Emotion 2006, Vol. 6, No.

3, 392 405

Copyright 2006 by the American Psychological Association 1528-3542/06/$12.00 DOI: 10.1037/1528-3542.6.3.392

Awareness of Subtle Emotional Feelings: A Comparison of Long-Term Meditators and Nonmeditators

Lisbeth Nielsen
National Institute on Aging

Alfred W. Kaszniak
University of Arizona

The authors explored whether meditation training to enhance emotional awareness improves discrimination of subtle emotional feelings hypothesized to guide decision-making. Long-term meditators and nonmeditators were compared on measures of self-reported valence and arousal, skin conductance response (SCR), and facial electromyography (EMG) to masked and nonmasked emotional pictures, and on measures of heartbeat detection and self-reported emotional awareness. Groups responded similarly to nonmasked pictures. In the masked condition, only controls showed discrimination in valence self-reports. However, meditators reported greater emotional clarity than controls, and meditators with higher clarity had reduced arousal and improved valence discrimination in the masked condition. These findings provide qualified support for the somatic marker hypothesis and suggest that meditation may influence how emotionally ambiguous information is processed, regulated, and represented in conscious awareness. Keywords: meditation, emotion, visual masking, subjective experience, somatic marker hypothesis

When selecting between two competing options, a rational tally of pros and cons often fails to identify a definitive choice. Rather, subtle conscious emotional intuitions may guide decisions. The somatic marker hypothesis (Bechara, Damasio, Tranel, & Damasio, 1997; Damasio, 1994) proposes that emotions affect decision-making processes by signaling the goodness or badness of possible future outcomes. This process may be particularly important when the best option is uncertain. When people experience emotions in judgment or decision tasks, they typically attribute their feelings to task context features, unless given good reasons to do otherwise (Clore & Parrott, 1991; Schwarz & Clore, 1983). However, evidence from a gambling task suggests that, even in an impoverished context in which information about costs, benefits, and probability is not explicitly represented, emotionally informative signals of the goodness or badness of particular choices can nonetheless arise in the form of physiologic responses and conscious hunches regarding anticipated outcomes (Bechara et al., 1997). Such signals are thought to arise through preattentive evaluative processing of emotionally relevant

Lisbeth Nielsen, Behavioral and Social Research Program, National Institute on Aging; Alfred W. Kaszniak, Department of Psychology, University of Arizona. We thank John Allen, Ken Forster, Lynn Nadel, and Richard Lane at the University of Arizona, Shinzen Young, and Laura Carstensen and members of the Life span Development Lab at Stanford University for their suggestions and comments on earlier versions of this paper. We would also like to thank research assistants Brian David, Lindsey Littrell, and Katja Harle for their tireless contributions to this project. This research was carried out as part of the dissertation research of LN at the University of Arizona and funded by a University of Arizona Center for Consciousness Studies/Fetzer Institute Small Research Grant to LN and AK. Correspondence concerning this article should be addressed to Lis Nielsen, Behavioral and Social Research Program, National Institute on Aging, National Institutes of Health, 7201 Wisconsin Avenue, Suite 533, Bethesda, MD 20892-9205. E-mail: nielsenli@nia.nih.gov 392

features of the context. These processes could influence behavior with or without the mediation of conscious feelings. On the one hand, an unconscious physiological or evaluative change may exert an implicit emotional bias on behavior (Winkielman, Berridge, & Wilbarger, 2005). Alternatively, such unconscious emotional processes may give rise to conscious feelings, which are then referenced when making a choice. Making volitional choices on the basis of subtle changes in conscious emotional feelings requires some means by which an options potential goodness versus badness is experienced. From a functional or evolutionary perspective, there is good reason to think that both the valence and the arousal of a feeling state provide important information to a behaving organism. Arousal signals the degree of self-relevance or uncertainty surrounding a choice or behavioral option, whereas valence signals whether a particular option is beneficial or harmful to ones interests. Thus, the dimensions of valence and arousal provide a framework for examining both the phenomenal structure and functional properties of emotions related to decision making. In addition, individual differences in the ability to discriminate among subtle emotional signals along valence and arousal dimensions may have important consequences for adaptive behavior. Enhanced emotional awareness may confer a greater ability to access, accurately identify, regulate, and act on information conveyed in ones feelings. The present study explores individual differences in emotional awareness at the very limits of consciousness, by probing awareness of the most subtle emotional feelings. The specific aims were as follows: first, to discover whether there is a difference in very subtle phenomenal qualities of good (approach it) and bad (avoid it) emotional intuitions elicited by masked pictures; and second, to determine whether there are individual differences in the ability to consciously discriminate among those feelings, and whether such differences are related to (a) emotional awareness cultivated through long-term meditation practice, (b) visceral perception ability as assessed in a heartbeat detection task, and/or (c) self-reported

AWARENESS OF SUBTLE EMOTIONAL FEELINGS

393 Background

emotional awareness as assessed by standardized questionnaires. Understanding the nature of individual differences in this ability may inform whether such skills can be cultivated through practices such as meditation or whether they represent underlying personality or perceptual differences.

Visual Masking
Despite claims about the advantages of enhanced emotional awareness, whether emotional feelings actually provide information about the valence of options remains an open question. Some evidence that preattentive emotional influences can bias behaviorwith or without the aid of conscious feelings comes from studies using visual masking. In visual masking paradigms, a target stimulus is presented very briefly ( 50 ms), followed and/or preceded by a visual mask (a stimulus appearing in the same location as the target) to block or degrade conscious perception of the target to the degree that participants are unable or unwilling to report explicit awareness. Thus, masking allows for some experimental control over the influences of reflective, conscious cognition on the measurements of interest, so that changes in physiologic responding, judgment, or emotion experience that vary along with features of an emotional target stimulus can be assumed to arise from the automatic activation of emotional processes based on a rough, preattentive analysis of those features. To be sure, as evidence from Maxwell and Davidsons (2004) studies with masked emotional facial expressions has made clear, absence of explicit awareness does not preclude higher order cognitive or perceptual processing (e.g., accurate discrimination in a forcedchoice paradigm), and establishing unconscious perception in the laboratory remains an empirical challenge. However, masking provides a useful technique for testing the question of whether emotional feelings can be evoked, accessed, and influence behavior in the absence of explicit awareness (for further discussion, see Nielsen & Kaszniak, in press).

Individual Differences Why Meditation?

Buddhist meditation has been claimed to enhance emotional awareness and control through extensive practice involving focus on the features and dynamics of emotional responses (Goleman, 2003). A common Buddhist meditation practice involves carefully noting the emergence, transformation, and manifestation in awareness of the discrete components that make up an emotional state. Other practices focus attention for extended periods on bodily phenomena such as the breath. Although there is no empirical evidence to suggest that meditators are better at using emotional information to inform decisions, years of training attention on the physiological and affective properties of emotional consciousness during meditation may result in heightened discrimination of emotional phenomenology in everyday life. Indeed, long-term Buddhist meditators have been described as possessing enhanced emotional awareness and improved emotional regulatory abilities (Brazier, 2001; Dalai Lama & Cutler, 1998). Some behavioral and physiological observations of long-term meditators are consistent with this description (see Goleman, 2003). For example, increased left-sided anterior electroencephalogram (EEG) activation (associated with greater positive affect) was observed after 8 weeks of training in mindfulness meditation (Davidson et al., 2003).

Visceral Perception Ability

Feldman Barrett (1998) has documented individual differences in the degree to which self-reports of emotion experiences emphasize valence or arousal and has linked these to individual differences in visceral perception ability, as measured by a heartbeat detection task (Feldman Barrett, Quigley, Bliss-Moreau, Aronson, 2004). Other studies have indicated that individuals with good heartbeat detection abilities report more intense emotion experiences (Wiens, Mezzacappa, & Katkin, 2000) and more accurately predict shocks contingent on backwardly masked fear-relevant stimuli (Katkin, Wiens, & Ohman, 2001). Consistent with theories of emotion that posit a central role of awareness of visceral states in the generation of emotional experience (Damasio, 1994; James, 1894/1994), visceral perception ability may be associated with enhanced awareness of subtle emotional feelings, and meditation practice may strengthen such abilities.

Preattentive Emotional Processing

Zajonc (1980) was an early advocate of the view that affective stimulus properties are rapidly processed and can exert global effects on preferences without the involvement of reflective, conscious cognition. Supporting research demonstrated that masked affective primes exert bivalent influences on preference judgments of neutral targets (Murphy, Monahan & Zajonc, 1995; Murphy & Zajonc, 1993; Wong & Root, 2003) and that mere exposure to neutral items (even when subliminal) increases subjective liking (Kunst-Wilson & Zajonc, 1980; Monahan, Murphy, & Zajonc, 2000;). Evidence of bivalent physiologic change in response to masked emotional faces has been found using facial electromyography (EMG; Dimberg, Thunberg, & Elmehed, 2000) and functional magnetic resonance imaging (fMRI; Whalen et al., 1998). At the negative end of the valence dimension, Ohman and colleagues have shown that masked, aversively conditioned, fearrelevant stimuli, for which we have a biological preparedness (Seligman, 1970) to develop strong conditioned responses, elicit larger skin conductance responses (SCRs) and are rated as more negative and more arousing than nonconditioned phobic or neutral stimuli (Ohman & Soares, 1994, 1998; Esteves, Dimberg, & Ohman, 1994; see Ohman, Flykt, & Lundqvist, 2000, for a review). This is true even when awareness of the conditioned stimulus unconditioned stimulus (CS-UCS) contingency is blocked using visual masking during acquisition of the fear response. However, participants report experiencing negative emotions and accurately report anticipatory gut feelings of shock expectancy during this acquisition phase (Ohman & Soares, 1998).

