Alcohol and Alcoholism Advance Access originally published online on May 9, 2008
Alcohol and Alcoholism 2008 43(5):551-558; doi:10.1093/alcalc/agm174
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
Relationships Between Alcohol-Related Memory Association and Changes in Mood: Systematic Differences Between High- and Low-Risk Drinkers
1 School of Social Science, The University of Queensland, Australia
2 School of Rural Health, Monash University, Victoria, Australia
* Author to whom correspondence should be addressed: Michie Building, School of Social Science, The University of Queensland, St Lucia, 4072, Australia. Tel.: +61 07 33656663; Fax: +61 07 33651544; E-mail a.kelly{at}uq.edu.au
Received 21 September 2007; first review notified 5 November 2007; in revised form 16 November 2007; accepted 23 November 2007
 |
ABSTRACT |
|---|
 TOP ABSTRACT IntroductionMethodResultsDiscussionReferences |
|---|
Heavy alcohol use is common in undergraduates and is associated with health-risk behaviors, negative consequences, and increased risk for future alcohol dependence. Alcohol-related memory associations (AMAs) and mood changes are independently related to student drinking, but more research on how these variables interact is needed. Aims: To examine (i) how AMAs predict drinking behavior after accounting for depression, and (ii) how changes in negative and positive mood predict AMAs among low- and high-risk drinkers. Methods: Positive and negative moods were manipulated using a musical mood induction procedure immediately prior to completion of memory association measures. A bootstrapped structural equation model was tested, permitting a sampling distribution free of the requirement of normality. Results: Negative mood changes predicted AMAs in high-risk drinkers but not in low-risk drinkers, and the opposite was found for positive mood changes. Conclusion: The negative mood–AMA association appeared related to risky drinking, and these subtle implicit cognitive processes may warrant a special focus in intervention programs for high-risk drinkers.
|
Introduction |
|---|
|
TOPABSTRACTIntroductionMethodResultsDiscussionReferences |
|---|
In Western societies, alcohol consumption is a common feature of university undergraduate experience (e.g., Pihl et al., 1993
; Maio et al., 1994; Roche and Watt, 1999; O’Malley and Johnston, 2002; Jones, 2003). Heavy episodic consumption (i.e., drinking five or more standard drinks per occasion; NHMRC, 1992) is also widespread among college students and is associated with a range of health-risk behaviors and negative consequences (Pihl et al., 1993; Maio et al., 1994; O’Malley and Johnston, 2002). In young people, heavy drinking is associated with injuries and motor vehicle accidents (McGinnis and Foege, 1993), unsafe sex (Weschler et al., 1994; Cooper, 2002), sexual and physical assault (Engs and Hanson, 1985), ethanol poisoning (Greenfield, 2001), smoking (Kelly and Jackson-Carroll, 2007; Kelly et al., 2006), and increased risk of developing alcohol dependence (Baer, 2002; Schulenburg and Maggs, 2002; Dawson et al., 2004).
Social-cognitive models of alcohol use have long emphasized the importance of explicit (conscious and considered) cognition in accounting for heavy drinking. A large body of literature supports the utility of alcohol expectancy and refusal self-efficacy models in explanations of drinking, though overall affects are modest (Leigh and Stacy, 1991). In more recent years, there have been challenges to the notion that explicit cognitive processes are the primary driver of drinking-related decisions (e.g., Stacy, 1995, 1997; Goldman, 1999; Kelly and Witkiewitz, 2003). Contemporary cognitive explanations of alcohol use emphasize the role of automatic information processing that may be more implicit (occurring outside awareness) than explicit in determining drinking outcomes (Tiffany, 1990; Greenwald and Banaji, 1995; Stacy, 1997; Tiffany and Conklin, 2000). Associative memory network theorists propose that alcohol-related information is interlinked in memory and that accessibility of these informational nodes is variable. A central assumption is that these alcohol-related networks contain representations of cues depending on the alcohol-related learning history of the individual (e.g., Stacy, 1995, 1997; Tiffany and Conklin, 2000). The informational nodes that increase the likelihood of drinking may vary from apparently irrelevant, or ambiguous, associations with alcohol (e.g., being tired, hot, or stressed) to strong associations with drinking (e.g., bars/pubs). People who drink heavily are proposed to be more likely to experience alcohol-related activation in response to ambiguous cues, compared to others.
To investigate the role of accessibility of alcohol-related memory associations (AMAs) in the prediction of drinking, researchers have previously used a cue-association paradigm (e.g., Stacy, 1995). In this paradigm, free associations are made to ambiguous alcohol-related homographs (e.g., pitcher, tap) embedded in a list of homographs not related to drinking (e.g., stair, field). Responses are then coded for alcohol-related references (see Stacy, 1997). This task is held to reflect implicit memory processes because respondents are not asked to introspect about outcomes and are not aware of the alcohol focus of the research. Studies utilizing cue-association measures have shown that AMAs cross-sectionally and longitudinally predict drinking in young adults (Stacy, 1995; Weingardt et al., 1996; Stacy, 1997; Palfai and Wood, 2001; Kelly et al., 2005). AMAs are associated with problem drinking among drug offenders (Ames and Stacy, 1998; Ames et al., 2002), and predict alcohol and marijuana use in high-risk adolescents (Ames et al., 2005).
