Appetite 76 (2014) 153–160

Contents lists available at ScienceDirect

Appetite journal homepage: www.elsevier.com/locate/appet

Research report

To eat or not to eat. The effects of expectancy on reactivity to food cues q Charlotte A. Hardman ⇑, Jade Scott, Matt Field, Andrew Jones Department of Psychological Sciences, University of Liverpool, Eleanor Rathbone Building, Bedford Street South, Liverpool L69 7ZA, UK

a r t i c l e

i n f o

Article history: Received 11 November 2013 Received in revised form 21 January 2014 Accepted 7 February 2014 Available online 13 February 2014 Keywords: Food cue reactivity Attentional bias Expectancy Learning Reward Motivation

a b s t r a c t Cue reactivity may be determined by the ability of cues to evoke expectations that a reward will be imminently received. To test this possibility, the current study examined the effects of manipulating expectations about the receipt of food (pizza) on self-reported and physiological responses to pizza cues, and attentional bias to pizza pictures. It was predicted that expecting to eat pizza would increase salivation, self-reported measures of motivation and attentional bias to pizza cues relative to conditions where there was no eating expectancy. In a within-subjects counterbalanced design, 42 hungry participants completed two pizza-cue exposures in a single experimental session during which their expectation of consuming the pizza was manipulated (i.e., expectancy of eating imminently vs. no eating expectancy). They also completed a computerised attentional bias task during which the probability of receiving pizza (0%, 50% or 100%) was manipulated on a trial-by-trial basis. Participants showed reliable increases in hunger and salivation in response to the pizza cues, as well as a bias in attentional maintenance on pizza pictures. However, these responses were not influenced by eating expectancy. Contrastingly, expectancy did influence early attentional processing (initial orientation of attention) in that participants directed their first gaze towards pizza pictures more often on 100% and 50% probability trials relative to 0% trials. Overall, our findings indicate that exposure to food cues triggers appetitive responses regardless of explicit expectancy information. Methodological features of the study that may account for these findings are discussed. Ó 2014 Elsevier Ltd. All rights reserved.

Introduction The consumption of food is highly rewarding and therefore supports the development of conditioned responses. Initially, food acts as an unconditioned stimulus (US) that elicits unconditioned responses (URs). Through a Pavlovian conditioning process, the rewarding properties of food become associated with external cues that are present at the time of consumption, such as the sight and smell of food. These external cues become conditioned stimuli (CS) that influence eating initiation and meal size and evoke a number of conditioned responses (CRs) (Weingarten, 1985). Accordingly, exposure to food cues in humans has been shown to reliably elicit CRs such as changes in subjective state (increased hunger, desire to eat and craving), physiological readiness to eat (increased salivation and heart rate) and cognitive changes, such as food-related attentional bias (Fedoroff, Polivy, & Herman, 1997; Ferriday & Brunstrom, 2011; Nederkoorn, Smulders, & Jansen, 2000; Rogers

q Acknowledgment: This study was funded by a University of Liverpool School of Psychology summer studentship awarded to Jade Scott. ⇑ Corresponding author. E-mail address: [email protected] (C.A. Hardman).

http://dx.doi.org/10.1016/j.appet.2014.02.005 0195-6663/Ó 2014 Elsevier Ltd. All rights reserved.

& Hill, 1989; Smeets, Roefs, & Jansen, 2009). There is also evidence that food cue exposure increases the amount of food that is subsequently consumed (Fedoroff et al., 1997; Ferriday & Brunstrom, 2008). Expectancy theory (Bolles, 1972) holds that experience of Pavlovian contingencies gives rise to explicit knowledge of the predictive relationships between environmental stimuli and reward. In this way, the cue (CS) signifies the availability of the reward and elicits an expectation that it will be imminently received. Expectancy may therefore be critical for the initial development of CRs (Field & Cox, 2008; Hogarth & Duka, 2006; Jedras, Jones, & Field, in press). This prospect is supported by human conditioning studies with addictive substances which indicate that nicotine CRs, such as salivary responses, attentional bias and subjective craving, depend on participants having explicit knowledge of, and hence an expectation about, the CS-US contingency (Field & Duka, 2001; Hogarth & Duka, 2006). Further evidence is provided by studies that experimentally manipulated participants’ expectations about substance availability and observed effects on cue reactivity (e.g., Carter & Tiffany, 2001). In one study, the probability that participants would receive beer (100%, 50%, or 0%) was manipulated on a trial-by-trial basis in an attentional bias eyetracking task (Field et al., 2011). Results indicated that, in light drinkers, attentional