Self-Reported Emotional Awareness

Finally, some people simply attend more closely to emotions and possess a greater ability to describe and classify their feelings. There is evidence that individual differences in self-reported emotional awareness predict how individuals make judgments in emotionally charged situations (Gohm, 2003). We hypothesized that meditation may influence self-reported emotional awareness and that individuals reporting heightened emotional awareness on standardized assessments would be more accurate at differentiating subtle feeling states.

394

NIELSEN AND KASZNIAK

More recently, Katkin, Wiens, and Ohman (2001) reported that good heartbeat detectors are better at discerning such gut feelings, and they proposed that superior visceral perception underlies individual differences in conscious awareness of subtle emotional somatic markers. These provocative findings fail to fully address the question of whether feeling states can inform decisions, because the use of only negative and neutral stimuli makes it impossible to ascertain whether participants experienced truly aversive experiences or just heightened arousalwhich they interpreted (in the context of receiving electric shocks) as aversive. Whether appetitive (i.e., positively valenced) stimuli can also elicit both physiologic and subjective emotional responses preattentively remains unknown. Evidence that preattentively elicited emotion gives rise to experiences of being emotionally aroused in a particular (valenced) direction is mixed. Monahan, Murphy, and Zajonc (2000) reported that repeated subliminal exposure to neutral stimuli can increase positive mood, but Winkielman, Zajonc, and Schwarz (1997) found no evidence of awareness of subtle positive or negative feelings elicited by masked stimuli in retrospective posttask reports. However, because subtle shifts in feeling may be very brief and memory for them short-lived, retrospective reports may be misleading. Only one study has reported bivalent experience changes in response to masked emotional stimuli. Robles, Smith, Carver, and Wellens (1987) inserted positive, negative, or neutral frames into three separate, but otherwise identical, film clips and found bivalent effects on anxiety ratings, suggesting that the response to positive primes was qualitatively different from the response to negative or neutral primes. To the best of our knowledge, however, no study to date has assessed whether masked pleasant and unpleasant stimuli can be distinguished from neutral stimuli in terms of both arousal and experienced positive or negative emotion. As argued earlier, if feelings cannot tell us whether to approach or avoid an option, they will be of little use in guiding decisions.

Awareness of Subtle Emotional Feelings

For the masked condition, we predicted that, when attention is focused on changes in emotion experience on a trial-by-trial basis, subtle changes in experienced valence and arousal may be discernable. We also predicted that, even in the absence of aversive conditioning (as used by Ohman & Soares, 1994, 1998), preattentive processing of cues with biological salience would give rise to discrimination in SCR and facial EMG. In accordance with evolutionary arguments for a biological predisposition to respond more strongly to aversive events (Cacioppo & Gardner, 1999; Rozin & Royzman, 2001), we anticipated stronger responses to unpleasant than to pleasant masked stimuli.

Individual Differences in Awareness

We anticipated that long-term meditators would be better at discriminating the subtle feelings evoked by masked emotional stimuli than controls. We also explored whether heartbeat detection accuracy or self-reported heightened awareness of emotion conferred additional discriminatory advantage.

Method Participants
Participants were 111 long-term meditators (9 female, 2 male) in the Buddhist tradition and 17 nonmeditating controls (15 female, 2 male)

The Present Study

In the present study, participants viewed pleasant, neutral, and unpleasant pictures with biologically salient content in both masked and nonmasked conditions. Self-reported valence and arousal and physiological change in electrodermal activity (SCR) and corrugator and zygomatic facial electromyography (EMG) were measured. ECG was also recorded, but will be reported elsewhere. In the nonmasked condition, discrimination on all measures was expected to replicate prior findings, with pleasant and unpleasant pictures eliciting increased self-reported pleasant or unpleasant valence, increased arousal and SCR, and increased corrugator (unpleasant) or zygomatic (pleasant) EMG compared with neutral pictures (see Bradley & Lang, 2000, for a review). Similar response patterns in the masked condition would be considered to be consistent with preattentive emotional processing. In contrast to other studies assessing experiential responses to masked visual primes, the present study examined responses to stimuli at both ends of the valence dimension for which there may be a biological predisposition to respond emotionally, and it incorporated concurrent trial-by-trial self-reports of experienced valence and arousal, along with physiological measures.

Initially 16 meditators were tested, with length of practice ranging from 4 29 years. Findings from videotaped structured interviews of emotional experiences revealed that this was not a homogeneous group as regards their approach to emotion experience. While a description of interview methods is beyond the scope of this paper, the essential finding was that participants with less than 10 years of meditation practice had a particularly hyper-bodily focused approach to describing their emotional experiences that clearly distinguished them from their longer-term counterparts. When comparing the percentage of statements containing bodily feeling reports that constituted descriptions of recent strong positive and negative emotional experiences, short term meditators (M 33%, SD .08) used bodily feeling descriptors more often than long-term meditators (M 16%, SD .10), F(1, 30) 7.524, p .005, 2 .334. Moreover, the percentage of bodily feelings included in emotional reports was negatively correlated with years of meditation practice (r .683, p .005, n 16). In addition, short term meditators had significantly more correct responses on the heartbeat detection task, a measure of visceral awareness, than their long-term counterparts, F(1, 13) 4.846, p .05, 2 .272 (short term M 30.60, SD 5.5; long term M 24.8, SD 4.4), suggesting a greater attention to bodily states. Based on these initial observations, we ran separate 2 Group (Short term, Long term) 3 Stimulus Type (Pleasant, Neutral, Unpleasant) repeated measures ANOVAs comparing short and long term meditators on our primary outcome variables: reports of experienced valence and arousal to masked pictures. These tests further confirmed that short term meditators were unlike their long term counterparts. There were significant differences in experienced arousal between the five short term and 11 long term meditators, Group Stimulus Type, F(2, 28) 4.554. p .05, 2 .245, with short term meditators reporting significantly higher levels of arousal

AWARENESS OF SUBTLE EMOTIONAL FEELINGS matched to the meditators on age (meditators: M 55.27 years, SD 10.43; controls: M 54.29, SD 10.0) and years of education (meditators: M 18.27 years, SD 3.37; controls: M 18.97, SD 2.49). Meditators were recruited from local meditation centers. Control participants were recruited from University of Arizona staff and the surrounding community. All meditators had a regular practice sustained for over 10 years (range 1229 years, M 21.18, SD 4.73), practicing an average of 8.2 times per week (SD 4.2), for an average of 47 min per session (SD 23).2 Length of practice was uncorrelated with age (r .13, ns). Participants had normal or corrected-to-normal vision and no history of psychiatric or neurologic illness. Informed consent was obtained in accordance with University of Arizona human participants protection guidelines. Participants received $30 in compensation.

395

ware (developed at Monash University and at the University of Arizona by K. I. Forster and J. C. Forster), which generated tones linked to the participants current R-wave and governed trial order using a randomization algorithm.

Emotional Stimuli
The experimental stimuli were 16 neutrallow arousal, 16 unpleasant high arousal, and 16 pleasant high arousal color pictures selected from the International Affective Picture System (IAPS; Center for the Study of Emotion & Attention, 1999), divided into two sets, matched for both content and normative ratings of valence and arousal. Emotional pictures were selected from categories likely to elicit biologically prepared responses and included babies, happy and angry faces, nudes, spiders, snakes, and vicious animals. Neutral stimuli were nonarousing, neutralvalenced pictures with biological content (humans, animals, and plants).4 In pilot studies, ratings of pleasantness were found to increase as a function of picture luminance. Consequently, the brightness of some of the original IAPS pictures was adjusted using image-processing software, so that the mean luminance of pleasant, neutral, and unpleasant experimental stimuli was equated. Masks were constructed with image-processing software by overlaying sections of six stimuli (two pleasant, two neutral, two unpleasant); they were brightened, and randomly placed white abstract shapes were inserted into each image, to create masks both complex and bright enough to interrupt visual processing of the targets. Five independent judges rated the masks on a 5-point version of the Self-Assessment Manikin (SAM; Bradley & Lang, 1994). Only masks rated on average as neutral on valence and low on arousal were included in the study. Masks for neutral, pleasant, and unpleasant stimuli were matched for luminance and mean red, blue, and green saturation.

