|USFDC Home | USF Electronic Theses and Dissertations||| RSS|
This item is only available as the following downloads:
Examining the Distinction and Concordan ce between Implicit Measures of Alcohol Expectancies: Toward Agreement on Their Meaning and Use by Maureen C. Below A thesis submitted in partial fulfillment of the requirements for the degree of Master of Arts Department of Psychology College of Arts and Sciences University of South Florida Major Professor: Mark S. Goldman, Ph.D. Thomas H. Brandon, Ph.D. Doug Rohrer, Ph.D. Date of Approval: August 1, 2007 Keywords: expectancy, memory, indi rect, drinking patterns, cognition Copyright 2007, Maureen C. Below
i Table of Contents List of Tables iii List of Figures iv Abstract v Introduction 1 Explicit Measurement 1 Implicit Measurement 3 What is Implicit? 4 Implicit Measures of Semantic Association in Alcohol Research 10 Priming 10 Stroop 11 False Memory 12 Implicit Association Task 13 The Current Study 17 Methods 21 Participants 21 Measures 22 Experimental Measures 22 Free Associates 22 Primed Recall 24 Additional Assessments 26 Alcohol Expectancy Questionnaire 26 Alcohol Expectancy Multiaxial Assessment 26 Daily Drinking Questionnaire 26 Demographics 27 Procedure 27 Results 29 Coding/Scoring 29 Sample Characteristics 30 Gender Analyses 37 Implicit Task Replication and Comparison 41 Intra-Session Reliability 52 Practice Effect/ Contamination Analyses 55 Regression Analyses 59 Discussion 65 Limitations 71 Directions for Future Research 74
iiReferences 75 Appendices 80 Appendix A: Primed Cue Stimuli by Word Type 80 Appendix B: Free Associate Cues 81 Appendix C: Alcohol Expectancy Questionnaire Items 82 Appendix D: Alcohol Expectancy Multiaxial Assessment Items 85 Appendix E: Demographics and Daily Drinki ng Questionnaire 86
iii List of Tables Table 1 Correlations Between Drinking Indices 31 Table 2 Correlations Between Dri nking Indices and AEQ Subscales 32 Table 3 Correlations Between Dri nking Indices and AEMax Subscales 33 Table 4 Drinking Descriptives by Gender 39 Table 5 Correlations Between Free Asso ciate Composites and Drinking Indices 43 Table 6 Correlations Between Free Associ ate Composites and AEQ and AEMax Factors 44 Table 7 Correlations Between Primed Re call Composites and Drinking Indices 45 Table 8 Table 9 Table 10 Table 11 Table 12 Table 13 Table 14 Table 15 Table 16 Table 17 Correlations Between Primed Reca ll Composites and AEQ and AEMax Subscales Correlations Between Time 1 Free Associate Composites and Time 1 Primed Recall Composites Correlations Between Time 1 Primed Recall Composites and Time 2 Free Associate Composites Correlations Between Time 1 Free Associate Composites and Time 2 Primed Recall Composites Correlations Between Time 1 and Time 2 Primed Recall Composites Correlations Between Time 1 and Ti me 2 Free Associate Composites Intra-Session Task Analysis: Changes Across Administrations Frequency and Percentage of Prim ed Recall Words Generated as Free Associates Linear Multiple Regression Analyses Predicting Drinking Indices from Separate AEQ, AEMax, and FA Models Linear Multiple Regression Analyses Predicting Drinking Indices from Implicit and Explicit Blocks 46 49 50 51 53 54 56 58 60 63
iv List of Figures Figure 1 Number of Drinks per Week 35 Figure 2 Figure 3 Diagram of Between-Condition Comparisons Number of Drinks per Week: Women 36 40 Figure 4 Number of Drinks per Week: Men 40
v Examining the Distinction and Concordan ce between Implicit Measures of Alcohol Expectancies: Toward Agreemen t on Their Meaning and Use Maureen C. Below ABSTRACT Alcohol expectancies have traditionally been measured with explic it self-report questionnaires, but in recent years implicit m easures have also been used to explore the tenets of expectancy theory. The basic psyc hometric properties of re liability and validity have not been established for most implicit task s, and the convergent validity of different implicit measures has not been explored. Despite these shortcomings, many researchers continue to treat implicit tasks as reliable and valid assessment tools. To address reliability and validity of implicit meas ures, 218 undergraduate women and men were recruited from the University of South Flor ida to examine the psyc hometric properties of and concordance between two previously esta blished implicit measures, Free Associates (FA) and a Primed Recall (PR) task. The FA task was replicated, demonstrating high concordance between FA responses and explic it measures and drinking. The PR task did not show a drinker-type effect as was previously reported. Though the relationship between the tasks could not be examined, an exploration of pract ice and contamination effects offers insight into how performa nce in similar comparison studies may be affected.
Implicit Expectancies 1 Introduction Alcohol expectancies are associations held in memory between stimuli, behavior, and outcome that affect alcohol-related behavi or. These associations vary according to individual differences in expe riences with alcohol and predic t future alcohol use. It has been shown, for example, that heavier dri nkers tend to endorse stronger positive and arousing expectancies than lig hter drinkers, who tend to en dorse sedating alcohol effects (Goldman, Reich, & Darkes, 2006; Reich & Goldman, 2005; Reich, Noll, & Goldman, 2005; Reich, Goldman, & Noll, 2004), and that such associations appear to develop in youth before drinking patterns do (Dunn & Goldman, 1998; Christiansen, Smith, Roehling, & Goldman; 1989). The alcohol expectancy literature using cognitive paradigms to probe these memory associa tions has grown tremendously over the past decade. With expanded methodology, however, has come growing debate over how best to capture alcohol-related memo ry associations. Use of explicit measurement tools and their drawbacks will be reviewed, followed by a discussion and review of implicit tools that have been used to unde rstand alcohol expectancies. Explicit Measurement Alcohol expectancies have most comm only been examined using explicit questionnaires, which ask individuals whether th ey concur with specific statements about the effects of alcohol. In recent years, th ere has been a rise in the use of implicit measurement, and numerous implicit cognitive research paradigms have been adapted to
Implicit Expectancies 2 probe alcohol expectancies. Use of implic it methods not only expands our assessment repertoire, but also allows researchers to address several lim itations of explicit measurement. For example, self-report of beliefs about alcohol may be distorted. They may reflect how one thinks one should feel about alcohol, instead of how one actually behaves in response to alcohol-related stimuli. Part icipants may also be sensitive to social desirability, which might vary in relation to reference group or experimental setting. For example, college students have been shown to use alcohol in greater amounts and more frequently than any other subgroup, and to hol d peers that can hold their liquor in high esteem. Thus college students may be subjec t to the normative influences of peers positive beliefs about the effects of alcohol when responding to questions about their own alcohol-related attitudes. The opposite may be true for individuals whose reference groups disapprove of alcohol use. Moreover, individuals may not actually be able to distinguish between their own beliefs about al cohol and those of th eir reference group. One who knows that alcohol is supposed to be good based on the reports and behaviors of her peers may endorse rela ted positive effects of alc ohol because she believes the experiences of others reflect truth, despite limited or negati ve personal experiences with alcohol. Another problem with self-report m easures may be rooted in the language used by such measures. Words used by researchers to describe the potential effects of alcohol may not accurately describe the subjectiv e effects experienced by each individual,
Implicit Expectancies 3 leading to a miscommunication of ideas and the possible repo rt of beliefs contrary to those actually held. Implicit Measurement There are two primary benefits that the im plicit measures may offer. First, the use of implicit expectancy measures may mi nimize some of the above-stated problems associated with explicit self -report. Implicit cognitions ar e thought to be automatic and immediate, and less influenced by conscious deliberation, if at al l (Goldman, Del Boca, & Darkes, 1999; Goldman, 1999; Roediger, 2003; Roediger & Amir, 2005; Fazio & Olson, 2003; Stacy, 1997). Therefore, it has been thought that finding a way to measure implicit cognitions about alcohol may circumve nt measurement difficulties intrinsic to the use of explicit asse ssment. Automatic cognitive proc esses are far less available to deliberation than purposeful processes, and participants would be significantly less capable of monitoring responses or consider ing the beliefs of a reference group while responding (Fazio & Olson, 2003; Roediger, 200 3; Roediger & Amir, 2005). Second, whereas the association between explicitly measured expectancies and alcohol use is strong, implicit measures may have unique predictive power (St acy, 1997; Weirs, van Woerden, Fren, & de Jong, 2002a; Jajoia & Earleywine, 2002; Palf ai & Ostafin, 2003). Thus, implicit measures do not appear to si mply be another way of tapping the same constructs explicit measures do, but provide unique information about human memory (Nosek, 2007). To the end of utilizing thes e benefits, many researchers have employed implicit paradigms to draw conclusions about the nature of alcohol expectancies.
Implicit Expectancies 4 However, several problems arise when th e study of implicit memory function is considered. What is implicit? First, the distinction between explicit and implicit types of memory must be understood. The frequency of pairing alc ohol behaviors and outcomes, either by observation or action, strengthens associative memory between alcohol-related concepts or between behavior (e.g. alc ohol use) and outcome. These associations appear to be formed at both explicit and implicit levels. Implicit cognition has b een described in the alcohol field as automatic activation of asso ciations in memory influenced by immediate motivational or situational factors (Stacy, 1997; Weirs, Stacy, Ames, Noll, Sayette, Zack, & Krank, 2002, Rather, Goldman, Roerich, & Brannick, 1992). According to this conceptualization, the distin ction between implicit and ex plicit lies in the complex interaction between contextual or motivati onal cues and memory activation. Explicit processes involve deliberative retrieval of information based on cues available to awareness and known goals. Implicit pro cesses occur before deliberation and interpretation, and under the influence of variables unidentified by the individual (Roediger & Amir, 2005; Roediger, 2003). Such processes are elusive by definition, and it has been argued that we cannot actually di rectly measure such implicit associations; implicit tasks can be employed, but it cannot be claimed that they fully reflect the memory content they attempt to measur e (Fazio & Olson, 2003; Roediger, 2003; Roediger & Amir, 2005). Implicit tasks do not ask participants to recall events, but
Implicit Expectancies 5 instead attempt to probe memory automatically activated by certain alcohol-related cues. Therefore, that which researchers want to understand (i.e. alcohol expectancies) is by definition a construct that can only be accessed by the measures, or cues, we select for the task. In this way, it has been argued that exploration of alcohol expectancies must be described as being through implicit means since we cannot verify that our measures truly quantify implicit alc ohol-related cognition (Fazio & Olson, 2003). Roediger (2003) defines implicit memo ry as being the after-effects of stimulation that occur in the absence of atte mpts at conscious recollection. While this definition seems to encompass almost all memo ry processes, he goes on to point out that the range of available measures that do not employ attempts at conscious recollection is actually much more limited. Additionally, De Houwer (2006) incorpor ates this idea into the three primary criteria he has established for a measure to qualify as implicit, one of which must be satisfied. These are: particip ant unawareness of the attitude or cognition of interest, lack of conscious access to the at titude or cognition of interest, or lack of participant control over meas urement outcome. The presence of any of these three criteria within the desi gn of an experimental task woul d block participants attempts at conscious recollection. A further understanding of the goals of implicit measurement can be gained by considering how memory guides behavior. In addition to purposeful action, behavior is unintentionally influenced by the implicit me mory of past events. Thus, implicit and explicit tasks reflect this conceptualization: explicit tasks are those in which participants
Implicit Expectancies 6 draw consciously upon memory of specific even ts, while implicit tasks are those in which participants are unaware of th e impact of past experiences on response. Explicit tasks are a straightforward and direct assessment of l earning based on previous events; implicit tasks are a more indirect assessment of th e influence of an individuals experiences. Despite these functional differences, e xplicit and implicit tasks can be quite similar. Roediger and Amir (2005) use th e comparison of word stem completion and cued word recall to demonstrate this. In bot h tasks, individuals are asked to study a list of words, perhaps including the word elephant . In an implicit word completion task, a participant would later be si mply asked to complete the stem ele-, with no further reference to the studied word list. Should th e instruction set specify that the participant complete the stem with a word previously stud ied, the task would then be an explicit cued recall test. How would the implicit primed wo rd stem completion task compare to a task in which an individual were to complete stem s without studying a priming word list first? Both tasks are considered implicit by Roediger & Amirs standards. Would these tasks be measuring different things? This is one of the most fundamental questions facing the uses of implicit methods. Until it is unde rstood how much information gathered from one measure is shared by other methods and how much is unique, we cannot accurately interpret our results. Another factor that complicates implic it measurement is that the relationship between implicit and explicit, or purposeful, memory func tion is not well understood. The question of how distinct or similar implicit and explicit memory processes are
Implicit Expectancies 7 complicates the understanding of implicit expectancy resear ch. Concordance between implicit and explicit tasks has been reported, wh ich has been interpre ted as validation for such tasks by comparison with an explic it gold standard, working under the assumption that explicit and implicit memo ry function work in harmony (Weirs, van Woerden, et al., 2002; Jajoda & Earleywine, 2003). It has also been determined that implicit measures have incremental validity; they explain varian ce in behavior that explicit reports do not (Stacy, 1997; Nosek, 2007; Reich, Below, & Go ldman, in preparation). This latter observation implies that though there may be concordance with explicit measures, implicit and explicit types of memory ma y actually have somewhat independent relationships to behavior. These disc repant findings cause disagreement among psychologists over how memory functions. So me researchers argue for a dual processing model of cognitive function in which implicit and explicit memories operate by different neural systems (e.g. Tulving, 1999). That discordance between implicit and explicit measures has been found may be evidence for this theory (Reich, Goldman, & Noll, 2004), although beliefs that theses systems ar e dissociable and thus separate abound. If we are unsure whether explicit and implicit memo ries are the same or different systems, how can we understand how explicit and implicit measures should relate to one another? And further, how can we be sure that our implicit tasks are actually probing the same memory system? It could be that implic it memory functions through multiple processes or systems instead of the monolithic entity that is commonly referenced by the term implicit memory.
