Measuring appetitive conditioned responses in humans

Physiology & Behavior 188 (2018) 140–150

Contents lists available at ScienceDirect

Physiology & Behavior journal homepage: www.elsevier.com/locate/physbeh

Measuring appetitive conditioned responses in humans a,⁎

a

Margaret C. Wardle , Paula Lopez-Gamundi , Shelly B. Flagel

T

b,c

a

Center for Neurobehavioral Research on Addiction, Department of Psychiatry and Behavioral Sciences, University of Texas Health Science Center at Houston, 1941 East Rd., BBSB 1st Floor CNRA, Houston, TX 77054, USA Department of Psychiatry, University of Michigan, USA c Molecular and Behavioral Neuroscience Institute, University of Michigan, 205 Zina Pitcher Place, Ann Arbor, MI 48109, USA b

A R T I C L E I N F O

A B S T R A C T

Keywords: Appetitive conditioning, translational research Psychophysiology Human subjects

Clinical and preclinical findings suggest that individuals with abnormal responses to reward cues (stimuli associated with reward) may be at risk for maladaptive behaviors including obesity, addiction and depression. Our objective was to develop a new paradigm for producing appetitive conditioning using primary (food) rewards in humans, and investigate the equivalency of several outcomes previously used to measure appetitive responses to conditioned cues. We used an individualized food reward, and multimodal subjective, psychophysiological and behavioral measures of appetitive responses to a conditioned stimulus (CS) that predicted delivery of that food. We tested convergence among these measures of appetitive response, and relationships between these measures and action impulsivity, a putative correlate of appetitive conditioning. 90 healthy young adults participated. Although the paradigm produced robust appetitive conditioning in some measures, particularly psychophysiological ones, there were not strong correlations among measures of appetitive responses to the CS, as would be expected if they indexed a single underlying process. In addition, there was only one measure that related to impulsivity. These results provide important information for translational researchers interested in appetitive conditioning, suggesting that various measures of appetitive conditioning cannot be treated interchangeably.

1. Introduction In classical appetitive conditioning, an initially neutral stimulus becomes associated with a rewarding unconditioned stimulus (US), such as food. This formerly neutral stimulus, now called the conditioned stimulus (CS), comes to elicit a conditioned response (CR), which often resembles the appetitive responses elicited by the US. Importantly, there is individual variation in the conditioned responses that emerge following appetitive conditioning, and this variation has been associated with the propensity to develop pathological behaviors including over-eating, addiction and depression [1–5]. For example, animals with heightened appetitive responses to CS associated with both drug and non-drug rewards (e.g. food) are at risk for certain addictive behaviors [6]. Similarly, obese individuals have heightened responses to CS that signal food reward [7,8]. Conversely, depressed patients show reduced activity in reward-related brain areas when viewing a CS previously paired with reward [9]. Together, these findings support the notion that differences in responses to appetitive CS may underlie reward-related pathological behaviors. The ability to consistently measure individual differences in appetitive conditioning in humans could help clinicians identify and develop treatment

⁎

strategies for individuals with abnormal appetitive conditioning. However, the field lacks standardized, established ways to measure appetitive conditioned responses in humans. Two aspects of previously used paradigms are particularly problematic: use of secondary reinforcers that are themselves conditioned stimuli (e.g. money), rather than biologically significant primary reinforcers (e.g. food) as US, and use of non-standardized outcomes to measure the strength of appetitive responses to the CS. While the human literature on aversive conditioning tends to utilize biologically significant US (i.e. painful shock, noise blasts, etc.) [10–14], the human literature on appetitive conditioning most often utilizes secondary reinforcers, such as money [11,13,15] and erotic pictures [16]. Secondary reinforcers engage different brain circuits than biologically significant primary reinforcers [11,17]. As the animal literature on appetitive conditioning generally uses biologically significant reinforcers (typically food), use of secondary reinforcers in the human literature raises questions about translational validity. The widespread use of secondary reinforcers may result from the difficulty of identifying universally-rewarding biologically significant US in humans [4]. There have certainly been studies utilizing food as a US in humans, but the majority of these have used sweets or chocolate

Corresponding author. E-mail address: [email protected] (M.C. Wardle).

https://doi.org/10.1016/j.physbeh.2018.02.004 Received 21 November 2017; Received in revised form 16 January 2018; Accepted 2 February 2018 0031-9384/ © 2018 Elsevier Inc. All rights reserved.

Physiology & Behavior 188 (2018) 140–150

M.C. Wardle et al.

2. Methods and materials

[2,3,8,18–24], with only a few exceptions using a limited selection of reinforcers (e.g. one sweet or one savory snack) [10,25]. These studies have often pre-selected participants that prefer sweets [2,8,22], or populations presumed to prefer sweets, such as women [3,23,24]. This approach is problematic when the goal is to create a generally applicable measure of individual differences. Thus, the first goal of the current study was to test whether a novel paradigm that allowed individuals to select a preferred food from a wide variety of sweet and savory snacks could produce robust appetitive conditioning in humans to a biologically significant US without the need to pre-select participants. A second issue with the appetitive conditioning literature in humans is the wide variety of measures of “appetitive responses” to the CS that have been used. These measures include subjective, behavioral, and physiological responses to the CS. Subjective liking of, attraction to, or arousal by the CS is typically measured through self-report, with appetitive CS typically rated as more positively valenced or arousing after conditioning [10,13,22,26–29]. Common behavioral measures of appetitive conditioning include approach tendency and attentional bias, with appetitive CS eliciting increased approach behavior and greater attention [13,26,30,31]. Appetitive CS can also elicit psychophysiological responses, which are typically measured by startle response suppression, skin conductance response, facial muscle responses via electromyography (EMG), and heart rate deceleration [10,16,18,32]. It remains to be determined whether or not these various multidimensional measures correlate with each other, or predict key external outcomes equivalently. Studies that have used several measures of appetitive conditioning have at times found inconsistent group-level (average) differences in the sensitivity of these measures to conditioning. For example, in one study a CS paired with appetitive food was rated as more positively valenced, yielded larger skin conductance response (SCR), and induced startle response attenuation [10], yet in another study a CS paired with sexual stimuli did not yield increased SCR [16]. In another example, a CS paired with food elicited a behavioral approach tendency, but not subjective craving [31]. However, these group-level, average differences do not directly address the question of whether various measures of appetitive conditioning correlate, as would be expected if they tap a unitary process or underlying individual difference in the strength of conditioned responses. To our knowledge, only one study has investigated relationships between appetitive responses to a CS, finding a correlation between electrophysiological responses to a CS and subjective ratings of that CS [19]. This lack of consideration for measure selection is problematic, because subjective, behavioral and psychophysiological responses often do not cohere in response to other emotional stimuli [33]. Thus, our second goal was to assess appetitive responses to the CS using a variety of outcomes, and test the extent to which appetitive responses to the CS on these various measures correlated with each other, and with a potential key external correlate, impulsivity. Impulsivity is thought to be related to increased sensitivity to appetitive rewards, and has been related to strength of appetitive conditioning in previous studies in both animals and humans [5,34–38]. Thus, our objective was to produce a robust and translational appetitive conditioning procedure in humans using food as a biologically significant US, and to use this procedure to examine the relationships between various measures of appetitive responses to the CS. We hypothesized that our individualized appetitive conditioning procedure would yield robust appetitive responses to the CS at a group-level (i.e. on average), across subjective, psychophysiological and behavioral measures. However, we had an open hypothesis about whether the various measures of appetitive response would be correlated with each other, and how they would relate to the external correlate of impulsivity.