Although research indicates a univariate association between AMAs and alcohol consumption, there is limited research exploring the potential role of affect in moderating/mediating the AMA–drinking behavior relationship. The basis for incorporation of affect into AMA models is strong given that negative and positive affect have long been implicated in drinking behavior. For example, depressed affect increases self-reported craving and motivation to drink among recreational drinkers (Willner et al., 1998). Nervous mood predicts increased drinking among social drinkers (Swendsen et al., 2000), and negative affect is a frequently endorsed antecedent to relapse in treated drinkers (Strowig, 2000). Drinking to enhance positive mood is a commonly endorsed motive for drinking among university undergraduates (e.g., Stewart et al., 1996). Positive mood enhancement motives have also predicted alcohol-related problems in college students (Carey and Correia, 1997) and are a frequent motivator for drinking in social situations (Kilty, 1990; Fromme and Dunn, 1992). In experimental research (involving the systematic manipulation of mood) there is good evidence of a main affect of mood on alcohol consumption. Negative mood induction produces higher ratings of urges to drink among alcohol-dependent people and undergraduates, relative to neutral mood (Cooney et al., 1997; Willner et al., 1998).
There is a strong theoretical basis for exploring the role of affect in mediating/moderating the AMA–drinking behavior association. According to the affect-priming principle (Bower, 1981), affective states have specific nodes in memory that are linked to other nodes containing memories of events where that emotion was aroused. Affect can therefore prime the kind of associations elicited by a stimulus, and the greater the availability of mood-consistent associations, the greater the constructive interpretation of ambiguous details (Bower, 1991; Clark and Waddell, 1983). For people with a history of alcohol consumption, reliable links between certain affect states and alcohol events are established. When ambiguous (potentially alcohol-related) stimuli are encountered during affect priming, the drinker is more likely to construct responses potentially related to alcohol, than otherwise. Consistent with this principle, negative mood-related words facilitated priming for alcohol targets in problem drinkers with high levels of psychiatric distress (Zack et al., 1999). In college students, priming with negative mood phrases reduced reaction time to alcohol target words while positive mood phrases did not (Zack et al., 2003).
Traditional linear modeling in alcohol-related associative memory research is hampered by the common non-normality of AMA and drinking variables (Kelly et al., in press). Ordinary least squares (OLS) regression and maximum likelihood methods assume that the errors are independently and identically distributed as a Gaussian (or normal) probability distribution, and that data are continuous and multivariate normal (Byrne, 2001). However, AMAs and young adult substance-abuse data are often in the form of counts and are zero inflated, so probability distributions are typically positive and non-normal (Kelly et al., in press; Kelly and Jackson-Carroll, 2007). In structural equation modeling, this can result in spuriously large 
2 values, failure to converge, and spuriously low standard errors (the latter resulting in regression paths that are statistically significant though these may be unreplicable; Byrne, 2001; Yung and Bentler, 1996). Bootstrapping is a way of increasing the robustness of SEM to violations of normality. Bootstrapping is a resampling procedure in which the original sample is presumed to represent the population, and multiple subsamples are randomly drawn with replacement, permitting an evaluation of the stability of parameter estimates and indices of fit (Zhu, 1997).
The overall aim of the present study was to explore the affect of altered mood state on accessibility of AMAs in undergraduate student drinkers. Given the covariance of various alcohol consumption indicators (e.g., frequency of drinking (days/week), quantity consumed per session, frequency of binge drinking, and alcohol-related problems), Hypothesis 1 was that a measurement model of drinking behavior that incorporates these indicators would show a good fit to the data and these indicators would each significantly load on the latent variable. Given established univariate associations between AMAs and drinking (e.g., Kelly et al., 2005; Stacy, 1995, 1997), Hypothesis 2 was that AMAs would significantly predict drinking behavior after accounting for trait-like depression and changes in mood following experimental manipulation. The core hypothesis of the study (Hypothesis 3) was that increases in negative mood would predict increased alcohol-related memory associations among high-risk drinkers compared to low-risk drinkers. Hypothesis 4 was that increases in positive mood would predict increased alcohol-related memory associations among high-risk drinkers compared to low-risk drinkers. Differential predictions of how changes in positive/negative mood might impact on AMAs were not made, given the evidence that induced changes in a variety of moods (happy, sad, anxious) induce craving among heavy drinkers (Rubonis et al., 1994).
|
Method |
|---|
|
TOPABSTRACTIntroductionMethodResultsDiscussionReferences |
|---|
Participants
The sample consisted of 109 university undergraduate drinkers from an introductory psychology course who participated in return for research credit points. All participants were of legal drinking age (18 years in Australia). There were 74 females and 35 males with a mean age of 20.5 years (SD = 4.18) from a range of 18–35 years. Three of these participants were excluded because they correctly identified the underlying focus of this implicit study (see Procedure), resulting in a final sample of 106. Five participants (4.8%) indicated that English was their second language. The majority of participants were current drinkers (96.2%) who indicated that they had drunk at least once over the past month. The remaining 3.8% identified themselves as current abstainers. Participants reported a mean frequency of drinking sessions of 6.1 per month (SD = 4.6), while mean quantity in number of standard drinks per session was 3.2 (SD = 2.8). Fifty percent of participants reported consuming a peak quantity of greater than five drinks on one occasion, while 28% drank in excess of 10 drinks in one session. While the majority of students (67%) reported this peak consumption on only one occasion during the past month, 13% drank this amount on at least three occasions. The present sample compares favorably with consumption norms for this student population. For example, a survey of 400 (209 male, 191 female) students aged 17–25 years conducted across three university campuses in the same geographical region as the present study showed that 94% of students were current drinkers and that 69% drank at risky levels (Roche and Watt, 1999
). Participants reported a mean score of 9.21 (SD = 5.2) on the Beck Depression Inventory (BDI; see Measures). Total scores on this measure indicated that 57% were not depressed, 31% had mild nonclinical depression, 10% had borderline clinical depression, and 7% had moderate clinical depression.