Implicit Expectancies 8 Herein lies the conundrum of implicit rese arch. Calls for development of better theory to explain implicit-e xplicit relations are plentiful (Nosek, 2007; Ames et al., 2007; Reich et al., in preparation). Yet implicit research is notoriously difficult to conduct due to difficulties with contamination, practice effects, and construct validity. Because implicit measures have no genuine gold standard against which to be measured, little definitive evidence exists for the mechanisms by which they function. Making attempts to refine the distinction between implicit and explicit alcohol expectancy even more difficult is the fact that the tasks employed to explore implicit expectancy have been varied, and replication has been scant. Wh en experimental research has been conducted more than once with the same implicit tasks for the study of alcohol expectancies, rarely have the same stimuli been used or the same procedures followed, as exemplified by use of different versions of the IAT (discussed below). Thus, the conclusions drawn from this body of literature are frag mentary and often conflicting. Over the past 15 years, multiple implicit measures have been adapted from cognitive psychology for the purpose of meas uring alcohol expectancies within the domain of clinical research. Many of these measures have shown promise in their ability to identify and predict patterns of drinking. However, most of these measures have begun to be used without having been subjected to the rigorous psychome tric tests that are applied to most clinical instruments at deve lopment. A long history of division between experimental and correlational (e.g. observa tional or clinical) research fields has maintained differences in their goals for psychological science (Cronbach, 1957).
Implicit Expectancies 9 Experimental, or cognitive, psychology has traditionally focused on generalization, or explaining behavioral phenomena across individuals. Correlati onal, or clinical, research has focused on individual differences. What experimental psychology has regarded as noise and attempted to minimize, clinical scie ntists have sought to understand. Thus the experimental measures brought to the alcohol expectancy field from cognitive research were developed for a different set of theoretic al goals, and judged by a differing criterion of acceptability. No matter how well established experimental methods may be in the cognitive field, their reliability and validity as measures of individual differences must be established separately within the realm of clinical research. Yet for the implicit measures adapted for alcohol expectancy research, results from empirical testing of each have been reported on few occasion, and on only one in many cases (e.g. Reich et al., 2005; Kramer & Goldman, 2003; McCarthy & Thompsen, 2006). Where multiple tests of one measur e have been conducted, variation in methodology (e.g. unipolar valence IAT versus se parate positive and negative versions), stimuli (such as different sets of words repr esenting similar construc ts; e.g. alcohol words or alcohol-related expectancy words), and samp le characteristics has been so great that true replication has been scant. In ad dition, the outcome indices of these implicit measures are greatly variable, and the use of different dependent variables across studies has rendered their results incomparable (e.g. quantity of drinking, frequency of drinking, alcohol-related problems, or measure of hea vy drinking; Reich et al., in preparation). Moreover, differences or deficits in theoretica l bases for research with implicit tasks have
Implicit Expectancies 10 resulted in a disjointed theo retical discussion throughout the expectancy lite rature. Thus, although interest in the predictiv e power of implicit measures of alcohol expectancies has been great, the field has yet to refine implic it measures into useful diagnostic tools. Implicit measures of semantic a ssociation in al cohol research Priming Several researchers have used priming pr ocedures to assess memory of alcohol concepts. One study examined the effects of two primes on later consumption of a beer placebo by female participants. One prime consis ted of sitcom scenes that took place in either a bar or an inn. The second prime was a Stroop task in which e ither expectancy or neutral words were embedded (Roehrich & Goldman, 1995). It was found that exposure to both types of alcohol-related cues (the bar sitcom scen e and the alcohol expectancy Stroop) resulted in greater consumption of th e placebo than exposure to neutral cues did. In another study of priming effects on cons umption (Carter, McNair, Corbin, & Black, 1998), participants cued with negative e xpectancy words consumed less non-alcoholic beer in a taste-test than participants in a neutral prime condition, while participants primed with positive expectancy words dra nk more. Another study (Stein, Goldman, & Del Boca, 2000) compared a verbal priming a pproach using expectancy or neutral words to mood induction using positive or neutral mu sic. In the positive expectancy word prime condition, participants drank significantl y more than in the neutral word condition, and within the positive word condition heavie r drinkers drank significantly more than lighter drinkers. In a priming study that did not include an ad lib drinking session (Reich,
Implicit Expectancies 11 Noll, & Goldman, 2005), participants were presented with one of two word lists containing food and alcohol expectancy words. The lists were iden tical except for the first word, which was either milk or beer. It was found that the beer prime resulted in a greater proporti on of recall of expectancy words than grocery words. In addition, there was an expectancy word type /drinker type inter action, with heavier drinkers recalling more positive expectancy words than lighter drinkers. Although free recall of wordlists is widely understood to be an explicit task, this design examined implicit effects of memory, which in this st udy were type of word recalled within each experimental condition (milk versus beer prime) and type of expectancy word recalled. Stroop In a Stroop task, interferen ce of alcohol-related memory primes with participants ability to report the ink color of expect ancy target words was examined (Kramer & Goldman, 2003). Participants were presented with one of eight priming words in black, four of which were neutral beverages, and four of which were alcoholic beverages. There were four categories of target words which we re presented in either blue, green, or red: arousing expectancy words, sedating expectancy words, negative expectancy words, and neutral words. Each priming word was pr esented four times, once preceding a target word from each of the categories. Particip ants were asked to name the color of each target words, and their reaction time was m easured. Lighter dri nkers were found to experience more memory interference with their recall of sedating expectancy words
Implicit Expectancies 12 following an alcohol prime, while heavier dr inkers experienced more interference with arousing expectancy word recall following al cohol beverage primes. These results indicate a strong association in memory between the alcohol primes and sedating expectancy words for light drinkers relative to heavy drinkers, and a stronger association in memory between alcohol and arousing expe ctancies for heavier drinkers relative to light drinkers. False Memory Implicit tasks have also been used to assess recall of alcohol-related expectancy words. The false memory paradigm (Deese, 1959; Roediger & McDermott, 1995) has also been used to examine activation of alcohol-related words (Reich et al., 2004). Heavy drinking participants in a bar contex t falsely remembered having studied positive alcohol expectancy words after studying an expectancy word list in a bar setting than light or moderate drinkers. Additionally, this effect was enhanced when participants were in a bar setting as o pposed to a neutral setting. Automatic activation of alcohol-related wo rds has also been examined using Free Associates (Reich & Goldman, 2005). It was found that the probability of using positive expectancy words (e.g. happy) in respons e to the statement alcohol makes me _________ increased across participants with higher reported quantity of alcohol consumed. The reverse pattern was found fo r negative expectancy words (e.g. sick), where probability of use increa sed as drinking level decreased.
Implicit Expectancies 13 Implicit Association Task The Implicit Association Task (IAT; Gr eenwald, McGhee, & Schwartz, 1998) has been used as an implicit measure of alcohol expectancies. The IAT asks participants to categorize stimuli, often based on valence, from two target groups (such as white faces and black faces) presented either together or individually and inte rmixed with stimuli from one of two attribute groups (such as white names and black names. Task performance, as measured by reaction time, ha s been said to reflect implicit connections in participants memory struct ures between a specific target group and a specific attribute group. The IAT effect is indicated by an ove rall difference in reacti on time to categorize the two different target categories across condi tion blocks, and is c onsidered to signify preference for a certain category over the other. Weirs and colleagues (Weirs, van Woer den, et al., 2002) created two alcohol expectancy-based versions of the IAT, one to examine valence and one for the examination of arousal. The target stimu li for both where the same, with categories consisting of either alcohol words or soda wo rds. Attribute categor ies were comprised of either valence-related words (positive or negative) or arousal-related words (active or passive), respectively for the valence and arousal versions. Weirs et al. reported that on the arousal IAT, heavy drinkers showed a stronger association between arousal and alcohol than light drinkers. On the valen ce IAT, it was found that both heavy and light drinkers held negative associati ons with alcohol. It was also reported that the results of the IATs added unique prediction to drinking at one month follow-up.
Implicit Expectancies 14 Jajodia and Earleywine (2003) also admini stered two IAT versions. Because it has been argued that positivity and negativity may not be true opposites, and because the predictive power of negative expectancies seems quite complex, th is study sought to measure positive and negative expectancies separately. The positive expectancy IAT included 12 alcohol and 12 mammal words as target categories, and 12 positive and 12 neutral adjectives as attribute categories. The negative expectancy IAT substituted negative adjectives for the positive ones used in the first IAT. Fi ndings were that the positive IAT had a positive relationship to drinking, while the relationship between performance on the negative IAT and drinking was nonsignificant. A third experiment also used two different versions of the IAT. The attitude IAT consisted of alcohol and soft drink target categories, and li ked and disliked food attribute categories (DeHouwer, Crombez, Koster, & De Beul, 2004). For the arousal IAT, target categories remained the same while the attrib ute categories were active and passive, as in Weirs, van Woerden, et al. (2002) Findings reflected those of Weirs et al., with evidence for more negative connotations with alcohol than soft drinks across drinker type, and stronger arousal connotations w ith alcohol for heavy drinkers than for light drinkers. Palfai and Ostafin (2003) used electricity and alcohol-related words as stimuli for their target categories, and what they term ed behavior categories of approach and avoidance-related words as attribute categorie s. With a sample of hazardous drinkers, strong approach-alcohol associa tions were found to relate to episodes of heavy drinking,
Implicit Expectancies 15 drinking quantity, drinking anticipation urges, and difficulty of consumption control, but not drinking frequency, drinking thoughts, or ba seline urge to drink. Most recently, McCarthy and Thomps en (2006) administered positive and negative IATs modeled after Jajodia & Ea rleywines versions. They found positive relationships between the IAT and an e xplicit measure, the Alcohol Expectancy Questionnaire, as well as with drinking. Additionally, good test-ret est reliability was established for this version of the IAT over a one-month period. Comparison across these results demonstrates that the IAT has thus far not been a reliable tool for the establishment and repli cation of meaningful e xploration of implicit expectancies. First, few of these tasks used the same stimuli. While the alcohol target category was consistent across studies, its comparison target category varied, including soft drinks, food, electricity, and mammals. Th ese pairings of IAT target categories may present different types of choi ces to individuals. For exam ple, the choice between soft drinks and alcohol is one that individuals may make in dail y life (Weirs, van Woerden, et al., 2002), and thus regularly assign valence or arousal-related meaning to each of those beverage categories. One may not be us ed to making choices between mammals and electricity-related concepts. Even though those categories may have valence or arousalrelated meaning to individuals, these m eanings (and the reaction times that may differentially reflect those meanings) may not be comparable to the meanings held in memory for alcohol. Palfai a nd Ostafin (2003) hold that the selection of a target category such as electricity, for which light and h eavy drinkers are like ly to have similar
Implicit Expectancies 16 associations, is preferable to the selection of a category that acts as an alternative to alcohol such as soft drinks. Because alcohol has no clear opposite, they argue, individual differences in contrast category associations may obscure measured associations to the alcohol category. Different attribute categories have also been chosen for alcohol IATs. Valence, indicating negative or positive associations, and arousal, indicating arousing or sedating associations, have been the most commonl y used. However, findings between IATs using similar attribute categories have led to differing conclusions. Using positive and negative attribute categories, Weirs, van Woerden, et al. (2002) found heavy and light drinkers alike to hold negative associations w ith alcohol as compared to soft drinks, as did DeHouwer et al. (2003). Jajoda & Earleywines (2003) positive IAT showed a significant association with h eavier drinking, but no significa nt relationship between the negative IAT and drinking. In contrast, McCarthy & Thompsen (2006) found significant relationships between both positive and negative IATs and drinking. Motivation as assessed by the IAT has also been evaluated inconsistently, with either active and passive or approach and avoidance attribute categories.1 While heavy drinkers show a stronger asso ciation between arousal and al cohol than light er drinkers (Weirs, Stacy, et al., 2002; DeHouwer et al 2003), approach-alcohol associations were 1 It should also be noted that these categories and the stimuli therein are different from those that have been identified as underlying arousing or sedating associatio ns with alcohol (Rather & Goldman, 1994; Rather at al., 1996; Goldman & Darkes, 2004). Wh ile there are certainly no restricti ons on the verbal stimuli used in alcohol expectancy research, we must take care to ensure that those we sel ect carry the meaning for participants that we expect they do. Additionally, utilization of stimuli that have an established relationship to expectancy or drinker type will serve to in crease consistency and agreement between studies.