2.1. Participants 90 healthy volunteers (59 female, 31 male) ages 18 to 35 were recruited through flyers, online advertisements, and word of mouth. Participants first completed a brief eligibility survey online or by phone. If a participant appeared likely to qualify, they attended a 2 h in-person screening consisting of a review of their medical history, a modified structural Clinical Interview for DSM-IV [39], drug use history form, and the Barratt Impulsiveness Scale (see Section 2.4.2. Measures of Impulsivity). Exclusion criteria were: Serious medical condition, past year DSM-IV Axis I Disorder (except Substance Abuse), lifetime Substance Dependence, smoking > 10 cigarettes per week, psychoactive medications, pregnancy, less than high-school education or poor English fluency. Prior to each session, participants were instructed to refrain from alcohol and over-the-counter drugs for 24 h, refrain from all recreational drugs for 48 h, maintain typical caffeine and nicotine intake for 24 h, and eat and sleep normally. Compliance with alcohol and drug requirements was verified using Alco-Sensor III breathalyzers (Intoximeters Inc., St. Louis, MO) and Readitest 6 Cassette urine drug screens (Redwood Toxicology, Santa Rosa, CA). Female participants completed a urine pregnancy test (Pro Advantage, National Distribution & Contracting, Inc., Nashville, TN) before each session. All participants provided informed consent, and all procedures were approved by the University of Texas Health Science Center at Houston Committee for the Protection of Human Subjects and carried out in accordance with The Code of Ethics of the World Medical Association (Declaration of Helsinki). 2.2. Overall design The study consisted of two 2 h sessions (see Fig. 1). In Session 1, participants completed an orientation, provided baseline picture ratings to select the conditioned stimulus (CS) and control picture, completed the first conditioning session, and did post-conditioning manipulation checks. In Session 2, participants completed the second conditioning session, post-conditioning manipulation checks, measures of appetitive responses to the CS vs. the control picture (“Rating Pictures Task”, “Chasing Pictures Task” and “Dot Probe Task”, presented in counterbalanced order) and the Stop Signal Task measure of impulsivity. Sessions were conducted 48–96 h apart, during typical working hours (8 am–6 pm) and both were required to be at the same time of day (within a 1 h window). 2.3. Procedures 2.3.1. Session 1 Fig. 1 shows a timeline of both sessions. At the beginning of Session 1, baseline criteria were confirmed and participants rated their hunger on a visual analog scale from 0 “Not hungry at all” to 100 “As hungry as I've ever been”. Psychophysiological sensors were then applied. Participants selected a preferred snack from a variety of savory and sweet snacks (Reese's Pieces, Peanut M&Ms, Gummie Bears, Cheez Its, Chex Mix, Pringles, microwave popcorn) and completed a brief orientation to the conditioning procedure and measures of appetitive responses using practice stimuli (these stimuli were never shown again). Participants then rated the subjective valence of 12 neutral pictures from −4 (very negative) to 4 (very positive) using the Evaluative Space Grid [40]. They rated the arousal of the same pictures using a one-item scale ranging from 1 (not at all arousing) to 9 (extremely arousing), per [41–43]. Two images with median valence ratings (typically 0 or 1) and the most similar arousal ratings were randomly assigned to be the CS and control picture. The control picture was not presented again until after conditioning. 141

Physiology & Behavior 188 (2018) 140–150

M.C. Wardle et al.

Fig. 1. Study timeline for Sessions 1 and 2. CS – conditioned stimulus.

Fig. 2. Conditioning apparatus and stimuli presentation. A. Overhead view, showing participant, location of CS presentation, snack delivery and research assistant B. Side view from participant side, showing participant, location of CS presentation and location of snack delivery. C. Depiction of a single conditioning trial. T1: ITI ranging from 6 to 90 s, T2: 6 s picture presentation, T3: First 15 s of 25 s consumption time, blank screen, T4: Last 10s of 25 s consumption time, countdown timer appears on screen.

The first conditioning session was then conducted. Participants were seated in front of a computer monitor, with a partition to their right. The research assistant was seated on the other side of the partition, wearing headphones (Fig. 2a and b show the apparatus). A rotating tray was inset into the partition with two clear glass food cups on opposite sides, such that one cup was always on the participant's side, and one was always on the research assistant's side. Participants were given the instructions “Sometimes I will rotate the tray to give you a piece of the snack food you chose. When that happens, pick the snack up immediately and eat it. You can't ‘save’ snacks for later, and we want you to eat all the pieces that are given to you. You will be given a 25 second period to consume your snack. You will be notified when you have 10 seconds remaining in your eating period.” Research assistants loaded a single piece of the selected snack food (e.g. one M&M, one piece of popcorn) into the cup on the research assistant side of the partition. Then the conditioning trials began (see Fig. 2c for a diagram of a conditioning trial). The CS picture was presented in the center of the monitor for 6 s. Immediately following CS presentation, the research assistant rotated the tray 180 degrees to deliver the loaded food cup to the participant's side of the partition, and the participant retrieved and ate the snack. Research assistants were notified to rotate the tray via a

headphone alert that was not audible to the participant. Research assistants reloaded the empty food cup on their side of the partition with a single piece of the snack food immediately after rotating the tray. A countdown timer appeared on screen during the last 10 s of the 25 s consumption time, prompting participants to finish consuming the snack. Each consumption period was followed by an inter-trial interval (ITI) during which the screen was blank (varying from 6 to 90 s). One block of conditioning consisted of 6 CS/snack pairings. Participants were given a 30s rest period between blocks. Each conditioning session consisted of four blocks, totaling 24 pairings each session, and 48 pairings across the entire study. After the first conditioning session, participants completed the postconditioning manipulation checks: 1. They were shown all 12 original neutral pictures (including both the CS and control picture) and asked to identify which were paired with the snack, to assess explicit knowledge; 2. They rated liking for the snack on a visual analog scale of 0–100 where 0 was “dislike very much”, 50 “neutral”, and 100 “like very much”; 3. They rated the amount of snack they received on a visual analog scale from 0 was “much too little”, 50 “the right amount”, and 100 “much too much”. Note, these amount ratings were not used to individually adjust the amount of food delivered. Participants received 142

Physiology & Behavior 188 (2018) 140–150

M.C. Wardle et al.

the same amount of food at Session 2 regardless of their Session 1 ratings. Psychophysiological sensors were then removed and participants were paid.

to be greater for CS than control pictures. This was quantified as HR during the 1 s fixation vs. HR during seconds 3, 4, and 5 of picture presentation.

2.3.2. Session 2 At the beginning of Session 2, baseline criteria were confirmed and participants rated their hunger on the same visual analog scale. Psychophysiological sensors were then applied. The second conditioning session was conducted, following the same procedures as the first. The measures of appetitive responses to the conditioned stimulus were then completed, in counterbalanced order. These measures were the Rating Pictures Task, Chasing Pictures Task and Dot Probe Task. (see Section 2.4.1. Measures of Appetitive Responses to the Conditioned Stimulus). After the measures of appetitive responses, participants again completed the post-conditioning manipulation checks (identifying which pictures were paired with the snack, rating their liking for the snack and rating the amount of snack they received). Psychophysiological sensors were then removed, and participants completed the Stop Signal Task (see Section 2.4.2. Measures of Impulsivity). At the conclusion of the second session participants were given a staged debriefing that included questions about the perceived purpose of the study, and then paid.

2.4.1.2. Chasing pictures task. This task measured behavioral approach tendency for the pictures. The task consisted of 8 blocks with 8 trials each. At the beginning of each block, the CS and the control picture were presented side by side on the screen and participants were instructed to “approach” one picture and “avoid” the other during that block. In each trial a figure appeared on the screen, followed by the appearance of either the CS or the control picture either above or below the figure. The participant had to move the figure in the necessary direction as quickly as possible by pressing the up or down key. For example, if the “approach” picture was presented below the figure on the screen, participants had to press the down key to move the figure towards the picture. Conversely, if the “avoid” picture was presented in the same position, participants had to press the up key to move the figure away from the picture. Which picture to approach/avoid, location of the picture relative to the figure, and whether the participant had to press the up or down key to execute the desired response was fully counterbalanced, with blocks and trials presented in random order. In this task, participants are faster to approach and slower to avoid pictures with motivational value [31,47]. Thus, we expected “approach” to the CS to be faster and “avoidance” slower, compared to the control picture. This was quantified as reaction time to avoid minus reaction time to approach a picture, with higher numbers indicating more approach bias.