Measures
Current mood
In line with recent mood induction research (Grant et al., in press) a visual analogue scale was used to assess current mood state. This type of scale allows for a more fine-grained capture of variability in mood than conventional Likert scales. The visual analogue scale consisted of eight adjectives assessing positive mood (e.g., cheerful, happy, pleased) and eight adjectives assessing negative mood (e.g., sad, despondent, depressed, blue). To complete the scale, participants were required to place a vertical mark on a 100-mm horizontal line with endpoints anchored at 0 (not at all) to 100 (extremely) that indicated to what extent they felt this way at the present moment. Coefficient alphas for the positive and negative visual analogue scale were 0.94 and 0.89 at baseline, and 0.97 and 0.89 at post–mood induction, respectively.
Alcohol-related memory associations (AMAs)
AMAs were assessed using a variation of the cue-association paradigm (see the paragraphs that precede; Stacy, 1995, 1997) developed for Australian populations in our previous research (see Kelly et al., 2005). In our variation, the Memory Association Test (MAT; Kelly et al., 2005), overall scores longitudinally predict drinking in young Australian adults after controlling for autocorrelational affects of previous drinking and memory associations (Kelly et al., 2005). The MAT makes no reference to alcohol use and consists of 38 homographs of which five were ambiguously associated with alcohol (e.g., port, shot). Students respond with the first word that comes to mind and responses are coded according to whether they are alcohol-related or not using a master list of unambiguous alcohol-related words. This procedure was conservative with words such as drink not coded as alcohol-related (Kelly et al., 2005). The derived measure is the total number of alcohol-related responses. The MAT begins with "For the first questions, you will be asked to write single words that immediately come to mind in response to other words. For these types of questions, please respond with the very first word that comes to mind, whatever it is. Remember your answers are totally anonymous, and there are no right or wrong answers. Write next to each word the first word it makes you think of. For example, if the word is doctor, you might write "nurse". Work quickly!" (adapted from Stacy, 1997, p. 64).
Drinking behavior
Drinking was assessed using an explicit self-report measure of quantity and frequency of consumption included at the end of the assessment session. To measure frequency of drinking over the past month, participants were asked to tick one of seven boxes for each day of the week that they usually had a drink, and the typical quantity consumed on each day. This approach accounts for weekend drinking that is characteristic of undergraduate consumption patterns (e.g., Weingardt et al., 1996). From this measure, three indicators of drinking behavior were derived: measures of drinking frequency (days/month), average daily amount of drinking, and number of binge drinking episodes in the last month (more than six drinks a day; NHMRC, 1992). Alcohol problems were measured using the 23-item Rutgers Alcohol Problem Index (RAPI; White and Labouvie, 1989), which has high internal consistency (Cronbach alpha.84; Grant et al., in press) and is moderately correlated with drinking intensity among young adults (White and Labouvie, 1989).
Depression
To test for overall differences between the three conditions on trait-like depression, the Beck depression inventory (BDI; Beck et al., 1964) was administered. The rationale for administering the BDI was to avoid the potential confound of trait-like depression and temporary shifts in mood, the latter of which was of central interest in this study. The BDI is a valid and reliable measure of depression and is highly correlated with other measures of negative affect (Watson and Clark, 1984).
Procedure
This study approximated the procedures used in earlier mood induction research (McKee et al., 2003; Goldstein et al., 2004). The study was advertised as a "health and musical preference study" to disguise the alcohol-related focus of the research and to preserve the integrity of the MAT. Testing took place in a quiet, air-conditioned laboratory setting in groups of six participants per session. Participants were randomly assigned to the positive, negative or neutral conditions. Each participant sat at a separate desk facing away from other participants.
Once informed consent was obtained, participants were given standardized instructions to complete Section 1 of the questionnaire that comprised a baseline positive/negative mood measure and the BDI. Participants were randomly assigned to three conditions within a musical mood induction procedure (MMIP) (see paragraphs that follow). Participants in the two experimental conditions (positive/negative MMIP) listened to the 10-minute selection of music through headphones, while recording any thoughts, feelings or images that came to them while they listened to the music (e.g., McKee et al., 2003). This task was included to focus attention on the affective content of the music and to employ multiple stimuli (e.g., auditory, visualization) that may improve the affectiveness of the mood manipulation (Goodwin and Sher, 1993).
Participants listened to either a 10-minute segment of classical music to induce a negative mood (Prokofiev's Russia Under the Mongolian Yoke and Sibelius’ Swan of Tuonela) or positive mood (Bach's Brandenburg Concertos Nos. 2 and 3 and Handel's Water Music). These selections have been found in previous studies to successfully alter mood (e.g., Goodwin and Sher, 1993; Hufford, 2001). In the control condition, participants were given a 10-minute break in the quiet, air-conditioned surroundings of the laboratory foyer. The control condition was similar to recent studies employing mood-induction procedures (e.g., Hufford, 2001; McKee et al., 2003; Goldstein et al., 2004) in which no changes in mood among control participants was found. We chose this control condition over other possible control conditions (e.g., listening to white or other noise) because we envisaged that the latter control conditions would likely produce unwanted changes in affect.
Once the mood induction finished, participants received recorded instructions to remove their headphones and complete Section 2 of the questionnaire booklet. Section 2 comprised a second positive/negative mood assessment, immediately followed by the MAT, a distracter task (rating of musical style preferences), and demographic questions. To check the degree to which mood changes were sustained during the completion of the MAT, a third positive/negative mood assessment was conducted at the end of this section of the assessment session. During the MMIP, participants in the control condition were asked to take a 10-minute break but to remain in the laboratory foyer area. All other instructions and measures were identical to those for participants in the mood induction conditions.