Implicit Expectancies 17 found to positively relate to indicators of hazardous drinking. Though these results are consistent, we cannot conclude that these resu lts have been replicated, as testing stimuli and procedures varied, as did sample char acteristics, and the hypothesized processes underlying the results. Thus, future research must be conducted to advance our understanding of implicit measures and the memory systems they reflect. For construct validity to be established, however, our experimental methods must unde rgo more rigorous testing (Smith, 2005). Indeed, there have been recent calls to establis h the reliability of implicit measures and to determine whether discrepant findings between them reflect error, or whether such discrepancies indeed reflect di fferent constructs or proces ses (Waters & Sayette, 2006). It is to this end that the present study strives. The Current Study Use of the many implicit measures documented in the extant literature may greatly enrich our understanding of alcohol expectancy operation, but the scientific significance of these measures has yet to be adequately established. The relationships between explicit measures are commonly examine d, in order to estab lish the criterion and construct validity of newly emerging measures For example, scale 3 of the Alcohol Expectancy Questionnaire (the social pleasure scale, which has been shown to be most predictive of drinking) has been shown to significantly corr elate to corresponding subscales of the Alcohol Expectancy Multi axial Questionnaire: .52 with the aroused subscale, .51 with the positive/aroused subscale, and .61 with the positive subscale
Implicit Expectancies 18 (Darkes, Greenbaum, & Goldman, 1996). Li kewise, implicit measures are often compared to explicit measures. For example, examination of the relationship of Free Associates to AEQ subscales has revealed th at the valence and arous al of generated FA words correlated significantly with AEQ subscal es from .18 to .46, with most correlations falling above .30 (Reich, Brandon, Morean, & Go ldman, 2005). It has been shown that explicit and implicit measures of expectancies differentially predict drinking (Stacy, 1997), indicating that low correlations betw een implicit and explicit measures (than between two explicit measures) may reflect so me qualitative differences in memory. Some implicit researchers discuss this uniqueness as evidence for a monolithic implicit memory store, without examining whether imp licit and explicit memory serve parallel or divergent functions or, more crucially, whether that which implicit tasks are purported to probe is actually homogenous. To understand how alcohol concepts are stored in memory, we need to understand how our measures repr esent memory. To do this, we must better understand how our implicit m easures function over time (e.g. their reliability) and in relation to one another (e.g. their construct validity). Only once we understand how our measures function and whet her they perform in accord with our theories about memory can we interpret them meaningfully. Thus, th e goal of this study is to examine the concordance between two diffe rent implicit measures of expectancy. In addition, the within-session test-re test reliability of each of th ese measures will be tested. Two implicit tasks were selected for th is project: Free Associ ates (FA; Reich & Goldman, 2005) and a Primed recall (PR; Re ich, Noll, & Goldman, 2005). The wordlist
Implicit Expectancies 19 consists of 30 words, 15 of which are grocer y-related words, and 15 of which are alcoholrelated words. The first word of this list, beer, is considered to be the prime. The type of alcohol-related words from this list recalled by participants has be en shown to vary in accord with drinking level. These two tasks were selected for three primary reasons. First, both have been shown to be effective in differentiating between drinker type, with h eavier drinkers either reporting more expectancy-related first asso ciates or remembering more expectancyrelated words from a grocery li st. In addition, both tasks have an established capacity to distinguish between drinker types by exp ectancy types, with heavier drinkers demonstrating more positive and arousing expe ctancies and lighter drinkers endorsing more negative and sedating expectancies. Seco nd, both tasks can be scored in a way that lends them to direct comparison (see deta ils below). Third, both tasks can be administered to groups of participants. A dditionally, because both of these tasks were originally designed and tested in the cont ext of alcohol expect ancy research, method adherence can be maximized. Because individuals are asked to comp lete straightforward self-referential statements (e.g. alcohol makes me _______), a free associate measure may at face seem explicit. Criteria that have been esta blished to consider a measure implicit clearly place Free Associates into this category, however (DeHouwer, 2006): a) free associate tasks do not direct individuals to retrieve information re garding past events (Fazio & Olson, 2003; Roediger & Amir, 2005), b) indivi duals completing such tasks are unaware
Implicit Expectancies 20 of the attitude or cognition of interest, wh ich in the case of the present study is the valence and level of arousal associated with specific be liefs about alcohol, and c) researchers across many cognitive domains have overtly classified free associate measurement as implicit (Fazio & Ols on, 2003; Roediger, 2003; DeHouwer, 2006; Nelson, 2000). Because inadvertent priming may be a strong influence on recall of alcohol and alcohol-expectancy-re lated information, steps will be take n to mask the nature of this study. Additional questions about other comm on activities will be added to the Free Associates task to reduce potential priming effects. Participants will be informed that they are participating in res earch assessing the processing of written information. These masking design elements will enable us to thoroughly examine the differential relationships of real world beha vior (i.e. drinking) and self-r eported expectancies to each of these tasks. Thus, this experiment will advance our understanding of the meaning of these tasks: if each task show s strong relationships to drinki ng and self-report but a weak relation to the other, we can conclude that we have evidence for alcohol-related memory processes that are more complex than a bina ry implicit/explicit m odel. Conversely, if each task seems to mirror the results of the other as well as drinking and self-report measures, we can present these findings as evidence for a dual implicit/explicit storage system of alcohol-related memory.
Implicit Expectancies 21 Method Participants A total of 218 participants (46 male ; 21.1%, 172 female; 78.9%) were recruited from ExperimenTrack, an electr onic participant pool, at the University of South Florida. The mean age of participants was 20.65 (SD = 4.32; range = 18-45). Participants were randomly assigned to two different experiment al groups in which the ordering of task administration was reversed (for FRF, n = 115, 26 (22.6%) male; for RFR, n = 103, 20 (19.4%) male). Course credit or extra credit for psychology courses was offered. Caucasians comprised 67.4% of the sa mple, African-Americans 15.1%, 1.8% were Asian, .9% Pacific Islander, and .5% Native Amer ican. 20.6% of the sample identified as being of Hispanic/Latino origin and having membership in anothe r racial group (10.5%) or as being of Hispanic/Latino origin with no other racial identifica tion (10.1%). 3.7% of the sample classified themselves as bei ng of other racial or ethnic descent. Means and standard deviations for dri nking indices can be found in Table 1. Three individuals were eliminated from an alyses because they reported drinking more than three standard deviations above the m ean consumed by drinkers in a normal week (57.23 drinks per week). These responses i ndicated that these i ndividuals did not fit within the parameters of a normal population of young social drinkers An additional two were eliminated because they reported ha ving reached excessive BACs (1.90 and 15.45).
Implicit Expectancies 22 These responses indicated that these two indi viduals either may not have been able to give accurate self report or did not respond to our questions truthfully. We felt that inclusion of these five cases would have co mpromised the integrity of our data and the normalcy of our sample. Thus for all reported analyses, n = 213. Measures Experimental Measures Free Associates (FA ; Reich & Goldman, 2005; Nelson, McEvoy, & Dennis, 2000)2. Participants were asked to free associ ate five words or phrases in response to each response stem, Alcohol makes me_______. Consistent with Reich and Goldman (2005), they were instructed as follows: In the blank spaces provided below, please write down the words or short phrases you would use to complete th e phrases Alcohol makes me _______, food makes me_______, exercise ma kes me _______, cooking makes me _______ and shopping makes me_______. If you do not drink alcohol, exercise, cook, or shop, please indicate what you think would happen if you did. Please write your res ponses in order, starting with the top blank and working down toward the bottom or last (fifth) blank. Please write whatever 2 Although this task is technically a sentence completion task, it is the most reliable method established to date for eliciting adjectives (alcohol expectancy words) in response to an alcohol cue. Because other words that have been used in pure alcohol free association ta sks have such large associative sets, the base rate of expectancy-specific responses tend to be low (Stacy, 1997). Nonetheless, we refer to the present task as a free associates task
Implicit Expectancies 23 first comes to mind. Do not think to o long. Respond as quickly as you can, but please write legibly. By nature, the FA task is a qualitativ e task, not a quantitative task. Thus, Free Associates were scored according to type of outcome they connoted. In keeping with the method used by Reic h and Goldman (2005), they were be categorized based on how each corresponds to an empirically validated (Rather & Goldman, 1994; Darkes & Goldman, 2004) two-dimensional re presentation of expectancies. Specifically, this model re flects two distinct continua: positivenegative, and sedating-arousing. Where e xpectancy words fall in two-dimensional space in respect to both of these conti nua has been shown to represent eight independent expectancy t ypes: negative, negative sedating, sedating, positive sedating, positive, positive arousing, and ne gative arousing. A large body of Free Associate responses have been normed pr eviously according to the ratings of valence and arousal (Reich & Goldman, 2005) Previously uncategorized responses were given to two independent undergradua te or post-baccalaureate raters be categorized. Any responses on which the two raters disagreed were given to a panel of 3 raters, whose instructions were to reach consensus on valence and arousal ratings for each response. Continuous Free Associates scores (ranging from .00 to 1.00) were calculated by examining the proportions of words produced. Thus, proportion of
Implicit Expectancies 24 positive words produced comprises a positive score, and proportion of sedating words comprises a sedating score. Positive and sedating dimensions were chosen because they have been shown to most eff ectively differentiate drinker level (Reich, Noll, & Goldman, 2005). Primed recall (PR ; Reich, Noll, & Goldman, 2005). Reich et al. (2005) developed two word lists consisting of pr eviously normed alcohol expectancy and grocery words. These lists were identical except for the first word, which was either milk or beer. It was shown that th e manipulation of the first word primed participants as to which type of word from the list to remember, so that those in the milk condition remembered more grocery wo rds, and those in the beer condition remembered more alcohol expectancy words. Additionally, the number of alcohol expectancy words remembered by those in the beer condition covaried with expectancy level so that heavy drinking participants re membered more expectancy words in this condition. With this evidence for the ability of the beer-headed word list alone to distinguish drinker type, here we will exclude the milk-headed word list. Following the established method (Reich, Noll, & Goldma n, 2005), participants were presented with 30-word lists, with 15 being grocery words and 15 being expectancy words. Also consistent with Reich et al. (2005), participan ts were instructed to remember as many words as possible prior to stimulus presenta tion, and words were presented individually on a screen in the front of the room at the rate of 3 seconds per stimulus 1-second interstimulus interval. Once the list was presen ted, participants were given 3 minutes to
Implicit Expectancies 25 record all remembered words. There were six differently ordered lists presented to prevent order effects. As noted above, while this type of cued recall task is widely recognized as an explicit task in and of itsel f, the outcome of interest, namely incidental encoding of expectancy words, is an implicit va riable. Thus, the PR measures of interest here will throughout be referred to as implic it measures, irrespective of how the task per se is classified. In order to create a continuous score for th e Primed recall task comparable to that created for the Free Associates task, the fi rst five expectancy words recalled were examined. A continuous PR score for both positive and sedating dimensions was be created. Because the wordlist was originally designed using 15 expectancy words, 5 of which were sedating, 5 of which were pos itive, and 5 of which were overlapping (neutral), this scoring met hod utilized the full range of alcohol-expectancy words embedded in the PR task (Reich, Noll, & Goldman, 2005). Additional scoring methods were utilized in order to replicate findings relating to this task as closely as possibl e. A proportion of total exp ectancy words recalled to total list words recalled was calculated, and ra w number of type of expectancy words (positive, sedating) was calculated. These scor es were correlated with explicit measures and drinking indices in order to establish ex pected parameters of the task and to duly replicate it.
Implicit Expectancies 26 Additional assessments Alcohol Expectancy Questionnaire (AEQ ; Brown, Christiansen, & Goldman, 1987). The AEQ asks participants to either agre e or disagree with a series of statements about the effects of alcohol. The subscales of the 68-item AEQ have coefficient alphas ranging from .72 to .92. It has been shown to account for 57% of variance in concurrent drinking, and 50% of variance in drinki ng over one year (Goldman & Darkes, 2004). Alcohol Expectancy Multiaxi al Assessment (AEMax; Goldman & Darkes, 2004). The 24-item AEMax assesses the strength of e xplicit alcohol-related expectancies along a continuum of valence (positive-negative) as well as along a continuum of arousal (aroused-sedated), thus allowing for the ma pping of expectancies in three-dimensional space (Rather, at al., 1992). Participants will be asked to rate the phrase alcohol makes one_______ for twenty-four items on a 7 point Likert scale ranging from 0=never to 6=always. Additional items beginning ci garettes make one _______ and exercise makes one_______ will be added to mask the nature of the questionnaire. Daily Drinking Questionnaire ( DDQ : Collins, Parks, & Marlatt, 1985). Participants were asked to i ndicate how many drinks they typically consumed each day of the week (Monday through Sunday) for the pr evious 3 months, and over what period of time these drinks were typically consumed. From this information, frequency of drinking (0-7) was calculated, as was typical quantity co nsumed (total weekly quantity/frequency). In addition, participants were asked to repor t the number of drinks they drank on their heaviest drinking day within the past 30 days, and the period of time over which
Implicit Expectancies 27 consumption took place. The DDQ also asked participants for their weight so that average BAC and 30-day peak BAC could be calculated (gender information was collected on the demographics questionnaire). Demographics The demographics questionn aire assessed age, gender, racial/ethnic background. Religious affiliati on and religious activity for the preceding 6 months were also assessed. Procedure All subjects were administered two implic it tasks in a group classroom setting at one time point. Participants were randomly assigned to one of two conditions: FA-PRFA or PR-FA-PR. Subjects in the FA-PRFA condition were administered the Free Associates task followed by the Primed recall task, and were again asked to complete the Free Associates task before administration of questionnaires. Likewi se, participants in the PR-FA-PR condition received the Prim ed recall first and third, and the Free Associates task second. This design offers several advantages for d ealing with potential sources of error in this study. First, it allows th e convergent and discriminant validities of th e two tasks to be assessed within subjects, while control ling for order effects. Second, it allows the within-session test-retest reliab ility of each task to be asse ssed. Third, this design is constructed to minimize contamination be tween tasks. Distractor tasks were administered between experimental measures and careful selection of stimuli for the experimental measures were intended to mask the nature of the study and to de-
Implicit Expectancies 28 emphasize the measures of interest, decreasing their salience and lik elihood of carry-over effects between tasks. Distractor tasks cons isted of a series of multiplication problems, and counting nouns within written paragraphs describing the construction of an outdoor gazebo and the function of computer progr amming syntax. The function of these paragraphs was to occupy part icipant attention in a verbal task to minimize withinsession memory of preceding experimental tasks, and to reduce the salience of experimental questions. To further mask the nature of this st udy, the alcohol Free As sociates task was embedded among four other Free Associates tasks (cooking makes me _______, exercise makes me _______, food makes me_______, and shopping makes me_______), which served to draw focus away from the alcohol -related task, and which all related in some way to the Primed recall wordlist, elimina ting singular priming by the alcohol-related words across tasks. Upon completion of the final implicit task, subjects will be asked to complete all additional explicit measures. The expectancy questionnaires were be administered in counter-balanced order, and demographics and the DDQ were administered last.