2.4. Measures 2.4.1. Measures of appetitive responses to the conditioned stimulus (CS) 2.4.1.1. Rating pictures task. This task measured subjective and psychophysiological responses to pictures. Participants viewed and rated 24 IAPS pictures, including 6 presentations of the CS, 6 presentations of the control picture, 6 positive images and 6 negative images. Positive and negative images were included to provide scale for responses. Each trial consisted of a 1 s pre-picture fixation and a 6 s picture presentation, followed by single item ratings of subjective valence and arousal [40]. Electromyography (EMG) of the corrugator (“frown”; corrugator EMG) muscle and zygomatic (“smile”; zygomatic EMG) muscle, skin conductance response (SCR) in the eccrine sweat glands of the palm [44] and heart rate (HR) were collected during the 1 s fixation and 6 s picture presentation. All physiological hardware and software were from Biopac Systems, Goleta CA. EMG was recorded using two 4 mm sensors per site, filled with conductive gel and attached using adhesive collars, amplified using an EMG100C amplifier with 500 hz low pass and 10 hz high pass filters, and rectified and smoothed over 20 ms. SCR was recorded using two 8 mm sensors filled with isotonic gel attached to the participant's palm with adhesive collars, amplified using an EDA100C amplifier with a 1.0 hz low pass filter and 0.05 hz high pass filter. HR was measured using disposable adhesive Ag/AgCl electrodes in a standard bilateral configuration on the chest, and amplified using an ECG100C amplifier with a 35 hz low pass filter and 0.05 hz high pass filter. All physiological signals were digitized with a MP150 system and sampled at 1000 hz with Acqknowledge software. This task yielded six measures of appetitive responses to the pictures: Subjective valence, subjective arousal, corrugator EMG, zygomatic EMG, SCR and HR. We expected subjective positivity and arousal to be higher for the CS compared to the control picture. Pleasurable stimuli decrease corrugator (“frown”) and increase zygomatic (“smile”) EMG compared to baseline [45], so we expected corrugator EMG to be lower and zygomatic EMG to be higher to CS compared to control pictures. This was quantified as mean EMG during the 1 s fixation subtracted from activity during the 6 s picture presentation. SCR amplitude is higher during more arousing pictures regardless of valence [44]. Thus, we expected SCR to be higher to CS vs. control pictures. This was quantified as maximum SCR during picture presentation. HR initially decelerates in an orienting response to stimuli, and then reaccelerates, with this acceleration component denoting the motivational properties of the stimuli [46]. Thus, we expected re-acceleration

2.4.1.3. Dot Probe Task. This task measured attentional bias for the pictures [48]. During this task, participants were positioned approximately 20 in. from the monitor. At the start of each trial, a fixation cross appeared in the center of a 22 in. computer monitor for 1200 to 1600 ms, followed by a 2 s presentation of the CS and control picture on either side of the cross, with side of presentation randomized. Pictures were 7.75in apart and measured 4.5 by 3.25 in. Picture presentation was followed by a distractor task in which one picture was replaced by either a circle or a square, and participants pressed a key to indicate the shape. Gaze during picture presentations was measured using electroculography (EOG), with two 4 mm sensors filled with conductive gel (Biopac System Inc.), attached 1.5 cm from the outer canthus of each eye using an adhesive collar. Signals were amplified and digitized using an EOG100C/MP150 system, and sampled at 1000 Hz with Acqknowledge software (all hardware and software from Biopac Systems, Goleta CA). Attentional bias towards drug-related cues over neutral cues in this type of task predicts subjective craving for drugs, prospective drug use, and likelihood of relapse [49–51], and has been proposed as a possible human equivalent to sign-tracking [52]. Thus, we expected attentional bias to be higher for the CS over the control picture. This was quantified as proportion of total fixation time spent gazing at each picture. 2.4.2. Measures of impulsivity Participants completed the Barratt Impulsiveness Scale (BIS-11) at screening [53] and the stop signal task (STOP-IT) at the end of Session 2 (Fig. 1; [54]). The BIS-11 is a 30-item self-report, with higher totals indicating greater self-reported impulsivity. The STOP-IT task is a validated measure of action impulsivity in which participants must inhibit pre-potent button-pressing responses [54]. Impulsivity in the STOP-IT task was quantified using the stop signal reaction time (SSRT), with longer stopping times indicating greater impulsivity. 2.5. Statistical analyses All measures were examined for normality and outliers, and transformations applied where appropriate. We assessed for possible gender differences in all analyses, but these are reported only when significant. 143

Physiology & Behavior 188 (2018) 140–150

M.C. Wardle et al.

CS alone (i.e. without subtracting control picture responses), but results did not lead to substantively different conclusions (results available upon request to the corresponding author). To test whether measures of appetitive response to the CS were related to each other, correlations were conducted. Following Pocock et al. [55], demographic (e.g. age, sex) or procedural (e.g. hunger, liking for the snack etc.) variables that significantly correlated with more than one measure of appetitive response were considered potential confounds and controlled for using regression in analyses including those measures. To test whether measures of appetitive responses to the CS were consistently related to the external correlate of impulsivity we correlated measures of appetitive responses to the CS with self-report (BIS) and behavioral (SSRT) measures of impulsivity. Demographic or procedural variables that significantly related to both a measure of appetitive response to the CS and an impulsivity measure were considered potential confounds and controlled for by using regression for analyses involving those variables.

Table 1 Participant demographics. M(SD) or N(%) Demographics Sex Female Race White Asian Black Other/more than one race Ethnicity Hispanic/Latino Age in years Education in years Body mass index Impulsivity Barratt impulsivity scale total Stop signal reaction time

59 (66%) 35 21 18 16

(39%) (23%) (20%) (18%)

23 (26%) 25.3 (4.8) 16.2 (2.4) 25.1 (5.8) 55.5 (9.2) 251.0 (25.9)

3. Results To test our first hypothesis, that our individualized procedure would produce robust appetitive conditioning, we conducted planned comparisons using paired sample t-tests to contrast response to the CS vs. control picture. For measures where positive and negative pictures were also presented (valence, arousal, Corrugator EMG, Zygomatic EMG, SCR and HR) we secondarily conducted planned comparisons of responses to the CS and control pictures relative to positive and negative pictures using paired sample t-tests, to establish a range of responses. Omnibus ANOVA were not conducted because the large established differences between positive and negative pictures would produce a significant omnibus effect of picture type on every measure, making the omnibus effect uninformative and necessitating the exact same set of planned comparisons proposed above to explicate the effect. Because all measures were taken during extinction (i.e. after conditioning rather than during conditioning), we also tested for effects of extinction in two ways: 1. Between-subjects t-tests comparing appetitive responses to the CS vs. control for individuals who received a given measure first vs. later in the task order. If extinction reduced responses over time we would expect stronger appetitive responses to the CS when a measure was administered first. 2. For variables derived from the Rating Pictures and Dot-Probe task, we also conducted a withinsubject ANOVA testing for significant trends across CS presentations (relative to control picture presentations) using only data from the subset of participants for whom the given measure was administered first in the counterbalanced order (approximately 1/3 of our sample for each measure). This allowed us to examine possible effects of extinction starting from the very first post-conditioning presentation of the CS. If extinction reduced responses over time we would expect stronger appetitive responses to the CS compared to the control picture in early presentations, but not later presentations. The Chasing Pictures Task had counterbalanced trial types, and thus was not suitable for this analysis, as the instruction set (approach vs. avoid) differed across sequential presentations of the CS from participant to participant. For all measures of appetitive responses – valence, arousal, Corrugator EMG, Zygomatic EMG, SCR, HR, approach bias and attentional bias – we then subtracted responses to the control picture from responses to the CS. This was done to control for potential individual differences in overall response style that could have artificially increased correlations between measures of appetitive responses to the CS, particularly within response modalities (e.g. a response bias for higher numbers on subjective scales, or overall tendency to more facial expressiveness in response to all pictures). This produced a single score for each measure that captured individual differences in appetitive responses to the CS while controlling for potential individual differences in overall response style/level. These difference scores were used in all of the following analyses. We also examined appetitive responses to the

3.1. Participant demographics and manipulation checks As seen in Table 1, our participants were majority female with a college education, and represented a variety of racial and ethnic groups. Manipulation checks indicated that our procedure was successful in selecting a rewarding snack and administering an acceptable amount of that snack. Participants reported high snack liking, averaging across the two sessions at 81.9 (SD = 15.5) points on a 0–100 scale. Further, they reported receiving a good amount of the snack, averaging across the two sessions at 46.6 (SD = 17.9) points on a 0–100 scale where 50 was “just the right amount”. Participants were not required to fast, but reported their last meal on average 5.6 h before (SD = 4.7), and a moderate average hunger score of 54.9 (SD = 18.6) on a 0 (not hungry at all) to 100 (as hungry as I've ever been) scale. There were no significant differences between Session 1 and Session 2 on any of these measures. 3.2. Effects of conditioning on appetitive responses 3.2.1. Valence ratings Valence ratings for CS vs. control pictures did not significantly differ before conditioning; Valence: t(89) = 0.38, p = 0.71, d = 0.04, (Fig. 3, Panel A, left side). After conditioning, participants unexpectedly rated the control picture more positively than the CS, t(89) = −2.27, p = 0.03, d = 0.24 (Fig. 3, Panel A, right side). It should be noted, however, that both CS and control pictures were rated less positively than positive pictures, CS vs. positive: t(89) = 12.53, p < 0.001, d = 1.31; control vs. positive: t(89) = 12.66, p < 0.001, d = 1.33, and less negatively than negative pictures, CS vs. negative: t(89) = 14.30, p < 0.001, d = 1.51; control vs. negative: t(89) = 20.09, p < 0.001, d = 2.12 (Fig. 3, Panel A, right side). Testing for potential impacts of extinction, participants who received the Rating Pictures Task first did not rate the CS significantly differently on valence (data not shown). In the subset of participants who received the Rating task first, there was a significant linear trend for valence ratings across repeated presentations, such that earlier presentations of the control picture were rated comparatively more positively than earlier presentations of the CS picture, t(32) = 2.09, p = 0.04, d = 0.38 (data not shown). The reason for this effect is unclear, but it does not indicate an effect of extinction decreasing appetitive responses to the CS over time (as CS liking actually increased over time relative to the control). 3.2.2. Arousal ratings Post-conditioning measures of arousal were square-root transformed to account for non-normality. Arousal ratings for CS vs. control pictures did not significantly differ before conditioning; Arousal: t(89) = 0.35, p = 0.36, d = 0.10 (Fig. 3, Panel B, left side). After conditioning, the CS 144