When Section 2 was concluded, all participants completed Section 3 of the questionnaire comprising the explicit alcohol use measure. Prior to beginning Section 3, however, participants also completed the following validity check question. "Please write below what you think the main emphasis and goals of the study were. In other words, tell us what you think the study was about". Three students identified alcohol use as a possible focus of the study and were dropped from the analysis given the focus of the study on implicit cognition.
|
Results |
|---|
|
TOPABSTRACTIntroductionMethodResultsDiscussionReferences |
|---|
In the first stage of the analysis, the aim was to examine the extent to which the mood manipulation was successful. This entailed testing differences between the three experimental groups in changed positive and negative mood (pre– and post–mood induction, and on potentially confounding variables such as BDI scores, and demographic variables (age, gender).
To investigate the affectiveness of the mood manipulation, a 3 (mood manipulation condition) by 3 (time) mixed model ANOVA was conducted with time (baseline, mood induction, and study completion) as the within subjects factor. The first analysis was on positive mood ratings. We expected that the positive mood manipulation would significantly increase positive mood, and the negative mood manipulation would significantly decrease positive mood. The main affect for time showed a trend towards significance, F(2, 102) = 3.26, P < 0.1, and the interaction of time and group was significant, F(4, 206) = 21.52, P < 0.001, partial 
2 = 0.30. At the group level, the control group showed no significant change over time. The positive mood manipulation group showed a significant increase from Time 1 to Time 2 (pre- to post-manipulation), M(T1) = 69.2 (95% CIs 63.1, 76.9), M(T2) = 78.7 (95% CIs 71.3, 86.1), and a significant decrease in positive mood from Time 2 to Time 3, M(T3) = 71.8 (95% CIs 64.0, 79.6). The negative mood manipulation group showed a significant decrease in positive mood from Time 1 to Time 2, M(T1) = 70.4 (95% CIs 63.4, 77.4), M(T2) = 51.2 (95% CIs 43.8, 58.6), and a significant increase in positive mood from Time 2 to Time 3, M(T3) = 59.6 (95% CIs 51.8, 67.4).
The second analysis was on negative mood ratings. The main affect for time showed a trend towards significance, F(2, 102) = 2.95, P < 0.1, and the interaction of time and group was significant, F(4, 206) = 21.48, P < 0.001, partial 2 = 0.29. At the group level, the control group showed no significant change over time. The negative mood manipulation group showed a significant increase in negative mood from Time 1 to Time 2, M(T1) = 33.8 (95% CIs 27.0, 40.6), M(T2) = 48.1 (95% CIs 41.5, 54.8), and a significant decrease in negative mood from Time 2 to Time 3, M(T3) = 34.6 (95% CIs 27.9, 41.4). The positive mood manipulation group showed a significant decrease in negative mood from Time 1 to Time 2, M(T1) = 35.7 (95% CIs 29.0, 42.5), M(T2) = 25.3 (95% CIs 18.6, 31.9), and a nonsignificant change in negative mood from Time 2 to Time 3. In all, the mood manipulation resulted in significant changes in mood in the expected directions from pre- to post-manipulation. A between-groups ANOVA analysis showed there were no significant differences between the positive, negative and control conditions at baseline on age, BDI scores, MAT scores, and mood ratings. Chi-square analysis showed no differences in gender across mood conditions. In most instances, affect ratings returned from post-manipulation elevations to near-baseline levels at Time 3. However, it took participants approximately 15 minutes to complete the study before responding to the final affect rating measure and mood induction affects may have dissipated over this time. However, the key memory association measures were completed in the first few minutes immediately following the mood induction. This ensured that memory accessibility measures were completed in the context of differential mood.
In the second stage of the analysis, a structural equation model of drinking behavior was evaluated (see Fig. 1). The distribution of the total number of alcohol-related associations was heavily skewed, justifying a bootstrapping analysis that compensated for the violation of normality for this variable. Specifically, 11 participants (10.4% of the sample) did not make an alcohol-related response. Forty-eight (45.3%) made one response, 36 (34%) made two responses, 11 (10.4%) made three responses. Because of the anticipated intercorrelations of frequency of average drinking, average quantity per session, frequency of binge drinking, and alcohol problems (Kelly et al., 2006), a latent variable drinking behavior was constructed, with each of these specific measures used as indicators of the latent variable. Maximum-likelihood estimation was used to assess the fit of the models, using the 2 goodness of fit statistic (P > 0.05) and the root mean square error of approximation (RMSEA).
|
The full model (Fig. 1) showed good overall fit,
2(13) = 16.46, P = 0.2, RMSEA = 0.05. Consistent with Hypothesis 1, the four alcohol indicators loaded highly on the latent variable (standardized regression weights 0.79 for frequency, 0.93 for quantity, 0.79 for frequency of binge drinking, and 0.64 for alcohol problems). Consistent with Hypothesis 2, AMAs were significantly predictive of the latent variable drinking behavior, B = 3.37, SE(B) = 0.65, CR = 5.15, P < 0.001. Change in negative mood showed a trend towards predicting drinking behavior significantly, B = 0.08, SE(B) = 0.05, CR = 1.71, P < 0.1. BDI scores were not related significantly to drinking behavior. A replication of this model with positive mood changes substituted for negative mood changes showed no meaningful differences in the regression weights for indicators of drinking behavior. These results were consistent with Hypothesis 2—AMAs predicted drinking behavior after accounting for trait-like depression and changes in mood following experimental manipulation.