Implicit Expectancies 29 Results Coding /Scoring Because responses on the two implicit tasks us ed in this study consisted of verbal responses that were qualitativ e in nature, responses on the FA and PR tasks were first coded and composite scores were created. For the FA task, all responses were scored for valence (positive, negative, or neutral) and arousal (arousing, sedating, or neutral in accord with the Alcohol Expectancy Free A ssociates norms established by Reich and Goldman (2005). Responses that were not or iginally part of the norm set were coded according to the same scheme separately by two independent judges (undergraduate and post-baccalaureate research a ssistants). Any associate for which judges disagreed on valence or arousal was submitted to a panel of three additional judges. The panel was instructed to reach consensus for each score. For each participant, positive, arousing, positive arousing, negative, and sedating compos ites were calculated. For the positive composite, the total number of positive respons es generated per task administration was summed and divided by the total number of FA responses calculat ed, resulting in a proportion of positive responses for that ad ministration. The same procedure was followed for responses from the other categories. For participants th at completed the FA task twice, composites were calculated for both time points. For the PR task, a similar scoring method was employed. The proportion of expectancy words to total number of words r ecalled was calculated. Positive, arousing,
Implicit Expectancies 30 positive arousing, negative, and sedating propo rtions were calculated by summing the number of each type of word recalled and dividing this number by the total number of words recalled. By adding recall for all word s as the denominator of these proportions, for memory for specific types of expectancy words could be examined while controlling for overall memory performance. Thus, five scales ranging from 0 to 1.0 were calculated for each task. These represent each end of valence and arousal continua and a positive arousing composite. A sixth scale representing the pr oportion of expectancy words recalled was calculated for the PR task. Sample Characteristics Before performance on the FA and PR ta sks could be explored, group differences and sample characteristics were examined. This process was necessary to ensure that the conditions did not differ on any demographic or experi mental characteristics. Additionally, we wanted to determine whet her basic sample parameters had been established. These included distributions of drinking and relationships between explicit expectancy and drinking indice s similar to those typically re ported for college students. The confirmation of these parameters was nece ssary to support the va lidity of additional analyses. Results yielded significant relationships between drinking indi ces (see Table 1) and between drinking and the AEQ (see Ta ble 2) and the AEMax (see Table 3). Additionally, the distribution of drinking levels among participants in this sample was as
Implicit Expectancies 31 Table 1. Correlations Between Drinking Indices ________________________________________________________________________ QUAN FREQ DPW TBAC MBAC Drinking Quantity ALL 1.0 Women 1.0 Men 1.0 Frequency ALL .540** 1.0 Women .609** 1.0 Men .395** 1.0 Drinks per week ALL .883** .729** 1.0 Women .874** .778** 1.0 Men .904** .622** 1.0 T-BAC ALL .869** .402** .726** 1.0 Women .897** .483** .755** 1.0 Men .935** .280 .846** 1.0 Max-BAC ALL .704** .480** .684** .775** 1.0 Women .712** .558** .705** .780** 1.0 Men .718** .342* .738** .756** 1.0 *p < .05; ** p < .01; QUAN = typical quantity of standard alc oholic drinks consumed pe r occasion; FREQ = frequency of drinking (per week); DPW = drinks consumed per week; TBAC = BAC reached during typical drinking occasion; MBAC = past 3 month maximum BAC reached
Implicit Expectancies 32 Table 2. Correlations Between Drinking Indices and AEQ Subscales ______________________________________________________________________________ QUAN FREQ DPW TBAC MBAC AEQ Global positive ALL .442** .381** .434** .342** .336** Women .376** .363** .354** .290** .296** Men .598** .420** .614** .588** .487** Sexual enhancement ALL .388** .310** .364** .295** .328** Women .333** .298** .303** .254** .283** Men .522** .316* .492** .517** .513** Social/Physical Pleasure ALL .480** .540** .431** .385** .388** Women .548** .557** .481** .439** .464** Men .301* .482** .290 .233 .136 Social Assertion ALL .459** .465** .413** .340** .361** Women .456** .439** .383** .330** .364** Men .488** .569** .524** .417** .362* Tension Reduction ALL .447** .457** .456** .369** .406** Women .454** .493** .449** .362** .403** Men .453** .367** .515** .458** .442** Aggression/Arousal ALL .384** .328** .360** .332** .307** Women .359** .281** .300** .309** .293** Men .449** .439** .502** .472** .377** ________________________________________________________________________________________________ *p < .05; ** p < .01 QUAN = typical quantity of standard al coholic drinks consumed per occasion; FREQ = frequency of drinking (per week); DPW = drinks consumed per week; TBAC = BAC reached during typical drinking occasion; MBAC = past 3 month maximum BAC reached
Implicit Expectancies 33 Table 3. Correlations Between Drinking Indices and AEMax Subscales ______________________________________________________________________________ QUAN FREQ DPW TBAC MBAC AEMax Sedating ALL -.298** -.199** -.281** -.202** -.215** Women -.246** -.200** -.238** -.190* -.185* Men -.391** -.135 -.322* -.352* -.356* Negative ALL -.242** -.211** -.208** -.179* -.246** Women -.281** -.281** -.265** -.206** -.280** Men -.138 .007 -.051 -.085 -.121 Positive/Arousing ALL .337** .393** .344** .295** .258** Women .372** .396** .384** .333** .345** Men .249 .369* .240 .218 -.011 Horny ALL .049 .120 .080 .074 .022 Women .094 .125 .137 .116 .095 Men -.093 .077 -.100 -.066 -.244 Egotistical ALL -.208** -.077 -.145* -.189** -.190** Women -.183* -.126 -.141 -.153 -.178* Men -.315* -.001 -.220 -.312* -.218 Sick ALL -.315** -.209** -.269** -.253** -.250** Women -.308* -.237** -.262** -.277** -.259** Men -.319* -.086 -.242 -.280 -.276 Woozy ALL -.231** -.171* -.224** -.121 -.103 Women -.171* -.137 -.166* -.105 -.059 Men -.353* -.170 -.293 -.300* -.305* Social ALL .344** .376** .333** .309** .304** Women .355* .355** .337** .329** .372** Men .335* .485** .374* .230 .085 Attractive ALL .378** .403** .373** .297** .268** Women .392** .412** .392** .315** .325** Men .348* .328* .292 .357* .116 Sleepy ALL -.187** -.109 -.206 -.116 -.173* Women -.097 -.098 -.135 -.055 -.112 Men -.378* -.114 -.340* -.364* -.378* Dangerous ALL -.221** -.277** -.218* -.134 -.242** Women -.308** -.354** -.316** -.211** -.311** Men .046 .011 .106 .131 -.011 ________________________________________________________________________ *p < .05; ** p < .01 QUAN = typical quantity of standard alcoholic drinks consumed pe r occasion; FREQ = frequency of drinking (per week); DPW = drinks consumed per week; TBAC = BAC reached during typical drinking occasion; MBAC = past 3 month maximum BAC reached
Implicit Expectancies 34 expected, and similar to that found for othe r college samples (see Figure 1; (Del Boca, Darkes, Greenbaum, & Goldman, 2004). Next, analyses were conducted to ensure that there were no differences between conditions on key demographic or experimental variables. T-tests were used to examine possible differences between the FRF and RFR conditions. There were no differences in demographic variables, drinking, or responses on explicit expectancy measures between the two conditions. T-tests were also used to look for possible differences on implicit task performance. Because each condition received one of the two tasks twice and the other only once, between-condition comparisons could only be made on the first administration of a task. In other words, performance co mparisons for the PR were between the first implicit task administration (t he first of two PR task administrations) for the RFR condition, and the second implicit task admi nistration for the FRF condition (the only time this condition completed the PR task), si nce this group completed the FA task before they were presented with the PR task for the first time. The reverse was true for FA task comparisons (see figure 2). No significant differences were found between the groups on FA task performance. However, differences be tween the groups on wordlist recall task performance were found. The RFR group recalled significantly more positive expectancy words from the PR wordlist (M = 4.70; SD = 4.50) than the FRF group (M = 4.51; SD = 1.91), and the RFR group recalled significantly more expectancy words overall at the
Implicit Expectancies 35 Figure 1. Number of Drinks per Week 10.00 20.00 30.00 40.00 Drinks per week 10% 20% 30% 40% 50%Percent
Implicit Expectancies 36 Figure 2. Diagram of Between-Condition Comparisons FRF Condition RFR Condition Note: Unidirectional arrows i ndicated order of task administration; bidi rectional arrows repr esent between-condition comparisons. FA Task PR Task PR Task FA Task FA Task FA Task PR Task
Implicit Expectancies 37 same time point (for RFR: M = 6.06; SD = 1.81, for FRF: M = 5.50; SD = 2.30). These differences appeared to be a result of the fact that the RFR group recalled significantly more words from the PR wordlist overall (M = 14.25; SD = 2.86) th an the FRF group (M = 12.70; SD = 4.37). Indeed, when overall li st recall was taken into account, these differences in recall of positive words and expectancy words disappeared. Thus, when memory performance was controlled for, type of word recalled did not vary significantly by group. Still, the significant difference in memory performance between the groups was unexpected. A likely explanation is that those in the FRF condition that had already completed the FA task and one round of dist ractor tasks were fatigued, and consequently recalled fewer words overall. It is also possible that participants in this condition experienced proactive interference on the PR task after having genera ted their own set of words for the FA task. Overall memory performance was controlled in all additional analyses as described above. Thus, we found that overall, the present sample resembled other college samples used in alcohol research, and no significan t differences between conditions were found. Gender Analyses Because this sample was comprised of disproportionate numbers of men and women, results could possibly have been in fluenced by this imbalance. Independent samples t-tests were conducted to explore gender differences on key variables, and correlations between implicit tasks, explic it tasks, and drinking were examined.
Implicit Expectancies 38 No significant differences on drinking indices were found, although differences on frequency of drinking approached signifi cance (see Table 4 for drinking descriptives by gender; see figures 3 and 4 fo r distribution of drinks per w eek by gender). Equivalent percentages of men and women in this samp le had not consumed alcohol in the three months preceding participation in this study (64.3% for women, 67.4% for men), had consumed alcohol in the previous three m onths but did not repor t drinking during a typical week (9.9% for women, 8.7% for me n), and reported drinking during a typical week (25.7% for women, 23.9% for men). Next, responses on the AEQ and AEMa x were examined. No significant differences were found on any AEQ subscales. Differences on several AEMax subscales were noted. Women had significantly hi gher scores on the Woozy, Dangerous, and Sedating factors, and lower scores on the At tractive factor. Since no analysis of gender differences on specific factors scores of the AEMax has been reported, it is unknown whether the differences found refl ect idiosyncrasies of our samp le or parameters expected from equivalent groups. Nonetheless, differential responding on th e AEMax by gender does not appear to be tied to differential responding on our experime ntal tasks. No si gnificant differences were found on Free Associate performance. One significant difference was found on PR task performance: women recalled more words from the word list. As described above, this overall memory difference was controlled for and thus di d not affect any additional measures and analyses.
Implicit Expectancies 39 Table 4. Drinking Descriptives by Gender ________________________________________________________________________ Women Men All M (SD) M (SD) M (SD) _______________________________________________________________________ Frequency 1.48 (1.47) 2.14(2.03) 1.62(1.62) Drinkers only 2.00 (1.38) 2.91(1.86) 2.19(1.53) Quantity 2.38 (2.66) 2.76 (3.68) 2.46 (2.90) Drinkers only 3.16 (2.52) 3.85 (3.87) 3.30 (2.85) Drinks/Week 5.89 (8.01) 8.92 (11.87) 6.52 (9.00) Drinkers only 7.78 (8.04) 12.24 (12.41) 8.71 (9.26) T-BAC .042 (.06) .028 (.05) .039 (.06) Drinkers only .058 (.063) .039 (.06) .054 (.06) Max BAC .112 (.14) .094 (.13) .108 (1.35) Drinkers only .152 (1.37) .129 (.14) .147 (.14) N (%) N (%) N (%) Abstinent in a typical week 61 (36.3%) 15 (34.1%) 76 (35.8%) Do not binge in a typical week 53 (31.5%) 16 (36.4%) 69 (32.5%) Binge in a typical week 54 (32.1%) 13 (29.5%) 67 (31.6%) Drinkers defined as individuals th at reported any drinking within th e 3 months prior to participation.
