 . Google Scholar CrossRef Search ADS Lindgren KP , Neighbors C , Gasser ML , et al. . ( 2017 ) A review of implicit and explicit substance self-concept as a predictor of alcohol and tobacco use and misuse . Am J Drug Alcohol Ab 423 : 237 – 46 . Google Scholar CrossRef Search ADS Lindgren KP , Neighbors C , Teachman BA , et al. . ( 2013 ) I drink therefore I am: validating alcohol-related Implicit Association Tests . Psychol Addict Behav 27 : 1 – 13 . Google Scholar CrossRef Search ADS PubMed Lindgren KP , Neighbors C , Teachman BA , et al. . ( 2016 a) Implicit alcohol associations, especially drinking identity, predict drinking over time . Health Psychol 35 : 908 – 18 . Google Scholar CrossRef Search ADS PubMed Lindgren KP , Ramirez JJ , Olin CC , et al. . ( 2016 b) Not the same old thing: establishing the unique contribution of drinking identity as a predictor of alcohol consumption and problems over time . Psychol Addict Behav 30 : 659 – 71 . Google Scholar CrossRef Search ADS PubMed Markus H , Wurf E . ( 1987 ) The dynamic self-concept: a social psychological perspective . Annu Rev Psychol 38 : 299 – 337 . Google Scholar CrossRef Search ADS Naimi TS , Brewer RD , Mokdad A , et al. . ( 2003 ) Binge drinking among US adults . J Am Med Assoc 289 : 70 – 5 . Google Scholar CrossRef Search ADS Nosek BA , Greenwald AG , Banaji MR . ( 2007 ) The Implicit Association Test at age 7: a methodological and conceptual review. In Bargh JA (ed) . Automatic Processes in Social Thinking and Behavior . New York, NY: Psychology Press , 265 – 92 . Nosek BA , Hawkins CB , Frazier RS . ( 2011 ) Implicit social cognition: from measures to mechanisms . Trend Cog Sci 15 : 152 – 9 . Google Scholar CrossRef Search ADS Open Science Collaboration . ( 2015 ) Estimating the reproducibility of psychological science . Science 349 : aac4716 . CrossRef Search ADS PubMed Ostafin BD , Palfai TP . ( 2006 ) Compelled to consume: the Implicit Association Test and automatic alcohol motivation . Psychol Addict Behav 20 : 322 – 7 . Google Scholar CrossRef Search ADS PubMed O’Connor RM , Lopez-Vergara HI , Colder CR . ( 2012 ) Implicit cognition and substance use: the role of controlled and automatic processes in children . J Stud Alcohol Drugs 73 : 134 – 43 . Google Scholar CrossRef Search ADS PubMed O’Neill SE , Parra GR , Sher KJ . ( 2001 ) Clinical relevance of heavy drinking during the college years: cross-sectional and prospective perspectives . Psychol Addict Behav 15 : 350 – 9 . Google Scholar CrossRef Search ADS PubMed Reich RR , Below MC , Goldman MS . ( 2010 ) Explicit and implicit measures of expectancy and related alcohol cognitions: a meta-analytic comparison . Psychol Addict Behav 24 : 13 – 25 . Google Scholar CrossRef Search ADS PubMed Rooke SE , Hine DW , Thorsteinsson EB . ( 2008 ) Implicit cognition and substance use: a meta-analysis . Addict Behav 33 : 1314 – 28 . Google Scholar CrossRef Search ADS PubMed Stacy AW , Wiers RW . ( 2010 ) Implicit cognition and addiction: a tool for explaining paradoxical behavior . Annu Rev Clin Psychol 6 : 551 – 75 . Google Scholar CrossRef Search ADS PubMed Van Der Vorst H , Krank M , Engels RCME , et al. . ( 2013 ) The mediating role of alcohol-related memory associations on the relation between perceived parental drinking and the onset of adolescents’ alcohol use . Addiction 108 : 526 – 33 . Google Scholar CrossRef Search ADS PubMed Wiers RW , Bartholow BD , van den Wildenberg E , et al. . ( 2007 ) Automatic and controlled processes and the development of addictive behaviors in adolescents: a review and a model . Pharmacol Biochem Be 86 : 263 – 83 . Google Scholar CrossRef Search ADS Wiers RW , Rinck M , Kordts R , et al. . ( 2010 ) Retraining automatic action-tendencies to approach alcohol in hazardous drinkers . Addiction 105 : 279 – 87 . Google Scholar CrossRef Search ADS PubMed Wiers RW , van Woerden N , Smulders FT , et al. . ( 2002 ) Implicit and explicit alcohol-related cognitions in heavy and light drinkers . J Abnorm Psychol 111 : 648 – 58 . Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Medical Council on Alcohol and Oxford University Press. All rights reserved. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices) http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Alcohol and Alcoholism Oxford University Press

A Declaration of Independence: Implicit Alcohol Associations Have Independent, not Interactive, Relationships with Alcohol Consumption and AUD Risk

Loading next page...
 
/lp/ou_press/a-declaration-of-independence-implicit-alcohol-associations-have-d8gk94pEtO
Publisher
Oxford University Press
Copyright
© The Author(s) 2018. Medical Council on Alcohol and Oxford University Press. All rights reserved.
ISSN
0735-0414
eISSN
1464-3502
D.O.I.
10.1093/alcalc/agy023
Publisher site
See Article on Publisher Site

