Introduction

Drug-related cues could induce subjective craving (i.e. appetitive motivation of drug seeking and drug taking) and drug abuse behaviours in drug users (Baker et al. 1987; Robinson and Berridge 1993, 2003; Stewart et al. 1984; Wise 1988), despite that they are fully aware of the harms (Wiers and Stacy 2006b). The incentive sensitization theory of addiction (Robinson and Berridge 1993, 2003) points out that chronic exposure to drugs causes sensitization of dopamine activity in the nucleus accumbens and associated structures within the mesolimbic dopamine system. It abnormally increases the incentive value of drugs and endues drug-related cues with conditioned high motivational salience, hence drug-related stimuli become highly attractive or attention grabbing and thus guide behaviours toward the incentive (drug) (Robinson and Berridge 1993) or related environments (Rooke et al. 2008). To describe how drug-related cues lead to addictive behaviours, the dual process model (e.g. Wiers and Stacy 2006a) proposed two semi-independent but collaborating systems: a relatively automatic “impulsive” system, in which stimuli are first evaluated for their motivational significance and subsequently trigger corresponding motivational orientation (e.g. approach or avoidance), is mediated by a slower “reflective” system, which regulates the impulsive processes when the reflective system has both sufficient resources and motivations (Grenard et al. 2008; Hofmann et al. 2008; Thush et al. 2008; Wiers et al. 2007). There is a growing body of evidence in drug users suggesting that drug-related cues can automatically capture attention (e.g. Bradley et al. 2003; Ehrman et al. 2002; Franken et al. 2000; Hogarth et al. 2003; Lubman et al. 2000) and attract approach responses (e.g. Bradley et al. 2004; Field et al. 2006, 2008; Mogg et al. 2003, 2005; Ostafin and Palfai 2006; Palfai and Ostafin 2003; Wiers et al. 2009). Combining with weak executive control, it may eventually lead to drug abuse behaviours (e.g. Grenard et al. 2008; Thush et al. 2008).

Recently, researchers started to put more emphasis on drug users' behavioural approach tendencies elicited by drug-related cues. The stimulus–response compatibility task (SRC; De Houwer et al. 2001) has been deployed in diverse substance-dependent populations and found consistent results that, compared to the control groups, nicotine dependents (Bradley et al. 2004; Mogg et al. 2003, 2005), cannabis-dependents (Field et al. 2006), and heavy social drinkers (alcohol; Field et al. 2008) all made significantly faster approach responses than avoidance responses towards respective drug-related stimuli. In these studies, drug-related stimuli (e.g. a smoking cigarette) and matched neutral stimuli (e.g. a lipstick) were displayed. Instruction required participants to complete two reversed assignments: (1) move a manikin (a carton figure) towards the drug-related stimuli and away from the neutral stimuli and (2) move the manikin away from the drug-related stimuli and towards the neutral stimuli. By comparing the participants' response speeds in different assignments, researchers were able to determine one's tendency to approach different categories of stimuli (drug related versus neutral). However, SRC task measures the given tendency based on symbolic rather than real movements (Field et al. 2008). It remains doubtful whether it can fully illustrate the actual behavioural tendencies in substance users.

In contrast, Wiers and colleagues (2009, 2010) have applied an Approach–Avoidance Task (AAT; Rinck and Becker 2007) based on real movements to directly measure alcoholics' behavioural approach tendencies. In the paradigm, the participants were instructed to pull (arm flexion) or push (arm extension) a joystick in accordance with the picture orientation (portrait versus landscape); meanwhile, the picture contents were manipulated as drug related or not. When the joystick was pulled, the stimulus picture grew bigger on the screen, and when it was pushed, the picture grew smaller; such a “zooming feature” was utilised to enhance the sensation of approach or avoidance (Neumann and Strack 2000; Wiers et al. 2009). By comparing the participants' response times in pulling versus pushing the joystick in relation to picture contents, one's automatic behavioural tendency towards certain stimulus category could be therefore determined. Using this task, Wiers et al. (2009) found that male heavy drinkers with a g-allele in the human μ opioid receptor (OPRM1) gene had an automatic approach tendency towards alcohol and other appetitive stimuli. More importantly, they employed the same task to implicitly retrain alcoholics' impulsive process of alcohol-related cues and as a result improved hazardous drinkers' drinking behaviours (Wiers et al. 2010, 2011).