Individual Difference Measures

Self-reported emotional awareness was assessed using the Toronto Alexithymia Scale (TAS; Taylor, Ryan, & Babgy, 1985), the Private Self-Consciousness Scale (PSC; Fenigstein, Scheier, & Buss, 1975), the Mood Awareness Scale (MAS; Swinkels & Giuliano, 1995), the Trait Metamood Scale (TMM; Salovey, Mayer, Goldman, Turvey, & Palfai, 1995), and the Emotional Creativity Scale (ECS; Averill & ThomasKnowles, 1991). On the basis of hierarchical cluster analyses of data collected from a sample of 151 college students, Gohm and Clore (2000) provided evidence that these scales constitute good measures of two trait dimensions of individual differences in emotion experience: attention to emotion, defined as the extent to which individuals monitor their emotions, value their emotions, and maximize their experience of emotion (p. 684), and clarity, or the ability to identify, distinguish, and describe specific emotions (p. 686). Gohm and Clore (2000) recommended the Labeling subscale of the MAS, the Clarity subscale of the TMM, and the Difficulty in Identifying Feelings subscale of the TAS (reverse scored) as good measures of clarity, while the ECS, the PSC, the Externally-Oriented Thinking subscale of the TAS (reverse scored), and the Monitoring subscale of the MAS were recommended as good measures of attention to emotion.3 Individual differences in visceral perception ability were assessed using a heartbeat detection task following the procedure outlined by Wiens, Mezzacappa, and Katkin (2000). The task consisted of 50 trials of 10 tones triggered by the R-wave of 10 consecutive beats of the participants heart. On half of the trials, the tones were presented 200 ms after the R-wave (typically perceived as synchronous); on the other half, there was a 500-ms delay (perceived as delayed; see Wiens & Palmer, 2000). In a practice session, participants judged synchrony between a series of tones and a graphically displayed beating heart (cf. Katkin et al., 2001). In the actual task, participants sat upright in a recliner and were instructed to attend to their heartbeats without feeling for their pulse and to verbally report whether tones on a given trial were in synch or delayed from actual heartbeats. We collected heartbeat data using a Biopac Systems ECG amplifier, with output to a PC running AcqKnowledge software (Biopac Systems, Inc., Santa Barbara, CA; see below for recording parameters). The AcqKnowledge program triggered another PC running DMDX soft-

Experimental Design
In each of two testing sessions, participants viewed one set of stimuli masked and the other nonmasked, with conditions reversed in Session 2 and set order counterbalanced across participants. Six alternating blocks of masked and nonmasked trials were presented in each session. Each stimulus valence was represented at least twice per block, with two pseudorandom stimulus orders counterbalanced across participants. Each session included eight catch trials with no emotional content (masked masks) to assess discrimination accuracy and bias. Stimulus presentation and psychophysiological data acquisition were synchronized and controlled by a PC running DMDX software. Stimuli were presented on a 17-in. (43.18-cm) computer monitor 30 in. (76.20 cm) from the participant. Trials began with a 4-s fixation cross. In the nonmasked condition, the stimulus then appeared for 3000 ms. In the masked

to unpleasant (M 2.34, SD .95) than to pleasant (M 1.99, SD .82) pictures ( p .05), while long term meditators did not differ (pleasant M 1.99, SD .70; unpleasant M 1.95, SD .66). In addition, comparisons of valence reports to masked pictures yielded a ns trend to a Group Valence interaction, F(2, 28) 2.933, p .07, 2 .173. Short term meditators tended to experience unpleasant pictures as more unpleasant than did long term meditators (short term M 3.16, SD .23; long term M 2.93, SD .23, p .08). Because the small number of shorter-term practitioners (n 5) enrolled in the study prohibited meaningful comparisons of both groups with our control participants, these individuals were excluded from analyses reported here.

2 One meditator (with 20 years of practice) reported no set schedule for formal practice, but that she had integrated mindfulness into her daily life. Of those providing information on their weekly practice, one meditator (with 29 years of practice) was unable to say how long her sessions lasted on average. 3 While other scales reviewed by Gohm and Clore (2000) measured factors like intensity of emotion, absorption in emotion, and emotional expressiveness, the clarity and attention factors were considered most likely to index the ability to attend to and distinguish subtle emotional feeling states. 4 The IAPS pictures used in this study were: Set I (pleasant: 2150, 2344, 1811, 4653, 4660, 2340, 2040, 2165; neutral: 5534, 2215, 2516, 5535, 5510, 2383, 7550, 5520; unpleasant: 1300, 2900, 1274, 1050, 1200, 1052, 2800, 3060). Set II (pleasant: 4670, 2550, 2160, 2058, 2050, 4220 (for males)/4520(for females), 2341, 1710; neutral: 2840, 2190, 5531, 9070, 1616, 2495, 9210, 5530; unpleasant: 3168, 1120, 1270, 3170, 1220, 1930, 1301, 1070).

396

NIELSEN AND KASZNIAK presentation and bioelectrical amplifiers for psychophysiological measurement, participants were seated in an upright recliner for all tasks. In Session 1, participants completed a structured interview about emotion experiences and provided measures of heart rate variability as part of a separate study. The electrodes for recording of skin conductance, ECG, and facial EMG were attached. Instructions for the picture-viewing task disclosed stimulus presentation parameters, mentioned prior findings of physiological responses to masked stimuli, and highlighted the researchers interest in whether emotion experiences might also be elicited under masked presentation conditions. Participants then viewed and rated a sample set of stimuli before being left alone to complete the experiment. In Session 2, participants completed the second session of the picture-viewing task following the procedure outlined earlier and then completed all psychometric measures and the heartbeat detection task.

condition, a 45-ms forward mask preceded a 45-ms presentation of the target stimulus, followed by a reappearance of the mask for 2910 ms.5 In both conditions, a blank screen appeared for 4 s after stimulus presentation while participants attended to their feeling states. Then, 5-point SAM valence and arousal scales appeared consecutively on the screen, and participants entered their experience ratings on a keyboard. Finally, on masked trials, participants responded either yes or no to a probe asking whether they thought they had seen some identifiable feature or figure in the briefly presented target stimulus. For the nonmasked trials, participants merely indicated by yes/no response whether they had seen the picture. This ended the trial and triggered the onset of the next trial. Variations in stimulus features and individual differences in perceptual thresholds and response biases are among the factors that can influence perception of masked stimuli (Maxwell & Davidson, 2004). By equating stimuli for luminance and providing an explicit set of criteria for reporting awareness, we hoped to minimize some of this variation. Our purpose was to degrade perception to the extent that participants were unable or unwilling to report seeing a stimulus, and yet discrimination on other measures (physiology, self-reported feelings) may be possible.

Data Analysis
Individual difference measures. Raw scores on each psychometric scale or subscale were standardized with reference to sample means and standard deviations, and subscale z scores were summed to generate composite attention and clarity scores for each participant. Each participant received a score for total number of correct trials on the heartbeat detection task (200-ms delay tones reported as in synch or 500-ms delay reported as delayed). Additionally, following the procedure of Katkin, Wiens, & Ohman (2001), individuals with 30 correct trials (binomial probability of 30 out of 50 correct .05) were classified as good heartbeat detectors. Separate analyses of variance (ANOVAs) assessed differences between meditators and controls in visceral perception ability, attention, clarity, and associated subscales. Correlation analyses explored relationships between age, length of meditation practice, self-reported emotion awareness, and visceral perception ability (number of correct trials on the heartbeat detection task). Self-report and psychophysiological data. Mean valence rating, arousal rating, SCR magnitude, and EMG change were computed for each participant for each stimulus type (pleasant, neutral, unpleasant) in each condition (masked, nonmasked), collapsed over the two viewing sessions. SCR variables were subjected to a square-root transformation to correct for skewed distributions. Repeated measures ANOVAs (2 [group] 3 [stimulus type] 2 [session]) revealed no significant main or interaction effects or trends related to day of testing in either condition; thus, order effects were not explored further. Arousal ratings were reverse coded to provide a more intuitive representation on a scale ranging from 1 (not aroused) to 5 (very aroused). We subjected all measures to 2 3 2 (Group Stimulus Type Condition) repeated measures ANOVAs, with alpha set at .05, using a GreenhouseGeisser correction for violations of the sphericity assumption. Corrected degrees of freedom are reported where appropriate. We conducted separate 2 3 (Group Stimulus Type) repeated measures ANOVAs for each condition, comparing the responses of meditators and controls. Separate stimulus type repeated measures ANOVAs were con-

Emotion Ratings
Participants rated their experienced valence and arousal in response to all stimuli on 5-point versions of the SAM scales. Instructions for ratings emphasized the focus on participants experienced emotion, regardless of the objective content of the stimuli and regardless of how participants thought they ought to feel. Participants were instructed to fully attend to each stimulus and then focus their attention on their own subsequent reaction, noting subtle changes in feelings and allowing these to inform ratings on the valence and arousal scales. Examples of appropriate (e.g., Ugh! Thats awful!) and inappropriate (e.g., That person must be feeling bad, so Ill rate it negative) criteria for making ratings were discussed during the practice session.