154

C.A. Hardman et al. / Appetite 76 (2014) 153–160

bias towards alcohol-related pictures was seen only on 100% probability trials (i.e., when alcohol was expected imminently), whereas heavy drinkers’ attentional bias was insensitive to alcohol expectancy. These findings were replicated using both chocolate and alcohol rewards (Jones et al., 2012) and, intriguingly, effects were not outcome-specific; that is, the expectation of receiving alcohol increased attentional bias for both alcohol and chocolate cues, and vice versa. This general transfer effect might reflect an increase in arousal when rewards are anticipated, which enhances attentional bias for a variety of motivationally-salient cues (Jones et al., 2012). There has been much less consideration of the role of expectancy in human eating studies. In the majority of existing food cue reactivity studies, it is not clear whether participants expected to be able to eat the cued foods. This is potentially problematic because, on the basis of the above evidence, eating expectancy would be predicted to strongly influence the magnitude of food cue reactivity. In order to address this important issue, some studies have manipulated expectations about whether the cued foods will be available for subsequent consumption. In a study with dietary restrained women, Higgs (2007) manipulated information about the post-task availability of a cued food (chocolate cake) and examined concurrent performance on a reaction time task. It was predicted that reaction time performance would be impaired when participants expected to subsequently eat the chocolate cake because consumption of a forbidden food would trigger diet-related anxieties. However, there was no evidence for an effect of post-task food availability on reaction time. More recently, Werthmann, Roefs, Nederkoorn, and Jansen (2013) found that the perceived availability of chocolate did not affect chocolate-related attentional bias, craving or chocolate intake in healthy weight female participants. These null findings stand in marked contrast to results from the addiction studies but it is possible that they are explained by methodological differences. For example, in the Werthmann et al. (2013) study, the time delay between giving participants the availability information and the actual opportunity to consume the food (approximately 15–20 min) may have resulted in the availability information losing its motivational impact. This is supported by a study by Field and Duka (2004) in which participants who expected to smoke had to wait around 20 min before being able do so and there were no effects on smoking cue reactivity (craving and physiological reactivity). On the basis of existing studies, it would appear that the effects of reward anticipation on cue reactivity are most prominent when expectancy is manipulated on an immediate, trial-by-trial basis. This ensures that participants expect to receive the reward (or not receive it) at the exact moment that the response is measured (Jedras et al., 2014). The current study examined the effects of expectancy information about the imminent receipt of an appetizing food (pizza) on reactivity to food cues. In addition to assessing attentional bias and self-reported responses to food cues, we also included a physiological measure of cue reactivity (salivation). Salivary responses are sensitive to food cue exposure (Brunstrom, Yates, & Witcomb, 2004; Ferriday & Brunstrom, 2011; Nederkoorn et al., 2000; Rogers & Hill, 1989), contextual appetitive conditioning (van den Akker, Jansen, Frentz, & Havermans, 2013) and smoking expectancy (Field & Duka, 2001); however, to our knowledge, salivation has not yet been examined in the context of explicit manipulations of eating expectancy. Our primary hypothesis was that expecting to eat pizza would increase salivation, attentional bias to pizza and self-reported measures of cue reactivity relative to comparable noeating expectancy conditions. We also anticipated that higher levels of cue reactivity as a result of explicit eating expectancy would, in turn, be expressed behaviourally as increased food intake. In this way, our secondary hypothesis predicted that increased cue reactivity during conditions of eating expectancy (relative to

conditions of no-expectancy) would be predictive of subsequent ad libitum intake of pizza. Our attentional bias task was adapted from that used by Field et al. (2011) and Jones et al. (2012) and critically enabled the manipulation of pizza expectancy on an imminent, trial-by-trial basis. The task also included alcohol-related pictures. This was to determine whether expectancy of receiving pizza would increase attentional bias towards other reward-relevant stimuli thus indicating a general transfer effect, as has recently been reported (Jones et al., 2012). Finally, divergent findings between the drug and food studies might also be explained by differences in the level of substance exposure. In the nicotine studies for example, participants were daily smokers (e.g., smoking at least 15 cigarettes per day in Field and Duka (2001)) whereas, in the eating studies, habitual consumption of the target food was less frequent (e.g., less than once a day in Werthmann et al. (2013)). This greater exposure might lead to stronger associations between drugs and cues, thus affecting both the magnitude of the CRs themselves and responses to expectancy information. In order to address this possibility, we examined whether individual differences in habitual consumption of pizza (i.e., more-frequent vs. less-frequent consumers) would moderate the effects of the expectancy manipulation on food cue reactivity. Method Participants Forty-two participants were recruited from the University of Liverpool (UK) through email and poster advertisements. A total of 28 females (67%) and 14 males (33%) participated; the average age of participants was 27.83 (SD = 8.26) years. Inclusion criteria were regular consumption of pizza (i.e., every 2–3 months, at least), consumption of alcohol (monthly, at least), and normal or corrected-to-normal vision. Participants were excluded if they wore glasses (due to eye-tracking apparatus) or if they had any food allergies or intolerances. The study was approved by the University of Liverpool research ethics committee. Participants were informed that the study was investigating the relationship between cognitive function and eating behaviour. Measures Salivation Total volume of salivation was measured over a period of 30-s. Participants placed a 3.5 cm dental roll horizontally under their tongue. The rolls were weighed before and after being placed in participants’ mouths and the difference in weight (g) was recorded as the amount of salivation. This technique is widely used in food studies (Brunstrom et al., 2004; Ferriday & Brunstrom, 2011; Rogers & Hill, 1989), is relatively non-invasive and provides a sensitive measure of whole-mouth saliva volume. Self-report measures Hunger, pizza pleasantness and desire to eat pizza were measured using 100-mm visual-analogue scales (VAS). Subjective arousal was measured using VAS ratings of excitedness (‘‘how excited do you feel right now?’’) and alertness (‘‘how alert do you feel right now?’’). All of the scales were anchored with the phrases ‘‘Not at all’’ and ‘‘extremely’’. Attentional bias Pictorial stimuli. We used 5 pizza-related pictures (e.g., a closeup of a piece of pizza), each of which was paired with a neutral picture (e.g., a close-up of a stationery item). To test for a general

