Whereas the concept of cognitive bias is logical and is empirically supported in many ways, the obstacles involved in confidently assessing the effect of inferred motivational/affective states on cognitive processes are significant. Ideally it would be most useful to utilize a cognitive task to implicitly assess or diagnose motivational issues on an individual basis. However at present we must deal with uncertainties in almost every direction. Inference of motivational state can not be considered a straightforward measure. In the most concrete of assessments we might measure behaviors that are logically related to a concept motivation or emotion. So for example, direct measures of food involvement would be a reasonably defensible way of inferring food-related motivations. While some motivational states may be relatively amenable for direct behavioral assessments, for many motivationally linked behaviors this is problematic. For example, illicit recreational drug taking behaviors as a measure of drug taking motivation would in many cases be difficult for ethical reasons as well as the necessarily covert nature of such activity. On the other hand, it is not unreasonable to accept that clinically anxious individuals are likely to have more intense anxiety-related motivations than the average non-clinical individual. A significant proportion of past studies of cognitive bias have utilized subjects who had clinically diagnosed conditions or syndromes that made them logically predisposed for certain directions and intensities of motivation/affect ( Williams, Mathews and MacLeod, 1996). For example a number of studies have reported evidence for cognitive bias for food/hunger or body-image related stimuli in eating disordered subjects (Smeets, Roefs, van Furth, and Jansen, 2008). However there are abundant reasons why accurate assessment of motivation and emotional states in nonclinical populations should be pursued, so the problem of measurement accuracy remains large. This problem is no doubt compounded in that many motivations will be relatively unique to individuals or small subsets of the general population. One way that we may begin to get around this potential problem is to initially target motivational states that are likely to be “common.” The use of the term “common” motivation here refers to the self-evident idea that some motivations are relatively strong and consistently present in almost all healthy adult humans. ( possible refs substantiating the commonality of different motivations?).
DIFFICULTIES WITH SELF-REPORT
Utilization of self-report measures to infer motivation and emotional states is ubiquitous. Unfortunately, and despite intense efforts to validate self-report measures (e.g.. Maisto et al., 1990; Martin et al., 1988), the issue remains at best controversial (Hoffman, Gawronski, Gschwendner, Le, and Schmitt, 2005). It can not be denied that there are a host of factors that may compromise the accuracy of self-report information ( O’Sullivan, 2008; Palen, Smith and Caldwell et al, 2008). Subjects may be defensive (Rosenberg, 1969), reluctant or incapable of reporting accurately for varied reasonsparticularly in regard to private motivational information (Greenwald et al., 2002). It is clear that self-knowledge may be limited or inaccurate (Epstein, 1994; Fazio, 1990; Greenwald et al., 2002; Wilson, Lindsley, & Schooler, 2000; Nisbett & Wilson, 1977). It is also known that low motivation, demand characteristics(Orne, 1962), social desirability, and faking (Cronbach, 1990; Viswesvaran &Ones, 1999; Holden, Wood, & Tomashewski, 2001; Holden, 2008) impact self-report measures. Consequently assumptions about intensity and type of motivation inferred from self-report are immediate concerns.
Implicit measures of motivational state attempt to assess state in ways that don’t primarily involve subjects responses to motivation questions. A growing number of techniques have been developed for this purpose. For example the Implicit associations test (Nosek, Banaji, & Greenwald, 2002) typically involvesmeasures of reaction times to conditional responses for positively valenced word stimuli associated with target stimuli (like faces of blacks vs whites) compared to RTs for the target stimuli conditionally associated with negatively valenced word stimuli. The idea is that attitudes may be covertly measured by differences in RT since underlying target stimulus valence will have closer associative nodes to either the positive or negative word stimuli. The use if the IAT has proliferated however it is not without controversy. Minimally, the relationship of behavior to IAT measuresrequires more validation (Arkes and Tetlock, 2004; Rudman, 2004; Fazio and Olson, 2003).