Heroin is one of the major illegal drugs abused worldwide and has created serious public health problem in many countries. Meanwhile, there is a lack of published studies directly concerning heroin abusers' behavioural approach tendencies towards drug-related stimuli. The present study intended to make contribution in this part of the literature.

A Pull/Push Task (PPT), which originated from social cognition research (Chen and Bargh 1999), was adopted in the present study. Similar with AAT's rationale, it assesses the participants' behavioural approach tendencies based on real movements because of the close relation between overt arm movements and emotional valence of stimuli. People tend to respond to aversive stimuli more efficiently with a “push” movement while respond to pleasant stimuli more efficiently with a “pull” movement (e.g. Chen and Bargh 1999; Marsh et al. 2005). In addition, arm movements can conversely influence evaluative reactions to the stimuli. Neutral stimuli presented during arm flexion could be subsequently evaluated more positive than those presented during arm extension (e.g. Cacioppo et al. 1993). Hence an association between arm movements (arm flexion versus arm extension) and approach/avoidance motivational orientations could be posited. In the PPT task, the participants pull or push a lever according to the contents of stimulus pictures (e.g. heroin related or neutral). Therefore the responding differences in regard to different contents of stimulus pictures in pulling or pushing the lever indicate corresponding behavioural approach or avoidance tendencies.

Methods

Participants

Twenty-two abstinent heroin abusers (AH participants; aged between 21 and 37 years, mean ± SD = 31.45 ± 4.42 years, all males) and 20 healthy normal controls (NC participants; aged between 20 and 25 years, mean ± SD = 22.20 ± 1.74 years, all males) participated in the present study. All were right-handed Chinese with normal vision.

The AH participants were inpatients from a legal rehabilitation centre in Yunnan Province, China; all met the Diagnostic Statistical Manual IV criteria for heroin dependence, with a history of exclusive illegal drug abuse of heroin.Footnote 1 They participated in the study voluntarily. Everyone at least finished primary education and was able to understand and react to the experiment materials correctly. At the time of experiment, they had been abstinent from heroin for 1–14 (9 ± 3.88) months. The duration of the abstinence was known based on the length of their stay in the rehabilitation centre as inpatients had no access to drugs and regular urine tests ensured abstinence. None of them were taking opioid substitution treatment or medication affecting the central nervous system. Neither did anyone report history of diagnosed mental diseases or brain injuries.

On the other hand, 20 students (undergraduates and postgraduates) were recruited as an NC group in Zhejiang University. Each of them received ¥20 for 1 h of psychological experiment session containing the present experiment. The NC group was screened, by self-report, to be free of any history of drug dependence, mental disease or brain injury.

This study was approved by the Research Ethics Board of the Department of Psychology and Behavioural Sciences at Zhejiang University and the legal rehabilitation centre where the AH participants in the present study came from. Prior to the experiment, both the AH and NC participants had been informed of the scientific purpose of the study as well as their right to withdraw at any time point. Every participant had signed on a printed consent form.

Materials and apparatus

Two categories of stimuli were used in the experiment: eight heroin-related pictures depicting items that are associated with heroin abuse such as tinfoil, white powder, syringe, etc. and eight complexity-matched neutral pictures (revised from Franken et al. 2000) depicting diverse vehicles (e.g. a SUV) (see Fig. 1 for examples). Their relatedness to heroin abuse were determined in a pilot study by 35 abstinent heroin abusers who were not included in the present study, rating each stimulus picture on a Likert scale from 1 (not at all heroin related) to 5 (extremely heroin related). The eight heroin-related pictures had an average score of 4.49 ± 0.24, whereas the eight neutral pictures had an average score of 1.27 ± 0.18.