C.A. Hardman et al. / Appetite 76 (2014) 153–160

transfer effect, we also used 5 alcohol-related pictures (e.g., a closeup of a bottle of beer) and each one was paired with a neutral stationery-related picture. The pizza picture pairs were specifically prepared for this study whereas the alcohol pictures were a subset of those used in Field et al. (2011) and Jones et al. (2012). Pictures in each pair were matched on complexity and brightness, and each individual picture was 90 mm high by 130 mm wide. Expectancy attentional bias task (based on Field et al. (2011) and Jones et al. (2012)). In this task, the expectation of receiving pizza (100%, 50%, or 0% probability) was manipulated on a trial-by-trial basis. Each trial began with the display of a picture of pizza (65  52 mm) for 1000 ms. Probability information (‘‘100%’’, ‘‘50%’’, or ‘‘0%’’) was presented directly below. Participants were explicitly informed that this probability represented their chance of winning a pizza point, with each point representing a small piece of pizza that they could consume after the task. This was immediately followed by a picture pair on the left and right sides of the screen, 120 mm apart, for 2000 ms. Picture pairs were either a pizza/neutral or an alcohol/neutral matched pair. Following this, the pictures were replaced with text stating ‘‘press space to try and win!’’. Pressing the spacebar immediately triggered a feedback screen, which was presented for 1000 ms. On all of the 100% trials and half of the 50% trials, feedback stated ‘‘you win a pizza point!’’. On half of the 50% trials and all of the 0% trials, feedback stated ‘‘you win nothing’’. The task began with 8 practice trials in which pairs of neutral stimuli were presented, in order to familiarise participants with the procedure. The main block consisted of 120 critical trials, which were presented in a random order. Pizza-neutral picture pairs were presented on half of the trials, and alcohol-neutral picture pairs were presented on the other half of trials. For each type of picture pair, the reward-related picture (i.e., pizza or alcohol) was presented on the left-hand side of the screen on half of the trials, and on the right-hand side of the screen on the remaining trials. For each picture position, the expectation of receiving pizza was manipulated. There were an equal number of 0%, 50%, and 100% probability trials. Therefore, for each type of picture pair, there were 60 trials in total, comprising 20 each of 100% pizza, 50% pizza, and 0% pizza expectancy trials. After 60 trials, participants were allowed to rest for one minute whilst a feedback screen displayed the number of pizza points they had accumulated so far. We opted to award pizza points at the end of each trial, as opposed to actual amounts of pizza, due to concerns that consumption of small pieces of pizza at the end of each trial would cause participants to quickly become sated which would diminish the overall motivational value of pizza. Importantly, Jones et al. (2012) and Field et al. (2011) found comparable effects of reward expectancy regardless of whether participants received points for consumption later or actual rewards for immediate consumption, respectively. Procedure In a fully within-subjects design, participants attended one 60min session held between 1 pm and 5 pm in the Department of Psychological Sciences, University of Liverpool. On the day of their session, participants were asked to eat their usual breakfast but to refrain from eating lunch. On arrival, they provided written informed consent. They recorded their usual frequency of pizza consumption on a 0–6 scale, where response options were ‘‘Never’’, ‘‘Less than once a year’’, ‘‘More than once a year’’, ‘‘Every 2– 3 months’’, ‘‘Once a month’’, ‘‘Fortnightly’’, and ‘‘Every week or more’’. They also recorded their usual consumption of alcohol on a 0–5 scale where response options were ‘‘Never’’, ‘‘Monthly or less’’, ‘‘2 to 4 times a month’’, ‘‘2 to 3 times a week’’, ‘‘4 or more times a week’’, and ‘‘Every day’’.

155

Participants then took part in two pizza cue exposure periods. The order of the cue exposures was counterbalanced across participants. Each cue exposure lasted for 1 min during which participants were presented with a 25-g (75 kcal) slice of freshly baked cheese and tomato pizza (‘Goodfella’s’ Thin Stonebaked Margherita Pizza, Green Isle Foods Ltd., Co Kildare, Ireland; 298 kcal/100 g). In one of the cue exposures, participants were told that they would be able to eat one mouthful of the pizza in 1 min time (eating expectancy (EE) condition). In the other cue exposure, they were told that they would not be able to eat any of the pizza (no eating expectancy (NEE) condition). Given that counterbalancing was in place, critically half of the participants had the EE condition first followed by the NEE condition (condition order 1) whereas the remaining participants had the NEE condition first followed by the EE condition (condition order 2). Participants were alternately allocated to the respective condition order groups, balanced across gender. Prior to each cue exposure, participants rated their level of hunger. A pre-exposure measure of salivation was also taken. Each cue exposure began with the presentation of the slice of pizza. The relevant expectancy instruction was given at this point; participants were either told ‘‘you will definitely eat a mouthful of this pizza in 1 minute’s time’’ (EE condition) or ‘‘you won’t get to eat a mouthful of this pizza’’ (NEE condition) (the order of cue exposure conditions was counterbalanced across participants). Participants were instructed to take the pizza in their hand and to smell and imagine eating it. At the same time, they completed VAS ratings of expected pizza pleasantness, desire-to-eat, excitedness, and alertness. Another hunger rating was also completed at this point. During the last 30 s of the cue exposure, a further salivation measure was taken. After the 1 min period had elapsed, participants in the EE condition ate one mouthful of pizza. Both cue exposure periods ended with the removal of the pizza. After completing the first cue exposure, participants worked on an irrelevant problem solving task for 10 min. This was to provide a ‘‘wash-out’’ period between the two cue exposure periods and was also consistent with the cover story, that the study was investigating the relationship between cognitive function and eating behaviour. They then completed the second cue exposure, after which they continued to work on the problem solving task for a further 5 min. Participants then completed the expectancy attentional bias task. They were instructed to pay close attention to the percentages that were presented at the start of each trial as this indicated the probability that they would ‘‘win’’ pizza points on that trial. Participants were seated approximately 23-inches from the computer screen with their chin on a chin-rest before calibration. Eye movements were continuously recorded using an Eye-Trac D6 desktop mounted camera (Applied Science Laboratories, Bedford, MA). The task lasted approximately 15 min. Participants were then provided with the pizza that they ‘‘won’’ during the attentional bias task. All participants were presented with 1700 kcal of the same brand of pizza used in the cue exposures. The pizza was cut into equal sized square pieces so that participants could not easily monitor the amount eaten and to reinforce the fact that they had won these pieces during the eye-tracking task. Participants were invited to eat as much or as little as they wished and were left alone. They rang a bell to indicate to the researcher that they had finished eating. The amount of pizza consumed was assessed by covertly weighing the plate before and after consumption. Participants completed VAS ratings of hunger before and after consumption. They then completed the restraint and disinhibition scales of the Three Factor Eating Questionnaire (TFEQ) (Stunkard & Messick, 1985) and their height and weight were measured. At the end of the session, they were debriefed and reimbursed £10 as compensation for their travel expenses and time.