Likewise the modified (emotional) Stroop test is now often used to infer motivation or emotional status based on differences in RT for naming word color associated with relatively neutral words vs target words (Williams, Mathews and MacLeod, 1996). Subjects are typically instructed to ignore the word and respond only to the color of the letter font. Differences in RT across different categories of semantically related word stimuliare taken to reflect the automatic tendency to process word meaning and that words are automatically more salient if they are relevant to the subjects motivational status. The Dot-probe task ( REF), The “dot-probe” task is another measure of attentional bias (Posner, Snyder, and Davidson, 1980) Subjects are required to fixate on a point in the center of a computer screen that is then replaced by a neutral and an “emotional” word, typically 5° above and below the fixation point. Placement of the words was randomly assigned. After a short presentation time (typically 500ms), the word pair disappears and a “dot” probe replaces one of the words. Subjects’ are required to press designated keys in response to the probe location. Their reaction time (RT) is measured as a reflection of cognitive processing time for these stimuli. Theoretically, if subjects were attending to both stimuli equally subjects should respond equally as fast to either position of the probe on average. However, emotionally relevant words have been found to faciliate reaction times. Thus reaction times on the dot-probe task are taken as an index of cognitive bias.Problems: Schmulke, 2005)
the RSVP (ref) the picture-word stroop (REF), and incidental learning tasks have all been utilized by researchers in efforts to measure underlying states in ways that bypass primary dependence on self-report or behavioral observations. In general the influence of motivational state as measured by these implicit techniques is now referred to as cognitive bias (attentional bias, memory bias and report bias are logically incorporated). Again, in clinical samples, the relatedness of cognitive bias with their clinical diagnostic measures are not troubling compared to efforts to correlate these measures in nonclinical samples. Evidence that these covert implicit measures reliably measure motivation or emotional states in nonclinical populations is debated
Rudman (2004),
Greenwald et al. (1998),
REFS on cog bias relatedness to self-report or behavior measures).
POTENTIAL ARTEFACTS IN COGNITIVE BIAS MEASURES
There remain a number of other problems for both RT measures and memory measures of cognitive bias. One common problem exists in the use of words as stimuli.It is quite clear that the majority of studies in the area of cognitive bias have utilized single word stimuli as independent variables. So for example threat words are utilized to assess cognitive bias in anxiety prone individuals. Other studies have utilized images in a similar fashion, and problems associated with image stimuli are in many cases parallel (O'Neill, 2005; Fernández-Rey, 2007; Bellhouse-King, 2007; Forsythe, 2008; Weaver and Stanny, 1984) to the difficulties in use of word stimuli that will be discussed.
Memory for and reaction times to word stimuli ( and probably for image stimuli where applicable (Bradley et al., 1992??) are affected by a number of variables that are independent of specific motivational states. “Imageability” (Paivio et al., 1994; Schwanenflugel et al., 1988), plurality of, and idiosyncratic nature of word meaning or interpretation (perhaps especially in regard to slang words; Braun and Kitziger, 2001);Klepousniotoua and Baum, 2007; Hertel and El-Messidi, 2006;Compeau, Lindsey-mullikin,Grewal, and Petty, 2004;Frisson and Pickering, 2001), changes in word meaning, use and acceptability over time; word length (see Campoy, 2008), part of speech(memory for verbs is not as good as memory fornouns, e.g., Earles & Kersten, 2000; Earles, Kersten,Turner, & McMullen, 1999; Engelkamp, Zimmer, & Mohr,1990; Reynolds & Flagg, 1976), number of syllables (may affect processing time; Clifton and Tash, 1973), general emotionality(Kensinger and Corkin, 2004; Buchanan and Adolphs, 2004; Buchanan, Etzel,Adolphs and Tranel, 2006), the malleable nature of word emotionality (for example the changes in acceptability of use of a word like “bitch”), frequency of word use (Arndt and Reder,2002; Borowskyand Masson, 1999), and the idiosyncratic and changing nature of word usemay all may influence memory and RT measures.