Implicit Expectancies 40 Figure 3. Number of Drinks per Week: Men 10.020.030.040.0 Drinks per week 10% 20% 30% 40% 50% Percent Figure 4. Number of Drinks per Week: Women 10.00 20.0030.0040.00Drinks per week 0% 20% 40% 60% Percent
Implicit Expectancies 41 Thus it is possible that the di sproportionate representation of each gender had little effect on the present results. Differences between the alcohol consumpti on of men and women in this sample were negligible, and perf ormances on the experimental tasks were comparable. Although there were significan t differences on explicit measures, these differences likely had little effect on our experimental ques tions. Correlations between drinking indices and between drinking and th e AEQ and AEMax for both genders can be found in Tables 1-3, respectively. Implicit Task Replic ation and Comparison Before the relationship between the PR a nd FA tasks could be examined, we first needed to determine whether previous tests of these measures had been replicated. This determination was carried out by examining the relationships between both of these tasks, participant-reported drinking behavior, and responding on explicit tasks. Since each experimental condition completed one of the tasks after completing the other, it was important for us only to examine each task only within the condition in which it was presented first. Using this approach, perf ormance on each task could be assessed in the absence of influence (contamination) of the ot her task. Thus we examined only the first administration of the Free Associates task within the FRF c ondition, and the first administration of the PR task only usi ng individuals from the RFR condition. Results showed strong relationships in th e expected directions between composite scores on the FA task and drinking and exp licit measures (positive relationships with drinking for positive or arousing scales and negative relationships with drinking for
Implicit Expectancies 42 negative or sedating scales; see Table 5 fo r details on relationships with drinking variables; AEMax and AEQ see Table 6). Consis tent with past findings (e.g. Reich et al., 2004; Reich et al., 2005b), the positive and positiv e arousing FA composites showed the most robust relationships with drinking indice s. Correlations between the FA composites and the AEQ and AEMax reflected a simila r pattern. These results demonstrate replication of the FA task. Next, relations hips between performance on the PR task and drinking indices were examined (within the RFR condition). All re lationships between the PR composites and drinking indices were nonsignificant (see Tabl e 7). Correlations between the PR and AEMax and AEQ (Table 8) were also calculat ed. Two significant relationships were found. The Aggression/Arous al subscale of the AEQ correlated with the proportion of expectancy wo rds recalled to all words reca lled (r = -.210; p < .05) and with the proportion of positive expectan cy words recalled (r = -.238, p < .05). No relationships were found between any of the PR composites and any of the subscales on the AEMax. The absence of any significant relationships between the PR task and drinking and the AEMax and its limited relations hip with the AEQ indicated a failure to replicate the PR task. Performance on the PR task was explored further using analyses of variance and t-tests. ANOVAs using Bonferr oni corrections compared drinker classes (abstainers, drinkers, and weekly bingers) on the probability of recalling specific words from the PR list. No significant differen ces were found. When the same analysis was done using a median split on drinks per week to create abstainer (M = 0; SD = 0), lighter drinker (M = 2.62; SD = 2.10), and heavier drinker classes (M = 16.30; SD = 9.46), a
Implicit Expectancies 43 Table 5. Correlations Between Free Associate Composites and Drinking Indices ________________________________________________________________________ FAPOS FAA FAPA FAS FAN ________________________________________________________________________ Free Associates Positive 1.0 Arousing .485** 1.0 Positive/ Arousing .632** .828** 1.0 Sedating -.552** -.519** -.459** 1.0 Negative -.615** -.321** -.388** .091 1.0 Drinking Quantity .291** .272** .262** -.228* -.254** Frequency .373** .296** .301** -.261** -.307** Drinks/Week .283** .206* .219* -.181 -.247** T-BAC .157 .207* .230* -.139 -.179 Max-BAC .223* .211* .257** -.178 -.151 ________________________________________________________________________ RFR condition only p < .05; ** p < .01 PRP = proportion of positive words recalled; FAPA = proportion of positive arousing words recalled; FAS = proportion of sedating words recalled; QUAN = typical quantity; FREQ = frequency of drinking (per week); DPW = drinks consumed per week; TBAC = typical BAC; MBAC = past 3 month maximum BAC reached
Implicit Expectancies 44 Table 6. Correlations Between Free Associate Composites and AEQ and AEMax Factors ________________________________________________________________________ FAPOS FAA FAPA FAN FAS ________________________________________________________________________ AEQ Global Positive .439** .249** .319** -.339** -.236* Sexual Enhancement .424** .260** .316** -.240* -.230* Social/Physical Pleasure .484** .318** .348** -.454** -.268** Social Assertion .403** .232* .261** -.418** -.238* Tension Reduction .463** .226* .275** -.399** -.221* Aggression/ Arousal 293** .200* .157 -.275** -.107 AEMax Sedating -.268** -.160 -.205* .149 .115 Negative -.128 -.074 -.071 .184 -.001 Positive Arousing .365** .374** .320** -.373** -.240* Horny .217* .248** .212* -.247** -.130 Egotistical -.035 -.085 -.044 .108 -.034 Sick -.250** -.080 -.136 .251** .106 Woozy -.224* -.236* -.219* .126 .118 Social .260** .285** .207* -.243** -.198* Attractive .365** .332** .319** -.370** -.230* Sleepy -.166 -.079 -.151 -.061 .049 Dangerous -.175 -.049 -.077 .204* .026 ________________________________________________________________________ FRF condition only; *p < .05; ** p < .01 FAP = proportion of positive free associates produced to al l; FAPA = proportion of positive arousing free associates to all; FAS = proportion of sedating free associates produced to all;
Implicit Expectancies 45 Table 7. Correlations Between Primed Recall Composites and Drinking Indices ________________________________________________________________________ EXP PRP PRPA PRA PRS PRN ________________________________________________________________________ Primed recall EXP 1.0 PRP .792** 1.0 PRPA .392** .584** 1.0 PRA .489** .610** .856** 1.0 PRS .474** .045 -.104 -.027 1.0 PRN .219* .022 -.031 -.053 -.084 1.0 Drinking QUAN -.137 -.115 .011 .045 -.017 -.002 FREQ -.150 -.166 .004 .016 .059 .108 DPW -.143 -.124 -.038 -.023 .014 .055 TBAC -.119 -.105 -.041 .004 .003 -.028 MBAC -.128 -.149 -.058 -.020 -.020 -.021 ________________________________________________ _________________________ RFR condition only p < .05; ** p < .01 EXP = proportion of expectancy words recalled; PRP = pr oportion of positive words recalled; PRPA = proportion of positive arousing words recalled; PRA = proportion of arousing words recalled; PRS = proportion of sedating words recalled; PRN = proportion of negative words recalled; QUAN = typical quantity; FREQ = fr equency of drinking (per week); DPW = drinks consumed per week; TBAC = typical BAC; MBAC = past 3 month maximum BAC reached
Implicit Expectancies 46 Table 8. Correlations Between Primed Recall Composites and AEQ and AEMax Subscales ________________________________________________________________________ EXP PRP PRA PRPA PRS PRN ________________________________________________________________________ AEQ Global -.203 -.136 -.010 -.033 -.103 .051 Positive Sexual -.156 -.090 -.011 -.048 -.062 -.004 Enhancement Social/Physical -.185 -.157 -.022 .019 -.121 .122 Pleasure Social Assertion -.168 -.187 -.025 -.048 .006 .006 Tension Reduction -.128 -.143 -.014 .045 -.036 .050 Aggression/ -.210* -.238* -.182 -.141 .028 .046 Arousal AEMax Sedating .135 .071 .052 .003 .019 .044 Negative .092 .056 -.091 -.063 .021 .051 Positive Arousing -.133 -.071 -.006 -.059 -.062 .017 Horny .064 .095 -.008 -.009 -.026 .066 Egotistical .008 -.051 -.120 -.099 -.044 .065 Sick .070 .039 -.075 -.052 -.034 .026 Woozy .119 .055 -.083 -.021 .048 .037 Social -.135 -.146 -.054 -.126 .025 .015 Attractive -.213* -.107 .037 -.010 -.122 -.033 Sleepy .166 .091 .029 .087 .043 .052 Dangerous .146 .135 -.050 -.022 -.003 .030 ________________________________________________________________________ FRF condition only p < .05; ** p < .01 FAP = proportion of positive free associates produced to all; FAPA = proportion of po sitive arousing free associates to all; FA S = proportion of sedating free associat es produced to all; QUAN = typical quantity; FREQ = frequency of drinking (per week); DPW = drinks consumed per week; TBAC = typical BAC; MBAC = past 3 month maximum BAC reached
Implicit Expectancies 47 significant difference was found on the likelihood of recalling the word sociable, whereby heavier drinkers were significantly less likely to recall it (l ikelihood of recall = .43) than abstainers (likeli hood of recall = .65; p = .042). This pattern was opposite of what would have been predicted from alcohol expectancy theory, th at heavier drinkers would be more likely to generate such a response. In a similar vein, t-tests were also used to examine differences on drinking variables between those that recalled each indi vidual expectancy word and those that did not. No differences were found in frequenc y of consumption based on recall of any PR list words. A difference was found between th ose that recalled the word slow and those that did not: those that recalled the word drank significantly fewer drinks per week (M = 4.85; SD = 5.84 for those that recalled, M = 7.18; SD = 9.86 for those that did not; p < .05). Additionally, those that did not reca ll the word beer (n = 5) had a significantly lower t-BAC (M = .009; SD = .01) than those that did recall the wo rd (n = 200, M = .04; SD = .06, p < .01). No other differences were found. In sum, the FA task was strongly rela ted both to drinking and to explicit measures, but the PR task was not. Additiona l post hoc analyses attempting to elucidate performance patterns on the PR ta sk returned conflicting result s, bringing us no closer to an explanation of the relationship between th is task and any other sample parameter. Therefore, any comparison between performan ce on this task and performance on the FA task is uninterpretable. If we cannot demonstr ate that the PR task is significantly related to explicit expectancy measures or to dr inking, we cannot consider it to be a valid
Implicit Expectancies 48 measure of alcohol expectancy, and therefore cannot use it to compare two measures of expectancy. Regardless, corre lations between these tasks we re calculated for exploratory purposes. Because these relationships diffe red by condition, correlations are presented by condition (within subjects) as well as for the entire sample (within and between subjects; see Table 9). The onl y two significant results obtained from this analysis were found within the FRF condition, and were nega tive relationships between the positive arousing FA composite and the arousing and positive arousing PR composites. These results contradicted the hypothesis that co rresponding composites from the implicit tasks would be positively related to one another. Correlations were also calculated betw een the second implicit task conducted and the second administration of the first tas k. Even in the abse nce of meaningful relationships between the two implicit tasks at the first time points (f irst administration of the first task and only admi nistration of the second), it was expected that these correlations would reflect contamination or practice effects. However, no correlations were significant (see Tables 10 and 11). This seemed to indicate that responses on the PR and FA tasks were not related, despite th e fact that exposure to the PR wordlist may have provided responses to the FA task, or that responses on the FA task may have increased the salience of related PR words for participants. In summary, performance on the FA task was highly related to drinking indices and responses on explicit measures, but the PR was not. Additionally, correlation analyses between the two tasks revealed few significant relationships and no meaningful
Implicit Expectancies 49 Table 9. Correlations Between Time 1 Free Associate Composites and Time 1 Primed Recall Composites ________________________________________________________________________ FAP FAA FAPA FAN FAS ________________________________________________________________________ Expectancy Prop ALL .014 -.026 -.074 .047 .004 FRF (.133) (.060) (-.030) (-.014) (-.139) RFR (-.109) (-.128) (-.123) (.117) (.150) PR Positive ALL -.005 -.020 -.064 .080 .030 FRF (.104) (.066) (-.002) (-.020) (-.050) RFR (-.097) (-.119) (-.131) (.195) (.101) PR Arousing ALL .035 -.046 -.130 -.019 .026 FRF (-.062) (-.115) (-.286**) (.009) (.124) RFR (.160) (.039) (.059) (-.057) (-.085) PR Positive Arousing ALL .048 -.044 -.125 -.024 -.013 FRF (.053) (-.053) (-.186) (-.028) (-.053) RFR (.054) (-.032) (-.042) (-.018) (.029) PR Negative ALL -.026 .002 -.018 -.013 .038 FRF (.009) (.021) (.011) (-.002) (.081) RFR (-.080) (-.023) (-.057) (-.022) (.002) PR Sedating ALL .008 -.042 -.019 .016 .038 FRF (.043) (-.022) (-.042) (.014) (-.071) RFR (-.034) (-.066) (.008) (.019) (.154) ________________________________________________________________________ p < .05; ** p < .01 FAP = proportion of positive free associates produced to all; FAA = proportion of arousing fre e associates to all; FAPA = proportion of positive arousing free associates to all; FAS = proportion of sedating free associates produced to all; FAN = proportion of negative free associat es to all; Expectancy Prop. = proportion of expectancy words recalled to all recalled in PR task
Implicit Expectancies 50 Table 10. Correlations between Time 1 Primed Recall Composites and Time 2 Free Associate Composites ________________________________________________________________________ FAP2 FAA2 FAPA2 FAN2 FAS2 ________________________________________________________________________ PR Expectancy Proportion .056 -.046 -.117 .056 -.130 PR Positive .035 -.082 -.134 .085 -.043 PR Arousing -.042 -.171 -.188 .042 .094 PR Positive Arousing .080 -.150 -.127 .050 -.049 PR Negative .072 -.096 -.038 -.051 .027 PR Sedating .004 .048 -.053 .056 -.126 _________________________________________________________________________ FRF condition only p < .05; ** p < .01 FAP = proportion of positive free associates produced to al l; FAPA = proportion of positive arousing free associates to all; FAS = proportion of sedating free associates produced to all; QUAN = typical quantity; FREQ = frequency of drinking (per week); DPW = drinks consumed per week; TBAC = typical BAC; MBAC = past 3 month maximum BAC reached
Implicit Expectancies 51 Table 11. Correlations Between Time 1 Free Associate Composites and Time 2 Primed Recall Composites ________________________________________________________________________ PREXP PRP2 PRA2 PRPA2 PRN2 PRS2 _______________________________________________________________________ FA Positive -.128 -.147 .044 .044 -.016 -.081 FA Arousing -.145 -.037 .049 .049 .036 -.144 FA Positive Arousing -.191 -.081 .105 .105 -.018 -.153 FA Negative .123 .154 .116 .116 .021 -.031 FA Sedating .128 .058 -.010 -.010 .065 .131 _________________________________________________________________________ RFR condition only p < .05; ** p < .01 FAP = proportion of positive free associates produced to al l; FAPA = proportion of positive arousing free associates to all; FAS = proportion of sedating free associates produced to all; QUAN = typical quantity; FREQ = frequency of drinking (per week); DPW = drinks consumed per week; TBAC = typical BAC; MBAC = past 3 month maximum BAC reached
Implicit Expectancies 52 ones. The absence of any meaningful fi ndings from the PR task was surprising, and precludes the completion of one of the major goals of this study: to directly compare implicit tasks. Despite the failure of the PR task to replicate, additional findings still provide us with valuable information about th e nature of these two tasks, and the nature of implicit tasks in general. Further, the experimental design used here allows us to address several other equally important que stions about the nature of inter-task contamination. Intra-Session Reliability Although we were unable to replicate previ ous findings from the PR task, a within subjects analysis of performance on the two separate administrations of this task can provide us with valuable information about the stability of this task, and likewise for the FA task. To this end, correlational an alyses were conducted between composites calculated at the first administration of each task and the corresponding composites calculated for the second administration to ex amine within-session reliability of the FA and PR (see Tables 12 and 13). Because part icipants in each condition only completed one of the implicit tasks twice, these analys es were carried out within the respective conditions. As expected, valence and arousal proportions for each task at time one were strongly correlated with the same tasks re spective proportions at time two. Exceptions to this were the PR sedating and PR negative measures; correlations for neither set of measures approached significance. This fi nding may reflect a ceili ng effect due to the