Physiology & Behavior 188 (2018) 140–150

M.C. Wardle et al. $ %

A. Valence

6

&

#

2

Arousal Rating (0 to 9)

Valence Rating (- 4 to 4)

3

*

1 0 -1 -2 -3

%

B. Arousal

#

&

5 4

*

3

CS Pictures Control Pictures Positive Pictures Negative Pictures

Included for scale

2 1

$

0

Pre-Conditioning

Post-Conditioning

Pre-Conditioning

Post-Conditioning

Fig. 3. Means and SEMs for CS and control pictures for valence and arousal ratings before and after conditioning, with responses to standardized positive and negative pictures included for scale. * - p < 0.05 difference between CS and control pictures, # - p < 0.05 difference between positive and CS pictures, & - p < 0.05 difference between positive and control pictures, % - p < 0.05 difference between negative and CS pictures, $ - p < 0.05 difference between negative and control pictures.

3.2.4. Zygomatic electromyography (zygomatic EMG) Two participants were missing Zygomatic EMG readings due to excessive movement, yielding an n of 88 for these analyses. Zygomatic EMG scores were log transformed to correct non-normality, and one to three outliers per picture type were reduced to three standard deviations from the mean. Zygomatic EMG (“smile”) EMG was higher to CS relative to control pictures, also indicating more positive responses to the CS, t(87) = 2.55, p = 0.01, d = 0.28 (Fig. 4, Panel B). Zygomatic EMG responses to CS pictures were similar to positive pictures, t (87) = 0.87, p = 0.39, d = 0.09, and higher than Zygomatic EMG responses to negative pictures, t(87) = 3.05, p = 0.003, d = 0.33 (Fig. 4, Panel B). Zygomatic EMG responses to control pictures were similar to negative pictures, t(87) = 1.23, p = 0.22, d = 0.13, and lower than Zygomatic EMG responses to positive pictures, t(87) = 3.43, p = 0.001, d = 0.37 (Fig. 4, Panel B). Testing for potential impacts of extinction, receiving the Rating Pictures Task first did not significantly affect Zygomatic EMG responses to the CS vs. the control picture (data not shown). As with Corrugator EMG, missing data led to presentations being divided into “first half of task” and “second half of task” and these averages analyzed within-subjects in the subset of participants who received the Rating task first. Using this procedure, there was no significant change in Zygomatic EMG activity from earlier to later presentations of the CS (data not shown). These analyses suggest no significant impact of extinction on Zygomatic EMG.

was rated as more arousing than the control picture, t(89) = 2.42, p = 0.02, d = 0.26 (Fig. 3, Panel B, right side). Similar to the valence ratings, both CS and control pictures were rated less arousing than either positive or negative pictures, CS vs. positive: t(89) = 13.66, p < 0.001, d = 1.44; control vs. positive: t(89) = 17.50, p < 0.001, d = 1.85; CS vs. negative: t(89) = 13.44, p < 0.001, d = 1.42; control vs. negative: t(89) = 16.67, p < 0.001, d = 1.76 (Fig. 3, Panel B, right side). There was a sex difference in arousal ratings, such that women rated the CS as relatively more arousing, compared to men, t (88) = −3.14, p = 0.002, d = 0.73 (data not shown). Testing for potential impacts of extinction, participants who received the Rating Pictures Task first did not rate the CS significantly differently on arousal (data not shown). In the subset of participants who received the Rating task first, there was no significant trend in arousal ratings across repeated presentations of the CS (data not shown). These analyses suggest no significant impact of extinction on arousal ratings. 3.2.3. Corrugator electromyography (corrugator EMG) Two participants were missing Corrugator EMG readings due to excessive movement, yielding an n of 88 for these analyses. One to three outliers per picture type were reduced to three standard deviations from the mean. Corrugator (“frown”) EMG was lower to CS relative to control pictures, indicating more positive responses to the CS, t (87) = −3.33, p = 0.001, d = 0.35 (Fig. 4, Panel A). In fact, CS pictures produced similar Corrugator EMG responses to positive pictures, t (87) = 0.96, p = 0.34, d = 0.10, and significantly lower Corrugator EMG responses than negative pictures, t(87) = 5.51, p < 0.001, d = 0.59 (Fig. 4, Panel A). Control pictures fell between positive and negative pictures, producing greater Corrugator EMG responses than positive pictures, t(87) = −3.54, p = 0.001, d = 0.38, but lower Corrugator EMG responses than negative pictures, t(87) = 2.66, p = 0.009, d = 0.28 (Fig. 4, Panel A). Testing for potential impacts of extinction, receiving the Rating Pictures Task first did not significantly affect Corrugator EMG responses to the CS vs. the control picture (data not shown). There was too much missing EMG data to conduct a trend analysis across each individual picture presentation. This is typical for psychophysiological measures as generally one or two presentations may be affected by movement artifacts or noise per participant. Thus, presentations were divided into “first half of task” and “second half of task” and these averages analyzed within-subjects in the subset of participants who received the Rating task first. Using this procedure, there was no significant change in Corrugator EMG activity from earlier to later presentations of the CS (data not shown). These analyses suggest no significant impact of extinction on Corrugator EMG.

3.2.5. Skin conductance response (SCR) Two participants were missing SCR due to excessive movement, yielding an n of 88 for these analyses. SCR scores were log transformed to correct non-normality, and one to three outliers per picture type were reduced to three standard deviations from the mean. SCR was higher to CS relative to control pictures, indicating more arousal in response to the CS, t(87) = 2.06, p = 0.04, d = 0.22 (Fig. 4, Panel C). SCR responses to CS pictures were similar to emotional pictures: CS vs. positive pictures, t(87) = 1.43, p = 0.16, d = 0.02; CS vs. negative pictures, t(87) = 0.52, p = 0.61, d = 0.05 (Fig. 4, Panel C). SCR responses to the control picture were similar to positive pictures, t (87) = 0.08, p = 0.94, d = 0.01, but lower than negative pictures, t (87) = 2.04, p = 0.04, d = 0.22 (Fig. 4, Panel C). Testing for potential impacts of extinction, receiving the Rating Pictures Task first did not significantly affect SCR responses to the CS vs. the control picture (data not shown). As with the EMG measures, missing data led to presentations being divided into “first half of task” and “second half of task” and these averages analyzed within-subjects in the subset of participants who received the Rating task first. Using this procedure, there was no 145

Physiology & Behavior 188 (2018) 140–150

M.C. Wardle et al.

B. Zygomatic

A. Corrugator %

Zygomatic EMG (uV)

Corrugator EMG (uV)

1.0

%

2.5

&

$

1.5

&

*

0.5 0.0 -0.5 -1.0

*

2.0 1.5 1.0 0.5 0.0

-1.5

Conditioned Control Pictures Pictures

Conditioned Control Pictures Pictures

Positive Negative Pictures Pictures (Included for Scale)

Positive Negative Pictures Pictures (Included for Scale)

C. Skin Conductance

SCR (usiemens)

0.10

$

*

CS Pictures Control Pictures Positive Pictures Negative Pictures

0.08 0.06 0.04 0.02 0.00

Conditioned Control Pictures Pictures

Positive Negative Pictures Pictures (Included for Scale)

Fig. 4. Means and SEMs for CS and control pictures for corrugator (“frown”) EMG, zygomatic (“smile”) EMG, and skin conductance after conditioning, with responses to standardized positive and negative pictures included for scale. * - p < 0.05 difference between CS and control pictures, # - p < 0.05 difference between positive and CS pictures, & - p < 0.05 difference between positive and control pictures, % - p < 0.05 difference between negative and CS pictures, $ - p < 0.05 difference between negative and control pictures.