Abstract

Abstract Aims The current study aimed to test for potential interactive effects of three implicit alcohol-related associations (drinking identity, alcohol approach and alcohol excitement) in predicting concurrent and prospective alcohol consumption and risk of alcohol use disorders (AUDs) in two samples of the US undergraduate drinkers and non-drinkers. Short summary We investigated the independent and interactive effects of three implicit associations on alcohol consumption and risk of AUD in two US undergraduate student samples. We found that implicit associations had independent but not interactive effects on concurrent and subsequent alcohol consumption and risk of AUD in two independent samples. Methods Implicit drinking identity, alcohol approach and alcohol excitement associations were assessed in two US undergraduate student samples (Sample 1: N = 300, 55% female; Sample 2: N = 506, 57% female). Alcohol consumption and risk of AUD were assessed at baseline (Samples 1 and 2) and 3 months later (Sample 2). We fit zero-inflated negative binomial models to test for independent and interactive effects of the three implicit associations on alcohol consumption and risk of AUD. Results Although we found multiple, unique main effects for alcohol associations, we found minimal evidence of interactions between implicit alcohol-related associations. There was no reliable evidence of interactions in models in predicting alcohol consumption or risk of AUD, concurrently or prospectively, in either sample. Conclusions Contrary to expectations, results from both studies indicated that implicit alcohol-related associations in the US undergraduate samples generally have independent, not interactive, relationships with alcohol consumption and risk of AUD. INTRODUCTION Although more than half of the US students attending college are under the minimum legal drinking age of 21 (American College Health Association, 2012), 78% report lifetime alcohol use and 40% report having been drunk in the past 30 days (Johnston et al., 2015). The US college years are a time of both initiation of drinking for a substantial proportion of student (Fromme et al., 2008) as well as escalation of alcohol consumption (see Naimi et al., 2003). Moreover, 36% of the US college students report engaging in heavy episodic drinking (i.e. five or more drinks) at least once in the previous 2 weeks (Johnston et al., 2015). Heavy drinking during the college years prospectively predicts risk of alcohol use disorders (AUDs; O’Neill et al., 2001), further cementing college-age drinking as a major public health concern. Recent theoretical models of drinking emphasize the joint contribution of explicit and implicit (automatic/reflexive/impulsive) cognitive factors in alcohol use and misuse (see Wiers et al., 2007; Stacy and Wiers, 2010). Multiple implicit cognitive factors, including implicit associations about alcohol and drinking (i.e. implicit alcohol-related associations), have found to be robust predictors of college students’ alcohol use and misuse even after controlling for explicit measures (see Wiers et al., 2002; Reich et al., 2010; Lindgren et al., 2013, 2016a). As research on implicit alcohol-related associations has matured, attention has turned to identifying potential moderators of the relationship between implicit associations and drinking, such as cognitive capacity or mood (see Hofmann et al., 2008). One set of potential moderators that has not been considered are implicit alcohol-related associations, themselves. That is, there are multiple, different implicit alcohol-related associations and whether they might have interactive effects on alcohol consumption and risk of AUD has not, to our knowledge, been evaluated. Thus, the current study tested for possible interactive effects among well-established implicit alcohol-related associations on measures of alcohol use and misuse. Implicit Alcohol-related Associations Multiple implicit alcohol-related associations have been studied (see Rooke et al., 2008; Reich et al., 2010), with implicit drinking identity, implicit alcohol excitement and implicit alcohol approach emerging as key predictors of alcohol outcomes among the US college students (Lindgren et al., 2016a). First, implicit drinking identity—or the extent to which one associates drinking with the self—has been found to be a consistent predictor of college student alcohol consumption and risk of AUD, both cross-sectionally (Lindgren et al., 2013) and over time (see Gray et al., 2011; Lindgren et al., 2016a, 2016b). The theoretical interest in implicit drinking identity stems from social and cognitive psychology theories emphasizing the importance of the constructs that become associated with the self (e.g. Greenwald et al., 2002; Back et al., 2009), such as drinking behaviors and/or drinking social groups, positing that those associations can become unique, important drivers of drinking behavior and risk of AUD (see Dingle et al., 2015; Lindgren et al., 2017). Second, implicit alcohol excitement—or the extent to which one associates alcoholic beverages with excitement—has been linked to college alcohol use previously (Lindgren et al., 2013) and is similar to Wiers et al.’s (2002) alcohol arousal Implicit Association Test (IAT). Theoretically, since enhancement of positive mood is cited as one of the most common reasons for drinking among young adults, alcohol-excitement associations may represent an implicit expression of intention to drink alcohol for this specific reason (see Houben et al., 2010; Lindgren et al., 2013). Third, implicit alcohol-approach associations—or the extent to which one associates alcoholic beverages with approach—are assumed to reflect an individual’s appetitive inclinations for acquiring and consuming alcohol (Ostafin and Palfai, 2006). Conceptually, alcohol-approach inclinations could be viewed as analogous to the ‘wanting’ (versus ‘liking’) alcohol that occurs during compulsive alcohol use (see Berridge et al., 2009). They have been shown to be associated with alcohol consumption among adolescents and young adults (Janssen et al., 2015a; Lindgren et al., 2016a), and have been successfully targeted in interventions (Wiers et al., 2010). There is also some evidence that measures of these associations are not redundant (they are only weakly correlated with one another) and they have unique effects on alcohol use and risk of AUD in the US undergraduates (see Lindgren et al., 2016a). Collectively, they may provide a comprehensive view of key implicit associations related to drinking for this population: these associations focus on associations about alcohol as a substance (implicit alcohol approach and excitement, can be viewed as analogous to wanting and liking) and behavior linked to the self (implicit drinking identity). Simultaneously, theories about what implicit associations are, and how they affect behavior, suggest the possibility that they may also have interactive effects. Implicit associations have been conceptualized as a network of associations stored in memory, with constructs (e.g. ‘alcohol’, ‘excitement’, ‘approach’ and ‘me’) representing nodes in that network (e.g. Greenwald et al., 2002; Lindgren et al., 2017; but see De Houwer, 2014, for an alternative view). Furthermore, the activation of one (or more) set of associations could spread to other associations (i.e. spreading activation). Hence, because both implicit alcohol approach and alcohol excitement associations focus on alcohol, one might expect that activation of one would lead to activation of the other (and vice versa) and they could have an increased effect on drinking behavior. Thus, these associations could interact and have multiplicative effects on alcohol consumption. Further justification for hypothesizing interactive effects stems from social psychologies of the self (e.g. Markus and Wurf, 1987; Greenwald et al., 2002), which suggest that the self is a key, organizing/central aspect of mental constructs and guides information processing (i.e. self-relevant info). Thus, the degree to which one has strong alcohol-related associations combined with a strong drinking identity might be expected to boost alcohol-related information processing (and spreading activation) and further increase inclinations towards drinking. Study Overview The goal of the current study was to evaluate whether scores on measures of implicit alcohol-related associations (i.e. drinking identity, alcohol excitement and alcohol approach associations) have interactive effects on college student alcohol consumption and risk of AUD. We tested the following hypothesis: implicit associations would amplify one other—i.e. have multiplicative effects—and thus have an even stronger relationship on alcohol consumption and risk of AUD beyond their previously shown individual effects. We did so via secondary data analysis of two independent samples of the US undergraduates (one cross-sectional, Lindgren et al., 2013 and one longitudinal, Lindgren et al., 2016a). We expected that each of the three two-way interactions would be significant and that the pattern of the interactions would be similar and indicate multiplicative effects in the prediction of alcohol consumption/risk status and amount