Implicit Expectancies 53 Table 12. Correlations between Time 1 and Time 2 Primed Recall Composites ________________________________________________________________________ PREXP2 PRP2 PRA2 PRPA2 PRN2 PRS2 _______________________________________________________________________ PREXP1 .382** .321** .244* .244* -.056 .215* PRP1 .316** .356** .303** .303** -. 141 .089 PRA1 .261** .204* .326** .326 ** .031 .038 PRPA1 .214* .154 .292** .292** .019 .055 PRN1 .194 .109 .030 .030 .100 .210* PRS1 .084 -.004 -.057 -.057 .063 .106 _________________________________________________________________________ RFR condition only p < .05; ** p < .01
Implicit Expectancies 54 Table 13. Correlations between Time 1 and Time 2 Free Associates Composites ________________________________________________________________________ FAP2 FAA2 FAPA2 FAN2 FAS2 _______________________________________________________________________ FA Positive 1 .691** .266** .450** -.508** -.335** FA Arousing 1 .365** .576** .595** -.288** -.369** FA Positive/ Arousing 1 .385** .502** .622** -.287** -.336** FA Negative 1 -.512** -.254** -.333** .688** .154 FA Sedating 1 -.431** -.375** -.430** .197* .599** _________________________________________________________________________ FRF condition only p < .05; ** p < .01
Implicit Expectancies 55 content of the word list (see below for further discussion of list cont ent). These findings indicate that the emotional c ontent of the Free Associates generated and words recalled in the PR task was largely consis tent across administrations. Practice Effect/Cont amination analyses Repeated measures analyses of variance we re used to assess change in each task composite across same-session administrations (see Table 14; all resu lts are based on the Greenhouse-Geisser correction to control for violations of sphericity). As with many analyses reported here, all repeated meas ures analyses were conducted within the respective conditions, as each condition only completed one imp licit task twice). For the free recall task, overall memory increased si gnificantly from the first administration to the second. This increase likely reflects a pr actice effect: exposure to the same list of words twice undoubtedly results in better recall for these word s than only one exposure. Additionally, the proportion of expectancy word s recalled to all word s recalled increased significantly across administrations (F = 22.01; p < .01). It is unlikely that this change occurred due only to overall increase in memo ry. A general memory effect would have resulted in a greater number of all words re called from the list, and the proportion of expectancy words recalled would not have in creased. Instead, this increase may have been driven by contamination; exposure to th e FA task may have increased the salience of alcohol expectancy words in general ove r and above grocery words. The proportions of positive (F = 11.90; p < .01) and sedating words (F = 7.87; p < .01) recalled also increased significantly, while the pr oportions of negative, arousing, or
Implicit Expectancies 56 Table 14. Intra-Session Task Analysis: Changes Across Administrations _____________________________________________________________________ Time 1 Time2 F M (SD) M (SD) _____________________________________________________________________ FA positive .335 (.283) .348 (.283) ns FA arousing .170 (.189) .168 (.200) ns FA positive arousing .118 (.174) .120 (.171) ns FA negative .334 (.289) .344 (.282) ns FA sedating .458 (.257) .457 (.257) ns PR expectancy .423 (.090) .465 (.067) 22.01** PR positive .330 (.091) .362 (.071) 11.91** PR arousing .152 (.064) .150 (.044) ns PR positive arousing .143 (.064) .150 (.044) ns PR negative .14 (.029) .030 (.026) ns PR sedating .102 (.067) .125 (.053) 7.87** FA analyses conducted only within FRF condition PR analyses conducted only within RFR condition ** p < .01
Implicit Expectancies 57 positive arousing words did not. It is probable that this pattern of composite change was a result of the content of the wordlist. On ly one negative word and four arousing words were on the PR list, leaving little room for improvement on these indices of performance, resulting in a ceiling effect. Increased recall of the more plentiful positive (11) and sedating (5) words was likely facilitated by the increase in overal l recall at the second administration. For the Free Associates task, no signifi cant increases in th e proportions of any type of associate generated were observed. This finding suggests that there was minimal contamination between implicit tasks. If the FA task had been affected by the content of the PR task, one might expect to see infl ated positive or sedating composite scores. To directly gauge effects of the PR word list on responses generated in the Free Associates task, an analysis of the frequency of expectancy words from the word list that participants generated as free associates was calculated (see Table 15; only the FRF condition was used for this analysis). This enabled us to assess FA performance both unaffected by the PR task and following the PR task, and allowed for an exploration of contamination effects of the recall task on the free associates. Four words from the wordlist were not generated as free associates at either time point: active, jolly, slow, and verbal. Four words were not generated at th e first time point, but were generated at time two following exposure to the wordlist: confident, foolish, mellow, and noisy. Additionally, large increases (50% or greater) in the generation of se veral associates were found: dizzy, drowsy, fun, horny, sociable, and wild. The only word to have been
Implicit Expectancies 58 Table 15. Frequency and Percentage of Primed Recall Words Generated as Free Associates _______________________________________________________________________ Time 1 Time 2 Increase _______________________________________________________________________ Active 0 (0) 0 (0) Confident 0 (0) 2 (.2%) 2 (-) Dizzy 6 (.6%) 15 (1.5%) 9 (150%) Drowsy 4 (.4%) 10 (1%) 6 (150%) Foolish 0 (0) 1 (.1%) 1 (-) Fun 8 (.8%) 9 (.9%) 1 (12.5%) Happy 34 (3.5%) 35 (3.6%) 1 (2.9%) Horny 6 (.6%) 9 (.9%) 3 (50%) Jolly 0 (0) 0 (0) Mellow 0 (0) 1 (.1%) 1 (-) Noisy 0 (0) 2 (.2%) 2 (-) Slow 0 (0) 0 (0) Sociable 1 (.1%) 5 (.5%) 4 (400%) Verbal 0 (0) 0 (0) Wild 2 (.2%) 5 (.5%) 3 (150%) FRF condition only All results expressed as whole numbers, or frequency of occurrence, and percenta ge of all free associates generated (in parentheses)
Implicit Expectancies 59 generated as an associate at time one that wa s not generated at an increased rate at time two was happy. Of all expectancy words from the wordlist, this word was most frequently generated as an associate at both tim e points. No direct analysis of the effect of the FA task on the PR task was possible. Regression Analyses Next, regression analyses were used to examine the incremental validity of explicit and implicit measures. Because no si gnificant relationships between the PR task and drinking variables were found, only FA i ndices were included in regressions. Again, in order to eliminate the potential infl uence of contamination effects caused by completion of both implicit tasks, these analys es were conducted using only participants from the FRF condition. Separate analyses were conducted examining each drinking measure as a dependent variable and explic it and implicit composites as the independent variables. Analyses were conducted entering al l AEQ subscales as a se t of predictors, all AEMax composites as another set, and FA composites as a thir d separate set. This was done to examine the predictive power of each measure as a whole, and to examine the predictability of each subscale in the ab sence of overlapping predictors from other measures. Subscales for each predictor set we re entered into equations simultaneously. Results for these analyses are presented in Table 16. With frequency of consumption as the depe ndent variable, the AEQ was the best overall predictor (adjusted R2 = .296; F = 8.76; p < .01). The So cial and Physical Pleasure ( = .372; p < .01) was the only subscale that was a significant predictor. Of the
Implicit Expectancies 60 Table 16. Linear Multiple Regression Analyses Predicting Drinking Indices from Separate AEQ, AEMax, and FA Models _____________________________________________________________ Drinking index Predictor B SE _____________________________________________________________ Frequency AEQ Social/Physical .221 .078 .372** Pleasure Full AEQ Model R2 = .296, F = 8.76** AEMax Social .109 .044 .244** AEMax Attractive .117 .044 .286** Full AEMax Model R2 = .237, F = 5.28** Full FA Model R2 = .131, F = 4.34** Quantity Full AEQ Model R2 = .239, F = 6.86** AEMax Social .198 .073 .266** AEMax Attractive .199 .073 .289** Full AEMax Model R2 = .245, F = 5.50** Full FA Model R2 = .080, F = 2.95* Drinks per Week AEQ Tension Reduction .798 .289 .281* Full AEQ Model R2 = .232, F = 6.60** AEMax Social .541 .223 .244* AEMax Attractive .594 .224 .292** Full AEMax Model R2 = .215, F = 4.77** Typical BAC Full AEQ Model R2 = .107, F = 3.17** AEMax Egotistical -.007 .002 -.513** AEMax Sick -.004 .002 -.300* AEMax Woozy .005 .002 .273* AEMax Social .003 .002 .214* Full AEMax Model R2 = .190, F = 4.17** Max BAC Full AEQ Model R2 = .151, F = 4.26** AEMax Woozy .113 .005 .328** AEMax Social .010 .004 .260* Full AEMax Model R2 = .220, F = 4.84** _____________________________________________________________ FRF condition only *p < .05; **p < .01
Implicit Expectancies 61 three predictor sets, the AEMax explaine d the second highest amount of variance (adjusted R2 =.237; F = 5.28; p < .01). The Social ( = .244; p < .05) and Attractive ( = .286; p < .01) factors both signi ficantly predicted frequency of drinking. Lastly, the Free Associates set explained the least variance in frequency (adjusted R2 = .131; F = 4.34; p < .01), and none of the FA composite s were significant predictors. Using typical quantity as a dependent variable, the AEMax model (adjusted R2 = .245; F = 5.50; p < .01) predicted more vari ance than the AEQ model (adjusted R2 = .239; F = 6.86; p < .01) or the FA model (adjusted R2 = .080; F = 2.95; p < .05). The Social ( = .266; p < .01) and Attractive ( = .289; p < .01) factors of the AEMax were significant predictors, but none of the subscales from the AEQ of FA were significant predictors of quantity. Drinks per Week (DPW) were signifi cantly predicted by the AEQ (adjusted R2 = .232; F = 6.60; p < .01) and AEMax (adjusted R2 = .215; F = 4.77; p < .01), but not the FA model. The AEQ Tension Reduction scale ( = .282; p < .05) and AEMax Social ( = .244; p < .05) and Attractive ( = .292; p < .01) Factors were all significant predictors. Using t-BAC as the criterion, again the AEQ (adjusted R2 = .107; F = 3.17; p < .01) and AEMax (adjusted R2 = .190; F = 4.17; p < .01) models were significant, while the FA model was not. While none of the AEQ subscales were significant predictors, the Egotistical ( = -.513; p < .01), Sick ( = -.300; p < .028), Woozy ( = .273; p < .05), and Social ( = .214; p < .05) factors of the AEMax were each significant predictors of tBAC.
Implicit Expectancies 62 Lastly, the AEQ (adjusted R2 = .151; F = 4.26; p < .01) and AEMax (adjusted R2 = .220; F = 4.84; p < .01) models were significan t in predicting past 3-month max-BAC, while again the FA model was not. Only the AEMax Woozy ( = .328; p < .01) and Social ( = .260; p < .01) factors were significa nt predictors; no AEQ subscales were. Next, to determine whether FA and th e explicit measures predicted unique variance in our drinking variab les, implicit and explicit measures were combined into one model (see Table 17). These analyses were conducted only for frequency and quantity of drinking, since these are the only two drinking indices that the FA mode l significantly predicted. Because no specific FA subscales were significant as individual predictors, all five of the FA composites were entered as a block. Subscales from the AEQ and AEMax that had reached significance in their respective regression models were entered into equations. For frequency, the AEQ Social a nd Physical Pleasure and AEMax Social and Attractive factors were entered simultaneously with the FA block. Since no specific subscales reached significance as individual predictors for quantity, the AEQ was also entered as a block. The AEMax Social and Attractive factors were entered into this model as well. With the frequency criterion, neith er the AEMax Social factor ( = .338; p < .01) nor the Free Associates added unique prediction over and above the AEQ-SPP ( = .338; p < .01) and the AEMax Attractive factor ( = .328; p < .01). In fact, the adjusted R2 for the model containing only the thr ee explicit subscales (adjusted R2 = .305; p < .01) was
Implicit Expectancies 63 Table 17. Linear Multiple Regression Analyses Predicting Drinking Indices from Implicit and Explicit Blocks ________________________________________________________________ Drinking index Predictor B SE ________________________________________________________________ Frequency AEQ Social and Physical .202 .061 .338** Pleasure AEMax Attractive .092 .040 .226* Explicit Model adj. R2 = .305, F = 17.19** AEQ Social and Physical .173 .066 .289** Pleasure Combined Explicit/Implicit Model adj. R2 = .288, F = 6.56** Quantity Explicit Model adj. R2 = .245, F = 5.49** Combined Explicit/Implicit Model adj. R2 = .216, F = 3.35** ____________________________________________________________ FRF condition only *p < .05; **p < .01
Implicit Expectancies 64 higher than that for the model with th e FA composites included (adjusted R2 = .288; p < .01). A similar pattern was found using typica l quantity as the criterion, where the adjusted R2 for the model including both implicit and explicit measures (adjusted R2 = .216; p < .01) decreased from the R2 fo r the explicit only model (adjusted R2 = .245; p < .01). No explicit subscales we re individuall y predictive. In review, the AEQ, AEMax, and FA were each entered into regression equations as individual predictor models separately for each drinki ng criterion. Results were that the AEQ and AEMax significantly predicted each of the five drinking variables, but that FA only significantly predicted frequency and quantity. Thus in order to determine whether our implicit measure added unique e xplanation of drinking, additional regression analyses were conducted, and included all signi ficant predictors for each criterion from the first set of analyses. Our findings were that scores on implicit measures did not explain unique variance in drinking variab les beyond that predicted by the explicit measures.
Implicit Expectancies 65 Discussion The use of implicit measures in social and clinical research has seen a sharp increase over the past decade. Most of th ese measures have been imported from the experimental cognitive field. Consequently, most implicit measures used as indicators of real-world behavior have not been subjected to rigorous psychometric testing as are most other clinical instruments. This pattern holds especially in the alcohol expectancy field. To address this shortcoming in the expectan cy literature, one goal of the present study was to examine the stability of implicit measures of alcohol expectancy and to examine the degree to which they probe the same underl ying construct. We proposed to assess the reliability of two implicit measures by comp aring our results to those reported in the literature (replication), and by measuring intra-individual in tra-session performance on each. We also hoped to compare performance across implicit tasks in order to determine the degree of concordance. Our design used two conditions which each received both implicit tasks, with each condition completing one task twice. We used multiplication problems and a verbal distractor task (counting nouns in paragraphs) between administrations of implicit tasks to prevent (or at least reduce) processing of alcohol expectancy information between tasks. Not only did this design en able us to address same-session test-retest reliability and concor dance between tasks, but it also allowed us to examine contamination and practice e ffects both between and within tasks.