3.2.8. Attentional bias Examining eye-tracking, there was no significant difference in attentional bias for the CS vs. control picture, t(89) = −1.00, p = 0.32, d = 0.10 (data not shown). Receiving the Dot Probe Task first did not significantly affect attentional bias for the CS picture vs. the control picture (data not shown). In the subset of participants who received the Dot Probe Task first, there were no significant trends in attentional bias across repeated presentations of the CS (data not shown). These analyses suggest no significant impact of extinction on attentional bias.

significant change in SCR activity from earlier to later presentations of the CS (data not shown). These analyses suggest no significant impact of extinction on SCR.

3.2.6. Heart rate (HR) Five participants were missing HR because this measure was added after these participants were enrolled, yielding a final n of 85 for these analyses. There were no significant differences in HR between CS and control pictures, t(84) = −1.06, p = 0.29 (data not shown). Testing for potential impacts of extinction, receiving the Rating Pictures Task first did not significantly affect HR responses to the CS vs. the control picture (data not shown). In the subset of participants who received the Rating task first, there was no evidence that HR changed across repeated presentations of the CS (data not shown). These analyses suggest no significant impact of extinction on HR.

3.3. Correlations among measures of appetitive responses to the CS As noted above, correlations between measures of appetitive response to the CS used difference scores (i.e. responses to the CS minus responses to the control picture) to account for possible individual differences in overall level of responding. Table 2 indicates where a potential confound was identified and controlled for using regression. As seen in Table 2, there were small to moderate correlations between individual measures of appetitive response to the CS, but no larger set of variables that was consistently related. For example, valence and arousal responses were correlated in expected directions with Corrugator EMG, such that greater arousal and a more positive subjective evaluation of the CS was related to lower Corrugator (“frown”) activity in response to the CS. However, although Corrrugator and Zygomatic EMG were also correlated in the expected direction such that higher Zygomatic (“smile”) responses related to lower Corrugator (“frown”) responses, subjective ratings of valence and arousal and Zygomatic

3.2.7. Approach bias One participant was unable to remember the directions for the Chasing Pictures task, so this participant's data were excluded, yielding a final n of 89 for this analysis. There was no significant difference in approach bias for CS vs. control pictures, t(88) = 0.10, p = 0.92, d = 0.01 (data not shown). Testing for potential impacts of extinction, receiving the Chasing Pictures Task first did not significantly affect approach bias for CS vs. control pictures (data not shown). As noted above, trends across repeated presentations of the CS could not be analyzed for this measure due to the task design. 146

Physiology & Behavior 188 (2018) 140–150

M.C. Wardle et al.

Table 2 Correlations between measures of appetitive response to the conditioned stimulus (CS; quantified as responses to the CS minus responses to the control picture). Arousal ratings

Corrugator EMG

Zygomatic EMG

SCR

Heart rate

Approach bias

Attentional bias

Arousal Ratings

r (88) = 0.20 p = 0.06 –

Corrugator EMG

–

r(86) = −0.21⁎ p = 0.05 r(86) = −0.37⁎ p < 0.01 –

Zygomatic EMG

–

–

r(86) = −0.10a p = 0.37 r(86) = 0.15 p = 0.18 r(85) = −0.22⁎ p = 0.04 –

SCR

–

–

–

r(86) = −0.12 p = 0.27 r(86) = 0.24⁎ p = 0.02 r(85) = −0.01b p = 0.95 r(85) = 0.19 p = 0.07 –

HR

–

–

–

–

r(83) = −0.04 p = 0.71 r(83) = 0.19 p = 0.09 r(82) = −0.04 p = 0.71 r(82) = 0.19 p = 0.09 r(82) = 0.28⁎ p = 0.01 –

Approach Bias

–

–

–

–

–

r(87) = 0.14 p = 0.18 r(87) = 0.26⁎ p = 0.01 r(85) = −0.10 p = 0.38 r(85) = 0.18 p = 0.11 r(85) = 0.10 p = 0.36 r(82) = 0.13 p = 0.26 –

r(88) = −0.15 p = 0.17 r(88) = −0.13 p = 0.22 r(86) = 0.002 p = 0.99 r(86) = −0.11 p = 0.30 r(86) = −0.02 p = 0.87 r(83) = 0.19 p = 0.09 r(87) = 0.09 p = 0.41

Valence Ratings

Corrugator EMG – Corrugator Electromyography, Zygomatic EMG – Zygomatic Electromyography, SCR – Skin Conductance Response, HR – heart rate. ⁎ p < 0.05. a Controlling for amount of snack. b Controlling for liking of snack.

impulsivity on the BIS, r(89) = 0.31, p = 0.003, (Fig. 5, Panel A). Greater attentional bias to the CS was also marginally related to greater behavioral impulsivity on the SSRT, r(88) = 0.19, p = 0.08, (Fig. 5, Panel B). No other significant relationships between measures of appetitive responses to the CS and impulsivity were detected.

responses were not significantly related. Further, the attentional bias measure was unrelated to any of the other measures of appetitive response. There was no obvious pattern to the correlations, such as stronger correlations within types of measures, e.g. among subjective measures, or among psychophysiological measures. We considered using factor analysis to empirically identify subsets of related measures, but this approach is typically implemented only when correlations above r = 0.30 are present [56], which was not generally the case here. Sex only moderated one of these relationships – the correlation between valence and arousal. Valence and arousal were positively and significantly correlated for women (r[57] = 0.32, p = 0.01), but negatively and non-significantly correlated for men (r[30] = −0.24, p = 0.19). As indicated in Table 2, when men and women were combined, the relationship between valence and arousal was positive, but only trended towards significance (r[88] = 0.20, p = 0.06).

4. Discussion We successfully developed a paradigm to pair presentation of a conditioned stimulus with an individually-selected, highly-liked food reward in humans, resulting in robust appetitive responses to the CS on psychophysiological measures. However, our various measures of appetitive response were both differentially sensitive to conditioning on a group (average) level, and comparatively uncorrelated at an individual level. Only one measure of appetitive response, attentional bias, was correlated with impulsivity, suggesting the various measures of appetitive response also differentially predicted key external outcomes. These findings were not consistently moderated by sex, and the failure of the measures to correlate did not appear to be due to differential influence of extinction over time, as there were few differences between initial and later presentations of the CS. Below, we discuss what our findings may demonstrate about appetitive conditioning in humans, and present recommendations for future studies seeking to measure appetitive conditioning in humans. The current paradigm is noteworthy for producing conditioned appetitive responses in humans using an individually-selected biologically significant reward. Our snack choice paradigm was successful in

3.4. Relationship between measures of appetitive response to the CS and impulsivity One individual's SSRT data was excluded because the individual did not perform the task according to instructions. As above, with measures of appetitive responses to the CS, analyses were conducted using difference scores (i.e. responses to the CS minus responses to the control picture) to remove potential individual differences in overall level of responding. One potential confound was identified; race related to both arousal ratings and BIS Scores and was controlled for using regression. Greater attentional bias for the CS was related to greater self-reported

Fig. 5. A. Scatterplot of relationship between gaze bias for the CS vs. control picture and self-reported impulsivity. B. Scatterplot of relationship between gaze bias for the CS vs. control picture and behavioral impulsivity.

147

Physiology & Behavior 188 (2018) 140–150

M.C. Wardle et al.