Psychophysiological Measures: Recording and Data Reduction

SCR and facial EMG data were recorded using Biopac MP100 Systems bio-amplifiers. Both measures were sampled at 1000 Hz for 4 s pre- and 11 s poststimulus onset. SCR was recorded using a constant voltage (0.5 V) amplifier and 8-mm Ag/AgCl electrodes filled with a pharmaceutically prepared electrolyte (0.05 molar NaCl in Unibase). Electrodes were attached to the medial phalanges of the first and second fingers of the nondominant hand. The skin conductance signal was low-pass filtered (10 Hz) and amplified 20 times. Before scoring, the signal was smoothed offline over a 100-ms moving window and resampled at 20 Hz. For each stimulus, the amplitude of the first SCR with minimum amplitude of 0.03 S and onset in the time window 1 4 s poststimulus onset was recorded. Trials with no measurable SCR were assigned zero amplitude and included in all analyses. EMG signals were recorded from left and right corrugator and zygomatic facial muscle sites, using electrode placements recommended by Fridlund and Cacioppo (1986). Two 4-mm Ag/AgCl electrodes were placed over each muscle site, with a ground placed in the center of the forehead. Impedances at all sites were less than 10 k . EMG signals were amplified 5000 times and notch filtered to remove 60 Hz noise. Signals were bandpass filtered offline over a range of 10 500 Hz, rectified, and smoothed over a 20-ms moving window. Change scores in microvolts from a 1-s prestimulus baseline were computed for all 100-ms epochs during the 3-s stimulus presentation period.

Procedure
Participants came to the laboratory on two separate occasions within a 2-week period. In a small testing room containing a computer for stimulus

5 It should be noted that the 45 ms target presentation is somewhat longer than the 30 33 ms typical in studies with emotional pictorial stimuli in which only backward masks have been employed. However, use of a forward and backward mask, effectively consuming visual processing capacity on both sides of the target, blocks explicit perceptual awareness, as indicated by self-report, at this slightly longer stimulus duration. Research from lexical masked priming studies has demonstrated that at durations shorter than 50 ms, using forward and backward masks, participants are surprised to learn that a target stimulus has been presented, while at durations from 60 67 ms, participants may be aware that something occurred, but are unable to identify the stimulus (Forster, Mohan, & Hector, 2003). The current study employed complex pictorial stimuli for which such systematic examinations of perception at varying exposure durations have not been explored.

AWARENESS OF SUBTLE EMOTIONAL FEELINGS

397

Table 1 Means and Standard Errors for Scores on the Clarity and Attention Dimensions of Self-Reported Emotional Awareness, and for Their Respective Subscales for 17 Controls and 11 Long-Term Meditators
Dimension Clarity Labeling (MAS) Clarity (TMM) Difficulty Identifying Feelings (TAS)* Attention Emotional Creativity Private Self Consciousness Externally-Oriented Thinking (TAS)* Monitoring (MAS) Note. MAS Mood Awareness Scale; TMM * Indicates reverse scoring. Subscale Controls .78 (.72) 22.35 (.85) 40.71 (1.33) 20.76 (1.20) .11 (.82) 103.44 (2.41) 54.65 (2.55) 25.24 (.87) 19.71 (.93) Meditators 2.1 (.99) 26.82 (.88) 46.45 (2.31) 22.0 (1.49) .13 (1.1) 97.40 (2.23) 46.73 (4.17) 25.18 (.81) 21.18 (1.44)

Trait Metamood Scale; TAS

Toronto Alexithymia Scale.

ducted for controls and meditators in each condition to explore withingroup effects on all measures. For meditators only, we included years of meditation practice as a covariate to explore whether length of practice enhanced discrimination. Planned pairwise comparisons between pleasant, neutral, and unpleasant pictures were performed using the DunnSidak correction for multiple comparisons. Clarity and attention scores were entered as covariates in separate subsequent analyses. We explored significant main effects or interactions with these variables using correlation analyses. Additional analyses compared good and nondiscriminating heartbeat detectors on all measures. Power considerations. Post hoc power analyses based on the present sample size for within-participants repeated measures ANOVA, with alpha set at .05, were conducted with GPower software (Buchner, Erdfelder, & Faul, 1997). Power was .52 to detect medium effects and .87 to detect large effects: Critical F(1, 52) 4.03. Less conservative compromise power analysis, recommended by Buchner, Erdfelder and Faul (1997) for dealing with limited samples, setting the / ratio at 1, indicated that our power was .78 to detect medium effects, critical F(1, 52) 1.58, and .91 for large effects, critical F(1, 52) 3.09.

lated with composite scores of Attention (r .05), Clarity (r .13) or their subscales. One-way ANOVAs comparing good (n 5) and nondiscriminating (n 21) heartbeat detectors revealed no significant differences in attention (good discrimination: M 0.76, SD 2.6; nondiscriminating: M 0.26, SD 3.4) or clarity (good discrimination: M 0.64, SD 1.0; nondiscriminating: M 0.49, SD 3.7). Among meditators, years of practice was positively correlated with clarity, although not significantly so (r .35, p .31), but not with attention (r .09, ns) or the heartbeat detection score (r .13, ns).6 The absence of relationships among these different measures of internal state awareness suggests that they tap different abilities. Younger age was associated with higher scores on the Attention scale (r .55, p .005) and its Externally Oriented Thinking (TAS, reverse scored; r .40, p .05) and Monitoring (MAS; r .50, p .01) subscales; age was unrelated to other measures.

Results Individual Differences in Emotional and Visceral Awareness

Self-reported emotional awareness. Meditators reported greater emotional clarity than controls, F(1, 26) 5.74, p .05, 2 .18, scoring significantly higher on the Labeling (MAS), F(1, 26) 12.21, p .005, 2 .32; and Clarity (TMM) subscales, F(1, 26) 5.38, p .05, 2 .17, with no differences in the Difficulty in Identifying Feelings subscale (TAS). Meditators and controls did not differ on overall attention scores or attention subscales. Table 1 shows means and standard errors on these measures. Heartbeat detection. Data for the heartbeat detection task was missing from 2 participants (1 meditator and 1 control) because of irregular R-wave amplitudes. Meditators (M 24.8 correct, SD 4.4) and controls (M 26, SD 5.3) did not differ in performance on the heartbeat detection task. In all, 5 participants1 meditator and 4 controlswere classified as good heartbeat detectors; all were female. Relations among individual difference measures. Attention and Clarity subscale scores were not significantly correlated (r .11, ns), nor was the heartbeat detection score significantly corre-

The Picture Viewing Task

Perception of masked stimuli. In all, 16 of 28 participants reported seeing some of the masked items, with these participants reporting a mean of 1.14 neutral (SD 2.01), 1.64 pleasant (SD 2.79), and 1.36 unpleasant (SD 1.36) items seen. There was no difference in the number of items reported seen on Day 1 (M 2.2, SD 3.8) versus Day 2 (M 1.9, SD 3.5) of testing, t(27) .78, p .1. Meditators and controls did not differ in the average number of masked items reported seen, F(1, 26) .22, p .65 (meditators: M 3.36, SD 3.93; controls: M 4.65, SD 8.5). Only masked items reported as unseen were included in calculation of individual participant means for all dependent variables. Nonmasked items that participants reported not seeing were likewise excluded. Equipment failures or experimenter errors resulted in missing left corrugator data from 1 control participant, and missing right corrugator data from 1 control and 2 meditators. Individual differences. Among meditators, years of practice was not significantly associated with any of the outcome measures

6 When short-term meditators were included, the relation between Clarity and years of meditation practice was stronger (r .485, p .06).

398

NIELSEN AND KASZNIAK

Figure 1. Mean valence (A. nonmasked condition; B. masked condition) and arousal (C. nonmasked condition; D. masked condition) ratings of meditators and controls, by stimulus type. Error bars represent standard errors.

in either task condition.7 Attention to emotion and heartbeat detection ability failed to account for substantial performance differences, with single exceptions in each case, whereas clarity scores related to several outcomes in the task, particularly among meditators. Age was not associated with valence discrimination on any measure. Significant findings are reported below.

Responses to Nonmasked Pictures

The nonmasked condition served as a manipulation check on the ability of the stimuli to evoke emotional responses and to illuminate any baseline group differences. The anticipated significant effects of condition were found on all measures, with no significant differences between meditators and controls in either selfreported emotion or psychophysiological responses to the nonmasked pictures. Self-reports of experienced arousal and valence. Figure 1 (A and C) shows valence and arousal ratings of meditators and controls in the nonmasked condition. A 2 3 (Group Stimulus Type) ANOVA yielded the predicted significant main effect of stimulus type on arousal ratings, F(2, 52) 54.36, p .001, 2 .68, and valence ratings, F(2, 52) 191.22, p .001, 2 .88, but no difference between groups. Pairwise comparisons (all ps