To test Hypotheses 3 and 4, a two-group bootstrapped structural model was tested (low- versus high-risk drinkers). To create the two groups, a median split on total alcohol consumption in the past month was conducted. The median split was 15 drinks/month. This cut-off for group membership conveniently approximated Australian national guidelines for low- versus high-risk drinking (low-risk: less than 2 and 4 drinks per day for females and males, respectively, and greater than 2 and 4 standard drinks per day for females and males, respectively, with two alcohol-free days per week; NHMRC, 1992). Because the multi-group model contained roughly half in each group, some changes to the overall model (Fig. 1) had to be made to reduce the number of sample moments to be estimated and hence increase statistical power. To achieve this, an observed variable was created using the drinking behavior indicator loadings determined above (Byrne, 2001). A single observed variable was computed by summing the four weighted drinking indicators.
The two-group negative-mood model consisted of two observed variables, changes in negative mood and AMAs (T2 mood rating minus T1 mood rating), each with error terms, and the regression path between the two indicators constrained and unconstrained to determine how this path varied across the two groups. The two-group negative-mood model showed good fit, 2(2) = 0.07, P = 0.9, RMSEA = 0.001. For low-risk drinkers, the path from change in negative mood to AMAs was not significant, B = –0.005, SE(B) =.008, P = 0.5. For high-risk drinkers, this path was significant, B = 0.015, SE(B) = 0.007, P < 0.025. This result was consistent with Hypothesis 3—B(high-risk drinkers) was greater than 1.96 standard errors from B(low-risk drinkers) and was therefore significantly greater in magnitude.
The same two-group model was run, with positive mood change as the observed predictor instead of negative mood change. The two-group positive-mood model showed a good fit: 2(2) = 0.84, P = 0.7, RMSEA = 0.001. The pattern of differences between high- and low-risk drinkers in this model was opposite to the pattern in the model of negative mood change. For low-risk drinkers, the path from change in positive mood to AMAs was significant, B = 0.013, SE(B) = 0.006, P < 0.05. For high-risk drinkers, this path was not significant, B = –0.008, SE(B) = 0.006, P = 0.2. This result was opposite to that predicted in Hypothesis 4—B(low-risk drinkers) was greater than 1.96 standard errors from B(high-risk drinkers).
|
Discussion |
|---|
|
TOPABSTRACTIntroductionMethodResultsDiscussionReferences |
|---|
The aim of the present study was to investigate the affects of mood on accessibility of AMAs among undergraduate low- versus high-risk drinkers. In preparation for this evaluation, an overall model incorporating a latent variable of drinking behavior was evaluated. Consistent with Hypothesis 1, the four indicators of drinking behavior loaded highly on the latent variable. Consistent with Hypothesis 2 and prior research (Kelly et al., 2005
, 2006), AMAs predicted the latent measure of drinking behavior significantly. The weightings obtained for these indicators were then used to create an observed variable for use in the bootstrapped two-group structural equation model. Consistent with Hypothesis 3, there was a stronger association of changes in negative mood and AMAs for high-risk drinkers versus low-risk drinkers. A novel finding was evident for changes in positive mood, where the pattern of results for negative mood was reversed (inconsistent with Hypothesis 4). There was a stronger association of changes in positive mood and AMAs for low-risk drinkers versus high-risk drinkers. The results for Hypotheses 3 and 4 were unlikely to be so because of systematic differences in the magnitude of mood changes across high- versus low-risk drinkers. For low-risk drinkers, the mean difference score (T2 – T1) for positive and negative mood was –1.4 (SD = 17.1) and –2.3 (SD = 15.7), respectively. For high-risk drinkers, the mean difference score for positive and negative mood was –4.2 (SD = 19.3) and –3.7 (SD = 15.2), respectively. Overall, the results suggest that changes in mood increase the accessibility of AMAs and that these changes varied according to the risk status of undergraduate drinkers.
The results are partially consistent with prior related research on implicit cognition and drinking severity. Zack et al. (2006) found that negative affect priming (but not positive affect priming) predicted increased beer consumption in a subsequent taste-testing session, and problem drinking predicted beer consumption in the negative affect prime condition. Zack et al. (2003) evaluated the affect of mood primes (conducted using mood-related sentence stems) on response times for reading alcohol target words. It was found that activation of alcohol words by negative mood cues were associated with greater severity of alcohol problems in young drinkers. However, the reverse association of positive mood change on AMAs (for low-risk drinkers) stands in contrast to the lack of reliable findings for the association of positive mood on associative memory (Wiers et al., 2002) and alcohol priming (Zack et al., 2003, 2006). The inconsistency between our result for positive mood and prior research may be because any significant affect was eroded by the inclusion of high-risk drinkers in early research. It is also possible that our group of high-risk drinkers varied systematically from low-risk drinkers in drinking motives potentially related to positive mood changes. For example, positive mood manipulations may only influence alcohol-related cognitions among drinkers who are motivated to enhance positive emotions (in contrast to the alleviation of negative emotions) (Birch et al., 2004). Our heavy-drinking subsample may have been generally motivated to use alcohol to alleviate negative mood, and so showed a greater association of negative mood change and AMAs than for positive mood change and AMAs. This confound is a difficult one to resolve, because drinking motives were not assessed in the present study, and Birch et al. (2004) did not control for alcohol problems in their study of drinking motives. However, it was noted in our sample that there were no significant differences between high- and low-risk drinkers in depression (BDI scores), M (low-risk) = 10.5 (SD = 6.6), M (high-risk) = 10.2 (6.25), and the latent construct drinking behavior was unrelated to depression (see Fig. 1). If the high-risk sample were systematically higher in a drinking motive of alleviating negative affect, we might anticipate that this group would have higher levels of depression.