Implicit Expectancies 66 Earlier findings from the Free Associates sentence completion task were successfully replicated, with responses on this task found to be highly related to selfreported alcohol consumption and to explicit reports of alcohol expectancies. The FA task also demonstrated high within-session test-retest reliabilit y. On the other hand, while most of the indices of PR task perf ormance demonstrated reliability across the experimental session, earlier findings rela ting to the PR free recall task were not successfully replicated. The present research found no signi ficant relationship between performance on this task and drinking, and only two significant relationships between two indices of performance on this task (p roportion of expectan cy words recalled and proportion of positive expectancy words recalled) and one AEQ subscale (Aggression/Arousal). That the Aggression/Arousal subscale of the AEQ was the only subscale to correlate with the PR task was rather surpri sing, since this subscale has been shown to be a relatively weak predictor of drinking (Goldman, Greenbaum, & Darkes, 1997), to have comparatively low internal consistency (G oldman, Brown, Christiansen, & Smith, 1991), and because it showed some of the weakest re lationships with drinking in this sample. Additionally puzzling was the fact that thes e relationships are bot h negative, indicating that recall of expectancy words and positive expectancy words on the PR task was related to low Aggression/Arousal scores. There may be several reasons for the failure of the PR task to replicate. First, it may not be a reliable probe of alcohol-relate d associations in memo ry. As demonstrated
Implicit Expectancies 67 by Reich et al. (2004), simply changing the firs t word of the list from milk to beer significantly affected the type of words th at participants reca lled, whereby more expectancy words were recalled when the first word of the list was beer. Additionally, heavier drinkers remembered more positive expectancy words when the first word on the list was beer. The fact that such significant changes took place as a result of such a slight change in stimuli is a testament to th e sensitivity of automatic memory processes. However, a stimulus change so small may al so lead to performance changes that are unreliable or due either to noise or context specificity. Cognitive responses to contextual change are so nuanced that these patterns ar e tricky or impossible to reliably identify. An other possible explanation is that implicit memory processes themselves developed to be highly respons ive to context, and thus may vary in accord with uncontrolled (or uncontrollable) elements of the environment. Our knowledge about the world is constantly updated by new experi ences and exposure to new contexts. Individuals have no criterion ag ainst which to measure the st ability or correctness of output implicit memory processes as we have awareness to modulate output of declarative information. Thus, continuous revision of the associations we hold in memory based on ever-changing contingencies a nd contextual cues lead to inconsistent responses to the same stimuli. Indeed, ev idence of this is found in responses on free association tasks, despite the strong reliability reported here and elsewhere (Reich et al., 2005, Ames et al., 2007). While individuals te nd to respond to alcohol expectancy Free Associates with words having similar propert ies, specific responses are impossible to
Implicit Expectancies 68 predict. In fact, an individuals responses on a free association task are better predicted by established norms than by that in dividuals own past performance. Although findings from earlier administrations of the FA task were replicated, the fact that earlier findings rela ted to the PR task were not precluded any direct comparison between the tasks. However, we were still ab le to assess intra-session reliability of each task and practice and contamin ation effects both within and between tasks. Both tasks showed good within session test-retest reliabil ity. This finding may have been influenced by practice effects, specifically double exposure to a task within a short time frame (less than one hour). Many participants in the FRF condition generated the same free associates at both time points, a phenome non which most likely would have been less frequent had the time interval between the two FA task administrations been longer. Overall memory improved on the free recall ta sk from the first administration to the second, increasing the likelihood that grocer y and expectancy words alike would be recalled from the list. Interestingly, we did not find as much contamination across tasks as one might expect; we expected that this would ha ve resulted in significant within-condition correlations between the second administration of the repeated task and whichever task was administered only once. In addition, whil e there seemed to be some direct influence of the PR task on the FA task, several exp ectancy words from the PR list were never generated as free associates (active, jolly , slow, and verbal), and in one other case (happy) there was a minimal increase in PR word use as free associates from the
Implicit Expectancies 69 first FA administration to th e second. This may indicate th at distractor tasks were somewhat successful in blocking processing of alcohol expectancy information between tasks. It may also be the case that the PR task actually had little effect on subsequent performance on the FA task; the fact that no significant changes in any of the FA composites were observed from time 1 to ti me 2 lends support to this suggestion. While individuals tend to respond to alc ohol expectancy Free Associates with words having similar properties, specific respons es are impossible to predict. In fact, an individuals responses on a fr ee association task are better predicted by established norms than by that individuals own past performance (Jenkins, in process of confirming date). Therefore, the mere fact that some of the wo rds from the PR word list either increased in frequency as free associates or appeared as associates for the first time at the second time point may be a reflection of associate fluctuat ion and not of contamination per se. Only comparing results from a procedure similar to th is one to another in which the FA task is administered twice with no other interven ing alcohol expectancy task can offer a definitive explanation. Practice effects were observed primarily in the PR task. Memory for all word types improved at the second administration. The positive, positive arousing, and sedating PR composites all increased from the first time point to the second. This was likely due to the content of the wordlist. Of 15 words, 11 were positive and 5 were sedating, while only one was negative and four arousing. No significant changes in any of the FA composites were observed from time 1 to time 2. A lthough the pattern of
Implicit Expectancies 70 composite scores for both tasks both remained consistent across the two administrations, it is notable that despite this consistenc y correlation coefficients between these composites and explicit measures and drinking were far from perfect, and the relationship between the tasks was negligible. The finding that both tasks are sufficientl y stable to produce similar results at administrations about 30 minutes apart may be evidence that instability of one task may not solely account for the lack of notewort hy relationships between the tasks. While same session test-retest reliability may be explained by a contamination effect, it is curious that no such effects were found between the tasks. It is possible that each task may measure a construct reliably, at least 30 mi nutes apart, but that the constructs they tap are not the same. Regression analyses were used to determine whether explicit and implicit measures predicted unique variance in drinki ng indices. Results indicated that the FA task did not contribute unique e xplanation of variance in drinki ng. This is in contrast to multiple studies that have demonstrated that implicit measures do seem to explain unique variance (Ames et al., 2007; Stacy. 1997; Pa lfai & Woods, 2001; Wiers et al., 2002; Jajodia & Earleywine, 2003; McCarthy & Thompsen, 2006; Kramer & Goldman, 2003; Reich et al., 2004), though the a dded explanation had tended to be small (Reich et al., in preparation). Our failure to find unique imp licit predictive power is likely due to the manner in which the FA task was scored. Th e five composites that were created were highly intercorrelated, which may have meant that they all indexed the same underlying
Implicit Expectancies 71 performance tendencies. Future examination of the differences in Free Associate versus explicit measure predictability may include ex amination of latent performance variables to parse out overall response patterns. Much of the present experiment was e xploratory. Although Free Associates have been shown to be reliable, we were unsure as to whether results as sociated with the PR task would replicate, and whether there woul d be reliability for either task from one administration to a second within the same experimental session. While we expected contamination and practice effects, we did not predict a specific pattern. The largest surprise was the failure of the PR to replicat e. The procedure used here was identical to that described by Reich et al (2004), using the same word s, identical timing, and the same instructions verbatim. The only variation was in the lists used to present the stimuli to participants. Both experiments used si x randomized lists, but we created our own for this experiment. It is possible that our lists had some systematic fl aws that suppressed the effects reported by Reich et al. (2004). Whether this is the case or not, our inability to replicate previous results indica tes that either the PR task is an unstable measure, or the phenomenon being measured is unstable. Perh aps both are true. The best way to address this uncertainty is to continue to explore previously established tasks and the conditions under which they do and do not replicate. Limitations While the present work brought us a step closer to understand ing the nature of implicit alcohol expectancy measures, there were several shortcomings. First, our sample
Implicit Expectancies 72 contained a significantly smalle r number of men (n = 46; 21. 1%) than women (n = 172; 78.9%). Our analyses indicated no major di fferences between the sexes on indices of drinking or on implicit task performance. Past research has consistently shown that men tend to drink more than women, with more me n identifying as current drinkers, drinking more frequently and in greater quantities th an female drinkers (York, Welte, & Hirsch, 2003; Substance Abuse and Mental Health Services Administ ration, 2006; National Institute on Alcohol Abuse and Alcoholism, 2002), although this gender gap has been narrowing among college-aged individuals in recent years (Young, Morales, McCabe, Boyd, & DArcy, 2005). The present failure to find significant differences between men and women on drinking variables ma y be due to a lack of power as a result of the smaller number of men in the sample, as is likely the case with the trend toward a significant difference in drinking frequency. However, since our results i ndicate that men and women are quite similar on the remaining indice s of drinking, it is po ssible that there are simply few sex-based differences in alcohol consumption in our sample. Some significant differences were f ound on the AEMax, with women scoring higher on the woozy, dangerous, and sedating fact ors, and lower on the attractive factor. Although no gender differences on these factor s have been published to date, these patterns seem to contrast those reported by Darkes, Greenbaum, & Goldman (1999) that womens alcohol use is best predicted by higher scores on positive and positive arousing factors and lower expectations of illness. Womens lower score on the Attractive factor
Implicit Expectancies 73 is consistent with Darkes & Goldmans finding that attractiveness was a stronger predictor of drinking for men. However, further examination of the parame ters of our sample confirmed that the drinking patterns observed here are similar to those reported by other samples. We found that the abstinence rate of 62.4% and the pe rcentage of students that reported binge drinking during a typical week (31.9%) were both consistent with other reports of college student drinking (Del Boca et al., 2004). These findings reassured us as to the similarity of the present sample to other undergradua te samples used in alcohol expectancy research. It is still possible that because we obtained so many fewer male participants, that this sample deviated in some other undetected way. The only way to completely eliminate the question of whether gender diffe rences influenced our final results would have been to include equal numbers of me n and women to provide sufficient power for separate gender difference analyses. The limited content of the PR word list ma y also have suppressed effects of this task. Significant effects were reported for this task in the past and one of the major goals of this research was to closely replicate pa st work, yet the unequal distribution of types (e.g. valence and arousal propert ies) of words surely increased the likelihood of memory for the more frequently occurring positive words and suppressed any effects for memory for negative words. Additionally, all grocer y words on the PR list were concrete nouns, while the expectancy words were abstract. Again, although Reich and colleagues (2004) found effects on this task in spite of this variation, matching ne utral and expectancy
Implicit Expectancies 74 words for concreteness in future incarnations of this task may help determine whether the findings we report here are a function of inc onsistency of the task or of the implicit construct it attempts to measure. Directions for Future Research We feel that the present work has great implications for future implicit alcohol expectancy research. First, we feel that the lack of significant results on the PR task underscore the necessity of re plicating implicit tasks before attempting to use them as diagnostic tools, or in lieu of well-establis hed explicit measures. Furthermore, we hope that future research addressing the psychometric properties of implicit tasks will elucidate the issue of whether inconsistent findings reported here and elsewhere are in fact a function of unreliable measures or the tr ansience of implicit memory states. In conclusion, the present research un derscores the complexity of implicit research and its interpretation. Although we were not able to answer each of the questions we set out to address, these findings provide us with valuable insight that will hopefully help inform implicit rese arch endeavors in the future.
Implicit Expectancies 75 References Ames, S.L., Grenard, J.L., Thush, C., Sussman, S., Wiers, R.W., & St acy, A.W. (2007). Comparison of indirect assessments of asso ciation as predictors of marijuana use among at-risk adolescents. Experimental and Clinical Psychopharmacology, 15 (2), 204-218. Carter, J.A., McNair, L.D., Corbin, W.R., & Black, D.H. (1998). Effects of Priming Positive and Negative Outcomes on Drinking Responses. Experimental and Clinical Psychopharmacology 6(4), 399-405. Christiansen, B.A., Smith, G.T., Roehling, P. V., & Goldman, M.S. (1989). Using alcohol expectancies to predict adolescen t drinking behavior after one year. Journal of Consulting and Clinical Psychology 57(1), 93-99. Cronbach, L.J. (1957). The Two Disc iplines of Scientific Psychology. American Psychologist 12, 671-684. Darkes, J., Greenbaum, P.E., & Goldman, M.S. (1996). Positive/Arousal and Social Facilitation Expectancies and Concurrent Alcohol Use. Paper presented at the 19th Annual Meeting of the Research Soci ety on Alcoholism, Washington, D.C. Darkes, J., Greenbaum, P.E., & Goldman, M.S. (1999). Gender and alcohol expectancy structure: Factorial invariance a nd predictive validity of the AEI-SF. Alcoholism: Clinical an d Experimental Research 23 (suppl), 34A. DeHouwer, J. (2006). What are implicit measures and why are we using them? In R.W. Weirs & A.W. Stacy (Eds.). Handbook of Implicit Cognition and Addiction. London and New Dehli: Sage Publications Inc. DeHouwer, J., Crombez, G., Koster, E.H.W ., & De Beul, N.. (2004). Implicit alcoholrelated cognitions in a clinical sample of heavy drinkers. Journal of Behavior Therapy and Experimental Psychiatry 35, 275-286. Del Boca, F.K., Darkes, J., Greenbaum, P.E., & Goldman, M.S. (2004). Up close and personal: Temporal variability in the drinking of individua l college students during their first year. Journal of C onsulting and Clinical Psychology, 72(2), 155164.
Implicit Expectancies 76 Dunn, M.E. & Goldman, M.S. (1998). Age and drinking-related di fferences in the memory organization if al cohol expectancies in 3rd-, 6th-, 9th-, and 12th-grade children. Journal of Consulting and Clinical Psychology 66(3), 579-585. Fazio, R.H. & Olson, M.A. (2003). Implic it Measures in Social Cognition: Their Meaning and Use. Annual Review of Psychology 54, 297-327. Goldman, M.S., Reich, R.R., & Darkes, J. (2006). Expectancy as a unifying construct in alcohol-related cognition. In R.W. We irs & A.W. Stacy (Eds.). Handbook of Implicit Cognition and Addiction. London and New Dehli: Sage Publications Inc. Goldman, M.S. & Darkes, J. (2004). Alcohol expectancy multiaxial assessment: a memory network-based approach. Psychological Assessment 1(4), 4-15. Goldman, M.S., Del Boca, F.K., & Darkes, J. (1999). Alcohol Expectancy Theory: The application of cognitive neuroscience. In K.E. Leonard & H.T. Blane (Eds.), Psychological Theories of Drinking and Alcoholism 2nd edition. New York, New York: Guildford Publications, Inc. Goldman, M.S., Greenbaum, P.E., & Darkes, J. (1997). A confirmatory test of hierarchical expectancy structure and pr edictive power: Discriminant validation of the Alcohol Expectancy Questionnaire. Psychological Assessment 9(2), 145-157. Goldman, M.S., Brown, S.A., Christiansen, B. A., & Smith, G. T. (1991). Alcoholism & memory: Broadening the scope of alcohol-expectancy research. Psychological Bulletin 110(1), 137-146. Greenwald, A.G., McGhee, D.E., & Schwartz, J.L.K. (1998). Meas uring individual differences in implicit cognition: the implicit association test. Journal of Personality and Social Psychology 74, 1464-1480. Jajodia, A. & Earleywine, M.. (2003). Measur ing alcohol expectancies with the implicit association task. Journal of Addictive Behaviors 17(2), 126-133. Kramer, D.A. & Goldman, M.S. (2003). Usi ng a modified Stroop task to implicitly discern the cognitive organizati on of alcohol expectancies. Journal of Abnormal Psychology 112(1), 171-175. McCarthy, D.M. & Thompsen, D.M. (2006). Imp licit and explicit measures of alcohol and smoking cognitions. Psychology of Addictive Behaviors, 20 (4), 436-444.