156

C.A. Hardman et al. / Appetite 76 (2014) 153–160

Data reduction and analysis Attentional bias Eye-movements were recorded at a sampling rate of 120 Hz. Fixations were defined as a stable eye movement within one degree of visual angle for 100 ms or longer. The critical dependent variable for attentional bias was gaze ‘‘dwell time’’ on reward-related (pizza or alcohol) and neutral pictures. Gaze dwell time is the total time spent fixating on each picture during the 2000 ms epoch when both pictures were displayed. This method is commonly used to assess the duration of eye movement fixations to specific areas of interest in attentional tasks, and the gaze dwell time index has good concurrent validity with reaction time measures of attentional bias (Field, Mogg, & Bradley, 2004; Mogg, Bradley, Field, & De Houwer, 2003). As a secondary measure of attentional bias, we examined biases in initial orienting of gaze towards the pizza and alcohol pictures relative to the neutral images. A gaze ‘‘direction bias’’ was computed based on the proportion of trials in which the first fixation was directed at the pizza picture or alcohol picture instead of the neutral picture. A bias score greater than 50% represents a higher proportion of first fixations directed towards the pizza or alcohol pictures. Conversely, a bias score of less than 50% indicates a higher proportion of first fixations directed towards the neutral pictures. This measure is increasingly being used to quantify early attentional allocation towards food stimuli (Werthmann et al., 2011, 2013). Results Participant characteristics Due to technical problems with the eye-tracker, data from two participants were lost. Therefore, we analysed 40 complete data sets. Gender, age, body mass index (BMI), TFEQ-restraint and –disinhibition are shown in Table 1. On average, participants reported eating pizza between ‘‘once a month’’ and ‘‘fortnightly’’ (mean score (SD) on the 0–6 point scale = 4.55 (0.93)). Fifty-five per cent of participants (N = 22) were classified as more frequent consumers (reported eating pizza ‘‘fortnightly’’ or ‘‘every week or more’’ on the 0–6 point scale). The remaining 45% (N = 18) were classified as less frequent consumers (reported eating pizza ‘‘every 2– 3 months’’ and ‘‘once a month’’). Average reported consumption of alcohol was between ‘‘2–4 times a month’’ and ‘‘2–3 times a week’’ (mean score on the 0–5 point scale was 2.25 (0.87)). The average time that participants had abstained from eating any food prior to the test session was 3.97 (1.69) hours. Salivation The total amount of saliva produced (g) was analysed using a 2  2  2 mixed analysis of variance (ANOVA). The within-subjects factors were expectancy condition (1. EE and 2. NEE) and time (1.

Table 1 Descriptive characteristics of the sample. Values are means with SDs in parentheses, unless otherwise specified. Characteristic

Value

N Age (y) BMI (kg/m2) TFEQ-Restraint (0–21) TFEQ-Disinhibition (0–16)

40 (27 F, 13 M) 27.83 (8.26) 23.72 (4.44) 8.98 (5.04) 7.75 (3.16)

F = female, M = male; TFEQ = Three Factor Eating Questionnaire.

Pre-exposure and 2. During exposure). The between-subjects factor was condition order (1. EE first, NEE second or 2. NEE first, EE second). There was no main effect of expectancy and no evidence for the critical expectancy by time interaction (all ps P .31). There was a main effect of time, F(1, 38) = 27.50, p < .001, indicating that, overall, participants salivated more during the pizza exposure relative to pre-exposure (means (SDs) = 0.42 (0.26) and 0.29 (0.18) g, respectively). There was no main effect of condition order and no evidence for an expectancy by time by order interaction (all ps P .197). Self-reported measures Self-reported hunger (100-mm VAS) was analysed using a 2  2  2 mixed analysis of variance (ANOVA). The within-subjects factors were expectancy condition (1. EE and 2. NEE) and time (1. Pre-exposure and 2. During exposure). The between-subjects factor was condition order (1. EE first, NEE second or 2. NEE first, EE second). There was no main effect of expectancy and no evidence for the critical expectancy by time interaction (all ps P .198). There was a main effect of time, F(1, 38) = 18.33, p < .001, indicating that, overall, participants reported higher levels of hunger during the pizza exposure relative to pre-exposure (means (SDs) = 73.19 (18.10) and 67.95 (17.84) mm, respectively). There was no main effect of condition order and no evidence for an expectancy by time by order interaction (all ps P .15). Ratings of pizza pleasantness, desire to eat the pizza, excitedness, and alertness were analysed using separate mixed ANOVAs. In each analysis, the within-subjects factor was expectancy condition (1. EE and 2. NEE) and the between-subjects factor was condition order (1. EE first, NEE second or 2. NEE first, EE second). There were no main effects of expectancy on pleasantness, desire to eat, or alertness (all ps P .509). There was, however, a main effect of expectancy on excitedness, F(1, 38) = 5.26, p = .027, whereby participants rated higher levels of excitement when they expected to eat the pizza relative to when they did not expect to eat it (Fig. 1). There was no main effect of condition order and no expectancy by order interaction for excitedness or alertness. However, there was an expectancy by order interaction for both pleasantness, F(1, 38) = 16.75, p < .001, and desire to eat, F(1, 38) = 7.13, p = .01. These interactions were investigated using paired t-tests. For participants in the condition order 1 group (i.e., EE first, NEE second), pleasantness and desire to eat ratings were higher in the NEE condition relative to the EE condition (means (SDs); pleasantness, 82.45 (11.14) and 76.23 (16.71) mm, respectively, t(19) = 2.62, p = .017; desire to eat, 78.90 (16.25) and 72.98 (17.47) mm, respectively, t(19) = 1.87, p = .08). However, the converse was true for participants in the condition order 2 group (i.e., NEE first, EE second) where pleasantness and desire to eat ratings were higher in the EE condition relative to the NEE condition (means (SDs); pleasantness, 74.80 (16.64) and 66.15 (16.12) mm, respectively, t(19) = 3.15, p = .005; desire to eat, 69.48 (25.83) and 63.65 (20.03) mm, respectively, t(19) = 1.91, p = .07). These results thus indicate a general effect for pleasantness and desire to eat ratings to be higher during the second cue exposure, regardless of expectancy condition. Attentional bias Gaze dwell time (ms) was analysed using a 3 (probability: 100%, 50% and 0%)  2 (picture pair: pizza/neutral and alcohol/neutral)  2 (picture type: pizza/alcohol and neutral) repeated measures ANOVA. There was no main effect of probability and no indication of significant interactions between probability, picture pair and picture type (all ps > .35). There was a main effect of picture pair, F(1, 39) = 30.21, p < .001, which indicated that, overall,