localizable, manipulable CS such as a lever, or willingness to work for a CS. Use of similar localizable, manipulable stimuli in humans is technically more difficult, especially if designing a paradigm that can be used during neuroimaging studies. Nonetheless, such a setup could be used in initial studies to validate other more accessible measures. For example, eye-gaze has been suggested as an index of approach in humans [52], and was the only measure in our study associated with the expected correlate of impulsivity. An additional potential advantage of gaze as a dependent variable is that advances in technology mean that gaze can now be collected in freely behaving individuals using wearable, head-mounted systems (e.g. Tobii Pro Glasses 2, Tobii AB, Stockholm, Sweden or SR Research Eye Link II, SR Research Ltd., Ottowa, Ontario, Canada, among others). Thus, we recommend eyetracking be collected in freely behaving individuals during conditioning with primary rewards and a manipulable CS, to firmly validate gaze as a translational index of appetitive responses. Limitations of the current study included choice of control conditions, collection of measures only after conditioning, lack of control over time since last meal, and the possible influence of demand characteristics. First, in contrast to most human studies, we did not use as our a control stimulus a picture presented an equal number of times during conditioning not accompanied by a reward (typically designated a CS-). In animals, individual differences in appetitive conditioning are usually measured during Pavlovian conditioning with only a CS that predicts appetitive rewards [38,59], with comparatively few reports using a CS- [60–64]. Use of a CS- engages inhibitory process [65], as the CS- becomes a conditioned predictor of non-reward, and so use of a CSas a control would open the possibility we were also capturing individual differences in inhibitory processes. However, use of a neutral stimulus not presented during conditioning also has drawbacks, including an inability to fully distinguish effects of conditioning from effects of mere exposure [66]. One recommended solution is to use a control stimulus presented an equal number of times in a random relationship to delivery of the US [65]. This would have resulted in an impractically long study in our paradigm, because we were also attempting to maintain a similar length of CS presentation, ratio between inter-trial intervals and CS presentations (around a 5–6 fold ratio), and number of CS-US pairings to the animal literature on individual differences in appetitive conditioning [67]. However, it may be an option for more typical human paradigms that tend to use very short inter-trial intervals and few conditioning trials. Second, our measures of appetitive responses to the CS were taken during extinction, because the retrieval and eating of food rewards was not compatible with several of our behavioral and psychophysiological measures. Measuring responses during extinction is common in the human literature [1,10,32,58,68], but not the animal literature. We did not see effects of extinction when contrasting responses to initial post-conditioning presentations of the CS vs. after many non-reinforced presentations of the CS. However, we would recommend future trials collect intra-conditioning measures as practicable, to improve translational fidelity. Further, with the exception of subjective ratings, we tested responses to the CS only after conditioning, rather than before and after conditioning. This was done to avoid an impact of latent inhibition on conditioned responses [4], but also reduced our ability to control for possible initial differences in our other response measures. We also chose not to fast our participants, even though some findings indicate fasting is important for appetitive conditioning to food rewards in humans [20]. We observed an average time since last food of 5.6 h, which is as long or longer than many comparable studies that did fast participants. But it may be that control over time since last meal is actually more important than simply length of fasting when the goal is to detect individual differences. Future studies should consider feeding participants at a pre-specified time prior to conditioning to reduce this potential source of extraneous variability. We also did not collect hunger ratings after the snack administration, only before, so we do not have complete information on potential individual differences in satiety produced by the procedure.

delivering highly liked rewards on an individual basis without pre-selecting participants. Pairing this reward with a previously neutral picture produced robust conditioning on psychophysiological measures of appetitive response (corrugator and zygomatic EMG) and both psychophysiological (skin conductance) and self-report measures of arousal. Indeed, these responses to the conditioned picture were not only significantly different from a control neutral picture, but generally comparable in magnitude to responses to validated positive pictures. Further, this paradigm produced only one significant sex difference, i.e. higher self-reports of conditioned arousal in women. Taken together, this appears to be a promising new procedure for producing appetitive conditioning to primary rewards in humans that does not require preselection of participants. This paradigm also allowed us to demonstrate dissociations between measures of conditioned appetitive responses that have not been previously reported on in humans. At a group (average) level, after conditioning, the CS evoked robust appetitive responses on psychophysiological measures of appetitive responses and psychophysiological and self-report measures of arousal. However, the CS was also subjectively rated as less positive than the control picture, and behavioral measures of approach showed no significant attraction or aversion. Measures of appetitive responses to the CS were also not strongly related at the individual level, with generally weak correlations between measures. These findings raise two possibilities. First, humans may respond idiographically to conditioning – i.e. some may be “psychophysioloigical responders” while others are “behavioral responders”. Alternately, different measures of appetitive response may be controlled by different systems, with only some systems activated in the current paradigm. For example, psychophysiological responses and arousal may have indexed dopamine-driven “wanting” for the CS, while subjective responses indexed opioid-mediated “liking” [57]. Regardless of cause, this finding is important. Most previous studies of appetitive conditioning in humans have used single measures of appetitive responses to the CS. Even when multiple measures were used, their relationship was generally not investigated [2,5,10,18,22,23,27,58]. Thus, our findings raise key questions about comparability across human conditioning studies utilizing different measures. They also suggest caution about identification of maladaptive appetitive conditioning based on a single response measure. In addition to comparatively small relationships among measures of appetitive conditioning, our measures were also differentially related to impulsivity, a potential external correlate of individual differences in appetitive conditioning. Attentional bias, as measured by eye-gaze, was the only measure that significantly related to impulsivity. While some studies have demonstrated associations between behavioral measures of impulsivity and increased bias to food cues, increased food cue attraction and overeating after contextual appetitive conditioning [5,34,35], others indicate that impulsivity is unrelated to the acquisition of appetitive conditioning [27]. Our results suggest these discrepancies may be due to use of different measures of appetitive response. Based on our results suggesting that measures of appetitive response cannot be considered interchangeable, the question arises of which measure is best to use in studies of appetitive responses in humans. Our results suggest that psychophysiological responses may be the most sensitive to appetitive conditioning, as these were the most reliably different from responses to the control stimulus. However, the most responsive measures may not be the best for detecting individual differences. This is suggested by the fact that attentional bias, which was not responsive to conditioning on an average level, was the only measure significantly related to the important external correlate of impulsivity. Thus, this is an area that needs more research before a firm recommendation can be made. One way to resolve this question may be by increasing translational fidelity to the animal literature. In animals, appetitive responses to a CS are generally identified using either actual approach behavior to a 148

Physiology & Behavior 188 (2018) 140–150

M.C. Wardle et al.

Last, we made no attempt to conceal the CS/US relationship, which may have introduced demand characteristics. The question of whether humans acquire conditioned appetitive responses in the absence of explicit knowledge of the CS/US contingency has been hotly debated [69–71], and a firm analogy based on the animal literature is difficult, so it is unclear whether more robust conditioning or stronger relationships between our measures would have been observed had we included a distractor task or used a probabilistic relationship to mask the contingency between CS and US. In conclusion, we obtained appetitive conditioning with a new procedure using individually-selected primary rewards in humans, and demonstrated important dissociations between potential measures of appetitive responses to conditioned cues in humans. These results strike a cautionary note for translational researchers interested in capturing individual differences in appetitive conditioned responses in humans, suggesting that simply obtaining conditioned appetitive responses on a single measure may be insufficient to detect meaningful individual differences. Studies such as this one, that more systematically investigate the correlates and translational validity of these various outcomes, will help us better understand the factors underlying maladaptive appetitive conditioning and associated psychopathologies such as overeating and addiction.