.001) revealed that pleasant pictures (controls: M 2.82, SD 0.73; meditators: M 2.68, SD 1.06) and unpleasant pictures (controls: M 3.30, SD 0.90; meditators: M 3.34, SD 0.94) were rated as more arousing than neutral pictures (controls: M 1.94, SD 0.52; meditators: M 1.90, SD 0.58). On the valence dimension, pleasant pictures were rated as more pleasant (controls: M 1.71, SD 0.44; meditators: M 1.83, SD 0.47) and unpleasant pictures as more unpleasant (controls: M 4.24, SD 0.42; meditators: M 4.04, SD 0.50) than neutral pictures (controls: M 2.76, SD 0.26; meditators: M 2.71, SD 0.40). Including clarity as a covariate in analyses of arousal ratings revealed an ns trend to a Group Clarity interaction, F(1, 24) 3.80, p .06, 2 .14. Among meditators, clarity score was
This was also the case when these analyses included the 5 short-term meditators, with one exception: Years of meditation interacted with Stimulus Type on Arousal ratings in the masked condition, F(2, 28) 3.421, p .05, 2 .20. Correlation analyses revealed that among meditators (n 16) years of meditation practice was negatively correlated with arousal ratings to masked neutral (r .21, ns) and unpleasant (r .28, ns), but not pleasant pictures (r .07).
7

AWARENESS OF SUBTLE EMOTIONAL FEELINGS

399

negatively correlated with arousal ratings to nonmasked pictures (pleasant: r .46, ns; neutral: r .58, p .06; unpleasant: r .15, ns). Among controls, these correlations were all positive, although nonsignificant (pleasant: r .35, ns; neutral: r .24, ns; unpleasant: r .25, ns). Skin conductance. Figure 2A shows square-root transformed SCRs of meditators and controls in the nonmasked condition. A 2 3 (Group Stimulus Type) ANOVA on SCRs yielded a significant quadratic effect of stimulus type, F(1, 26) 11.03, p .005, 2 .298, with larger responses for both pleasant and unpleasant pictures, compared with neutral pictures, but no group main effect or Group Stimulus Type interaction, consistent with predictions. However, separate ANOVAs on the two groups revealed a stronger quadratic effect of stimulus type in the SCRs of controls, F(1, 16) 9.16, p .005, 2 .30, than meditators, F(1, 10) 3.10, p .1, 2 .25. Skin conductance and individual differences. When Clarity scale scores were included as a covariate in the analysis of SCRs, the significant quadratic effect of stimulus type remained, F(1, 24) 7.17, p .05, 2 .23, but there was a significant main effect of clarity, F(1, 24) 4.18, p .05, 2 .15, qualified by a Group Clarity interaction, F(1, 24) 8.93, p .01, 2 .27. Among meditators only, clarity was significantly negatively correlated with SCRs for all stimulus types (pleasant: r .77, p .01; neutral: r .74, p .01; unpleasant: r .68, p .05), indicating a reduction in physiological arousal as clarity increases. For controls, all correlations between Clarity subscale scores and SCRs were nonsignificant, rs .25. In the only case in which visceral perception ability influenced performance on the task, a 2 3 (Group Stimulus Type) ANOVA comparing good and nondiscriminating heartbeat detectors on SCR in the nonmasked condition revealed a main effect of Stimulus Type, F(2, 48) 5.01, p .01, 2 .17, and a significant Group Stimulus Type interaction, F(2, 48) 5.85, p .005, 2 .20. Good heartbeat detectors had larger SCRs to unpleasant (good discriminators: M 0.32, SD 0.22; nondiscriminating: M 0.14, SD 0.13; p .05) and neutral pictures (good discriminators: M 0.21, SD 0.13; nondiscriminating: M 0.09, SD 0.12; p .07) than nondiscriminating heartbeat

detectors, while their responses to pleasant pictures (good discriminators: M 0.19, SD 0.11; nondiscriminating: M 0.17, SD 0.15; ns) did not differ. Facial EMG. Figure 3 (Panels A and C) illustrate left corrugator and zygomatic responses of meditators and controls in the nonmasked condition. Separate 2 3 (Group Stimulus Type) repeated measures ANOVAs on mean EMG change during the 3 s of picture viewing revealed significant main effects of stimulus type at corrugator sites on both the left, F(1.44, 35.93) 9.99, p .005, 2 .29, and right, F(1.08, 26.02) 4.57, p .05, 2 .16. Pairwise comparisons revealed that pleasant pictures elicited less corrugator activation than unpleasant pictures (left: p .01; right: ns trend p .1) and neutral pictures (left: p .01; right: p .05). Corrugator responses to neutral and unpleasant pictures did not significantly differ. Stimulus type also influenced zygomatic activation on both the left, F(1.13, 29.28) 11.26, p .005, 2 .30, and right, F(1.10, 28.47) 10.92, p .005, 2 .30. Pairwise comparisons revealed that pleasant pictures elicited significantly greater zygomatic activation than neutral pictures (left: p .05; right: p .01) and unpleasant pictures (left: p .005; right: p .01). There were no differences between controls and meditators on any of these measures.

Responses to Masked Pictures

Self-reports of experienced arousal and valence. A 2 3 (Group Stimulus Type) ANOVA comparing meditators and controls on arousal ratings to masked pictures yielded a significant quadratic effect of stimulus type, F(1, 26) 6.91, p .05, 2 .21. Pairwise comparisons revealed nonsignificant trends for neutral pictures to be rated as less arousing than both pleasant and unpleasant pictures (pleasant: M 1.86, SD 0.64; neutral: M 1.75, SD 0.61; unpleasant: M 1.84, SD 0.62; ps .1). There was no main effect of group or Group Stimulus Type interaction on arousal ratings in the masked condition (see Figure 1D). A 2 3 (Group Stimulus Type) ANOVA on valence ratings in the masked condition yielded a significant Group Stimulus Type interaction, F(2, 52) 4.15, p .05, 2 .14. Pairwise comparisons revealed that controls rated masked unpleasant pic-

Figure 2. Square-root transformed SCRs of meditators and controls to pleasant, neutral, and unpleasant stimuli presented A) nonmasked and B) masked. Error bars represent standard errors.

400

NIELSEN AND KASZNIAK

Figure 3. Facial EMG activation change (in microvolts) in both meditators and controls in response to pleasant, neutral, and unpleasant pictures, measured as the difference between mean activation during the 3 seconds poststimulus onset and mean activation during the 1 second prestimulus period: A) left corrugator EMG change to nonmasked pictures; B) left corrugator EMG change to masked pictures; C) left zygomatic EMG change to nonmasked pictures; D) left zygomatic EMG change to masked pictures. Error bars represent standard errors.

tures as significantly more unpleasant than meditators ( p .05). Separate three stimulus type ANOVAs on valence ratings in controls and meditators indicated that controls, F(2, 32) 4.65, p .05, 2 .23, were sensitive to valence information contained in the masked pictures, but meditators were not, F(2, 20) 0.94, p .41, 2 .09. Planned pairwise comparisons indicated that controls rated unpleasant masked pictures as significantly more unpleasant than pleasant masked pictures ( p .05) with a tendency to rate them as more unpleasant than neutral masked pictures ( p .11), although ratings for neutral and pleasant pictures did not differ (pleasant: M 2.99, SD 0.19; neutral: M 3.04, SD 0.11; unpleasant: M 3.13, SD 0.17). This pattern in controls is consistent with accurate valence discrimination (see Figure 1B). In contrast with controls, and counter to hypotheses, meditators did not discriminate among masked pictures on the valence dimension (pleasant: M 3.04, SD 0.24; neutral: M 2.97, SD 0.35; unpleasant: M 2.93, SD 0.23). Individual differences. When including clarity as a covariate in the analysis of arousal ratings, the main effect of stimulus type remained, F(2, 48) 6.17, p .05, 2 .20, but a main effect of

group also emerged, F(1, 24) 4.09, p .05, 2 .15, qualified by an interaction of Group Clarity, F(1, 24) 5.93, p .05, 2 .20. Higher clarity scores were associated with lower arousal ratings to all stimulus types in meditators (pleasant: r .74, p .01; neutral: r .76, p .01; unpleasant: r .81, p .005), but not controls (all rs .2, ns). When clarity was included as a covariate in the analysis of valence ratings, the Group Stimulus Type interaction remained, F(2, 48) 7.22, p .005, 2 .23, but a Stimulus Type Clarity interaction emerged, F(2, 48) 3.18, p .05, 2 .12. Among meditators only, Clarity scale scores were significantly positively correlated with valence ratings to unpleasant pictures in the masked condition (r .76, p .01), which may be interpreted as a tendency to rate these stimuli as more unpleasant as clarity increases. All other correlations between clarity scale scores and valence ratings of masked items by either controls or meditators were low, .18 r .18, ns. Influences of masked stimulus awareness on ratings. One control participant reported seeing many more masked items than other participants (30 of 48 masked items; 17 on Day 1, 13 on Day