157

C.A. Hardman et al. / Appetite 76 (2014) 153–160

90

Eating expectancy

No eating expectancy

80

VAS rating (0-100 mm)

70

* 60 50 40 30 20 10 0 Pleasantness

Desire to eat

Excitedness

Alertness

alcohol pictures. However, this bias was not moderated by the expectancy of receiving pizza points. Gaze direction bias was analysed using a 3 (probability: 100%, 50% and 0%)  2 (picture pair: pizza/neutral and alcohol/neutral) repeated measures ANOVA. There was a main effect of probability, F(2, 78) = 4.41, p = .015 (Fig. 3). However, there was no evidence for an interaction between probability and picture pair, indicating that probability had similar effects on responses to both pizza and alcohol pictures (i.e., a general effect). As shown in Fig. 3, there was a greater direction bias towards the pizza and alcohol pictures on 100% probability trials relative to 0% probability trials, t(39) = 2.81, p = .008, and on 50% trials relative to 0% trials, t(39) = 2.40, p = .02, but there was no difference between the 100% and 50% trials (p = .36). One-sample t tests indicated that none of the gaze direction bias scores were significantly different from the criterion level of 50% (ps > .08) (which would indicate a higher proportion of first fixations directed towards the pizza or alcohol pictures relative to the neutral pictures). Thus there was little evidence for an overall attentional bias for reward pictures relative to neutral pictures on gaze direction.

Fig. 1. Mean 100-mm visual analogue scale (VAS) ratings during pizza cue exposure in the eating expectancy (EE) condition and the no eating expectancy (NEE) condition. Values are mean ± SEM. Significant difference between EE and NEE conditions, p = .027.

Moderating factors

participants maintained their gaze on pizza/neutral pictures for longer than alcohol/neutral pictures (Means (SDs) = 286.29 (77.14) and 257.62 (76.44) ms, respectively). There was also evidence of a picture pair by picture type interaction, F(1, 39) = 3.79, p = .059. This interaction is shown in Fig. 2. Paired samples t-tests indicated that participants maintained their gaze on pizza pictures for longer than on neutral pictures, t (39) = 2.41, p = .021. There was no difference between alcohol and neutral pictures, t(39) = 0.75, p = .46 (Fig. 2). This indicates an overall attentional bias for pizza pictures (relative to neutral pictures), but not for

To examine potential moderating effects of habitual pizza consumption, the above analyses were re-run with pizza frequency group (i.e., more frequent consumers vs. less frequent consumers) entered as a between-subjects factor. We found no evidence for an interaction between expectancy and pizza frequency group for salivation, attentional bias (dwell time and direction bias), self-reported hunger, desire-to-eat or alertness (ps > .1). However, for both self-reported pizza pleasantness and excitedness, there was an interaction between expectancy and pizza frequency group (F(1, 38) = 3.96, p = .05, and F(1, 38) = 4.98, p = .03, respectively). Further exploration of the data indicated higher levels of pizza pleasantness and excitedness in the EE condition compared to the NE condition only in less frequent consumers (t(17) = 2.08,

320

Pizza/Alcohol

Neutral

60

* 55

280

Direction bias (%)

Mean dwell time (ms)

300

260

240

* *

50

45

40

35

220

30

200 Pizza

Alcohol

Picture pair Fig. 2. Mean dwell time (ms) on pizza and alcohol pictures relative to neutral pictures averaged across probability trials. Values are mean ± SEM. Significantly different from respective neutral picture, p = .02.

Pizza 100%

Pizza 50%

Pizza 0%

Trial type Fig. 3. Mean gaze direction bias averaged across pizza and alcohol pictures as a function of perceived probability of receiving pizza. Values are mean ± SEM.  Significantly different from 0% trials, p 6 .02.

158

C.A. Hardman et al. / Appetite 76 (2014) 153–160

Table 2 Pearson correlation coefficients between absolute values and ad libitum pizza intake. Absolute values were computed by averaging responses across EE and NE conditions (or across 100%, 50% and 0% trials for attentional measures). Variable

1.

2.

1. Pizza intake (kcal) 2. Salivation (g) 3. Hunger (mm)

750.07 (360.11) .09 .44*

.42(.26) .00

4. Pleasantness (mm)

.08

.11

73.19 (18.10) .45*

5. Desire to eat (mm)

.39*

.05

.82*

.13

*

5.

74.91 (14.73) .57*

6.

.47

.15

.23

.20

54.24 (24.57) .43*

.13

.11

.25

.22

.20

.03

.00

.29

7. Alertness (mm)

.01

.05

8. Pizza dwell time bias (ms)

.02 .22

4.