of alcohol consumption/risk, concurrently and prospectively. Because the undergraduate years in the US are associated with both initiation and escalation of drinking and there is evidence that implicit alcohol associations can predate the initiation of drinking (see Van Der Vorst et al., 2013; Janssen et al., 2015a), both samples included drinkers and non-drinkers. Therefore, we also expected these effects to differentiate drinkers and non-drinkers, in the same way that they differentiate greater versus smaller alcohol consumption and risk for AUD. The studies used identical measures of implicit alcohol-related associations and alcohol consumption and risk of AUD. METHODS Procedures All procedures were approved by the university’s (a large Pacific Northwest public university) Institutional Review Board. Sample 1 A randomly sampled list of 18- to 25-year-old undergraduates was obtained from the Registrar’s Office. Participants were recruited via email but completed the study in the laboratory. Written informed consent was obtained during the lab session. Up to four participants shared a room at a time. Partitions separated individuals and privacy screens were placed over the laptops to minimize risks to privacy and confidentiality. All measures were computer-based and presented in random order. Participants were compensated $30. Sample 2 A randomly sampled list of full-time students in their first or second undergraduate year, ages 18–20, was obtained from the university’s Registrar’s Office. Students were recruited via email for a 2-year online study of alcohol and cognition. Informed consent and assessments were administered online, on participants’ choice of computer. Assessments consisted of the measures of implicit alcohol associations and alcohol outcomes. Participants were compensated $25. Data for the current study come from the first (enrollment) assessment and the first follow-up after 3 months. Eighty-six percent (n = 437) of participants completed this follow-up assessment. Missing the follow-up assessment was not associated with any baseline sociodemographic characteristic (gender, age, race and ethnicity) but was significantly predicted by greater baseline risk of AUD (mean Alcohol Use Disorders Identification Test (AUDIT, Babor et al., 2001) score not missing: 4.25, mean AUDIT score missing: 6.49, t(1,502) = −3.260, P < 0.01); see Lindgren et al., 2016a, for more details on attrition. Participants Sample 1 Participants (N = 300, 55% women) were between the ages of 18 and 25 (M = 20.47, SD = 1.52). Fifty-seven percent identified as White/Caucasian, 30% as Asian, 9% as multiracial and the remaining 4% as other, or declined to answer. Sample 2 Participants (N = 506, 57% women) were first- and second-year undergraduates, between the ages of 18 and 20 (M = 18.57, SD = 0.69). Fifty-two percent identified as White/Caucasian, 31% as Asian, 11% as multiracial and the remaining 6% as other, or declined to answer. Measures and Materials Implicit alcohol associations We used three IATs (Greenwald et al., 1998) to evaluate implicit alcohol associations. IATs are computer-based reaction time tasks used to measure the strength of associations between concepts, relative to an alternative (see target constructs below). A series of words and/or pictures are presented center-screen, and participants are instructed to press the corresponding key (d and k in Sample 1; e and i in Sample 2) to classify each stimulus into one of two contrasting target or attribute categories, or pairs of categories, as quickly and accurately as possible. Scores were calculated using the D score algorithm (Greenwald et al., 2003). IAT data were screened (per Nosek et al., 2007) and IATs with ≥10% trials faster than 300 ms or ≥30% trials with errors were considered invalid. Implementation of the drinking identity IAT (Lindgren et al., 2013), the alcohol approach IAT (Ostafin and Palfai, 2006) and the alcohol excitement IAT (Lindgren et al., 2011) was previously described for Sample 1 (Lindgren et al., 2013) and Sample 2 (Lindgren et al., 2016a). Higher scores indicated stronger associations between ‘alcohol’ and ‘me’ (versus ‘not me’), between ‘alcohol’ and ‘approach’ (versus ‘avoid’), and between ‘alcohol’ and ‘excite’ (versus ‘depress’), respectively. Internal consistencies (calculated by correlating D scores from Blocks 3 and 6 with D scores from Blocks 4 and 7) were: drinking identity, 0.51 (Sample 1), 0.58, (Sample 2); alcohol approach, 0.48 (Sample 1), 0.55, (Sample 2) and alcohol excitement, 0.52 (Sample 1), 0.57, (Sample 2). Alcohol consumption The Daily Drinking Questionnaire (DDQ; Collins et al., 1985) assesses individuals’ typical alcohol consumption over the past 3 months. Participants were asked to report how many standard drinks they consumed each day of the week. Participants were provided with a list of US standard drink equivalencies. Daily consumption was summed to reflect total drinks per week. Risk of AUD The AUDIT (Babor et al., 2001) is a 10-item measure used to assess alcohol-related problems and individuals’ risk of AUDs, considering the past year. Items were summed. Higher scores indicate a greater risk of an alcohol use disorder. Cronbach’s alphas were: 0.80, Sample 1 and 0.83, Sample 2. Analytic Strategy Analyses were run separately for each sample so that we could evaluate whether results were consistent. Both samples included drinkers and non-drinkers, and preliminary analyses showed a large number of zero scores on alcohol outcomes (Sample 1: 21.7% and Sample 2: 50.3% at enrollment, 47.4% at follow-up). We elected to retain non-drinkers in analyses because of preliminary evidence that implicit associations predate initiation of drinking and this is a period when initiation and escalation of drinking occurs. Further, their distributions were positively skewed. Therefore, data were modeled using a zero-inflated negative binomial (ZINB) distribution for alcohol consumption and risk of AUD, consisting of a count portion and an inflated portion. This approach addresses both the large number of zeros and the positively skewed distributions. In ZINB, two simultaneous regressions are conducted. The first—the inflated portion—predicts the probability of being one of the large number of zeros the distribution, which one can conceptualize of predicting the probability of being a ‘true’ non-drinker (or not at risk of AUD) versus the probability of being a drinker (or at risk of AUD). The second—the count portion–predicts the remainder of the distribution (i.e. the full range of alcohol consumption and risk of AUD). (For a detailed and informative overview of count distributions, see Atkins et al., 2013). We ran a series of stepwise zero-inflated negative binomial regression models predicting concurrent alcohol consumption and risk of AUD (Samples 1 and 2) and prospective alcohol consumption and risk of AUD (Sample 2), investigating the interactive effect of the three implicit associations. In Step 1, we entered the main effects of the three implicit associations. In Step 2, we entered the three possible two-way interactions between the three associations. All analyses controlled for gender and race. We evaluated the significance of interactive effects and significance of changes in model fit (based on two indexes comparing the model with main effects to the model with interactions included: loglikelihood chi-square and sample size-adjusted Bayesian Information Criterion, sBIC; Desmarais and Harden (2013)). All analyses were performed in Mplus 7.2. Missing data was handled using Full Information Maximum Likelihood under assumption of missing at random (Enders, 2011). Finally, for longitudinal models using Sample 2, we investigated prospective changes in alcohol consumption and risk of AUD by predicting Time 2 alcohol consumption and risk of AUD, while controlling for Time 1 alcohol consumption and risk of AUD, respectively, along with all previously included covariates. RESULTS Table 1 shows the means and correlations between implicit associations and alcohol outcome variables for each sample, as well as descriptive details on sociodemographic and alcohol consumption details of each sample. Table 2 shows the regression results for the stepwise ZINB models with DDQ and AUDIT scores as outcome, for both samples. While we note main effects, our main focus is on the presence or absence of interaction as those tests are the primary aim of the current study. Table 1. Descriptive statistics and correlations for variables for Sample 1 and Sample 2 Measure 1 2 3 4 5 6 7 8 9 10 M SD 1. Drinking identity IAT – 0.13* 0.21*** 0.32*** 0.34*** −0.02 −0.23*** 0.09 −0.03 −0.08 0.03 0.39 2. Alcohol approach IAT 0.22*** – 0.44*** 0.15* 0.23*** −0.03 −0.02 0.08 −0.03 −0.08 –0.15 0.35 3. Alcohol excitement IAT 0.13* 0.39*** – 0.24*** 0.26*** −0.03 −0.07 0.06 0.00 −0.09 –0.01 0.41 4. Alcohol consumption 0.38*** 0.20** 0.16** – 0.76*** −0.12* −0.20** 0.25** −0.23** −0.06 8.22 9.96 5. Risk of AUD 0.44*** 0.28*** 0.20*** 0.79*** – −0.05 0.15** 0.23** −0.22** −0.03 6.39 5.53 6. Age 0.05 −0.03 −0.02 0.19*** 0.22*** – – 0.07 −0.06 −0.02 20.47 1.52 7. Gender −0.13** −0.11* −0.12* 0.00 0.01 0.06 – −0.08 0.01 0.11 55% F – 8. Race: White 0.10* 0.02 0.11* 0.15** 0.18** −0.06 −0.03 – – – 57% – 9. Race: Asian-American −0.09 −0.07 −0.17** −0.18 −0.21 −0.04 0.00 – – – 30% – 10. Race: Othera −0.03 0.05 0.06 0.02 0.02 0.12** 0.03 – – – 14% – M −0.08 −0.21 −0.10 5.10 4.37 18.57 57% F 52% 31% 17% – – SD 0.45 0.40 0.46 8.54 5.07 0.69 – – – – – – Measure 1 2 3 4 5 6 7 8 9 10 M SD 1. Drinking identity IAT – 0.13* 0.21*** 0.32*** 0.34*** −0.02 −0.23*** 0.09 −0.03 −0.08 0.03 0.39 2. Alcohol approach IAT 0.22*** – 0.44*** 0.15* 0.23*** −0.03 −0.02 0.08 −0.03 −0.08 –0.15 0.35 3. Alcohol excitement IAT 0.13* 0.39*** – 0.24*** 0.26*** −0.03 −0.07 0.06 0.00 −0.09 –0.01 0.41 4. Alcohol consumption 0.38*** 0.20** 0.16** – 0.76*** −0.12* −0.20** 0.25** −0.23** −0.06 8.22 9.96 5. Risk of AUD 0.44*** 0.28*** 0.20*** 0.79*** – −0.05 0.15** 0.23** −0.22** −0.03 6.39 5.53 6. Age 0.05 −0.03 −0.02 0.19*** 0.22*** – – 0.07 −0.06 −0.02 20.47 1.52 7. Gender −0.13** −0.11* −0.12* 0.00 0.01 0.06 – −0.08 0.01 0.11 55% F – 8. Race: White 0.10* 0.02 0.11* 0.15** 0.18** −0.06 −0.03 – – – 57% – 9. Race: Asian-American −0.09 −0.07 −0.17** −0.18 −0.21 −0.04 0.00 – – – 30% – 10. Race: Othera −0.03 0.05 0.06 0.02 0.02 0.12** 0.03 – – – 14% – M −0.08 −0.21 −0.10 5.10 4.37 18.57 57% F 52% 31% 17% – – SD 0.45 0.40 0.46 8.54 5.07 0.69 – – – – – – Note. Correlations for Sample 1 (N = 300) are presented above the diagonal; correlations for Sample 2 (N = 506) are presented below the diagonal. Sample 2 drinking outcomes were assessed at baseline. N’s for each correlation vary slightly due to missing data. Means and standard deviations for Study 1 are presented in vertical columns, and those for Study 2 are presented in horizontal rows. IAT = Implicit Association Test; higher scores = stronger associations between the concepts in the IAT’s name. Alcohol consumption = number of drinks consumed in a typical week (assessed in the Daily Drinking Questionnaire). Risk of AUD = score on the Alcohol Use Disorders Identification Test; higher scores = greater risk of an alcohol use disorder. Gender is coded 0 = male; 1 = female. aIn Sample 1, the composition of this ‘other’ group was 1% African-American, 0.7% Native Hawaiian, 1.0 American Indian/Alaska Native and 9.3% multiracial, with 0.7% declining to answer. In Sample 2, this composition was 1.4% African-American, 0.8% American-Indian/Alaska Native, 11.3% multiracial and 3.5% unknown/declined to answer. *P < 0.05, **P < 0.01 and ***P < 0.001. Table 1. Descriptive statistics and correlations for variables for Sample 1 and Sample 2 Measure 1 2 3 4 5 6 7 8 9 10 M SD 1. Drinking identity IAT – 0.13* 0.21*** 0.32*** 0.34*** −0.02 −0.23*** 0.09 −0.03 −0.08 0.03 0.39 2. Alcohol approach IAT 0.22*** – 0.44*** 0.15* 0.23*** −0.03 −0.02 0.08 −0.03 −0.08 –0.15 0.35 3. Alcohol excitement IAT 0.13* 0.39*** – 0.24*** 0.26*** −0.03 −0.07 0.06 0.00 −0.09 –0.01 0.41 4. Alcohol consumption 0.38*** 0.20** 0.16** – 0.76*** −0.12* −0.20** 0.25** −0.23** −0.06 8.22 9.96 5. Risk of AUD 0.44*** 0.28*** 0.20*** 0.79*** – −0.05 0.15** 0.23** −0.22** −0.03 6.39 5.53 6. Age 0.05 −0.03 −0.02 0.19*** 0.22*** – – 0.07 −0.06 −0.02 20.47 1.52 7. Gender −0.13** −0.11* −0.12* 0.00 0.01 0.06 – −0.08 0.01 0.11 55% F – 8. Race: White 0.10* 0.02 0.11* 0.15** 0.18** −0.06 −0.03 – – – 57% – 9. Race: Asian-American −0.09 −0.07 −0.17** −0.18 −0.21 −0.04 0.00 – – – 30% – 10. Race: Othera −0.03 0.05 0.06 0.02 0.02 0.12** 0.03 – – – 14% – M −0.08 −0.21 −0.10 5.10 4.37 18.57 57% F 52% 31% 17% – – SD 0.45 0.40 0.46 8.54 5.07 0.69 – – – – – – Measure 1 2 3 4 5 6 7 8 9 10 M SD 1. Drinking identity IAT – 0.13* 0.21*** 0.32*** 0.34*** −0.02 −0.23*** 0.09 −0.03 −0.08 0.03 0.39 2. Alcohol approach IAT 0.22*** – 0.44*** 0.15* 0.23*** −0.03 −0.02 0.08 −0.03 −0.08 –0.15 0.35 3. Alcohol excitement IAT 0.13* 0.39*** – 0.24*** 0.26*** −0.03 −0.07 0.06 0.00 −0.09 –0.01 0.41 4. Alcohol consumption 0.38*** 0.20** 0.16** – 0.76*** −0.12* −0.20** 0.25** −0.23** −0.06 8.22 9.96 5. Risk of AUD 0.44*** 0.28*** 0.20*** 0.79*** – −0.05 0.15** 0.23** −0.22** −0.03 6.39 5.53 6. Age 0.05 −0.03 −0.02 0.19*** 0.22*** – – 0.07 −0.06 −0.02 20.47 1.52 7. Gender −0.13** −0.11* −0.12* 0.00 0.01 0.06 – −0.08 0.01 0.11 55% F – 8. Race: White 0.10* 0.02 0.11* 0.15** 0.18** −0.06 −0.03 – – – 57% – 9. Race: Asian-American −0.09 −0.07 −0.17** −0.18 −0.21 −0.04 0.00 – – – 30% – 10. Race: Othera −0.03 0.05 0.06 0.02 0.02 0.12** 0.03 – – – 14% – M −0.08 −0.21 −0.10 5.10 4.37 18.57 57% F 52% 31% 17% – – SD 0.45 0.40 0.46 8.54 5.07 0.69 – – – – – – Note. Correlations for Sample 1 (N = 300) are presented above the diagonal; correlations for Sample 2 (N = 506) are presented below the diagonal. Sample 2 drinking outcomes were assessed at baseline. N’s for each correlation vary slightly due to missing data. Means and standard deviations for Study 1 are presented in vertical columns, and those for Study 2 are presented in horizontal rows. IAT = Implicit Association Test; higher scores = stronger associations between the concepts in the IAT’s name. Alcohol consumption = number of drinks consumed in a typical week (assessed in the Daily Drinking Questionnaire). Risk of AUD = score on the Alcohol Use Disorders Identification Test; higher scores = greater risk of an alcohol use disorder. Gender is coded 0 = male; 1 = female. aIn Sample 1, the composition of this ‘other’ group was 1% African-American, 0.7% Native Hawaiian, 1.0 American Indian/Alaska Native and 9.3% multiracial, with 0.7% declining to answer. In Sample 2, this composition was 1.4% African-American, 0.8% American-Indian/Alaska Native, 11.3% multiracial and 3.5% unknown/declined to answer. *P < 0.05, **P < 0.01 and ***P < 0.001. Table 2. Results from concurrent zero-inflated negative binomial regression Sample 1 Sample 2 Count B e^B [95% CI] P Zero-inflated P Count B e^B [95% CI] P Zero-inflated P OR [95% CI] OR [95% CI] Risk of AUD  Main effects   Intercept 1.70 5.45 [4.54–6.54] <0.001 0.41 [0.25–0.66] <0.001 2.03 7.61 [6.50–8.90] <0.001 0.06 [0.02–0.15] <0.001   Excite 0.13 1.13 [1.00–1.28] <0.05 0.92 [0.64–1.31] 0.64 0.07 1.08 [0.97–1.19] 0.15 0.55 [0.32–0.94] <0.05   Identity 0.33 1.40 [1.26–1.55] <0.001 0.56 [0.41–0.78] <0.001 0.21 1.23 [1.12–1.35] <0.001 0.91 [0.55–1.52] 0.72   Approach 0.05 1.06 [0.94–1.19] 0.38 0.56 [0.39–0.81] <0.001 0.12 1.12 [1.01–1.25] <0.05 0.78 [0.42–1.45] 0.43   Gender (Female) 0.04 1.04 [0.84–1.27] 0.74 0.47 [0.26–0.84] <0.01 −0.12 0.89 [0.74–1.08] 0.23 1.11 [0.39–3.17] 0.85  Race (Ref: White)   Asian −0.28 0.76 [0.59–0.97] <0.05 2.03 [1.10–3.73] <0.05 −0.30 0.74 [0.59–0.93] <0.01 3.31 [1.18–9.25] <0.05   Other −0.04 0.96 [0.73–1.25] 0.75 0.71 [0.26–1.92] 0.50 −0.09 0.92 [0.67–1.24] 0.57 0.70 [0.05–9.26] 0.79  Interactions   Excite*Identity −0.10 0.91 [0.61–1.34] 0.63 0.91 [0.80–1.03] 0.14 0.03 1.03 [0.93–1.14] 0.60 1.19 [0.72–1.98] 0.49   App*Identity 0.17 1.19 [0.76–1.86] 0.45 1.02 [0.90–1.16] 0.76 −0.06 0.94 [0.86–1.03] 0.17 0.79 [0.46–1.37] 0.41   App*Excite −0.62 0.54 [0.35–0.84] <0.01 0.95 [0.86–1.06] 0.36 −0.09 0.91 [0.83–1.00] 0.06 1.13 [0.67–1.90] 0.66 Alcohol consumption  Main effects   Intercept 2.13 8.40 [6.57–10.73] <0.001 1.23 [0.80–1.89] 0.35 2.42 11.22 [9.24–13.63] <0.001 0.10 [0.05–0.23] <0.001   Excite 0.10 1.10 [0.95–1.28] 0.21 0.82 [0.62–1.07] 0.15 0.06 1.06 [0.93–1.21] 0.40 0.44 [0.27–0.70] <0.001   Identity 0.38 1.46 [1.26–1.70] <0.001 0.49 [0.38–0.65] <0.001 0.22 1.25 [1.10–1.41] <0.001 0.74 [0.50–1.12] 0.16   Approach −0.05 0.95 [0.82–1.11] 0.52 0.55 [0.42–0.72] <0.001 0.15 1.17 [1.02–1.33] <0.05 1.12 [0.71–1.76] 0.62   Gender (Female) 0.00 1.00 [0.75–1.32] 0.98 0.54 [0.32–0.89] <0.05 −0.34 0.71 [0.56–0.90] <0.01 0.80 [0.35–1.81] 0.59  Race (Ref: White)   Asian −0.28 0.76 [0.54–1.05] 0.09 1.41 [0.82–2.45] 0.22 −0.40 0.67 [0.50–0.90] <0.01 4.32 [1.78–10.48] <0.001   Other −0.11 0.90 [0.63–1.28] 0.55 0.82 [0.40–1.69] 0.59 0.06 1.06 [0.71–1.59] 0.78 3.01 [0.98–9.28] 0.06  Interactions   Excite*Identity −0.18 0.84 [0.70–1.00] 0.05 1.04 [0.77–1.41] 0.80 0.03 1.03 [0.90–1.18] 0.67 0.43 [0.26–0.72] <0.001   App*Identity 0.01 1.01 [0.85–1.19] 0.95 0.97 [0.71–1.32] 0.84 −0.10 0.91 [0.81–1.02] 0.11 0.74 [0.46–1.20] 0.23   App*Excite −0.06 0.94 [0.83–1.07] 0.36 0.80 [0.59–1.07] 0.13 −0.06 0.94 [0.83–1.07] 0.36 1.37 [0.81–2.31] 0.25 Sample 1 Sample 2 Count B e^B [95% CI] P Zero-inflated P Count B e^B [95% CI] P Zero-inflated P OR [95% CI] OR [95% CI] Risk of AUD  Main effects   Intercept 1.70 5.45 [4.54–6.54] <0.001 0.41 [0.25–0.66] <0.001 2.03 7.61 [6.50–8.90] <0.001 0.06 [0.02–0.15] <0.001   Excite 0.13 1.13 [1.00–1.28] <0.05 0.92 [0.64–1.31] 0.64 0.07 1.08 [0.97–1.19] 0.15 0.55 [0.32–0.94] <0.05   Identity 0.33 1.40 [1.26–1.55] <0.001 0.56 [0.41–0.78] <0.001 0.21 1.23 [1.12–1.35] <0.001 0.91 [0.55–1.52] 0.72   Approach 0.05 1.06 [0.94–1.19] 0.38 0.56 [0.39–0.81] <0.001 0.12 1.12 [1.01–1.25] <0.05 0.78 [0.42–1.45] 0.43   Gender (Female) 0.04 1.04 [0.84–1.27] 0.74 0.47 [0.26–0.84] <0.01 −0.12 0.89 [0.74–1.08] 0.23 1.11 [0.39–3.17] 0.85  Race (Ref: White)   Asian −0.28 0.76 [0.59–0.97] <0.05 2.03 [1.10–3.73] <0.05 −0.30 0.74 [0.59–0.93] <0.01 3.31 [1.18–9.25] <0.05   Other −0.04 0.96 [0.73–1.25] 0.75 0.71 [0.26–1.92] 0.50 −0.09 0.92 [0.67–1.24] 0.57 0.70 [0.05–9.26] 0.79  Interactions   Excite*Identity −0.10 0.91 [0.61–1.34] 0.63 0.91 [0.80–1.03] 0.14 0.03 1.03 [0.93–1.14] 0.60 1.19 [0.72–1.98] 0.49   