AWARENESS OF SUBTLE EMOTIONAL FEELINGS

401

2). Excluding data from this participant did not alter the findings of a significant main effect of stimulus type on arousal ratings, F(2, 50) 4.22, p .05, 2 .14, or the significant Group Stimulus Type interaction on valence ratings, F(2, 50) 3.31, p .05, 2 .12, with controls still showing a significant main effect of stimulus type, F(2, 30) 3.23, p .05, 2 .18. It was not possible to directly compare ratings of seen and unseen items, as the average number of seen items was so small. However, including seen items in analyses eliminated effects of stimulus type on valence ratings, suggesting that awareness of seen items was limited and possibly inaccurate. Rather than strengthening the affective signal, beliefs or experiences of having seen an item may merely have added perceptual and/or affective noise. Some evidence that the experiences of seen and not- seen trials were qualitatively different comes from the following finding. In contrast to our findings for unseen masked items, when all seen masked items were pooled across participants, and ratings of pleasant (n 47), neutral (n 32), and unpleasant (n 38) seen items were compared, there was no significant influence of stimulus type on either valence, F(2, 111) 0.19, p .83, or arousal ratings, F(2, 111) 0.54, p .58. Skin conductance and facial EMG. As illustrated in Figure 2, SCR magnitude in the masked condition was unaffected by stimulus type in either meditators or controls. There were likewise no significant differences in facial muscle activation change from baseline in either corrugator or zygomatic muscle groups as a function of stimulus type during the 3 s of stimulus presentation (see Figure 3). Separate analyses of EMG change in 500-ms epochs (cf. Dimberg et al., 2000) also failed to reveal significant activation differences as a function of stimulus type. The only influence of Attention scale scores was on SCRs to masked pictures. Covariance analysis yielded a significant main effect of attention, F(1, 24) 4.51, p .05, 2 .16. Attention score was positively correlated with SCRs to all stimulus types, (pleasant: r .31, ns; neutral: r .37, p .05; unpleasant: r .36, p .06), suggesting, perhaps not surprisingly, that individuals inclined to attend to emotion were more physiologically aroused when asked to attend to difficult-to-discern emotional states. When Clarity was included as a covariate in the analysis of SCRs, there was a significant Group Clarity interaction, F(1, 24) 5.37, p .05, 2 .18. Correlational analyses revealed that, among meditators, clarity scores were negatively correlated with SCRs to masked items (pleasant; r .77, p .005; neutral: r .47, p .15; unpleasant; r .58, p .06). Among controls, these relations were not obtained (all rs .3, ns). In addition, there was a significant Group Stimulus Type Clarity interaction on left zygomatic responses, F(2, 48) 5.03, p .01, 2 .17, and a Group Clarity interaction on right zygomatic responses, F(1, 24) 5.77, p .05, 2 .19. Among meditators, clarity scores were positively correlated with left and right zygomatic responses to masked pleasant pictures (left: r .54, p .09; right: r .60, p .05) and neutral pictures (left: r .72, p .05; right: r .73, p .05) but only on the right for unpleasant pictures (left: r .28, ns; right: r .77, p .01), whereas among controls, no significant correlations emerged (all rs .3, ns). Thus, meditators with higher clarity had less physiological arousal to all stimuli and more zygomatic activity in the masked condition.

Exploratory Analyses of Valence Discrimination: A Signal Detection Approach

Although the controls pattern of valence ratings in the masked condition was in the predicted direction, specific accuracy (e.g., actually rating pleasant stimuli as pleasant) could not be assessed in the above analyses of mean differences. Recognizing the potential for the following analyses to capitalize on chance, we were nonetheless interested in exploring whether unique individual differences were associated with accuracy in valence discrimination, suggesting hypotheses for future study. Therefore, using a signal detection approach, we sought to identify individual good valence discriminators to determine whether, as a group, they shared any common features. Two separate signal detection analyses were conducted for pleasantness discrimination and unpleasantness discrimination, yielding separate measures of sensitivity (d ) and willingness to endorse the presence of an emotional state (criterion) of pleasantness or unpleasantness. To assess valence discrimination, all pleasant or unpleasant masked items that participants reported not seeing were modeled as signals. Of these, pleasant items rated as 1 or 2 on the 5-point valence scale were classified as pleasant hits; unpleasant items rated as 4 or 5 were classified as unpleasant hits. Catch items (masked masks) were modeled as noise. The same 16 catch items were included in assessments of both pleasantness and unpleasantness discrimination. Catch items rated as 1 or 2 were classified as false alarms for the pleasantness discrimination analysis, and catch items rated as 4 or 5 were classified as false alarms for the analysis of unpleasantness discrimination. Data from the one control participant reporting a high number of seen masked items were excluded from these analyses, because of the unusually low number of items classified as signals in that particular case. Among the remaining 27 participants, actual numbers of pleasant hits ranged from 0 to 10 (M 2.41, SD 2.58) and unpleasant hits ranged from 0 to 7 (M 2.78, SD 1.97) out of a maximum possible 16 items in each case.8 Pleasantness discrimination. Seven participants had positive pleasantness d values, ranging from 0.03 to 1.61. Of these only 4 (all controls) had d values of 0.5 or greater, an arbitrarily chosen demarcation value for distinguishing good discriminators from nondiscriminators. Good discriminators had more pleasantness hits (M 5.5, SD 3.7) and fewer false alarms (M 1.8, SD 2.2) than nondiscriminators (hits: M 2.4, SD 2.6; false alarms: M 2.9, SD 2.9). Unpleasantness discrimination. Fifteen participants had positive unpleasantness d values, ranging from 0.17 to 1.94. Of these, 5 (4 controls, 1 meditator) had d values 0.5. By these criteria, only 1 control participant was classified as good at both pleasantness and unpleasantness discrimination. Good unpleasantness discriminators had, on average, nominally more unpleasantness hits (M 3.8, SD 2.6) and fewer false alarms (M 1.2, SD 1.3) than nondiscriminators (hits: M 2.6, SD 1.8; false alarms: M 2.9, SD 2.2). Pleasantness and unpleasantness d values were not significantly correlated with heartbeat detection scores or self-reported attention or clarity. Likewise, good and nondiscrimi8 While the number of actual hits is small, it may underestimate the number of items that support a discriminatory pattern, that is, pleasant neutral unpleasant on the valence dimension. This pattern could obtain without any pleasant items actually being rated as pleasant.

402

NIELSEN AND KASZNIAK

nators (pleasantness or unpleasantness) did not differ on these measures. Relating awareness to valence discrimination. Pleasantness and unpleasantness discrimination were uncorrelated with the number of items of either valence reported as seen, which suggests that participants with good valence discrimination were not simply more perceptually accurate. Moreover, there were no differences between good and nondiscriminators (pleasantness or unpleasantness) in hit rates for seen items, which further suggests that reporting seeing a masked item did not translate into accurately assessing its valence. There were also no significant differences between participants high and low in pleasantness or unpleasantness discrimination on SCR or EMG in the masked condition. Group differences in valence discrimination. Controls were better discriminators of pleasantness than meditators, F(1, 25) 5.07, p .05, 2 .17, but not unpleasantness, F(1, 25) 1.93, p .1, 2 .07. In addition, controls criterion for endorsing pleasantness was higher, F(1, 25) 10.23, p .005 (controls: M 1.75, SD 1.0; meditators: M 0.64, SD 0.63). Still, meditators were willing to endorse valenced experiences in the masked condition, making significantly more pleasantness false alarms than controls, F(1, 25) 8.44, p .01, 2 .25, with an average of 4.6 (SD 3.1) pleasantness false alarms, compared with an average of 1.6 for controls (SD 2.2). Of note, meditators with higher clarity scores had fewer pleasantness false alarms in the masking task (r .54, p .09), although not better discrimination. Meditators with higher pleasantness false alarms had higher arousal ratings to masked items of all types (rs .66, ps .05), whereas controls with higher false alarms (pleasant or unpleasant) reported higher arousal to masked unpleasant items only (rs .5, ps .05).

pleasant, pleasant masked stimuli elicited experiences that were rated as arousing and neutral, and neutral masked stimuli elicited experiences that were rated as unarousing and neutral. Long-term meditation practice, heightened emotional attention, and enhanced visceral perception ability were not specifically associated with improved discrimination on the valence dimension. However, in meditators only, emotional clarity was associated with more unpleasant ratings of unpleasant masked pictures. Exploratory signal detection analyses of participants ability to extract specific pleasantness or unpleasantness information from their reactions to masked stimuli revealed that controls had a greater sensitivity to pleasantness information than meditators did, whereas meditators were more likely to make pleasantness false alarms.

Lack of Physiological Discrimination Among Masked Stimuli

There was no discrimination among masked pictures on either of the physiological measures used in this study, despite discrimination in emotional self-report. It is not uncommon for the different components of emotion to dissociate, with individual differences revealed on one, but not all, of the measures of the emotional response (Ferguson & Katkin, 1996; Lang, 1994; Ohman & Soares, 1998). However, the present findings stand in contrast to prior studies using masked aversively conditioned stimuli (Katkin et al., 2001; Ohman & Soares, 1993, 1994, 1998) or masked pictures of facial affect (Dimberg et al., 2000) in which physiological discrimination was observed. Design features may account largely for these differences. Aversive conditioning and pictures of facial affect may be more effective at eliciting autonomic and somatic responses, respectively, under masked conditions than the unconditioned biological stimuli selected for this study. More sensitive physiological measures may yield different findings.