71.25 (19.35) .46*

6. Excitedness (mm)

9. Pizza direction bias (%)

3.

.43

*

.09

.07

.02

7.

65.75 (17.12) .17 .19

8.

23.86 (62.63) .41*

9.

51.06 (8.40)

Note. Off-diagonal shows correlation coefficients (r); Diagonal shows means (standard deviations in parentheses). Significant correlations (p < .05).

*

p = .05, and t(17) = 3.07, p = .007, respectively). There was no effect of eating expectancy in more frequent consumers (t(21) = .84, p = .41, and t(21) = .31, p = .76, respectively). Post hoc, we also examined whether sensitivity to the expectancy manipulation was moderated by individual differences in BMI and overall desire for pizza. Median splits were conducted on BMI (high BMI group P 22.57 kg/m2) and desire to eat the pizza (averaged across the EE and NE conditions; high desire group P 77.25 mm). The above analyses were re-run first with BMI group (i.e., high vs. low) and then with desire group (high vs. low) as a between-subjects factor. Across all the dependent variables, there was little evidence for interactions between expectancy and BMI (ps > .07) or between expectancy and desire (ps > .06). Associations with pizza intake In order to determine whether increased responses in the expectancy conditions relative to the no-expectancy conditions predicted pizza intake (our secondary hypothesis), we computed difference scores for all outcome measures. For salivation and the self-report measures, difference scores were computed by subtracting the value recorded during the NEE condition from that recorded during the EE condition. Difference scores were also computed for the attentional measures. For pizza gaze dwell time, we first calculated bias scores by subtracting attention to the neutral pictures from attention to the respective pizza pictures within probability trials (a positive score indicates an attentional bias toward pizza pictures). We then computed a difference score by subtracting the bias score on 0% trials from the bias score on 100% trials. Similarly, for pizza direction bias, the difference score was computed by subtracting the mean values recorded during the 0% trials from those recorded during the 100% trials. In all cases, a positive difference score indicated an increased response in the EE condition (or 100% trials) relative to the NEE condition (or 0% trials). Each difference score was then correlated with pizza intake (kcal). There was no evidence that change in any of the variables was associated with pizza intake (all ps P .08). Given the general lack of difference between expectancy conditions, we also examined correlation coefficients between pizza intake and absolute responses for all outcome measures. Absolute values were computed by averaging responses across EE and NE conditions for salivation and the self-report measures, and across

100%, 50% and 0% trials for the attentional measures. As shown in Table 2, pizza intake correlated positively with absolute levels of hunger and desire to eat pizza. Salivation did not correlate with any of the other variables. Pizza dwell time bias correlated positively with pizza direction bias but neither measure correlated significantly with any other variables. Discussion The aim of the current study was to examine the effects of expectancy information about the imminent receipt of pizza on physiological and subjective responses to pizza cues and attentional bias to those cues. Contrary to our primary hypothesis, we found a general lack of effect of expectancy on salivation, self-reported measures and our primary measure of attentional bias (total gaze dwell time). We did, however, find evidence that expectancy influenced the early orientation of attention (gaze direction bias) towards reward pictures. Furthermore, participants reported higher levels of arousal (excitedness) when they expected to imminently consume pizza. There was no evidence that a heightened response to expectancy information predicted subsequent pizza intake. When examining absolute responses (i.e., averaged across expectancy conditions), we found that pizza intake was positively associated with self-reported hunger and desire to eat pizza. However, we found an overall lack of correspondence between the self-report measures, salivation and the attentional measures. The generally null results in our study are at odds with findings from the addiction literature (Carter & Tiffany, 2001; Field et al., 2011; Jones et al., 2012), but are consistent with previous food studies (Higgs, 2007; Werthmann et al., 2013). The lack of expectancy effects in the study by Werthmann et al. (2013) might have been due to the time delay between the delivery of the expectancy information and the opportunity to consume the foods. For this reason, in the current study it was deemed critical to manipulate expectancy on an immediate, trial-by-trial basis to more closely mirror studies from the addiction literature. In the relevant cue exposure condition, participants expected to be able to eat the pizza imminently. We acknowledge, however, that in future studies it would be useful to conduct a believability check to determine the extent to which participants believed that they would be able to eat the pizza (or, in the no-eating expectancy condition, that they would not be able to eat it). In the attentional bias task,