App*Identity 0.17 1.19 [0.76–1.86] 0.45 1.02 [0.90–1.16] 0.76 −0.06 0.94 [0.86–1.03] 0.17 0.79 [0.46–1.37] 0.41   App*Excite −0.62 0.54 [0.35–0.84] <0.01 0.95 [0.86–1.06] 0.36 −0.09 0.91 [0.83–1.00] 0.06 1.13 [0.67–1.90] 0.66 Alcohol consumption  Main effects   Intercept 2.13 8.40 [6.57–10.73] <0.001 1.23 [0.80–1.89] 0.35 2.42 11.22 [9.24–13.63] <0.001 0.10 [0.05–0.23] <0.001   Excite 0.10 1.10 [0.95–1.28] 0.21 0.82 [0.62–1.07] 0.15 0.06 1.06 [0.93–1.21] 0.40 0.44 [0.27–0.70] <0.001   Identity 0.38 1.46 [1.26–1.70] <0.001 0.49 [0.38–0.65] <0.001 0.22 1.25 [1.10–1.41] <0.001 0.74 [0.50–1.12] 0.16   Approach −0.05 0.95 [0.82–1.11] 0.52 0.55 [0.42–0.72] <0.001 0.15 1.17 [1.02–1.33] <0.05 1.12 [0.71–1.76] 0.62   Gender (Female) 0.00 1.00 [0.75–1.32] 0.98 0.54 [0.32–0.89] <0.05 −0.34 0.71 [0.56–0.90] <0.01 0.80 [0.35–1.81] 0.59  Race (Ref: White)   Asian −0.28 0.76 [0.54–1.05] 0.09 1.41 [0.82–2.45] 0.22 −0.40 0.67 [0.50–0.90] <0.01 4.32 [1.78–10.48] <0.001   Other −0.11 0.90 [0.63–1.28] 0.55 0.82 [0.40–1.69] 0.59 0.06 1.06 [0.71–1.59] 0.78 3.01 [0.98–9.28] 0.06  Interactions   Excite*Identity −0.18 0.84 [0.70–1.00] 0.05 1.04 [0.77–1.41] 0.80 0.03 1.03 [0.90–1.18] 0.67 0.43 [0.26–0.72] <0.001   App*Identity 0.01 1.01 [0.85–1.19] 0.95 0.97 [0.71–1.32] 0.84 −0.10 0.91 [0.81–1.02] 0.11 0.74 [0.46–1.20] 0.23   App*Excite −0.06 0.94 [0.83–1.07] 0.36 0.80 [0.59–1.07] 0.13 −0.06 0.94 [0.83–1.07] 0.36 1.37 [0.81–2.31] 0.25 Note. Risk of AUD is assessed as the sum of items on the AUDIT scale. Alcohol Consumption is assessed as the sum of drinks per day on the Daily Drinking Questionnaire. Both outcomes are assumed zero-inflated negative binomial distributed. For the prediction of the zero-inflated portion, the coefficients reflect the change in likelihood of being an excess zero. In other words, scores under 1 reduce the likelihood of having lifetime risk of problems (top rows) or the likelihood of being a lifetime drinker (bottom rows). Excite: Implicit Alcohol-Excitement Associations. Identity: Implicit Alcohol-Identity Associations. App: Implicit Alcohol-Approach Associations. Table 2. Results from concurrent zero-inflated negative binomial regression Sample 1 Sample 2 Count B e^B [95% CI] P Zero-inflated P Count B e^B [95% CI] P Zero-inflated P OR [95% CI] OR [95% CI] Risk of AUD  Main effects   Intercept 1.70 5.45 [4.54–6.54] <0.001 0.41 [0.25–0.66] <0.001 2.03 7.61 [6.50–8.90] <0.001 0.06 [0.02–0.15] <0.001   Excite 0.13 1.13 [1.00–1.28] <0.05 0.92 [0.64–1.31] 0.64 0.07 1.08 [0.97–1.19] 0.15 0.55 [0.32–0.94] <0.05   Identity 0.33 1.40 [1.26–1.55] <0.001 0.56 [0.41–0.78] <0.001 0.21 1.23 [1.12–1.35] <0.001 0.91 [0.55–1.52] 0.72   Approach 0.05 1.06 [0.94–1.19] 0.38 0.56 [0.39–0.81] <0.001 0.12 1.12 [1.01–1.25] <0.05 0.78 [0.42–1.45] 0.43   Gender (Female) 0.04 1.04 [0.84–1.27] 0.74 0.47 [0.26–0.84] <0.01 −0.12 0.89 [0.74–1.08] 0.23 1.11 [0.39–3.17] 0.85  Race (Ref: White)   Asian −0.28 0.76 [0.59–0.97] <0.05 2.03 [1.10–3.73] <0.05 −0.30 0.74 [0.59–0.93] <0.01 3.31 [1.18–9.25] <0.05   Other −0.04 0.96 [0.73–1.25] 0.75 0.71 [0.26–1.92] 0.50 −0.09 0.92 [0.67–1.24] 0.57 0.70 [0.05–9.26] 0.79  Interactions   Excite*Identity −0.10 0.91 [0.61–1.34] 0.63 0.91 [0.80–1.03] 0.14 0.03 1.03 [0.93–1.14] 0.60 1.19 [0.72–1.98] 0.49   App*Identity 0.17 1.19 [0.76–1.86] 0.45 1.02 [0.90–1.16] 0.76 −0.06 0.94 [0.86–1.03] 0.17 0.79 [0.46–1.37] 0.41   App*Excite −0.62 0.54 [0.35–0.84] <0.01 0.95 [0.86–1.06] 0.36 −0.09 0.91 [0.83–1.00] 0.06 1.13 [0.67–1.90] 0.66 Alcohol consumption  Main effects   Intercept 2.13 8.40 [6.57–10.73] <0.001 1.23 [0.80–1.89] 0.35 2.42 11.22 [9.24–13.63] <0.001 0.10 [0.05–0.23] <0.001   Excite 0.10 1.10 [0.95–1.28] 0.21 0.82 [0.62–1.07] 0.15 0.06 1.06 [0.93–1.21] 0.40 0.44 [0.27–0.70] <0.001   Identity 0.38 1.46 [1.26–1.70] <0.001 0.49 [0.38–0.65] <0.001 0.22 1.25 [1.10–1.41] <0.001 0.74 [0.50–1.12] 0.16   Approach −0.05 0.95 [0.82–1.11] 0.52 0.55 [0.42–0.72] <0.001 0.15 1.17 [1.02–1.33] <0.05 1.12 [0.71–1.76] 0.62   Gender (Female) 0.00 1.00 [0.75–1.32] 0.98 0.54 [0.32–0.89] <0.05 −0.34 0.71 [0.56–0.90] <0.01 0.80 [0.35–1.81] 0.59  Race (Ref: White)   Asian −0.28 0.76 [0.54–1.05] 0.09 1.41 [0.82–2.45] 0.22 −0.40 0.67 [0.50–0.90] <0.01 4.32 [1.78–10.48] <0.001   Other −0.11 0.90 [0.63–1.28] 0.55 0.82 [0.40–1.69] 0.59 0.06 1.06 [0.71–1.59] 0.78 3.01 [0.98–9.28] 0.06  Interactions   Excite*Identity −0.18 0.84 [0.70–1.00] 0.05 1.04 [0.77–1.41] 0.80 0.03 1.03 [0.90–1.18] 0.67 0.43 [0.26–0.72] <0.001   App*Identity 0.01 1.01 [0.85–1.19] 0.95 0.97 [0.71–1.32] 0.84 −0.10 0.91 [0.81–1.02] 0.11 0.74 [0.46–1.20] 0.23   App*Excite −0.06 0.94 [0.83–1.07] 0.36 0.80 [0.59–1.07] 0.13 −0.06 0.94 [0.83–1.07] 0.36 1.37 [0.81–2.31] 0.25 Sample 1 Sample 2 Count B e^B [95% CI] P Zero-inflated P Count B e^B [95% CI] P Zero-inflated P OR [95% CI] OR [95% CI] Risk of AUD  Main effects   Intercept 1.70 5.45 [4.54–6.54] <0.001 0.41 [0.25–0.66] <0.001 2.03 7.61 [6.50–8.90] <0.001 0.06 [0.02–0.15] <0.001   Excite 0.13 1.13 [1.00–1.28] <0.05 0.92 [0.64–1.31] 0.64 0.07 1.08 [0.97–1.19] 0.15 0.55 [0.32–0.94] <0.05   Identity 0.33 1.40 [1.26–1.55] <0.001 0.56 [0.41–0.78] <0.001 0.21 1.23 [1.12–1.35] <0.001 0.91 [0.55–1.52] 0.72   Approach 0.05 1.06 [0.94–1.19] 0.38 0.56 [0.39–0.81] <0.001 0.12 1.12 [1.01–1.25] <0.05 0.78 [0.42–1.45] 0.43   Gender (Female) 0.04 1.04 [0.84–1.27] 0.74 0.47 [0.26–0.84] <0.01 −0.12 0.89 [0.74–1.08] 0.23 1.11 [0.39–3.17] 0.85  Race (Ref: White)   Asian −0.28 0.76 [0.59–0.97] <0.05 2.03 [1.10–3.73] <0.05 −0.30 0.74 [0.59–0.93] <0.01 3.31 [1.18–9.25] <0.05   Other −0.04 0.96 [0.73–1.25] 0.75 0.71 [0.26–1.92] 0.50 −0.09 0.92 [0.67–1.24] 0.57 0.70 [0.05–9.26] 0.79  Interactions   Excite*Identity −0.10 0.91 [0.61–1.34] 0.63 0.91 [0.80–1.03] 0.14 0.03 1.03 [0.93–1.14] 0.60 1.19 [0.72–1.98] 0.49   App*Identity 0.17 1.19 [0.76–1.86] 0.45 1.02 [0.90–1.16] 0.76 −0.06 0.94 [0.86–1.03] 0.17 0.79 [0.46–1.37] 0.41   App*Excite −0.62 0.54 [0.35–0.84] <0.01 0.95 [0.86–1.06] 0.36 −0.09 0.91 [0.83–1.00] 0.06 1.13 [0.67–1.90] 0.66 Alcohol consumption  Main effects   Intercept 2.13 8.40 [6.57–10.73] <0.001 1.23 [0.80–1.89] 0.35 2.42 11.22 [9.24–13.63] <0.001 0.10 [0.05–0.23] <0.001   Excite 0.10 1.10 [0.95–1.28] 0.21 0.82 [0.62–1.07] 0.15 0.06 1.06 [0.93–1.21] 0.40 0.44 [0.27–0.70] <0.001   Identity 0.38 1.46 [1.26–1.70] <0.001 0.49 [0.38–0.65] <0.001 0.22 1.25 [1.10–1.41] <0.001 0.74 [0.50–1.12] 0.16   Approach −0.05 0.95 [0.82–1.11] 0.52 0.55 [0.42–0.72] <0.001 0.15 1.17 [1.02–1.33] <0.05 1.12 [0.71–1.76] 0.62   Gender (Female) 0.00 1.00 [0.75–1.32] 0.98 0.54 [0.32–0.89] <0.05 −0.34 0.71 [0.56–0.90] <0.01 0.80 [0.35–1.81] 0.59  Race (Ref: White)   Asian −0.28 0.76 [0.54–1.05] 0.09 1.41 [0.82–2.45] 0.22 −0.40 0.67 [0.50–0.90] <0.01 4.32 [1.78–10.48] <0.001   Other −0.11 0.90 [0.63–1.28] 0.55 0.82 [0.40–1.69] 0.59 0.06 1.06 [0.71–1.59] 0.78 3.01 [0.98–9.28] 0.06  Interactions   Excite*Identity −0.18 0.84 [0.70–1.00] 0.05 1.04 [0.77–1.41] 0.80 0.03 1.03 [0.90–1.18] 0.67 0.43 [0.26–0.72] <0.001   App*Identity 0.01 1.01 [0.85–1.19] 0.95 0.97 [0.71–1.32] 0.84 −0.10 0.91 [0.81–1.02] 0.11 0.74 [0.46–1.20] 0.23   App*Excite −0.06 0.94 [0.83–1.07] 0.36 0.80 [0.59–1.07] 0.13 −0.06 0.94 [0.83–1.07] 0.36 1.37 [0.81–2.31] 0.25 Note. Risk of AUD is assessed as the sum of items on the AUDIT scale. Alcohol Consumption is assessed as the sum of drinks per day on the Daily Drinking Questionnaire. Both outcomes are assumed zero-inflated negative binomial distributed. For the prediction of the zero-inflated portion, the coefficients reflect the change in likelihood of being an excess zero. In other words, scores under 1 reduce the likelihood of having lifetime risk of problems (top rows) or the likelihood of being a lifetime drinker (bottom rows). Excite: Implicit Alcohol-Excitement Associations. Identity: Implicit Alcohol-Identity Associations. App: Implicit Alcohol-Approach Associations. Similar to reports in previous work (Lindgren et al., 2013, 2016a), results from the main effects only model (Step 1) indicated that implicit drinking identity was significantly and positively associated with the full range of alcohol consumption and risk of AUD across both samples (as observed in the count portion of the ZINB), and greater drinking identity was significantly associated with a decreased likelihood of being an abstainer (from alcohol consumption) and having never been at risk of AUD in Sample 2 (as observed in the inflated portion of the ZINB). The findings for implicit approach and excitement associations were mixed, both in terms of outcomes and for which Sample they offered significant predictive validity (see Table 2 for details). Significant prediction by drinking identity did not extend to prospective prediction of alcohol consumption and risk of AUD after controlling for baseline alcohol consumption (Notably, differences appear between main effects reported here and those reported in Lindgren and colleagues (2016a) because of differences in the analytic approach. Specifically, in the results from Lindgren and colleagues, data from multiple time points were used in a model that also included a time covariate.). Interactive Effects In Step 2, we added the three two-way interaction terms to the prediction of both the count- and inflated-portions of the models (see Tables 2 and 