Discussion Summary of Findings

In an emotion picture-viewing paradigm, in which stimuli were presented both masked and nonmasked, differences between longterm meditators and control participants were assessed on measures of self-reported emotion experience (valence and arousal) and emotional physiology (skin conductance and facial EMG). Individual differences associated with visceral perception ability (as assessed by a heartbeat detection task), and self-assessed emotional attention and clarity (on standardized questionnaires), were also explored. In the nonmasked condition of the picture-viewing task, there were no differences in physiological responses or in self-reported emotion experiences of valence or arousal as a function of long-term meditation practice, suggesting a lack of differences in emotional responses between the two groups under standard visual stimulus emotion-elicitation conditions. There was no evidence of physiological discrimination in the masked condition. However, both groups showed the expected pattern of arousal ratings. Contrary to hypotheses, only control participants were sensitive to valence information available in the masked stimuli. The combined pattern of valence and arousal discrimination among masked stimuli observed in controls supports the assertion that subtle emotional feelings possess the qualities necessary to guide behavior. For each stimulus type, there was a unique pattern of self-reported experience: Unpleasant masked stimuli elicited experiences that were rated as arousing and un-

Individual Differences
Heartbeat detection. Heartbeat detection was unrelated to selfassessments of emotional awareness, suggesting that these measures tap different abilities or constructs, the former perhaps a more visceral and the latter a more reflective form of awareness. Long-term meditators and controls did not differ in heartbeat detection ability, although more controls than meditators were classified as good heartbeat detectors in this study. Good heartbeat detectors had larger SCRs to nonmasked unpleasant and neutral pictures than nondiscriminators, which was consistent with previously reported positive correlations between heartbeat detection scores and several self-report measures of negative emotionality (Critchley et al., 2004). However, in the present study, this negativity bias was not reflected in self-reported affect to the same stimuli. Heartbeat detection ability has been linked with more pronounced facial expression of emotion (Ferguson & Katkin, 1996). Comparisons of EMG activations of good and nondiscriminating heartbeat detectors in the present study did not corroborate this finding. It should be noted that these analyses are compromised by the small number (n 5) of good heartbeat detectors in this study. Meditation and emotional clarity. Meditators rated themselves higher than controls in emotional claritythe ability to accurately

AWARENESS OF SUBTLE EMOTIONAL FEELINGS

403

discriminate among and label ones feeling statesand length of meditation practice was positively correlated with clarity score. Higher clarity scores in meditators were associated with the following: (a) a tendency to report lower arousal to all types of nonmasked stimuli, (b) lower arousal ratings to all masked stimuli, (c) more unpleasant ratings of valence experience after exposure to the masked unpleasant stimuli, (d) fewer pleasantness false alarms in the masked condition, (e) lower SCRs in both the masked and nonmasked conditions, and (f) increased zygomatic activity to masked pictures. In controls, of whom only 3 had clarity scores exceeding the mean score of meditators, none of these relations were obtained. In summary, lower physiological and experienced arousals were associated with higher clarity in meditators in both the masked and nonmasked conditions, whereas greater zygomatic activity and improved valence detection were apparent only in the masked condition. One possible interpretation of these findings is that the higher levels of clarity achieved by some meditators are accompanied by emotional regulatory skills honed through meditation practice that serve to dampen physiological and experienced arousal and enhance positive facial expressions while simultaneously permitting increased accuracy in valence discrimination. Long-term meditation practice. Our findings suggest that nonmeditators are more sensitive to affective pleasant (approach it) and unpleasant (avoid it) feelings, raising questions about why long-term meditation practice did not more readily facilitate access to such potentially behaviorally useful information. Only longterm meditators with very high levels of emotional clarity were sensitive to subtle negative feelings. This initially seemed puzzling because, according to Buddhist psychology, every conscious state has a feeling tonea quality of pleasantness, unpleasantness, or neutralitythat the experienced meditator is trained to discriminate (Goleman, 2003; Nyanaponika, 2000). These considerations were the motivation for our original, disconfirmed, hypotheses. However, Buddhist psychology itself may also help to explain our seemingly counterintuitive findings. First, in tandem with efforts to cultivate emotional awareness, meditators strive to adopt an equanimous attitude of observing and letting go of their emotions in an effort to avoid destructive desires and states of consciousness that arise from holding on to negative feelings and thoughts (Goleman, 2003). The result of these efforts may be the development of emotion regulation strategies that reduce arousal, enhance positive affect, and prohibit feeling states especially those of uncertain origin, such as feelings evoked by masked stimulifrom being elaborated by cognitive appraisals of valence. According to Young (2003), When feelings are experienced with equanimity they assure their proper function as motivators and directors of behavior as opposed to driving and distorting behavior. In contrast, controls less actively regulated emotional responsivity may permit the elaboration of emotional states in response to masked stimuli, resulting in the better valence discrimination observed in our sample. As predicted, controls were more sensitive to unpleasant than pleasant information in the masked stimuli, supporting the notion of a negativity bias (Cacioppo & Gardner, 1999; Rozin & Royzman, 2001). Meditation practice, on the other hand, gave rise to an apparent positivity effect reflected in more positive facial affect, a failure to rate ambiguous stimuli as negative, and a higher number of positive false alarms in the masked condition (except among meditators with very high clarity scores). The positivity effecta tendency to focus on positive information in attention and mem-

ory has been found in older adults in numerous studies (see Mather & Carstensen, 2005, for a review) and has been proposed to arise from a motivation to maintain well-being through emotional regulation (Carstensen, Fung, & Charles, 2003). Meditators, through practice, may develop similar regulatory skills, although at an earlier age.9 Although meditators seem to retain the ability to allow feelings to evolve under some circumstancesas in the nonmasked condition when asked to explore and report on the nature of their feeling statesperhaps they develop, through practice, a disposition toward equanimity and positivity when stimuli are ambiguous. In summary, long-term meditation practice may entrain automatic emotional regulatory mechanisms, fundamentally changing the way in which practitioners respond to ambiguous emotional events and altering the quality of experiences through changes in motivation and attention. When combined with heightened clarity, such emotion regulation may permit accurate valence discrimination against a background of equanimity. Clarity was positively correlated with years of meditation practice and with patterns of emotional responding consistent with an emotional regulatory skill. Perhaps these two skills discrimination and regulation develop separately in meditators and vie for priority, resulting in the observed differences between meditators and controls in this study. One may speculate that in exceptionally highly trained meditators, such as Buddhist monks, both skills may be more finely honed.

Can Feelings Inform Choice?

Our initial premise was that affective discrimination between pleasant and unpleasant pictures in the masked condition would provide evidence of a basis for subtle conscious emotional feelings that could serve as behavioral guides. Although prior research has demonstrated that negative affect can be elicited preattentively, the present study offers evidence that visually masked biologically relevant stimuli of different valences can give rise to feeling states that differ on arousal and valence dimensions, and that there are individual differences in the ability to discriminate among those feelings. In natural environments, preattentive responses to negative emotional events may support species survival (LeDoux, 1996; Ohman, Flykt, & Lundqvist, 2000). There are also plausible evolutionary reasons why preattentively elicited positive feelings might be adaptive, including the importance of emotion contagion for social interaction (Lundqvist & Dimberg, 1995), and role of positive emotional states in fostering seeking behavior and cognitive elaboration (Fredrickson, 1998; Panksepp, 1998). A better understanding of the role of emotions in guiding behavior will arise only from studies capable of exploring, in parallel, the various channels that carry emotional information, so that we can better assess which of the many components of emotion is doing the functional work in any given decision context. Although scales representing valence and arousal provide a useful starting point, they do not capture many of the qualities of emotional feeling states that may nonetheless play an important role in decision making. Future studies using concurrent measures of physiology, behavior, and emotion experience in a variety of
9 The authors thank Laura Carstensen for discussions suggesting this interpretation of the emotional changes associated with meditation.