C.A. Hardman et al. / Appetite 76 (2014) 153–160

participants received pizza points at the end of relevant trials and were explicitly informed that these accumulated points would be converted to actual quantities of pizza. This is entirely consistent with the version of the task that was used by Jones et al. (2012) in which clear effects of expectancy on attentional bias for food (chocolate) were observed and participants indicated high levels of believability that points would be converted to actual quantities of chocolate at the end of the study. In addition, it has been shown that drug-paired stimuli control drug seeking via a representation or expectation of the drug rather than through a direct stimulus– response association (Hogarth, Dickinson, Wright, Kouvaraki, & Duka, 2007). On this basis, the provision of probability information as opposed to actual quantities of pizza should have been sufficient to trigger an appropriate representation of food reward in our study. A potential explanation for the divergent findings relates to differences in the deprivation state of the participants. In the study by Jones et al. (2012), participants were not given explicit instructions to abstain from eating prior to the experiment. In our study, as in previous food cue reactivity studies (Fedoroff et al., 1997; Ferriday & Brunstrom, 2011; Nederkoorn et al., 2000), our participants were instructed to abstain from eating prior to attending their session. This is important because the magnitude of cued responses, such as salivation, are considerably lessened when participants are satiated compared to when they are hungry (Brunstrom et al., 2004). On average, our participants had abstained from eating for 4 h and accordingly they reported high levels of hunger (mean absolute hunger = 73.19 (SD = 18.10) on a 100 mm VAS). This suggests that the motivational value of food was already high and therefore exposure to food cues may have triggered appetitive responses regardless of the expectancy manipulation. Support for this interpretation comes from the significant increases in salivation and hunger ratings that were seen in response to pizza exposure, indicating that participants were highly reactive to the cue. In addition, participants had longer gaze dwell times on pizza pictures relative to neutral pictures regardless of the expectancy manipulation, indicating an overall attentional bias in maintained attention to the pizza pictures. We suggest that this bias in maintained attention was strongly influenced by the deprivation state of the participants and the related incentive value of food, and that this masked any effects of the expectancy information. This prospect is supported by the study of Field et al. (2011) in which heavy drinkers showed a significantly longer dwell time on alcohol cues relative to neutral cues which was insensitive to the expectancy manipulation. Importantly, the incentive value of alcohol is higher in heavier compared to lighter drinkers (MacKillop et al., 2010). In another study, gaze duration on smoking-related cues was strongly influenced by changes in nicotine deprivation and craving, however the initial orientation of attention was not (Field et al., 2004). Werthmann et al. (2013) similarly found a greater initial gaze duration bias to chocolate cues in high trait chocolate cravers relative to non-cravers, but no difference on initial attentional orientation (direction bias). Notably, our study did find that participants directed their first gaze towards pizza pictures more often on 100% and 50% probability trials relative to 0%, trials, which indicates that reward expectancy affected only the early allocation of attention. Taken together with previous research, this provides tentative support for the possibility that early and later attentional processes might be differentially influenced by expectancy and motivational factors. We did find some evidence to suggest that sensitivity to the expectancy manipulation was moderated by participants’ habitual consumption of pizza. Less frequent consumers reported higher levels of pizza pleasantness and excitedness when they expected to imminently eat the cued pizza compared to when there was no eating expectancy; however, there was no effect of eating

159

expectancy on these responses in more frequent consumers. This would appear consistent with the findings of Field et al. (2011) that only relatively light drinkers were sensitive to an alcohol expectancy manipulation, whereas heavier drinkers showed an overall attentional bias to alcohol cues regardless of expectancy. The lack of expectancy effects in more frequent consumers might reflect a ceiling effect in that the incentive value of the drug or food may already be very high. Alternatively, conditioned responses to drug or food cues may be increasingly automatized in more frequent consumers thus rendering them insensitive to a cognitive manipulation (see Field et al., 2011 for a broader discussion of this issue). However, we note that the moderating effect of pizza frequency was expressed on only two variables and there were no effects on salivation, attentional bias or the other self-reported measures. Hence this finding should be treated with caution. In order to further investigate this issue and to enhance consistency with the addiction literature, future studies should examine the effect of expectancy in participants who are more frequent (i.e., daily) consumers of pizza. Another noteworthy finding was that the effect of pizza expectancy on gaze direction bias was not specific to pizza pictures. That is, on trials where the probability of receiving a pizza point was 100% and 50%, participants more often directed their first gaze towards pizza or alcohol picture relative to 0% probability trials. A similar general transfer effect, albeit only on the maintenance of attention, was reported by Jones et al. (2012) who proposed that these non-specific effects may reflect a generalised increase in arousal as a result of reward expectancy. Consistent with this idea, our participants reported higher levels of excitedness during cue exposure when they imminently expected to eat the pizza compared to when there was no eating expectancy. However, this change in arousal was not correlated with attentional bias though this could be because these responses were measured at different time points within our study. More generally, there is considerable theoretical debate around whether reward-paired cues elicit general or specific motivational effects (for a broader discussion of these issues see Jones et al., 2012). Our findings also address questions about the effects of reward uncertainty on attentional processing in that gaze direction bias towards reward pictures was higher on 50% (i.e., uncertain) probability trials relative to 0% trials. Attention may resolve predictive uncertainty and, according to this account, attention to a conditioned stimulus should increase with the uncertainty with which the CS predicts the US (Pearce & Hall, 1980). While partly in line with our findings, this theory would also predict a higher attentional bias on 50% probability trials relative to 100% trials whereas we found a linear increase in bias with increasing expectancy. We also note that findings from existing studies are somewhat conflicting. For example, Carter and Tiffany (2001) found a linear increase in tobacco craving as cigarette availability increased (using 0%, 50% and 100% availability trials). In contrast, Field et al. (2011) did not find greater maintained attention to alcohol stimuli under conditions of uncertainty. Future studies should consider whether the effects of expectancy on cue reactivity are moderated by individual differences, such as trait impulsivity. A recent contextual appetitive conditioning study indicated that highly impulsive participants showed increased intake in the CS+ (i.e., the food-associated environment) relative to the CS , whereas low-impulsive participants did not (van den Akker et al., 2013). Impulsive individuals are biased towards immediate gratification and also show poor inhibitory control (Olmstead, 2006) and might therefore be particularly susceptible to direct manipulations of reward expectancy. Accordingly, there is evidence that expectancy influenced alcohol cue reactivity most strongly in impulsive individuals (Papachristou, Nederkoorn, Corstjens, & Jansen, 2012). Given that food stimuli