3). Table 3. Results from prospective zero-inflated negative binomial regression Sample 2 Count B e^B [95% CI] P Zero-inflated P OR [95% CI] Risk of AUD after 3 months  Main effects   Intercept 1.07 2.90 [2.41–3.49] <0.001 1.89 [0.86–4.16] 0.12   T1 alcohol problems 0.11 1.12 [1.10–1.14] <0.001 0.04 [0.01–0.21] <0.001   Excite 0.06 1.06 [0.97–1.17] 0.22 1.15 [0.73–1.83] 0.55   Identity 0.01 1.01 [0.92–1.09] 0.89 1.23 [0.76–2.00] 0.39   Approach −0.03 0.98 [0.89–1.07] 0.60 0.70 [0.42–1.17] 0.17   Gender (female) −0.12 0.88 [0.75–1.05] 0.15 1.26 [0.50–3.15] 0.62  Race (Ref: White)   Asian −0.03 0.97 [0.80–1.17] 0.74 1.78 [0.70–4.56] 0.23   Other −0.24 0.79 [0.62–0.99] <0.05 0.87 [0.20–3.78] 0.86  Interactions   Excite*Identity −0.08 0.92 [0.84–1.01] 0.09 0.68 [0.32 – 1.47] 0.33   App*Identity −0.02 0.98 [0.90–1.07] 0.70 1.82 [0.88–3.76] 0.11   App*Excite −0.05 0.95 [0.88–1.03] 0.21 1.04 [0.65–1.67] 0.86 Alcohol consumption after 3 months  Main effects   Intercept 1.55 4.70 [3.77–5.85] <0.001 2.14 [1.16–3.96] <0.05   T1 alcohol consumption 0.05 1.05 [1.04–1.06] <0.001 0.19 [0.07–0.54] <0.001   Excite 0.10 1.11 [0.98–1.24] 0.09 0.73 [0.49–1.08] 0.11   Identity 0.01 1.01 [0.91–1.13] 0.83 0.77 [0.52–1.14] 0.19   Approach 0.05 1.05 [0.92–1.19] 0.45 0.94 [0.63–1.41] 0.78   Gender (female) −0.08 0.92 [0.74–1.14] 0.45 1.30 [0.64–2.64] 0.47  Race (Ref: White)   Asian −0.33 0.72 [0.56–0.93] <0.01 1.63 [0.73–3.61] 0.23   Other −0.08 0.92 [0.68–1.24] 0.59 1.81 [0.59–5.58] 0.30  Interactions   Excite*Identity −0.09 0.91 [0.79–1.05] 0.20 0.87 [0.56–1.36] 0.54   App*Identity 0.04 1.04 [0.91–1.19] 0.59 0.96 [0.61–1.52] 0.87   App*Excite −0.05 0.96 [0.86–1.06] 0.40 1.03 [0.71–1.49] 0.87 Sample 2 Count B e^B [95% CI] P Zero-inflated P OR [95% CI] Risk of AUD after 3 months  Main effects   Intercept 1.07 2.90 [2.41–3.49] <0.001 1.89 [0.86–4.16] 0.12   T1 alcohol problems 0.11 1.12 [1.10–1.14] <0.001 0.04 [0.01–0.21] <0.001   Excite 0.06 1.06 [0.97–1.17] 0.22 1.15 [0.73–1.83] 0.55   Identity 0.01 1.01 [0.92–1.09] 0.89 1.23 [0.76–2.00] 0.39   Approach −0.03 0.98 [0.89–1.07] 0.60 0.70 [0.42–1.17] 0.17   Gender (female) −0.12 0.88 [0.75–1.05] 0.15 1.26 [0.50–3.15] 0.62  Race (Ref: White)   Asian −0.03 0.97 [0.80–1.17] 0.74 1.78 [0.70–4.56] 0.23   Other −0.24 0.79 [0.62–0.99] <0.05 0.87 [0.20–3.78] 0.86  Interactions   Excite*Identity −0.08 0.92 [0.84–1.01] 0.09 0.68 [0.32 – 1.47] 0.33   App*Identity −0.02 0.98 [0.90–1.07] 0.70 1.82 [0.88–3.76] 0.11   App*Excite −0.05 0.95 [0.88–1.03] 0.21 1.04 [0.65–1.67] 0.86 Alcohol consumption after 3 months  Main effects   Intercept 1.55 4.70 [3.77–5.85] <0.001 2.14 [1.16–3.96] <0.05   T1 alcohol consumption 0.05 1.05 [1.04–1.06] <0.001 0.19 [0.07–0.54] <0.001   Excite 0.10 1.11 [0.98–1.24] 0.09 0.73 [0.49–1.08] 0.11   Identity 0.01 1.01 [0.91–1.13] 0.83 0.77 [0.52–1.14] 0.19   Approach 0.05 1.05 [0.92–1.19] 0.45 0.94 [0.63–1.41] 0.78   Gender (female) −0.08 0.92 [0.74–1.14] 0.45 1.30 [0.64–2.64] 0.47  Race (Ref: White)   Asian −0.33 0.72 [0.56–0.93] <0.01 1.63 [0.73–3.61] 0.23   Other −0.08 0.92 [0.68–1.24] 0.59 1.81 [0.59–5.58] 0.30  Interactions   Excite*Identity −0.09 0.91 [0.79–1.05] 0.20 0.87 [0.56–1.36] 0.54   App*Identity 0.04 1.04 [0.91–1.19] 0.59 0.96 [0.61–1.52] 0.87   App*Excite −0.05 0.96 [0.86–1.06] 0.40 1.03 [0.71–1.49] 0.87 Note. Risk of AUD is assessed as the sum of items on the AUDIT scale. Alcohol consumption is assessed as the sum of drinks per day on the Daily Drinking Questionnaire. Both outcomes are assumed zero-inflated negative binomial distributed. Excite: Implicit Alcohol-Excitement Associations. Identity: Implicit Alcohol-Identity Associations. App: Implicit Alcohol-Approach Associations. Table 3. Results from prospective zero-inflated negative binomial regression Sample 2 Count B e^B [95% CI] P Zero-inflated P OR [95% CI] Risk of AUD after 3 months  Main effects   Intercept 1.07 2.90 [2.41–3.49] <0.001 1.89 [0.86–4.16] 0.12   T1 alcohol problems 0.11 1.12 [1.10–1.14] <0.001 0.04 [0.01–0.21] <0.001   Excite 0.06 1.06 [0.97–1.17] 0.22 1.15 [0.73–1.83] 0.55   Identity 0.01 1.01 [0.92–1.09] 0.89 1.23 [0.76–2.00] 0.39   Approach −0.03 0.98 [0.89–1.07] 0.60 0.70 [0.42–1.17] 0.17   Gender (female) −0.12 0.88 [0.75–1.05] 0.15 1.26 [0.50–3.15] 0.62  Race (Ref: White)   Asian −0.03 0.97 [0.80–1.17] 0.74 1.78 [0.70–4.56] 0.23   Other −0.24 0.79 [0.62–0.99] <0.05 0.87 [0.20–3.78] 0.86  Interactions   Excite*Identity −0.08 0.92 [0.84–1.01] 0.09 0.68 [0.32 – 1.47] 0.33   App*Identity −0.02 0.98 [0.90–1.07] 0.70 1.82 [0.88–3.76] 0.11   App*Excite −0.05 0.95 [0.88–1.03] 0.21 1.04 [0.65–1.67] 0.86 Alcohol consumption after 3 months  Main effects   Intercept 1.55 4.70 [3.77–5.85] <0.001 2.14 [1.16–3.96] <0.05   T1 alcohol consumption 0.05 1.05 [1.04–1.06] <0.001 0.19 [0.07–0.54] <0.001   Excite 0.10 1.11 [0.98–1.24] 0.09 0.73 [0.49–1.08] 0.11   Identity 0.01 1.01 [0.91–1.13] 0.83 0.77 [0.52–1.14] 0.19   Approach 0.05 1.05 [0.92–1.19] 0.45 0.94 [0.63–1.41] 0.78   Gender (female) −0.08 0.92 [0.74–1.14] 0.45 1.30 [0.64–2.64] 0.47  Race (Ref: White)   Asian −0.33 0.72 [0.56–0.93] <0.01 1.63 [0.73–3.61] 0.23   Other −0.08 0.92 [0.68–1.24] 0.59 1.81 [0.59–5.58] 0.30  Interactions   Excite*Identity −0.09 0.91 [0.79–1.05] 0.20 0.87 [0.56–1.36] 0.54   App*Identity 0.04 1.04 [0.91–1.19] 0.59 0.96 [0.61–1.52] 0.87   App*Excite −0.05 0.96 [0.86–1.06] 0.40 1.03 [0.71–1.49] 0.87 Sample 2 Count B e^B [95% CI] P Zero-inflated P OR [95% CI] Risk of AUD after 3 months  Main effects   Intercept 1.07 2.90 [2.41–3.49] <0.001 1.89 [0.86–4.16] 0.12   T1 alcohol problems 0.11 1.12 [1.10–1.14] <0.001 0.04 [0.01–0.21] <0.001   Excite 0.06 1.06 [0.97–1.17] 0.22 1.15 [0.73–1.83] 0.55   Identity 0.01 1.01 [0.92–1.09] 0.89 1.23 [0.76–2.00] 0.39   Approach −0.03 0.98 [0.89–1.07] 0.60 0.70 [0.42–1.17] 0.17   Gender (female) −0.12 0.88 [0.75–1.05] 0.15 1.26 [0.50–3.15] 0.62  Race (Ref: White)   Asian −0.03 0.97 [0.80–1.17] 0.74 1.78 [0.70–4.56] 0.23   Other −0.24 0.79 [0.62–0.99] <0.05 0.87 [0.20–3.78] 0.86  Interactions   Excite*Identity −0.08 0.92 [0.84–1.01] 0.09 0.68 [0.32 – 1.47] 0.33   App*Identity −0.02 0.98 [0.90–1.07] 0.70 1.82 [0.88–3.76] 0.11   App*Excite −0.05 0.95 [0.88–1.03] 0.21 1.04 [0.65–1.67] 0.86 Alcohol consumption after 3 months  Main effects   Intercept 1.55 4.70 [3.77–5.85] <0.001 2.14 [1.16–3.96] <0.05   T1 alcohol consumption 0.05 1.05 [1.04–1.06] <0.001 0.19 [0.07–0.54] <0.001   Excite 0.10 1.11 [0.98–1.24] 0.09 0.73 [0.49–1.08] 0.11   Identity 0.01 1.01 [0.91–1.13] 0.83 0.77 [0.52–1.14] 0.19   Approach 0.05 1.05 [0.92–1.19] 0.45 0.94 [0.63–1.41] 0.78   Gender (female) −0.08 0.92 [0.74–1.14] 0.45 1.30 [0.64–2.64] 0.47  Race (Ref: White)   Asian −0.33 0.72 [0.56–0.93] <0.01 1.63 [0.73–3.61] 0.23   Other −0.08 0.92 [0.68–1.24] 0.59 1.81 [0.59–5.58] 0.30  Interactions   Excite*Identity −0.09 0.91 [0.79–1.05] 0.20 0.87 [0.56–1.36] 0.54   App*Identity 0.04 1.04 [0.91–1.19] 0.59 0.96 [0.61–1.52] 0.87   App*Excite −0.05 0.96 [0.86–1.06] 0.40 1.03 [0.71–1.49] 0.87 Note. Risk of AUD is assessed as the sum of items on the AUDIT scale. Alcohol consumption is assessed as the sum of drinks per day on the Daily Drinking Questionnaire. Both outcomes are assumed zero-inflated negative binomial distributed. Excite: Implicit Alcohol-Excitement Associations. Identity: Implicit Alcohol-Identity Associations. App: Implicit Alcohol-Approach Associations. Concurrent results in Sample 1 Interactions did not significantly predict risk of AUD in the count- or inflated-portion of the model. Models featuring the interactions did not significantly improve model fit (Δ–2LL (6): 3.44, P = 0.75; sBIC main: 1529.8, sBIC interactions: 1537.5). Interactions also did not significantly predict alcohol consumption in count- or inflated-portions and model fit was not improved by their addition (Δ–2LL (6): 3.29, P = 0.77; sBIC main: 1606.2, sBIC interactions: 1614.2). Concurrent results in Sample 2 The interaction between implicit approach and excitement associations significantly predicted the count of AUD risk indications in Sample 2 (e^b = 0.54, 95% confidence interval (CI): [0.35–0.84], P < 0.01). No other interactions were significant. Furthermore, models predicting risk of AUD featuring the interactions did not significantly improve model fit (Δ–2LL (6): 6.85, P = 0.34; sBIC main: 1951.4, sBIC interactions: 1954.7). Interactions also did not significantly predict alcohol consumption in the count- or inflated-portions, and model fit was not improved by their addition (Δ–2LL (6): 4.40, P = 0.62; sBIC main: 1752.9, sBIC interactions: 1761.1). Prospective results in Sample 2 Interactions did not significantly predict risk of AUD in the count- or inflated-portion of the model. Models featuring the interactions did not significantly improve model fit (Δ–2LL (6): 5.38, P = 0.50; sBIC main: 1449.1, sBIC interactions: 1454.6). Interaction also did not significantly predict alcohol consumption in count- or inflated-portions, and model fit was not improved by their addition (Δ–2LL (6): 1.70, P = 0.95; sBIC main: 1325.5, sBIC interactions: 1338.3) (We exploratively tested three-way interactions between all three implicit predictors, and found that none significantly predicted any of the concurrent or prospective outcomes in either sample.). DISCUSSION The purpose of