404

NIELSEN AND KASZNIAK Fredrickson, B. L. (1998). What good are positive emotions? Review of General Psychology, 2, 300 319. Fridlund, A. J., & Cacioppo, J. T. (1986). Guidelines for human electromyographic research. Psychophysiology, 23, 567589. Gohm, C. L. (2003). Mood regulation and emotional intelligence: Individual differences. Journal of Personality and Social Psychology, 84, 594 607. Gohm, C. L., & Clore, G. L. (2000). Individual differences in emotional experience: Mapping available scales to processes. Personality & Social Psychology Bulletin, 26, 679 697. Goleman, D. (2003). Destructive emotions. How can we overcome them? A scientific dialogue with the Dalai Lama. New York: Bantam Books. James, W. (1994). The physical basis of emotion. Psychological Review, 101, 205210. (Originally published in 1894) Katkin, E. S., Wiens, S., & Ohman, A. (2001). Nonconscious fear conditioning, visceral perception, and the development of gut feelings. Psychological Science, 12, 366 370. Kunst-Wilson, W. R., & Zajonc, R. B. (1980). Affective discrimination of stimuli that cannot be recognized. Science, 207, 557558. Lang, P. J. (1994). The varieties of emotional experience: A meditation on JamesLange theory. Psychological Review, 101, 211221. LeDoux, J. (1996). The emotional brain. New York: Simon & Schuster. Lundqvist, L.-O., & Dimberg, U. (1997). Facial expressions are contagious. Journal of Psychophysiology, 9, 203211. Mather, M., & Carstensen, L. L. (2005). Aging and motivated cognition: The positivity effect in attention and memory. Trends in Cognitive Sciences, 9, 496 502. Maxwell, J. S., & Davidson, R. J. (2004). Unequally masked: Indexing differences in the perceptual salience of unseen facial expressions. Cognition and Emotion, 18, 1009 1029. Monahan, J. L., Murphy, S. T., & Zajonc, R. B. (2000). Subliminal mere exposure: Specific, general, and diffuse effect. Psychological Science, 11, 462 466. Murphy, S., Monahan, J., & Zajonc, R. (1995). Additivity of nonconscious affect: Combined effects of priming and exposure. Journal of Personality and Social Psychology, 69, 589 602. Murphy, S. T., & Zajonc, R. B. (1993). Affect, cognition, and awareness: Affective priming with optimal and suboptimal stimulus exposures. Journal of Personality and Social Psychology, 64, 723739. Nielsen, L. & Kaszniak, A. W. (in press). Conceptual, theoretical, and methodological issues in inferring subjective emotion experience: Recommendations for researchers. In J. A. Coan & J. J. B. Allen (Eds.), The handbook of emotion elicitation and assessment. New York: Oxford University Press. Nyanaponika, T. (2000). The vision of Dhamma. Onalaska, WA: BPE Pariyatti. Ohman, A., Flykt, A., & Lundqvist, D. (2000). Unconscious emotion: Evolutionary perspectives, psychophysiological data and neuropsychological mechanisms. In R. D. Lane, L. Nadel, G. A. Ahern, J. J. B. Allen, A. W. Kaszniak, S. Z. Rapcsak, & G. E. Schwartz (Eds.), Cognitive neuroscience of emotion (pp. 296 327). New York: Oxford University Press. Ohman, A., & Soares, J. J. F. (1993). Backward masking and skin conductance responses after conditioning to nonfeared but fear-relevant stimuli in fearful subjects. Psychophysiology, 30, 460 466. Ohman, A., & Soares, J. J. F. (1994). Unconscious anxiety: Phobic responses to masked stimuli. Journal of Abnormal Psychology, 103, 231 240. Ohman, A., & Soares, J. J. F. (1998). Emotional conditioning to masked stimuli: Expectancies for aversive outcomes following nonrecognized fear-relevant stimuli. Journal of Experimental Psychology: General, 127, 69 82. Panksepp, J. (1998). Affective neuroscience: The foundations of human and animal emotions. New York: Oxford University Press. Robles, R., Smith, R., Carver, C. S., & Wellens, A. R. (1987). Influence of

decision contexts and using a variety of reporting mechanisms may illuminate the specific ways that both positive and negative feelings states interact with cognition to guide choice.

References
Averill, D., & Thomas-Knowles, C. (1991). Emotional creativity. In Strongman, K. T. (Ed.), International review of studies on emotion (Vol. 1, pp. 269 299). New York: Wiley. Bechara, A., Damasio, H., Tranel, D., & Damasio, A. R. (1997). Deciding advantageously before knowing the advantageous strategy. Science, 275, 12931295. Bradley, M. M., & Lang, P. J. (1994). Measuring emotion: The SelfAssessment Manikin and the semantic differential. Journal of Behavior Therapy & Experimental Psychiatry, 25, 49 59. Bradley, M. M., & Lang, P. J. (2000). Measuring emotion: Behavior, feeling, and physiology. In R. D. Lane, L. Nadel, G. A. Ahern, J. J. B. Allen, A. W. Kaszniak, S. Z. Rapcsak, & G. E. Schwartz (Eds.), Cognitive neuroscience of emotion (pp. 242276). New York: Oxford University Press. Brazier, D. (2001). The feeling Buddha: A Buddhist psychology of character, adversity, and passion. New York: Palgrave. Buchner, A., Erdfelder, E., & Faul, F. (1997). How to use G*Power. [Retrieved from http://www.psycho.uni-duesseldorf.de/aap/projects/ gpower/how_to_use_gpower.html] Cacioppo, J. T., & Gardner, W. L. (1999). Emotion. Annual Review of Psychology, 50, 191214. Carstensen, L. L., Fung, H. L., & Charles, S. T. (2003). Socioemotional selectivity theory and emotion regulation in the second half of life. Motivation and Emotion, 27, 103123. Center for the Study of Emotion and Attention. (1999). The International Affective Picture System: Digitized photographs. Gainesville, FL: Center for Research in Psychophysiology, University of Florida. Clore, G., & Parrott, W. G. (1991). Moods and their vicissitudes: Thoughts and feelings as information. In J. P. Forgas (Ed.), Emotion and social judgments (pp. 107123). Oxford, England: Pergamon Press. Critchley, H. D., Wiens, S., Rotshtein, P., Ohman, A., & Dolan, R. J. (2004). Neural systems supporting interoceptive awareness. Nature Neuroscience, 7, 189 195. Dalai Lama, & Cutler, H. C. (1998). The art of happiness: A handbook for living. New York: Riverhead Books. Damasio, A. R. (1994). Descartes error: Emotion, reason and the human brain. New York: Putnam. Davidson, R. J., Kabat-Zinn, J., Schumacher, J. Rosenkranz, M., Muller, D., Santorelli, S. F., et al. (2003). Alterations in brain and immune function produced by mindfulness meditation. Psychosomatic Medicine, 65, 564 570. Dimberg, U., Thunberg, M., & Elmehed, K. (2000). Unconscious facial reactions to emotional facial expressions. Psychological Science, 11, 86 89. Esteves, F., Dimberg, U., & Ohman, A. (1994). Automatically elicited fear: Conditioned skin conductance responses to masked facial expressions. Cognition and Emotion, 8, 393 413. Feldman Barrett, L. (1998). Discrete emotions or dimensions? The role of valence focus and arousal focus. Cognition and Emotion, 12, 579 599. Feldman Barrett, L., Quigley, K. S., Bliss-Moreau, E., & Aronson, K. R. (2004). Interoceptive sensitivity and self-reports of emotional experience. Journal of Personality and Social Psychology, 87, 684 697. Fenigstein, A., Scheier, M. F., & Buss, A. H. (1975). Public and private self-consciousness: Assessment and theory. Journal of Consulting and Clinical Psychology, 43, 522527. Ferguson, M. L., & Katkin, E. S. (1996). Visceral perception, anhedonia, and emotion. Biological Psychology, 42, 131145. Forster, K. I., Mohan, K., & Hector, J. (2003). The mechanics of masked priming. In S. Kinoshita & S. Lupker (Eds.), Masked priming: State of the art (pp. 337). Hove, UK: Psychology Press.

AWARENESS OF SUBTLE EMOTIONAL FEELINGS subliminal visual images on the experience of anxiety. Personality & Social Psychology Bulletin, 13, 399 410. Rozin, P., & Royzman, E. B. (2001). Negativity bias, negativity dominance, and contagion. Personality and Social Psychology Review, 5, 296 320. Salovey, P., Mayer, J. D., Goldman, S. L., Turvey, C., & Palfai, T. P. (1995). Emotional attention, clarity, and repair: Exploring emotional intelligence using the trait meta-mood scale. In J. Pennebaker (Ed.), Emotion, disclosure, and health (pp. 125154). Washington, DC: American Psychological Association. Schwarz, N., & Clore, G. (1983). Mood, misattribution, and judgments of well-being; Informative and direct functions of affective states. Journal of Personality and Social Psychology, 45, 513523. Seligman, M. E. P. (1970). On the generality of the laws of learning. Psychological Review, 77, 406 418. Swinkels, A., & Giuliano, T. A. (1995). The measurement and conceptualization of mood awareness: Monitoring and labeling ones mood states. Personality and Social Psychology Bulletin, 21, 934 949. Taylor, G. J., Ryan, D., & Babgy, R. M. (1985). Toward the development of a new self-report Alexithymia scale. Psychotherapy and Psychosomatics, 44, 191199. Whalen, P. J., Rauch, S. L., Etcoff, N. L., McInerny, S., Lee, M. B., & Jenike, M. A. (1998). Masked presentation of emotional facial expres-

405

sions modulate amygdala activity without explicit knowledge. Journal of Neuroscience, 18, 411 418. Wiens, S., Mezzacappa, E. S., & Katkin, E. S. (2000). Heartbeat detection and the experience of emotions. Cognition and Emotion, 14, 417 427. Wiens, S., & Palmer, S. N. (2000). Quadratic trend analysis and heartbeat detection. Biological Psychology, 58, 159 175. Winkielman, P., Berridge, K. C., & Wilbarger, J. (2005). Unconscious affective reactions to masked happy versus angry faces influence consumption behavior and judgments of value. Personality and Social Psychology Bulletin, 1, 121135. Winkielman, P., Zajonc, R. B., & Schwarz, N. (1997). Subliminal affective priming resists attributional interventions. Cognition and Emotion, 11, 433 465. Wong, P. S., & Root, J. C. (2003). Dynamic variations in affective priming. Consciousness and Cognition, 12, 147168. Young, S. (2003). Equanimity. [Available at http://www.shinzen.org/ shinsub3/artEquanimity.pdf] Zajonc, R. B. (1980). Feeling and thinking: Preferences need no inferences. American Psychologist, 35, 151175.

Received July 7, 2005 Revision received January 31, 2006 Accepted March 7, 2006