The present study bears integration with more recent work by Grant et al. (in press) that extended their earlier seminal work on generic negative mood to examine the affects of musically induced positive and anxious mood on an alcohol Stroop task. A sample of 48 undergraduate students were selected on the basis of high scores on an enhancement and coping motive questionnaire. Because of pathways to participation, the sample on average had considerable alcohol problems (RAPI total score of 17.4 (SD = 10.4)). It was found that students whose motive to drink was to cope with anxiety showed an attentional bias towards alcohol-related stimuli, but this affect was not present when exposed to manipulation of positive-mood state. There were several obvious differences in sample and methodology between ours and Grant et al.'s study. Notably, our sample had substantially fewer alcohol problems, M(RAPI) = 5.5 (SD = 5.7), we used a generic negative mood measure, and we used the MAT rather than an RT-based attentional bias paradigm. However, the two sets of results are not inconsistent. Among our heavier drinkers, negative (and not positive) mood was related to AMAs. In our study, it was not possible to examine anxiety-specific state changes, because our mood ratings were heavily biased towards depressive-type states (e.g., sad, gloomy, unhappy, down). It remains for further research to examine whether the results of this study generalize to anxiety-specific manipulation. In any case, our findings suggest that changes in negative mood may have greater potential clinical or public health significance than changes in positive mood, because changes in negative mood were related exclusively to AMAs for high-risk drinkers.
The findings of this study contribute to our understanding of how certain mood states provide a context for implicit alcohol-related processes that may bias vulnerable individuals towards drinking. Mood change may be contextual to undergraduate drinking given that it indirectly (and not directly) predicted drinking behavior. While mood changes may cue drinking behavior via AMAs, it is clear that the AMA–drinking behavior association is strong and independent of positive/negative mood change (see Fig. 1). Positive and negative mood change is undoubtedly one of many internal (and external cues) that increase AMAs. On the basis of associative network models of alcohol consumption, we would anticipate that the two groups of drinkers (high/low) had historical differences in their experiences of drinking. High-risk students may have memory networks that link negative mood and drinking because of a negative reinforcement learning history (where drinking reduces negative feelings), or an operant conditioning history, where heavy drinking creates negative outcomes (e.g., negative mood) and the two constructs become paired in memory, or both. The data were insufficient to address these mechanisms, and future researchers might disentangle these affects by longitudinally evaluating the links between affect change, implicit memory associations, and explicit alcohol expectancies. The present study was partly experimental and correlational in its design (mood was systematically manipulated but AMAs and drinking were not). The present results would be strengthened by an evaluation of how manipulations in mood and AMAs impact on actual alcohol consumption in the laboratory setting.
Research on alcohol-related implicit cognition has pointed to the utility of interventions designed to influence risky implicit processes among heavy drinkers (Kelly and Witkiewitz, 2003; Kelly et al., 2005, 2006; Wiers and Stacy, 2006; Wiers et al., 2006). Most current clinical research has focused on changing explicit cognitive processes, and it may be the case that changes in explicit processing moderate implicit alcohol-related cognitive systems (Wiers et al., 2006). Results of the current research point to the potential utility of increasing the "mindfulness" of heavy drinkers to very subtle contextual changes (i.e., mood changes) that trigger alcohol-related associations. Common intervention strategies (e.g., coping with high-risk situations) could be supplemented by strategies that target high-risk processes that precede exposure to explicit alcohol-related cues.
Because of the nature of the implicit measure and the anonymity of data, the study may be less subject to explicit reporting biases, but ultimately relies on the accuracy of self-reports of drinking. A methodological strength-of-sorts was that mood ratings were taken after the MAT was administered to check for maintained affect elevations. However, the T3 mood rating was probably administered too late in the experimental procedure. The results indicated a return to baseline mood ratings except for negative mood change in the positive MMIP. Because the T2 mood rating was closer in real time to the MMIP than the T3 mood rating (1–3 min versus 10–15 min after the MMIP), we are confident that mood remained elevated during completion of the implicit measure, but ultimately this is unverified, and only supported by the statistical association of mood changes and AMAs. It is possible that the T3 mood rating was particularly susceptible to production of unwanted changes in affect (e.g., third administration producing boredom). However, we are confident that unwanted changes in affect resulting from the T2 mood rating were minimal, because in the neutral condition, there was no significant change in affect from T1 to T2. Of course, causality cannot be determined from these data, and the positive/negative mood change–AMA association may be epiphenomenal to changes in other constructs in response to the MMIP. The findings may not generalize to young adults outside university or to those that drink at more severe levels.
In conclusion, the results demonstrated that changes in negative and positive mood are predictive of concurrent AMAs, and that the strength of this prediction varied according to the alcohol-related risk status of the individual. Risky drinking was associated with a significant association of negative mood change and AMAs. Longitudinal research on how mood changes, alcohol consumption, and AMAs influence each other, is needed. Awareness-oriented intervention strategies (e.g., mindfulness training) may assist heavy drinkers in identifying the very early cognitive precursors to heavy-drinking episodes.
|
ACKNOWLEDGEMENTS |
|---|
This study was funded in part by Australian Research Council Discovery Grant DP0342587 and a NHMRC Career Development Award to the first author.
|
References |
|---|
|
TOPABSTRACTIntroductionMethodResultsDiscussionReferences |
|---|
Ames SL, Stacy AW. Implicit cognition in the prediction of substance use among drug offenders. Psychol Addict Behav (1998) 12:272–81.[CrossRef][Web of Science]
Ames SL, Sussman S, Dent CW, et al. Implicit cognition and dissociative experiences as predictors of adolescent substance use. Am J Drug Alcohol Abuse (2005) 31:129–34.[Web of Science][Medline]
Ames SL, Zogg JB, Stacy AW. Implicit cognition, sensation seeking, marijuana use and driving behaviour among drug offenders. Pers Individ Differ (2002) 33:1055–72.[CrossRef]
Baer JS. Student factors: Understanding individual variation in college drinking. J Stud Alcohol (2002) 14:40–53.