There is ample evidence that sexual motivations are common and potent in college-aged individuals (see Regan & Berscheid, 1999; Beck, Bozman, & Qualtrough, 1991; Gagnon, 1977; Kinsey, Pomeroy, Martin and Gebhard, 1953). Likewise a number of studies have reported cognitive bias for sexual stimuli occurs in non-clinical populations. For example,Geer and Bellard ( 1996) used a speeded lexical decision task and found slower processing of sexual word stimuli as compared to control word stimuli. This effect has been demonstrated in a number of different studies (see Conaglen, 2004) and is referred to as the “sexual content induced delay” (SCID) effect. Similarly, Geer and Melton (1997) presented subjects with sentences that ended in double-entendre words. Subjects were asked to make lexical decisions about these target words. Results showed that RTs were slower when sexual content was salient and supported the hypothesis that sexual words evoke a more complex processing sequence. Bias toward sexual stimuli has also been found in picture categorization tasks involving erotic stimuli (Spiering, 2002; Conaglen, 2006). In Spierings’task (2002), subjects were told either to ignore or focus on a sexual, threatening or neutral prime picture. Subjects were then asked to categorize sexual and neutral target pictures. Experimenters found that subjects were slower to respond in the unprimed condition when making decisions related to erotic stimuli.
It is prudent however to consider the possibility that measures of cognitive bias may be subject to methodological influences. In fact, most studies of cognitive bias for sexual stimuli have utilized lexical stimuli and therefore may be subject to the issues described above. Some studies of cognitive bias for sexual word stimuli have implemented controls across a broader range of these problems. For example Conaglen (2004) assessed bias for sex word stimuli in a rapid lexical decision task that included romantic and household item control words. The study also made efforts to control for word length, frequency of use, and social acceptability. Results from Conaglens study (2004) replicated previous reports that subjects typically have longer RTs to sexual word stimuli (Geer and Bellard, 1996); and in a non-speeded rating task for pictorial stimuli; Conaglen, 2006) and indicate that the sexual content induced delay (SCID) is a reliable phenomena. Although it is tempting and logical to conclude that the SCID reflects cognitive bias related to sexual motivations, there remain a number of ambiguities in such an interpretation. One of the stated goals in Conaglens study (2004) was to replicate the work of Geer and Bellard (1996). This replication utilized most of the sexual word stimuli used in the earlier study, and matched these words with neutral words and “romantic words” along the dimensions of word length, frequency of use ( based on Francis and Kucera, 1982). The study also assessed the potential influence of word familiarity, word emotionality and social acceptability.However, the sexual word stimuli used have a number of potential issues associated with them. Many of the sexual words used had ambiguous meaning so for example “screw,” “balls,” “prick,” and “pussy” are clearly double-entendre and slang terms. Word ambiguity can have effects on perceptual selectivity, paired-associate learning, recognition memory, and language. And concrete words are generally processed faster than ambiguous words in lexical decision tasks (Schwanenflugel, Harnishfeger, & Stowe, 1988;Rodd, Gaskell, & Marslen-Wilson,2002).
The sexual word stimuli used in these studies also had a greater number of syllables than the neutral stimuli which may have contributed to slower responses (Clifton and Tash, 1973).Moreover, the sexual word stimuli were highly overlapping in their primary intended meanings (screw/intercourse, vagina/pussy, testicles/balls) which could produce an unintended priming effect. The same issue applies to the romantic word stimuli (Embrace/hugging/cherish, darling/valentine/beloved, Tender/caring) used in Conaglens’(2004) study. Similar issues are associated with most cognitive studies that utilize sexual word stimuli. For example Geer and Robertson examined the relationship of self-reported sexuality (sexual opinions survey; Fisher et al., 1988) with a sexually oriented IAT. The sexual words used in part of their assessment were:
Low acceptability sex words
Balls, Cock, Crotch, Cum, Cunt, Fuck, Pussy, Screw, Tits, Whore;
Nonsexual: Burner, Cup, Grate, Meat, Orange, Oven, Plug, Salt, Spoon, Tongs
High Social Acceptability
Sexual words: Clitoris, Ejaculate, Intercourse, Masturbate,
Nipples, Orgasm, Penis, Semen, Testicles, Vagina
Nonsexual: Catalog, Envelope, Microwave, Oven, Recipe,
Restaurant, Salt, Sandwich, Spoon, Telephone
Evaluative Words
Positive: Diamond, Happy, Health, Heaven, Honest,
Laughter, Love, Lucky, Peace, Sunrise
Negative: Abuse, Accident, Ashamed, Hatred, Poison,
Poverty, Rotten, Sickness, Ugly, Vomit
Hakan
brownies dictionary alarm canteen food dishes levitate fixation abundance stomach 72= 7.2 letters per word; possible nouns = 8, possible verbs=3; avg syllables = 2.4; multiple meanings? Brownies, alarm, dishes, stomach
Surgery recovery medication hospital injection pulmonary treatment doctor healing benign 79= 7.9; possible nouns = 7, possible verbs =4; avg syllables =2.8; multiple meanings?