1016/J.BBR.2010.04.019. [14] G.S. Brunborg, B.H. Johnsen, S. Pallesen, H. Molde, R.A. Mentzoni, H. Myrseth, The relationship between aversive conditioning and risk-avoidance in gambling, J. Gambl. Stud. 26 (2010) 545–559, http://dx.doi.org/10.1007/s10899-010-9178-0. [15] D.J. Levy, P.W. Glimcher, Comparing apples and oranges: using reward-specific and reward-general subjective value representation in the brain, J. Neurosci. 31 (2011) 14693–14707, http://dx.doi.org/10.1523/jneurosci.2218-11.2011. [16] T. Klucken, J. Schweckendiek, C.J. Merz, K. Tabbert, B. Walter, S. Kagerer, D. Vaitl, R. Stark, Neural activations of the Acquisition of Conditioned Sexual Arousal: effects of contingency awareness and sex, J. Sex. Med. 6 (2009) 3071–3085, http:// dx.doi.org/10.1111/j.1743-6109.2009.01405.x. [17] G. Sescousse, X. Caldú, B. Segura, J.-C. Dreher, Processing of primary and secondary rewards: a quantitative meta-analysis and review of human functional neuroimaging studies, Neurosci. Biobehav. Rev. 37 (2013) 681–696, http://dx.doi.org/10. 1016/j.neubiorev.2013.02.002. [18] K.C. Armel, C. Pulido, J.T. Wixted, A.A. Chiba, The smart gut: tracking affective associative learning with measures of “liking”, facial electromyography, and preferential looking, Learn. Motiv. 40 (2009) 74–93, http://dx.doi.org/10.1016/j. lmot.2008.06.003. [19] I.H.A. Franken, J. Huijding, I.M.T. Nijs, J.W. van Strien, Electrophysiology of appetitive taste and appetitive taste conditioning in humans, Biol. Psychol. 86 (2011) 273–278 http://www.scopus.com/inward/record.url?eid=2-s2.0-79952315357& partnerID=40&md5=f9630d3947b68912e66551c52f9e1dce. [20] R.S. Astur, A.W. Carew, B.E. Deaton, Conditioned place preferences in humans using virtual reality, Behav. Brain Res. 267 (2014) 173–177, http://dx.doi.org/10. 1016/j.bbr.2014.03.018. [21] R.S. Astur, A.N. Palmisano, E.C. Hudd, A.W. Carew, B.E. Deaton, F.S. Kuhney, R.N. Niezrecki, M. Santos, Pavlovian conditioning to food reward as a function of eating disorder risk, Behav. Brain Res. 291 (2015) 277–282, http://dx.doi.org/10. 1016/j.bbr.2015.05.016. [22] I. van den Bosch, J.M. van Delft, R.A. de Wijk, C. de Graaf, S. Boesveldt, Learning to (dis)like: the effect of evaluative conditioning with tastes and faces on odor valence assessed by implicit and explicit measurements, Physiol. Behav. 151 (2015) 478–484, http://dx.doi.org/10.1016/j.physbeh.2015.08.017. [23] K. van den Akker, R.C. Havermans, A. Jansen, Effects of occasional reinforced trials during extinction on the reacquisition of conditioned responses to food cues, J. Behav. Ther. Exp. Psychiatry 48 (2015) 50–58, http://dx.doi.org/10.1016/j.jbtep. 2015.02.001. [24] J.M. Brunstrom, S. Higgs, Exploring evaluative conditioning using a working memory task, Learn. Motiv. 33 (2002) 433–455, http://dx.doi.org/10.1016/S00239690(02)00007-3. [25] I.M. Balodis, K.P. Lockwood, S.A. Magrys, M.C. Olmstead, Preference conditioning in healthy individuals: correlates with hazardous drinking, Alcohol. Clin. Exp. Res. 34 (2010) 1006–1012, http://dx.doi.org/10.1111/j.1530-0277.2010.01175.x. [26] D. Van Gucht, F. Baeyens, D. Hermans, T. Beckers, The inertia of conditioned craving. Does context modulate the effect of counterconditioning? Appetite 65 (2013) 51–57, http://dx.doi.org/10.1016/j.appet.2013.01.019. [27] H. Papachristou, C. Nederkoorn, S. Beunen, A. Jansen, Dissection of appetitive conditioning. Does impulsivity play a role? Appetite 69 (2013) 46–53, http://dx. doi.org/10.1016/j.appet.2013.05.011. [28] R.M. Kamal, A.J.M. Loonen, B.A.G. Dijkstra, C.A.J. De Jong, Baclofen as relapse prevention in the treatment of gamma-hydroxybutyrate dependence: a case series, J. Clin. Psychopharmacol. 35 (2015) 313–318, http://dx.doi.org/10.1097/JCP. 0000000000000315. [29] I.S. Johnsrude, A.M. Owen, W.V. Zhao, N.M. White, Conditioned preference in humans: a novel experimental approach, Learn. Motiv. 30 (1999) 250–264, http:// dx.doi.org/10.1006/lmot.1999.1031. [30] J.S. Cavallo, N.A. Ruiz, H. de Wit, Extinction of conditioned responses to methamphetamine-associated stimuli in healthy humans, Psychopharmacology 233 (2016) 2489–2502, http://dx.doi.org/10.1007/s00213-016-4297-7. [31] D. Van Gucht, D. Vansteenwegen, O. Van den Bergh, T. Beckers, Conditioned craving cues elicit an automatic approach tendency, Behav. Res. Ther. 46 (2008) 1160–1169, http://dx.doi.org/10.1016/j.brat.2008.05.010. [32] C. Prévost, M. Liljeholm, J.M. Tyszka, J.P. O'Doherty, Neural correlates of specific and general pavlovian-to-instrumental transfer within human amygdalar subregions: a high-resolution fMRI study, J. Neurosci. 32 (2012) 8383–8390, http://dx. doi.org/10.1523/jneurosci.6237-11.2012. [33] I.B. Mauss, R.W. Levenson, L. McCarter, F.H. Wilhelm, J.J. Gross, The tie that binds? Coherence among emotion experience, behavior, and physiology, Emotion 5 (2005) 175–190 http://www.apa.org. [34] A.C. Tetley, J.M. Brunstrom, P.L. Griffiths, The role of sensitivity to reward and impulsivity in food-cue reactivity, Eat. Behav. 11 (2010) 138–143, http://dx.doi. org/10.1016/j.eatbeh.2009.12.004. [35] R. Hou, K. Mogg, B.P. Bradley, R. Moss-Morris, R. Peveler, A. Roefs, External eating, impulsivity and attentional bias to food cues, Appetite 56 (2011) 424–427, http:// dx.doi.org/10.1016/j.appet.2011.01.019. [36] M. Spoelder, J.P. Flores Dourojeanni, K.C.G. de Git, A.M. Baars, H.M.B. Lesscher, L.J.M.J. Vanderschuren, Individual differences in voluntary alcohol intake in rats: relationship with impulsivity, decision making and Pavlovian conditioned approach, Psychopharmacology 234 (2017) 2177–2196, http://dx.doi.org/10.1007/ s00213-017-4617-6. [37] A. Tomie, K.L. Grimes, L.A. Pohorecky, Behavioral characteristics and neurobiological substrates shared by Pavlovian sign-tracking and drug abuse, Brain Res. Rev. 58 (2008) 121–135, http://dx.doi.org/10.1016/j.brainresrev.2007.12.003. [38] S.B. Flagel, H. Akil, T.E. Robinson, Individual differences in the attribution of incentive salience to reward-related cues: implications for addiction,

Funding This work was supported by the National Institutes of Health National Institute on Drug Abuse [grant number R03 DA036649 to MCW]. The funding source had no role in study design; in the collection, analysis and interpretation of data; in the writing of the report; or in the decision to submit the article for publication. Declarations of interest None. References [1] P. Watson, R.W. Wiers, B. Hommel, S. de Wit, Working for food you don't desire. Cues interfere with goal-directed food-seeking, Appetite 79 (2014) 139–148, http:// dx.doi.org/10.1016/j.appet.2014.04.005. [2] J. Blechert, G. Testa, C. Georgii, W. Klimesch, F.H. Wilhelm, The Pavlovian craver: neural and experiential correlates of single trial naturalistic food conditioning in humans, Physiol. Behav. 158 (2016) 18–25, http://dx.doi.org/10.1016/j.physbeh. 2016.02.028. [3] K.S. Burger, E. Stice, Greater striatopallidal adaptive coding during cue–reward learning and food reward habituation predict future weight gain, NeuroImage 99 (2014) 122–128, http://dx.doi.org/10.1016/j.neuroimage.2014.05.066. [4] C. Martin-Soelch, J. Linthicum, M. Ernst, Appetitive conditioning: neural bases and implications for psychopathology, Neurosci. Biobehav. Rev. 31 (2007) 426–440, http://dx.doi.org/10.1016/j.neubiorev.2006.11.002. [5] K. van den Akker, A. Jansen, F. Frentz, R.C. Havermans, Impulsivity makes more susceptible to overeating after contextual appetitive conditioning, Appetite 70 (2013) 73–80, http://dx.doi.org/10.1016/j.appet.2013.06.092. [6] T.E. Robinson, K.C. Berridge, The psychology and neurobiology of addiction: an incentive-sensitization view, Addiction 95 (2000). [7] E. Stice, S. Spoor, J. Ng, D.H. Zald, Relation of obesity to consummatory and anticipatory food reward, Physiol. Behav. 97 (2009) 551–560, http://dx.doi.org/10. 1016/j.physbeh.2009.03.020. [8] M.D. Meyer, V.B. Risbrough, J. Liang, K.N. Boutelle, Pavlovian conditioning to hedonic food cues in overweight and lean individuals, Appetite 87 (2015) 56–61, http://dx.doi.org/10.1016/j.appet.2014.12.002. [9] P. Kumar, G. Waiter, T. Ahearn, M. Milders, I. Reid, J.D. Steele, Abnormal temporal difference reward-learning signals in major depression, Brain 131 (2008) 2084–2093, http://dx.doi.org/10.1093/brain/awn136. [10] M. Andreatta, P. Pauli, Appetitive vs. aversive conditioning in humans, Front. Behav. Neurosci. 9 (2015) 128, , http://dx.doi.org/10.3389/fnbeh.2015.00128. [11] M.R.R. Delgado, R.L.L. Jou, E.A.A. Phelps, Neural systems underlying aversive conditioning in humans with primary and secondary reinforcers, Front. Neurosci. 5 (2011) 71, http://dx.doi.org/10.3389/fnins.2011.00071. [12] M.R. Delgado, J. Li, D. Schiller, E.A. Phelps, The role of the striatum in aversive learning and aversive prediction errors, Philos. Trans. R. Soc. Lond. Ser. B Biol. Sci. 363 (2008) 3787–3800, http://dx.doi.org/10.1098/rstb.2008.0161. [13] A. Austin, T. Duka, Mechanisms of attention for appetitive and aversive outcomes in Pavlovian conditioning, Behav. Brain Res. 213 (2010) 19–26, http://dx.doi.org/10.

149

Physiology & Behavior 188 (2018) 140–150

M.C. Wardle et al.