Beck AT, Ward CH, Mendelson M, et al. An inventory for measuring depression. Arch Gen Psychiatry (1964) 4:561–71.
Birch CD, Stewart SH, Wall A, et al. Mood-induced increases in alcohol expectancy strength in internally motivated drinkers. Psychol Addict Behav (2004) 18:231–8.[CrossRef][Web of Science][Medline]
Bower GH. Mood and memory. Am Psychol (1981) 36:129–148.[CrossRef][Medline]
Bower GH. Mood congruity of social judgments. In: Emotion and Social Judgments.—Forgas JP, ed. (1991) Elmsford, NY: Pergamon Press. 31–53.
Byrne BM. Structural Equation Modeling with AMOS: Basic Concepts, Applications, and Programming (2001) Mahwah, NJ: Lawrence Erlbaum.
Carey KB, Correia CJ. Drinking motives predict alcohol-related problems in college students. J Stud Alcohol (1997) 58:100–5.[Web of Science][Medline]
Clark MS, Waddell BA. Effects of moods on thoughts about helping, attraction and information acquisition. Soc Psychol Q (1983) 46:31–5.[CrossRef][Web of Science]
Cooney NL, Litt MD, Morse PA, et al. Alcohol cue reactivity, negative mood reactivity, and relapse in treated alcoholic men. J Abnorm Psychol (1997) 106:243–50.[CrossRef][Web of Science][Medline]
Cooper ML, Russell M, Skinner JB, et al. Development and validation of a 3-dimensional measure of drinking motives. Psychol Assess (1992) 6:117–28.
Dawson DA, Grant BF, Stinson FS, et al. Another look at heavy episodic drinking and alcohol use disorders among college and noncollege youth. J Stud Alcohol (2004) 65:477–88.[Web of Science][Medline]
Engs RC, Hanson DJ. Drinking patterns and problems of college students. J Alcohol Drug Educ (1985) 31:65–82.[Web of Science]
Forgas JP. Mood and judgement: the affect infusion model. Psychol Bull (1995) 117:39–66.[CrossRef][Web of Science][Medline]
Fromme K, Dunn ME. Alcohol expectancies, social and environmental cues as determinants of drinking and perceived reinforcement. Addict Behav (1992) 17:166–77.
Goldman MS. Risk for substance abuse: memory as a common etiological pathway. Psychol Sci (1999) 10:196–8.[CrossRef][Web of Science]
Goldstein AL, Wall A, McKee SA, et al. Accessibility of alcohol expectancies from memory: impact of mood and motives in college students. J Stud Alcohol (2004) 65:95–104.[Web of Science][Medline]
Goodwin AH, Sher KJ. Effects of induced mood on diagnostic interviewing: evidence for a mood and memory effect. Psychol Assess (1993) 5:197–202.[CrossRef]
Grant VV, Stewart SH, Birch CD. Impact of positive and anxious mood on implicit alcohol-related cognitions in internally motivated undergraduate drinkers. Addict Behav. in press.
Greenwald AG, Banaji MR. Implicit social cognition: Attitudes, self-steem and stereotypes. Psychol Rev (1995) 102:4–27.[CrossRef][Web of Science][Medline]
Greenfield TK. Individual risk of alcohol related disease and problems. In: International Handbook of Alcohol Dependence and Problems.—Heather N, Peters TJ, Stockwell T, eds. (2001) Brisbane: John Wiley. 413–37.
Hufford MR. An examination of mood effects on positive alcohol expectancies among undergraduate drinkers. Cogn Emotion (2001) 15:593–613.[CrossRef]
Jones BT. Alcohol consumption on the campus. Psychologist (2003) 16:523–5.[Web of Science]
Kelly AB, Haynes M, Marlatt GA. The impact of adolescent tobacco-related associative memory on smoking trajectory: an application of negative binomial regression to highly skewed longitudinal data. Addict Behav. in press.
Kelly AB, Jackson-Carroll C. Equifinality and interactivity of risks for adolescent smoking. J Child Adolesc Subst Abuse (2007) 17:51–64.[CrossRef]
Kelly AB, Masterman PW, Marlatt GA. Alcohol related associative strength and drinking behaviours: concurrent and prospective relationships. Drug Alcohol Rev (2005) 24:489–98.[CrossRef][Web of Science][Medline]
Kelly AB, Masterman PM, Marlatt GA. Tobacco-related associative strength in adolescents: a cross-sectional and contextual analysis. Nicotine Tob Res (2006) 8:49–55.
Kelly AB, Witkiewitz K. Accessibility of alcohol-related beliefs in young adults: a cross lag panel study. Alcoholism: Clin Exp Res (2003) 27:1–10.[Web of Science][Medline]
Kilty KM. Drinking styles of adolescents and young adults. J Stud Alcohol (1990) 51:556–64.[Web of Science][Medline]
Leigh BC, Stacy AW. On the scope of alcohol expectancy research: remaining issues of measurement and meaning. Psychol Bull (1991) 110:147–54.[CrossRef][Web of Science][Medline]
Maio RF, Portnoy J, Blow FC, et al. Injury type, injury severity, and repeat occurrence of alcohol related trauma in adolescents. Alcohol: Clin Exp Res (1994) 18:261–4.[CrossRef][Web of Science][Medline]
McGinnis JM, Foege WH. Actual cases of death in the United States. In: JAMA (1993) 270:2207–12.