the current work was to examine the potential interactive effects of implicit drinking identity, excitement and approach associations in predicting the US undergraduates’ alcohol consumption and risk of AUD. Contrary to predictions, results indicated that interactive effects did not contribute significantly to the prediction of drinking outcomes in any of the models. We found almost no evidence that measures of implicit alcohol associations had interactive effects in predicting alcohol consumption and risk of AUD cross-sectionally or prospectively. We offer three possible interpretations for the lack of interactions in our measures of implicit associations. First, it may be that the notion of spreading activation in associative networks did not influence behavior in our samples to the degree that it might influence behavior in populations exhibiting riskier, unregulated drinking behavior. Since our samples included a substantial number of young adults who appeared not to have a had lot of experience with alcohol use, implicit associations may not have been formed sufficiently to meaningfully interact with each other. Second, implicit measures, including the IAT, are not process-pure (e.g. in O’Connor et al., 2012). That is, implicit measures are used to infer underlying processes (Nosek et al., 2011), and they may not be direct representations of the process intended to be represented (i.e. actual implicit associations) and responses to those measures likely calls for at least some use of controlled/reflective processes (see De Houwer, 2014). For instance, higher scores on an alcohol-approach IAT may reflect a strong approach tendency or may reflect a failure of executive functions aimed at suppressing automatic approach tendencies. Alternatively, they may reflect individual differences in the motivation to suppress such automatic tendencies, since implicit measures do not presuppose that individuals are necessarily unaware of their associations. Third, it has been suggested that implicit cognitions are not necessarily associative (De Houwer, 2014). In other words, what we called associations (e.g. alcohol approach) may, in fact, be a propositional statement (e.g. ‘I want to approach alcohol’) expressed in a way that is less subject to suppression by executive functions. Note that a propositional model does not necessarily negate the utility of implicit measures, but rather suggests that underlying associative theory is flawed. Thus, the lack of observed interactions could be viewed as evidence in support of assertions that associative cognitive models are flawed. Despite not finding support for the interaction hypothesis, investigation of interactive effects of implicit associations remains an important endeavor. From a theoretical vantage, theories and research on implicit associations tends not to disambiguate the function of different associations in terms of their unique impact on behavior (particularly, ‘when’, ‘where’, ‘for whom’, and ‘why’ they influence drinking behavior). This represents a core weakness to theories explicating implicit associations. Although we did not find interactions in the current study, the fact that no prior study had sought to investigate them highlights that the current thinking on different alcohol associations tends to be less nuanced. By investigating whether implicit associations interact, we seek to start a conversation on further disambiguating their functioning. This study has several strengths worth noting. First, this study evaluated these models using two independent samples, thereby increasing confidence about the lack of observed interactive effects. Moreover, the samples were somewhat different in age range (Sample 1 included students from all years of college between the ages of 18–25, whereas Sample 2 only included students in the first or second year of college between the ages of 18–20), which suggest that the null findings may apply broadly to the college years. Whether these findings will generalize to non-college populations is, of course, unknown. Second, this work examined these associations as concurrent and prospective predictors of consumption and risk of AUDs and found little evidence of interactive effects cross-sectionally or over time. This is important due to the consistent lack of findings across both time points and samples. Furthermore, due to the current ‘replication crisis’ in science (see Open Science Collaboration, 2015), we felt that it was important to evaluate the extent to which similar findings were observed across two different samples. This is particularly important as the current work represents secondary data analyses and used data from studies that were not originally designed to address these questions. While this study advances our understanding of the independence of implicit processes, findings should be considered in light of several limitations. First, it is important to note that both these samples are college student samples, which may limit generalizability to other populations, potentially excluding those at greatest risk of AUD. Second, participants in Sample 2 completed the assessment battery outside the laboratory setting. Although multiple steps were taken with the intent of minimizing participant’s distraction and maximizing the quality of responses, i.e. asking participants to complete the assessment at a time and place when they could give it their complete concentration, requiring participants to respond to periodic check questions, it is possible that participants may have been distracted. While we cannot rule out that possibility, we note that the sample means and SDs for the IATs—arguably the measures most sensitive to distraction—for both samples were roughly equivalent, and that other studies have demonstrated the predictive validity of implicit measures in settings outside the laboratory (see Houben and Wiers, 2008; Janssen et al., 2015a, 2015b). While it might have been informative to investigate the interaction of these measures of implicit associations with measures of executive functions as well as each other, this investigation was outside the scope of the current study. Furthermore, IAT measures of implicit associations most likely, at best, represent associations held in the moment, which may not represent associations as they occur in naturalistic contexts of young adult alcohol use. It is important to emphasize that there are distinct levels of analysis underlying our assessment approach and our theoretical expectations. Specifically, we can only infer the process (alcohol-related associations) by way of the procedure (IAT). Thus, in the sense of De Houwer’s (2014) work, our procedures may be elucidating propositionally held concepts that may not match the process that was intended to be measured. Finally, other important measures of implicit associations exist that have previously been related to alcohol use (e.g. alcohol and valence, see Wiers et al., 2002), but were not included in the current samples. Ecological momentary assessment studies could assess naturally occurring associations in one’s actual drinking context, providing a promising next step. CONCLUSIONS Results from two independent US-based college samples suggests that when evaluating implicit measures of drinking identity, alcohol approach and alcohol excitement, there is little support for interactive effects among these variables. Furthermore, the current findings are consistent with existing literature suggesting that implicit drinking identity is the most consistent predictor of alcohol-related outcomes when considered among other implicit alcohol variables. Thus, future work must aim to disambiguate the functioning of different implicit associations further in order more thoroughly establish their unique role in the development of addictive behaviors. FUNDING This work was supported by the National Institute on Alcohol Abuse and Alcoholism at the National Institutes of Health (R01AA021763, R01AA024732 and R00AA017669 to K.P.L. and T32AA007459 to Peter Monti). NIAAA had no role in the study design, collection, analysis or interpretation of the data, writing the manuscript or the decision to submit the paper for publication. Conflict of Interest Statement None declared. REFERENCES American College Health Association . ( 2012 ) American College Health Association-National College Health Assessment II: Undergraduate reference group executive summary Spring 2012 . Hanover, MD : American College Health Association . Atkins DC , Baldwin SA , Zheng C , et al. . ( 2013 ) A tutorial on count regression and zero-altered count models for longitudinal substance use data . Psychol Addict Behav 27 : 166 – 77 . Google Scholar CrossRef Search ADS PubMed Babor TF , Higgins-Biddle JC , Saunders JB , et al. . ( 2001 ) The Alcohol Use Disorders Identification Test (AUDIT): Guidelines for Use in Primary Care , 2nd ed . Geneva, Switzerland : World Health Organization, Department of Mental Health and Substance Dependence . Back MD , Schmukle SC , Egloff B . ( 2009 ) Predicting actual behavior from the explicit and implicit self-concept of personality . J Pers Soc Psychol 97 : 533 – 48 . Google Scholar CrossRef Search ADS PubMed Berridge KC , Robinson TE , Aldridge JW . ( 2009 ) Dissecting components of reward: ‘liking’, ‘wanting’, and learning . Curr Opin Pharmacol 9 : 65 – 73 . Google Scholar CrossRef Search ADS PubMed Collins RL , Parks GA , Martlatt GA . ( 1985 ) Social determinants of alcohol consumption: the effects of social interaction and model status on the self-administration of alcohol . J Consult Clin Psych 53 : 189 – 200 . Google Scholar CrossRef Search ADS De Houwer J . ( 2014 ) A propositional model of implicit evaluation . Soc Personal Psychol Compass 8 : 342 – 53 . Google Scholar CrossRef Search ADS Desmarais BA , Harden JJ . ( 2013 ) Testing for zero inflation in count models: bias correction for the Vuong test . Stata J 13 : 810 – 35 . Dingle GA , Cruwys T , Frings D . ( 2015 ) Social identities as pathways into and out of addiction . Front Psychol 6 : 1795 . Google Scholar CrossRef Search ADS PubMed Enders CK . ( 2011 ) Analyzing longitudinal data with missing values . Rehabilitation Psychology 56 : 267 – 88 . Google Scholar CrossRef Search ADS PubMed Fromme K , Corbin WR , Kruse MI . ( 2008 ) Behavioral risks during the transition from high school to college . Dev Psychol 44 : 1497 – 1504 . Google Scholar CrossRef Search ADS PubMed Gray HM , Laplante DA , Bannon BL , et al. . ( 2011 ) Development and validation of the Alcohol Identity Implicit Associations Test (AI-IAT) . Addict Behav 36 : 919 – 26 . Google Scholar CrossRef Search ADS PubMed Greenwald AG , Banaji MR , Rudman LA , et al. . ( 2002 ) A unified theory of implicit attitudes, stereotypes, self-esteem, and self-concept . Psychol Rev 109 : 3 – 25 . Google Scholar CrossRef Search ADS PubMed Greenwald AG , McGhee DE , Schwartz JK . ( 1998 ) Measuring individual differences in implicit cognition: the implicit association test . J Pers Soc Psychol 74 : 1464 – 80 . Google Scholar CrossRef Search ADS PubMed Greenwald AG , Nosek BA , Banaji MR . ( 2003 ) Understanding and using the implicit association test: I. An improved scoring algorithm . J Pers Soc Psychol 85 : 197 – 216 . Google Scholar CrossRef Search ADS PubMed Hofmann W , Friese M , Wiers R . ( 2008 ) Impulsive versus reflective influences on health behavior: a theoretical framework and empirical review . Health Psychol Rev 2 : 111 – 37 . Google Scholar CrossRef Search ADS Houben K , Wiers RW. ( 2008 ) Measuring implicit alcohol associations via the Internet: Validation of Web-based implicit association tests . Behav Res Methods 40 : 1134 – 43 . Google Scholar CrossRef Search ADS PubMed Houben K , Nosek BA , Wiers RW . ( 2010 ) Seeing the forest through the trees: a comparison of different IAT variants measuring implicit alcohol associations . Drug Alcohol Depend 106 : 204 – 11 . Google Scholar CrossRef Search ADS PubMed Janssen T , Larsen H , Vollebergh WA , et al. . ( 2015 a) Longitudinal relations between cognitive bias and adolescent alcohol use . Addict Behav 44 : 51 – 7 . Google Scholar CrossRef Search ADS PubMed Janssen T , Wood MD , Larsen H , et al. . ( 2015 b) Investigating the joint development of approach bias and adolescent alcohol use . Alcohol Clin Exp Res 39 : 2447 – 54 . Google Scholar CrossRef Search ADS PubMed Johnston LD , O’Malley PM , Bachman JG , et al. . ( 2015 ) Monitoring the Future National Survey Results on Drug Use, 1975-2014: Volume 2, College Students and Adults Ages 19-55 . Ann Arbor : The University of Michigan . Lindgren KP , Hendershot CS , Neighbors C , et al. . ( 2011 ) Implicit alcohol motives predict unique variance in drinking in Asian American college students . Motiv Emotion 35 : 435 – 43 . Google Scholar CrossRef Search ADS Lindgren KP , Neighbors C , Gasser ML , et al. . ( 2017 ) A review of implicit and explicit substance self-concept as a predictor of alcohol and tobacco use and misuse . Am J Drug Alcohol Ab 423 : 237 – 46 . Google Scholar CrossRef Search ADS Lindgren KP , Neighbors C , Teachman BA , et al. . ( 2013 ) I drink therefore I am: validating alcohol-related Implicit Association Tests . Psychol Addict Behav 27 : 1 – 13 . Google Scholar CrossRef Search ADS PubMed Lindgren KP , Neighbors C , Teachman BA , et al. . ( 2016 a) Implicit alcohol associations, especially drinking identity, predict drinking over time . Health Psychol 35 : 908 – 18 . Google Scholar CrossRef Search ADS PubMed Lindgren KP , Ramirez JJ , Olin CC , et al. . ( 2016 b) Not the same old thing: establishing the unique contribution of drinking identity as a predictor of alcohol consumption and problems over time . Psychol Addict Behav 30 : 659 – 71 . Google Scholar CrossRef Search ADS PubMed Markus H , Wurf E . ( 1987 ) The dynamic self-concept: a social psychological perspective . Annu Rev Psychol 38 : 299 – 337 . Google Scholar CrossRef Search ADS Naimi TS , Brewer RD , Mokdad A , et al. . ( 2003 ) Binge drinking among US adults . J Am Med Assoc 289 : 70 – 5 . Google Scholar CrossRef Search ADS Nosek BA , Greenwald AG , Banaji MR . ( 2007 ) The Implicit Association Test at age 7: a methodological and conceptual review. In Bargh JA (ed) . Automatic Processes in Social Thinking and Behavior . New York, NY: Psychology Press , 265 – 92 . Nosek BA , Hawkins CB , Frazier RS . ( 2011 ) Implicit social cognition: from measures to mechanisms . Trend Cog Sci 15 : 152 – 9 . Google Scholar CrossRef Search ADS Open Science Collaboration . ( 2015 ) Estimating the reproducibility of psychological science . Science 349 : aac4716 . CrossRef Search ADS PubMed Ostafin BD , Palfai TP . ( 2006 ) Compelled to consume: the Implicit Association Test and automatic alcohol motivation . Psychol Addict Behav 20 : 322 – 7 . Google Scholar CrossRef Search ADS PubMed O’Connor RM , Lopez-Vergara HI , Colder CR . ( 2012 ) Implicit cognition and substance use: the role of controlled and automatic processes in children . J Stud Alcohol Drugs 73 : 134 – 43 . Google Scholar CrossRef Search ADS PubMed O’Neill SE , Parra GR , Sher KJ . ( 2001 ) Clinical relevance of heavy drinking during the college years: cross-sectional and prospective perspectives . Psychol Addict Behav 15 : 350 – 9 . Google Scholar CrossRef Search ADS PubMed Reich RR , Below MC , Goldman MS . ( 2010 ) Explicit and implicit measures of expectancy and related alcohol cognitions: a meta-analytic comparison . Psychol Addict Behav 24 : 13 – 25 . Google Scholar CrossRef Search ADS PubMed Rooke SE , Hine DW , Thorsteinsson EB . ( 2008 ) Implicit cognition and substance use: a meta-analysis . Addict Behav 33 : 1314 – 28 . Google Scholar CrossRef Search ADS PubMed Stacy AW , Wiers RW . ( 2010 ) Implicit cognition and addiction: a tool for explaining paradoxical behavior . Annu Rev Clin Psychol 6 : 551 – 75 . Google Scholar CrossRef Search ADS PubMed Van Der Vorst H , Krank M , Engels RCME , et al. . ( 2013 ) The mediating role of alcohol-related memory associations on the relation between perceived parental drinking and the onset of adolescents’ alcohol use . Addiction 108 : 526 – 33 . Google Scholar CrossRef Search ADS PubMed Wiers RW , Bartholow BD , van den Wildenberg E , et al. . ( 2007 ) Automatic and controlled processes and the development of addictive behaviors in adolescents: a review and a model . Pharmacol Biochem Be 86 : 263 – 83 . Google Scholar CrossRef Search ADS Wiers RW , Rinck M , Kordts R , et al. . ( 2010 ) Retraining automatic action-tendencies to approach alcohol in hazardous drinkers . Addiction 105 : 279 – 87 . Google Scholar CrossRef Search ADS PubMed Wiers RW , van Woerden N , Smulders FT , et al. . ( 2002 ) Implicit and explicit alcohol-related cognitions in heavy and light drinkers . J Abnorm Psychol 111 : 648 – 58 . Google Scholar CrossRef Search ADS PubMed © The Author(s) 2018. Medical Council on Alcohol and Oxford University Press. All rights reserved. This article is published and distributed under the terms of the Oxford University Press, Standard Journals Publication Model (https://academic.oup.com/journals/pages/about_us/legal/notices)

Journal

Alcohol and AlcoholismOxford University Press

Published: Mar 27, 2018

There are no references for this article.

You’re reading a free preview. Subscribe to read the entire article.


DeepDyve is your
personal research library

It’s your single place to instantly
discover and read the research
that matters to you.

Enjoy affordable access to
over 18 million articles from more than
15,000 peer-reviewed journals.

All for just $49/month

Explore the DeepDyve Library

Search

Query the DeepDyve database, plus search all of PubMed and Google Scholar seamlessly

Organize

Save any article or search result from DeepDyve, PubMed, and Google Scholar... all in one place.

Access

Get unlimited, online access to over 18 million full-text articles from more than 15,000 scientific journals.

Your journals are on DeepDyve

Read from thousands of the leading scholarly journals from SpringerNature, Elsevier, Wiley-Blackwell, Oxford University Press and more.

All the latest content is available, no embargo periods.

See the journals in your area

DeepDyve

Freelancer

DeepDyve

Pro

Price

FREE

$49/month
$360/year

Save searches from
Google Scholar,
PubMed

Create lists to
organize your research

Export lists, citations

Read DeepDyve articles

Abstract access only

Unlimited access to over
18 million full-text articles

Print

20 pages / month

PDF Discount

20% off