Implicit Expectancies 77 National Institute on Alcohol Abuse and Alcohol ism. (2002). A call to action: Changing the culture of drinking at U.S. colleges. NIH Publication No. 02-5010. Bethesda, MD. Nelson, D.L., McEvoy, C.L., Dennis, S. (2000). Wh at is free association and what does it measure? Memory & Cognition 28, 887-889. Nosek, B. A. (2007). Implicit-Explicit Relations. Current Directions in Psychological Science 16(2), 65-69. Palfai, T.P. & Ostafin, B.D. (2003). Alcohol-related motivational tendencies in hazardous drinkers: Assessing implicit response tendencies using the modified-IAT. Behaviour Research and Therapy 41(10), 1149-1162. Rather, B.C.; Goldman, M.S. (1994). Drinki ng-related differences in the memory organization of alcohol expectancies. Experimental and Clinical Psychopharmacology 2(2), 167-183. Rather, B.C.; Goldman, M.S.; Roehrich, L.; Br annick, M. (1992). Em pirical Modeling of an Alcohol Expectancy Memory Network Using Multidimensional Scaling Journal of Abnormal Psychology 101(1), 174-183. Reich, R.R; Goldman, M.S. (2005) Exploring the alcohol expe ctancy memory network: The utility of Free Associates. Psychology of Addictive Behaviors 19(3), 317325. Reich, R.R., Goldman, M.S., & Noll, J.A. (2004). Using the False Memory Paradigm to Test Two Key Elements of Alcohol Expectancy Theory. Experimental and Clinical Psychopharmacology, 12(2), 102-110. Reich, R.R.; Noll, J.A.; Goldman, M.S. (2005) Cue patterns and al cohol expectancies: how slight differences in stimuli can measurably change cognition. Experimental and Clinical Psychopharmacology 13(1), 65-71. Reich, R.R.; Brandon, K.O.; Morean, M.E.; Gold man, M.S. (2005). A psychometric test of the free associate approach to studying alcohol expectancies. Poster presented at the 28th Annual Scientific Meeting of the Research Society on Alcoholism. Santa Barbara, CA. Roediger, H.L. (1990). Implicit memory : retention without remembering. American Psychologist 45, 1043-1056.
Implicit Expectancies 78 Roediger, H.L. (2003). Reconsidering implicit me mory. In J.S. Bowers & C.J. Marsolek (Eds.). Rethinking Implicit Memory Oxford; New York: Oxford University Press. Roediger, H.L. & Amir, N. (2005). Implicit me mory tasks: retention without conscious recollection. In A. Wenzel & D.C. Rubin (Eds.). Cognitive Methods and the Application to Clinical Research Washington, D.C.: American Psychological Association. Roediger, H.L. & Geraci, L. (2005). Implicit me mory tasks: retention without conscious recollection. In A. Wenzel & D.C. Rubin (Eds.). Cognitive Methods and the Application to Clinical Research Washington, D.C.: American Psychological Association. Roediger, H.L., Buckner, R.L., & McDermott, K.B. (1999). Components of Processing. In J.K. Foster & M. Jelicic (Eds.). Memory: systems, process, or function? Oxford and New York: Oxford University Press. Roediger, H.L.; McDermott, K.B. (1995). Cr eating false memories: remembering words not presented in lists. Journal of Experimental Psyc hology: Learning, Memory, & Cognition 21, 803-814. Roehrich, L., & Goldman, M.S. (1995). Im plicit Priming of Alcohol Expectancy Memory Processes and Subsequent Drinking Behavior. Experimental and Clinical Psychopharmacology 3(4), 402-410. Stein, K.D.; Goldman, M.S.; Del Boca, F.K. (2000). The Influence of Alcohol Expectancy Priming and Mood Mani pulation on Subsequent Alcohol Consumption. Journal of Abnormal Psychology 109(1), 106-115. Substance Abuse and Mental Health Services Administration. (2006). Results from the 2005 National Survey on Drug Use and Health: National Findings (Office of Applied Studies, NSDUH Series H-30, DHHS Publication No. SMA 06-4194). Rockville, MD. Tulving, E. (1999). Study of Memo ry: processes and systems. In J.K. Foster & M. Jelicic (Eds.). Memory: systems, process, or function? Oxford; New York: Oxford University Press. Waters, A.J. & Sayette, M.A. (2006). Implicit cognition and tobacco addiction. In R.W. Weirs & A.W. Stacy (Eds.). Handbook of Implicit Cognition and Addiction.
Implicit Expectancies 79 London and New Dehli: Sage Publications Inc. Weinberger, A. H. (2005). Context effects in alcohol expectancies: Influence of context and specificity of items. Dissertation Ab stracts International: Section B: The Sciences and Engineering, 65(7-B), 3732. Weirs, R.W., van de Luitgaarden, J., van den Wildenberg, E., Smulders F.T.Y. (2005). Challenging implicit and explicit alc ohol-related cognitions in young heavy drinkers. Addiction 100, 806-819. Weirs, R.W., van Woerden, S., Fren, T.Y., de Jong, P.J. (2002). Implicit and explicit alcohol-related cognitions in heavy and light drinkers. Journal of Abnormal Psychology 111(4), 648-658. Wiers, R.W., Stacy, A.W, Ames, S.L., Noll, J.A., Sayette, M.A., Zack, M., Krank, M. (2002). Implicit and explicit alcohol-related cognitions. Alcoholism : Clinical and Experimental Research 26(1), 129-137. Young, A.M., Morales, M., McCabe, S.E., Boyd, C.J., & DArcy, H. (2005). Drinking like a guy: frequent binge drinking among undergraduate women. Substance Use & Misuse 40, 241-267.
Implicit Expectancies 80 Appendices Appendix A. Primed recall Stimuli by Word Type Grocery words Expectancy words 1. Beer 16. Active 2. Apples 17. Confident 3. Beans 18. Dizzy 4. Bread 19. Drowsy 5. Butter 20. Foolish 6. Catsup 21. Fun 7. Cereal 22. Happy 8. Cheese 23. Horny 9. Eggs 24. Jolly 10. Flour 25. Mellow 11. Granola 26. Noisy 12. Jelly 27. Slow 13. Mustard 28. Sociable 14. Pasta 29. Verbal 15. Sugar 30. Wild
Implicit Expectancies 81 Appendix B. Free Associate Cues 1. Alcohol makes me _________ 2. Cooking makes me _________ 3. Exercise makes me _________ 4. Food makes me _________ 5. Shopping makes me _________
Implicit Expectancies 82 Appendix C. Alcohol Expectancy Questionnaire Items 1. Some alcohol has a pleasant, cleansing, tingly taste. 2. Drinking adds a certain warmth to social occasions. 3. When I'm drinking, it is easier to open up and express my feelings. 4. Time passes quickly when I'm drinking. 5. Drinking makes me feel flushed. 6. I feel powerful when I drink, as if I can really influence others to do what I want. 7. Drinking gives me more confidence in myself. 8. Drinking makes me feel good. 9. I feel more creative after I've been drinking. 10. Having a few drinks is a nice way to celebrate special occasions. 11. When I'm drinking I feel freer to be myself and do whatever I want. 12. Drinking makes it easier to concentrate on the good feelings I have at the time. 13. Alcohol allows me to be more assertive. 14. When I feel "high" from drinking, everything seems to feel better. 15. I find that conversing with members of the op posite sex is easier for me after I've had a few drinks. 16. Drinking is pleasurable because it's enjoyable to join in with people who are enjoying themselves. 17. I like the taste of some alcoholic beverages. 18. If I'm feeling restricted in any way, a few drinks make me feel better. 19. Men are friendlier when they drink. 20. After a few drinks, it is easier to pick a fight. 21. If I have a couple of drinks, it is easier to express my feelings. 22. Alcohol makes me need less attent ion from others than I usually do. 23. After a few drinks, I feel more self-reliant than usual. 24. After a few drinks, I don't worry as much about what other people think of me.
Implicit Expectancies 83 25. When drinking, I do not consider myself to tally accountable or responsible for my behavior. 26. Alcohol enables me to have a better time at parties. 27. Drinking makes the future seem brighter. 28. I often feel sexier after I've had a couple of drinks. 29. I drink when I'm feeling mad. 30. Drinking alone or with one other person makes me feel calm and serene. 31. After a few drinks, I feel brave and more capable of fighting. 32. Drinking can make me more satisfied with myself. 33. My feelings of isolation a nd alienation decrease when I drink. 34. Alcohol helps me sleep better. 35. I'm a better lover after a few drinks. 36. Alcohol decreases muscular tension. 37. Alcohol makes me worry less. 38. A few drinks makes it easier to talk to people. 39. After a few drinks I am usually in a better mood. 40. Alcohol seems like magic. 41. Women can have orgasms more easily if they've been drinking. 42. Drinking helps get me out of a depressed mood. 43. After I've had a couple of drinks, I feel I'm more of a caring, sharing person. 44. Alcohol decreases my feelings of guilt about not working. 45. I feel more coordinated after I drink. 46. Alcohol makes me more interesting. 47. A few drinks makes me feel less shy. 48. Alcohol enables me to fall asleep more easily. 49. If I'm feeling afraid, alcohol decreases my fears. 50. Alcohol can act as an anesthetic, that is, it can deaden pain. 51. I enjoy having sex more if I've had some alcohol. 52. I am more romantic when I drink. 53. I feel more masculine/feminine after a few drinks. 54. Alcohol makes me feel better physically.
Implicit Expectancies 84 55. Sometimes when I drink alone or with one ot her person it is easy to feel cozy and romantic. 56. I feel like more of a happy-go-lucky person when I drink. 57. Drinking makes get-togethers more fun. 58. Alcohol makes it easier to forget bad feelings. 59. After a few drinks, I am more sexually responsive. 60. If I'm cold, having a few drinks will give me a sense of warmth. 61. It is easier to act on my feelings after I've had a few drinks. 62. I can discuss or argue a point more forcefully after I've had a drink or two. 63. A drink or two makes the humorous side of me come out. 64. Alcohol makes me more outspoken or opinionated. 65. Drinking increases female aggressiveness. 66. A couple of drinks make me more aroused or physiologically excited. 67. At times, drinking is like permission to forget problems. 68. If I am tense or anxious, having a few drinks makes me feel better.
Implicit Expectancies 85 Appendix D. Alcohol Expectancy Multiaxial Assessment Items 0 1 2 3 4 5 6 Never Very Rarely Occasionally Frequently Very Always Rarely Frequently "DRINKING ALCOHOL MAKES ONE ." 1. Dizzy ________ 13. Attractive ________ 2. Arrogant ________ 14. Ill ________ 3. Horny ________ 15. Sleepy ________ 4. Light-headed ________ 16. Lustful ________ 5. Erotic ________ 17. Social ________ 6. Appealing ________ 18. Cocky ________ 7. Deadly ________ 19. Sick ________ 8. Beautiful ________ 20. Dangerous ________ 9. Sociable ________ 21. Outgoing ________ 10. Egotistical ________ 22. Hazardous ________ 11. Tired ________ 23. Drowsy ________ 12. Woozy ________ 24. Nauseous ________
Implicit Expectancies 86 Appendix E. Demographics a nd Daily Drinking Questionnaire 1. How old are you? ___________________ 2. Gender (please circle): Male Female 3. What is your class standing? (1) Freshman (2) Sophomore (3) Junior (4) Senior (5) Non-matriculating (6) Other (Please spec ify): __________________ 4. Which of the following best describes you? (0) Native American/American Indian (1) Asian (2) Pacific Islander (3) African-American/Black, not of Hispanic origin (4) African-American/Black, and of Hispanic origin (5) Caucasian/White, not of Hispanic origin (6) Caucasian/White, and of Hispanic origin (7) Hispanic/Latino origin (8) Other (please specif y) _________________________ 5. What is your religious preference? _________________________ 6. How many times in the past 6 months have you attended religious services?_____ 7. Below, please write below the number of st andard drinks on average that you had each day of the week for the past 3 months (how many standard drinks did you have on a typical Monday, Tuesday, etc.; see standard drink guide below ). After you have done so, please specify the amount of time in which you typically consume alcohol each day of the week for the past three months (how much time you usually spend drinking on a typical Monday, Tuesday, etc.)
Implicit Expectancies 87 Standard Drink Guide : Monday Tuesday Wednesday Thursday Friday Saturday Sunday Number of Standard Drinks BEER : 12 oz. (1 bottle or can) = 1 drink 40 oz. = 3 drinks 1 pitcher = 6 drinks HARD LIQUOR and MIXED DRINKS: (Vodka, Rum, Whiskey, Bourbon, Scotch) 1 oz. of liquor (1 shot) = 1 drink mixed drink with 1 shot = 1 drink 375 ml. (1 pint) = 8 drinks 750 ml. (fifth or quart) = 17 drinks LIQUEUR : (schnapps, Kaluah, Baileys) 1 oz. (1 shot) = drink 3 oz. (2 shots) = 1 drink MALT LIQUOR: 40 oz. = 4 drinks WINE : 5 oz. (1 glass) = 1 drink 25 oz. / 750 ml. (standard bottle) = 5 drinks Wine cooler = 1 drink
Implicit Expectancies 88 Number of Hours 8. What is your weight in pounds? _________________ 9. Below, please write a number indicating the maximum number of drinks you had on your heaviest drinking occasion during the last six months. After you have done so, please write a number indicating how many hours you spent drinking on your heaviest drinking occasion. Max Drinks Past 6 months Hours Spent Drinking Max drinks
xml version 1.0 encoding UTF-8 standalone no
record xmlns http:www.loc.govMARC21slim xmlns:xsi http:www.w3.org2001XMLSchema-instance xsi:schemaLocation http:www.loc.govstandardsmarcxmlschemaMARC21slim.xsd
leader nam Ka
controlfield tag 001 001935445
007 cr mnu|||uuuuu
008 080423s2007 flu sbm 000 0 eng d
datafield ind1 8 ind2 024
subfield code a E14-SFE0002270
Below, Maureen C.
Examining the distinction and concordance between implicit measures of alcohol expectancies :
b toward agreement on their meaning and use
h [electronic resource] /
by Maureen C. Below.
[Tampa, Fla.] :
University of South Florida,
ABSTRACT: Alcohol expectancies have traditionally been measured with explicit self-report questionnaires, but in recent years implicit measures have also been used to explore the tenets of expectancy theory. The basic psychometric properties of reliability and validity have not been established for most implicit tasks, and the convergent validity of different implicit measures has not been explored. Despite these shortcomings, many researchers continue to treat implicit tasks as reliable and valid assessment tools. To address reliability and validity of implicit measures, 218 undergraduate women and men were recruited from the University of South Florida to examine the psychometric properties of and concordance between two previously established implicit measures, Free Associates(FA) and a Primed Recall (PR) task. The FA task was replicated, demonstrating high concordance between FA responses and explicit measures and drinking. The PR task did not show a drinker-type effect as was previously reported. Though the relationship between the tasks could not be examined, an exploration of practice and contamination effects offers insight into how performance in similar comparison studies may be affected.
Thesis (M.A.)--University of South Florida, 2007.
Includes bibliographical references.
Text (Electronic thesis) in PDF format.
System requirements: World Wide Web browser and PDF reader.
Mode of access: World Wide Web.
Title from PDF of title page.
Document formatted into pages; contains 88 pages.
Advisor: Mark S. Goldman, Ph.D.
t USF Electronic Theses and Dissertations.