McKee SA, Wall A, Hinson RE, et al. Effects of an implicit mood prime on the accessibility of smoking expectancies in college women. Psychol Addict Behav (2003) 17:219–25.[CrossRef][Web of Science][Medline]
NHMRC. Is There a Safe Level for Daily Consumption of Alcohol for Men and Women? (1992) National Health and Medical Research Council. Australian Government Publishing Service, Canberra: Australia.
O’Malley PM, Johnston LD. Epidemiology of alcohol and other drug use among American college students. J Stud Alcohol (2002) 14:23–39.
Palfai T, Wood MD. Positive alcohol expectancies and drinking behaviour: the influence of expectancy strength and memory accessibility. Psychol Addict Behav (2001) 15:60–7.[CrossRef][Web of Science][Medline]
Pihl RO, Petersen JB, Lau MA. A biosocial model of the alcohol aggression relationship. J Stud Alcohol (1993) 11:128–39.
Roche AM, Watt K. Drinking and university students: from celebration to inebriation. Drug Alcohol Rev (1999) 18:389–400.[CrossRef][Web of Science]
Rubonis AV, Colby SM, Monti PM, et al. Alcohol cue reactivity and mood induction in male and female alcoholics. J Stud Alcohol (1994) 55:487–94.[Web of Science][Medline]
Schulenburg J, Maggs JL. A developmental perspective on alcohol sue and heavy drinking during adolescence and the transition to young adulthood. J Stud Alcohol (2002) 54–70.
Stacy AW. Memory association and ambiguous cues in models of alcohol and marijuana use. Exp Clin Psychopharmacol (1995) 3:183–94.[CrossRef]
Stacy AW. Memory activation and expectancy as prospective predictors of alcohol and marijuana use. J Abnorm Psychol (1997) 106:61–73.[CrossRef][Web of Science][Medline]
Stewart SH, Zeitlin SB, Samoluk SB. Examination of a three dimensional drinking motives questionnaire in a young adult university student sample. Behav Res Ther (1996) 34:61–71.[CrossRef][Web of Science][Medline]
Strowig AB. Relapse determinants reported by men treated for alcohol addiction. The prominence of depressed mood. J Subst Abuse Treat (2000) 19:464–74.
Swendsen JD, Tennen H, Carney MA, et al. Mood and alcohol consumption: an experience sampling test of the self medication hypothesis. J Abnorm Psychol (2000) 109:198–204.[CrossRef][Web of Science][Medline]
Tiffany ST. A cognitive model of drug urges and drug use behaviour. Role of automatic and non-automatic processes. Psychol Rev (1990) 97:147–68.[CrossRef][Web of Science][Medline]
Tiffany ST, Conklin CA. A cognitive processing model of alcohol craving and compulsive alcohol use. Addiction (2000) 95:145–53.[CrossRef]
Watson D, Clark LA. Negative affectivity: the disposition to experience aversive emotional states. Psychol Bull (1984) 96:465–90.[CrossRef][Web of Science][Medline]
Weingardt KR, Stacy AW, Leigh BC. Automatic activation of alcohol concepts in response to positive outcomes of alcohol use. Alcohol: Clini Exp Res (1996) 20:25–30.[CrossRef]
Weschler H, Davenport A, Dowdall G. Health and behavioural consequences of binge drinking in college. JAMA (1994) 272:1672–1677.
White HR, Labouvie EW. Towards the assessment of adolescent problem drinking. J Stud Alcohol (1989) 50:30–7.[Web of Science][Medline]
Wiers RW, Stacy AW, Ames SL, et al. Implicit and explicit alcohol-related cognitions. Alcohol: Clini Exp Res (2002) 26:129–37.
Wiers RW, Stacy AW. Implicit cognition and addiction. Curr Dir Psychol Sci (2006) 15:292–6.[CrossRef]
Wiers RW, Cox M, Field M, et al. The search for new ways to change implicit alcohol-related cognitions in heavy drinkers. Alcohol: Clin Exp Res (2006) 30:320–31.[CrossRef][Web of Science][Medline]
Willner P, Field M, Pitts K, et al. Mood, cue and gender influences on motivation, craving and liking for alcohol in recreational drinkers. Behav Pharmacol (1998) 9:631–42.[CrossRef][Web of Science][Medline]
Yung Y-F, Bentler PM. Bootstrapping techniques in analysis of mean and covariance structures. In: Advanced Structural Equation Modeling: Issues and Techniques—Marcoulides GA, Schumacker RE, eds. (1996) Mahwah, NJ: Lawrence Erlbaum. 195–226.
Zack M, Poulos CX, Fragopulos F, et al. Effects of negative and positive mood phrases on priming of alcohol words in young drinkers with high and low anxiety sensitivity. Exp Clin Psychopharmacol (2003) 11:176–85.[CrossRef][Web of Science][Medline]
Zack M, Toneatto T, MacLeod CM. Implicit activation of alcohol concepts by negative affective cues distinguishes between problem drinkers with high and low psychiatric distress. J Abnor Psychol (1999) 198:518–31.
Zack M, Poulos CX, Fragopoulos F, et al. Effects of negative and positive mood phrases on priming of alcohol words in young drinkers with high and low anxiety sensitivity. Exp Clin Psychopharmacol (2003) 11:176–85.[CrossRef][Web of Science][Medline]
Zack M, Poulos CX, Fragopoulos F, et al. Negative affect words prime beer consumption in young adults. Addict Behav (2006) 31:169–73.[CrossRef][Web of Science][Medline]
Zhu W. Making bootstrap statistical inferences: a tutorial. Res Q Exerc Sport (1997) 68:44–55.[Web of Science][Medline]
CiteULike 
Connotea 
Del.icio.us What's this?
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||