160

C.A. Hardman et al. / Appetite 76 (2014) 153–160

are highly abundant in Westernised environments, it is important to better understand the psychological mechanisms and individual differences that underpin motivational responses to these cues. Chronic dieters would seem another target group for further research given their history of habitually restricting food intake which might mean that the predictive ability of food cues is weakened. In conclusion, we found little evidence that eating expectancy influenced food cue reactivity in hungry participants. While this might appear at odds with theoretical accounts and related studies in the addiction literature, we would urge future researchers to consider the moderating effects of deprivation status because expectancy might exert a stronger influence on physiological and cognitive responses in participants who are non-food deprived. References Bolles, R. C. (1972). Reinforcement, expectancy, and learning. Psychological Review, 79(5), 394–409. Brunstrom, J. M., Yates, H. M., & Witcomb, G. L. (2004). Dietary restraint and heightened reactivity to food. Physiology & Behavior, 81(1), 85–90. Carter, B. L., & Tiffany, S. T. (2001). The cue-availability paradigm. The effects of cigarette availability on cue reactivity in smokers. Experimental and Clinical Psychopharmacology, 9(2), 183–190. Fedoroff, I. D. C., Polivy, J., & Herman, C. P. (1997). The effect of pre-exposure to food cues on the eating behavior of restrained and unrestrained eaters. Appetite, 28(1), 33–47. Ferriday, D., & Brunstrom, J. M. (2008). How does food-cue exposure lead to larger meal sizes? British Journal of Nutrition, 100(6), 1325–1332. Ferriday, D., & Brunstrom, J. M. (2011). ‘I just can’t help myself’. Effects of food-cue exposure in overweight and lean individuals. International Journal of Obesity, 35(1), 142–149. Field, M., & Cox, W. M. (2008). Attentional bias in addictive behaviors. A review of its development, causes, and consequences. Drug and Alcohol Dependence, 97(1– 2), 1–20. Field, M., & Duka, T. (2001). Smoking expectancy mediates the conditioned responses to arbitrary smoking cues. Behavioral Pharmacology, 12(3), 183–194. Field, M., & Duka, T. (2004). Cue reactivity in smokers. The effects of perceived cigarette availability and gender. Pharmacology Biochemistry and Behavior, 78(3), 647–652. Field, M., Hogarth, L., Bleasdale, D., Wright, P., Fernie, G., & Christiansen, P. (2011). Alcohol expectancy moderates attentional bias for alcohol cues in light drinkers. Addiction, 106(6), 1097–1103. Field, M., Mogg, K., & Bradley, B. P. (2004). Eye movements to smoking-related cues. Effects of nicotine deprivation. Psychopharmacology, 173(1–2), 116–123. Higgs, S. (2007). Impairment of cognitive performance in dietary restrained women when imagining eating is not affected by anticipated consumption. Eating Behaviors, 8(2), 157–161.

Hogarth, L., Dickinson, A., Wright, A., Kouvaraki, M., & Duka, T. (2007). The role of drug expectancy in the control of human drug seeking. Journal of Experimental Psychology. Animal Behavior Processes, 33(4), 484–496. Hogarth, L., & Duka, T. (2006). Human nicotine conditioning requires explicit contingency knowledge. Is addictive behaviour cognitively mediated? Psychopharmacology, 184, 553–566. Jedras, P., Jones, A., & Field, M. (2014). The role of anticipation in drug addiction and reward. Neuroscience and Neuroeconomics, 3, 1–10. Jones, A., Hogarth, L., Christiansen, P., Rose, A. K., Martinovic, J., & Field, M. (2012). Reward expectancy promotes generalized increases in attentional bias for rewarding stimuli. Quarterly Journal of Experimental Psychology, 65(12), 2333–2342. MacKillop, J., Miranda, R., Jr., Monti, P. M., Ray, L. A., Murphy, J. G., Rohsenow, D. J., et al. (2010). Alcohol demand, delayed reward discounting, and craving in relation to drinking and alcohol use disorders. Journal of Abnormal Psychology, 119(1), 106–114. Mogg, K., Bradley, B. P., Field, M., & De Houwer, J. (2003). Eye movements to smoking-related pictures in smokers. Relationship between attentional biases and implicit and explicit measures of stimulus valence. Addiction, 98(6), 825–836. Nederkoorn, C., Smulders, F. T. Y., & Jansen, A. (2000). Cephalic phase responses, craving and food intake in normal subjects. Appetite, 35(1), 45–55. Olmstead, M. C. (2006). Animal models of drug addiction. Where do we go from here? Quarterly Journal of Experimental Psychology, 59(4), 625–653. Papachristou, H., Nederkoorn, C., Corstjens, J., & Jansen, A. (2012). The role of impulsivity and perceived availability on cue-elicited craving for alcohol in social drinkers. Psychopharmacology, 224(1), 145–153. Pearce, J. M., & Hall, G. (1980). A model for Pavlovian learning. Variations in the effectiveness of conditioned but not of unconditioned stimuli. Psychological Review, 87(6), 532–552. Rogers, P. J., & Hill, A. J. (1989). Breakdown of dietary restraint following mere exposure to food stimuli. Interrelationships between restraint, hunger, salivation, and food intake. Addictive Behaviors, 14(4), 387–397. Smeets, E., Roefs, A., & Jansen, A. (2009). Experimentally induced chocolate craving leads to an attentional bias in increased distraction but not in speeded detection. Appetite, 53(3), 370–375. Stunkard, A. J., & Messick, S. (1985). The three-factor eating questionnaire to measure dietary restraint, disinhibtion and hunger. Journal of Psychosomatic Research, 29, 71–83. van den Akker, K., Jansen, A., Frentz, F., & Havermans, R. C. (2013). Impulsivity makes more susceptible to overeating after contextual appetitive conditioning. Appetite, 70, 73–80. Weingarten, H. P. (1985). Stimulus control of eating. Implications for a two-factor theory of hunger. Appetite, 6(4), 387–401. Werthmann, J., Roefs, A., Nederkoorn, C., & Jansen, A. (2013). Desire lies in the eyes. Attention bias for chocolate is related to craving and self-endorsed eating permission. Appetite, 70, 81–89. Werthmann, J., Roefs, A., Nederkoorn, C., Mogg, K., Bradley, B. P., & Jansen, A. (2011). Can(not) take my eyes off it. Attention bias for food in overweight participants. Health Psychology, 30(5), 561–569.