[39]

[40]

[41]

[42]

[43]

[44]

[45]

[46]

[47]

[48]

[49]

[50]

[51]

[52]

[53] [54]

Neuropharmacology 56 (2009) 139–148, http://dx.doi.org/10.1016/j.neuropharm. 2008.06.027. M.B. First, Structured Clinical Interview for DSM-IV Axis I Disorders SCID-I: Clinician Version, Administration Booklet, 1997 American Psychiatric Press, Washington, DC, 1956. J.T. Larsen, C.J. Norris, A.P. McGraw, L.C. Hawkley, J.T. Cacioppo, The evaluative space grid: a single-item measure of positivity and negativity, Cognit. Emot. 23 (2009) 453–480, http://dx.doi.org/10.1080/02699930801994054. G.J. Norman, J.T. Cacioppo, J.S. Morris, K. Karelina, W.B. Malarkey, A.C. DeVries, G.G. Berntson, Selective influences of oxytocin on the evaluative processing of social stimuli, J. Psychopharmacol. 25 (2011) 1313–1319, http://dx.doi.org/10. 1177/0269881110367452. J.T. Cacioppo, C.J. Norris, J. Decety, G. Monteleone, H. Nusbaum, In the eye of the beholder: individual differences in perceived social isolation predict regional brain activation to social stimuli, J. Cogn. Neurosci. 21 (2009) 83–92, http://dx.doi.org/ 10.1162/jocn.2009.21007. G.G. Berntson, G.J. Norman, A. Bechara, D. Tranel, J. Bruss, J.T. Cacioppo, The insula, the amygdala and evaluative processes, Psychol. Sci. 22 (2011) 80–86, http://dx.doi.org/10.1177/0956797610391097. M.E. Dawson, A.M. Schell, D.L. Filion, The electrodermal system, in: J.T. Cacioppo, L.G. Tassinary, G.G. Berntson (Eds.), Handb. Psychophysiology, 3rd Ed., Cambridge University Press, New York, NY, 2007, pp. 159–181. J.T. Larsen, C.J. Norris, J.T. Cacioppo, Effects of positive and negative affect on electromyographic activity over zygomaticus major and corrugator supercilii, Psychophysiology 40 (2003) 776–785, http://dx.doi.org/10.1111/1469-8986. 00078. M.M. Bradley, M. Codispoti, B.N. Cuthbert, P.J. Lang, Emotion and motivation I: defensive and appetitive reactions in picture processing, Emotion 1 (2001) 276–298, http://dx.doi.org/10.1037/1528-3542.1.3.276. K. Mogg, B.P. Bradley, M. Field, J. De Houwer, Eye movements to smoking-related pictures in smokers: relationship between attentional biases and implicit and explicit measures of stimulus valence, Addiction 98 (2003) 825–836, http://dx.doi. org/10.1046/j.1360-0443.2003.00392.x. C. MacLeod, A. Mathews, P. Tata, Attentional bias in emotional disorders, J. Abnorm. Psychol. 95 (1986) 15–20, http://dx.doi.org/10.1037/0021-843X.95. 1.15. B.T.T. Saunders, T.E.E. Robinson, A cocaine cue acts as an incentive stimulus in some but not others: implications for addiction, Biol. Psychiatry 67 (2010) 730–736, http://dx.doi.org/10.1016/j.biopsych.2009.11.015. W.M. Cox, L.M. Hogan, M.R. Kristian, J.H. Race, Alcohol attentional bias as a predictor of alcohol abusers' treatment outcome, Drug Alcohol Depend. 68 (2002) 237–243, http://dx.doi.org/10.1016/S0376-8716(02)00219-3. M. Field, W.M. Cox, Attentional bias in addictive behaviors: a review of its development, causes, and consequences, Drug Alcohol Depend. 97 (2008) 1–20, http:// dx.doi.org/10.1016/j.drugalcdep.2008.03.030. D.N. Stephens, T. Duka, H.S. Crombag, C.L. Cunningham, M. Heilig, J.C. Crabbe, Reward sensitivity: issues of measurement, and achieving consilience between human and animal phenotypes, Addict. Biol. 15 (2010) 145–168, http://dx.doi.org/ 10.1111/j.1369-1600.2009.00193.x. J.H. Patton, M.S. Stanford, E.S. Barratt, Factor structure of the Barratt impulsiveness scale, J. Clin. Psychol. 51 (1995) 768–774. F. Verbruggen, G.D.D. Logan, M.A.A. Stevens, STOP-IT: windows executable

[55]

[56] [57]

[58]

[59]

[60] [61]

[62]

[63] [64]

[65] [66] [67] [68]

[69]

[70]

[71]

150

software for the stop-signal paradigm, Behav. Res. Methods 40 (2008) 479, http:// dx.doi.org/10.3758/BRM.40.2.479. S.J. Pocock, S.E. Assmann, L.E. Enos, L.E. Kasten, Subgroup analysis, covariate adjustment and baseline comparisons in clinical trial reporting: current practice and problems, Stat. Med. 21 (2002) 2917–2930. B.G. Tabachnick, L.S. Fidell, Using Multivariate Statistics, 4th ed., Allyn and Bacon, Needham Heights, MA, 2001. K.C. Berridge, T.E. Robinson, J.W. Aldridge, Dissecting components of reward: “liking”, “wanting”, and learning, Curr. Opin. Pharmacol. 9 (2009) 65–73, http:// dx.doi.org/10.1016/j.coph.2008.12.014. D. Van Gucht, F. Baeyens, D. Vansteenwegen, D. Hermans, T. Beckers, Counterconditioning reduces cue-induced craving and actual cue-elicited consumption, Emotion 10 (2010) 688–695, http://dx.doi.org/10.1037/a0019463. P.J. Meyer, V. Lovic, B.T. Saunders, L.M. Yager, S.B. Flagel, J.D. Morrow, T.E. Robinson, Quantifying individual variation in the propensity to attribute incentive salience to reward cues, PLoS One 7 (2012) e38987http://dx.doi.org/10. 1371%2Fjournal.pone.0038987. S.E. Chang, P.C. Holland, Effects of nucleus accumbens core and shell lesions on autoshaped lever-pressing, Behav. Brain Res. 256 (2013) 36–42. S.E. Chang, K.S. Smith, An omission procedure reorganizes the microstructure of sign-tracking while preserving incentive salience, Learn. Mem. 23 (2016) 151–155, http://dx.doi.org/10.1101/lm.041574.115. S.E. Chang, T.P. Todd, D.J. Bucci, K.S. Smith, Chemogenetic manipulation of ventral pallidal neurons impairs acquisition of sign-tracking in rats, Eur. J. Neurosci. 42 (2015) 3105–3116. S.E. Chang, D.S. Wheeler, P.C. Holland, Effects of lesions of the amygdala central nucleus on autoshaped lever pressing, Brain Res. 1450 (2012) 49–56. J.M.M. Uslaner, J.M.M. Dell'Orco, A. Pevzner, T.E.E. Robinson, The influence of subthalamic nucleus lesions on sign-tracking to stimuli paired with food and drug rewards: facilitation of incentive salience attribution[quest], Neuropsychopharmacology 33 (2007) 2352–2361, http://dx.doi.org/10.1038/sj. npp.1301653. R.A. Rescorla, Pavlovian conditioning and its proper control procedures, Psychol. Rev. 74 (1967) 71. R.B. Zajonc, Mere exposure: a gateway to the subliminal, Curr. Dir. Psychol. Sci. 10 (2001) 224–228. E. Hearst, H.M. Jenkins, Sign-tracking: the stimulus-reinforcer relation and directed action, Psychon. Soc. (1974) 1–49. R.S. Astur, A.N. Palmisano, A.W. Carew, B.E. Deaton, F.S. Kuhney, R.N. Niezrecki, E.C. Hudd, K.L. Mendicino, C.J. Ritter, Conditioned place preferences in humans using secondary reinforcers, Behav. Brain Res. 297 (2016) 15–19 SreeTestContent1 https://doi.org/10.1016/j.bbr.2015.09.042. P.F. Lovibond, D.R. Shanks, The role of awareness in Pavlovian conditioning: empirical evidence and theoretical implications, J. Exp. Psychol. Anim. Behav. Process. 28 (2002) 3. J. De Houwer, S. Thomas, F. Baeyens, Association learning of likes and dislikes: a review of 25 years of research on human evaluative conditioning, Psychol. Bull. 127 (2001) 853, http://dx.doi.org/10.1037//0033-2909.127.6.853. A.P. Field, I like it, but I'm not sure why: can evaluative conditioning occur without conscious awareness? Conscious. Cogn. 9 (2000) 13–36, http://dx.doi.org/10. 1006/ccog.1999.0402.