Panties foreplay ejaculate orgasm condom porno masturbate slut nipples lust 66= 6.6; possible nouns = 7, possible verbs = 5; avg syllables = 2.1; multiple meanings? Ejaculate, nipples
Celebrate excitement elation friend vacation love dream compliment birthday hugging74= 7.4; possible nouns = 8, possible verbs = 6; avg syllables = 2.2; multiple meanings? Vacation, dream, compliment
ArgumentBetrayalHumiliateEmbarrassCriticizeLonelyMucousCrueltyVomitGossip 7 .3; possible nouns = 6, possible verbs = 6; avg syllables = 2.6
The purpose of the present studies was to determine if cognitive bias for sexual stimuli could be observed in a lexical task where word stimuli were controlled for word length, part of speech, number of syllables, plurality of meaning, imageability, emotional valence, contemporary familiarity, and current frequency of encounter and use. Many studies utilizing word or pictorial stimuli have relied on previously published lists of words that have been rated on various dimensions such as emotionality, frequency of use, and imageability (Rubin and Friendly; Francis and Kucera, 1982 ).
However, logic dictates that the interpretation of word meaning is likely to be affected by numerous influences ( see for example Braun and Kitziger, 2001; Klepousniotoua and Baum, 2007; Hertel and El-Messidi, 2006;Compeau, Lindsey-mullikin,Grewal, and Petty, 2004;Frisson and Pickering, 2001). For example the offensiveness of words may be particularly sensitive to subject variables and context (e.g., Jay & Janschewitz, 2008; Mabry, 1974;Wells, 1989). Similarly, the use of words changes over time, conceivably over very short periods of time (see Jay, 2009).
STUDY 1
Therefore, to begin this study a “word library” was created. in an attempt to control for factors that may influence recall-ability of words (Rubin & Friendly, 1986). A list of 480 emotionally neutral, emotionally negative, emotionally positive, categorically related and sexual words was created first by rational criteria and then validated based on ratings provided by 113 participants. Words were rated on imageability, emotionality and frequency of use. Ratings were collected using the population of immediate interest, college-aged subjects, thus controlling for cultural or generational differences. The reliability of these ratings was assessed within and between subjects. Additionally, the reliability of these word ratings was cross-validated with the results of other word rating studies (Rubin and Friendly, 1986). Words were then matched on these ratings and utilized in the series of studies described here. Further validation for the characteristics of the word stimuli were assessed following procedures used in study 2 (see below), that also required subjects to rate the selected word stimuli for imageability, emotional valence and frequency of use/encounter.
Subjects
One hundred forty-nine subjects were recruited using a sign-up board for introductory psychology students who received credit for participating. After elimination criteria (see results), 113 subjects were included in analysis (37 males and 76 females). The mean age of the subjects was 18.3, with a standard deviation of 0.8.
Materials
Four hundred eighty words were chosen by rational criteria in a collaborative effort by researchers. Words were generated based on a number of criteria. These included efforts to: minimize social offensiveness;to minimize semantic ambiguity; control for general familiarity, control for word length, and efforts to control for affective intensity and direction were determinants of inclusion in our initial word list. Neutral words for example included “camera” and “number.” Negative emotional words included “criticize” and “humiliate.” Positive emotional words included “vacation” and “birthday.” Sexual words included “orgasm” and “sexual.”
Words were presented on individual slides in a Microsoft Power Point presentation through a classroom multimedia projector. Subjects responded with a numbered answer sheet on three rating scales for each word: imageability, emotionality and frequency of use. Definitions for scales were provided verbally and on the answer sheet. Imageability referred to “how well the word can be concretely visualized or represented by a mental image.” Emotionality was defined as “the positive or negative way you think most people might respond to the word if heard in public.” Frequency of use was defined as “how often you use or encounter the word by hearing, reading, or speaking it.” Each rating was on a 7-point scale, with 1 corresponding to low and 7 to high. Also on the answer sheet was a space to provide a synonym for later assessment of semantic interpretation.