Fig. 1
figure 1

The examples of heroin-related pictures (a) and neutral pictures (b)

Full size image

Procedure

Stimulus pictures were displayed on a Pentium IV portable computer with a 14-inch colour LED monitor. Participants were instructed to pull or push a 56-cm-long lever (which was connected to the computer via a switchbox) by their right hand in response to the stimulus categories. The task consisted of two blocks. In one block participants were asked to pull the lever when they saw a heroin-related picture but to push the lever when they saw a vehicle picture. The stimulus–reaction relationship was reversed in the other block and the order of the presentation of the two blocks (with a 30-minute break between) was counterbalanced within each participant group.Footnote 2 Prior to each block, participants familiarised themselves with the response direction by using black/white colours as stimuli in 20 practise trials. Each block consisted of 32 test trials (with each stimulus picture presented twice).

Each trial began with a fixation at the centre of the screen for 500 ms followed by a stimulus picture (subtended a visual angle of 19.7º × 14.8º) on a grey background. The programme automatically proceeded to the next trial if a response was detected (i.e. the participant moved the lever a sufficient distance to trigger a signal) or no response for more than 2,000 ms. Response time (RT) and direction were recorded for each trial.

Results

As error rates were low (<3%), only RT data were subjected to further analyses. Incorrect responses, non responses and RTs shorter than 300 ms or longer than 3 standard deviations (SD) above the mean were discarded based on each participant (none of the participants had removed trials exceeding 20% of the total). The descriptive statistics are shown in Table 1.

Table 1 Mean response time in milliseconds to approach and avoid heroin-related and neutral pictures for AH and NC participants
Full size table

A mixed 2 × 2 × 2 Analysis of Variance (ANOVA) was performed on the original RT data with group (AH versus NC) as a between-subjects factor, stimulus category (heroin related versus neutral) and response direction (approach versus avoidance) as within-subjects factors. Results can be seen in Table 2.

Table 2 ANOVA results for the original RT data
Full size table

In keeping with previous researches on relevant tests (e.g. Wiers et al. 2009, 2010), participants' behavioural tendencies for approaching and avoiding heroin-related stimuli were calculated as the mean RT for responding to neutral pictures minus the mean RT for responding to heroin-related pictures (RTneutral − RTheroin). The approach score was calculated based on pulling responses while the avoidance score was calculated based on pushing responses. Thus a positive score always indicates a tendency to approach or avoid heroin-related stimuli rather than neutral stimuli, respectively; a negative score, in contrast, indicates the opposite.

The means of the approach and avoidance scores for AH and NC participants are illustrated in Fig. 2. Independent t tests revealed that AH participants had significantly higher approach scores (M ± SE = 86.51 ± 26.10 ms) than NC participants (M ± SE = −21.88 ± 8.43 ms; t(40) = 3.796, p < 0.001); whereas the difference of avoidance scores was marginally significant between the two groups (AH group: M ± SE = −69.56 ± 31.11 ms; NC group: M ± SE = −8.31 ± 9.91 ms; t(40) = −1.802, p = 0.079).

Fig. 2
figure 2

Approach and avoidance scores (RTneutral – RTheroin) for AH and NC participants. Note ***p = 0.000; †p = 0.079

Full size image

Discussion

The results in the present study indicated that, compared to NC participants, AH participants showed a significantly heightened behavioural tendency to approach, but a (marginally) significantly blunted tendency to avoid, heroin-related stimuli. In other words, AH individuals exhibited a special bias to behaviourally approach heroin-related cues as well as a bias to resist avoiding them.

A number of studies that employed the SRC task have demonstrated that users of a variety of substances, such as alcohol (Field et al. 2008), tobacco (Bradley et al. 2004; Mogg et al. 2003) and cannabis (Field et al. 2006), were generally faster to approach than avoid respective drug-related stimuli via moving a manikin, while no such difference was found in non-users. Our present finding was consistent with those results, notwithstanding these studies did not calculate separate approach and avoidance indices as we did. Noteworthy, PPT used in the present study improved the SRC task by substituting real movements for symbolic movements to reflect one's behavioural approach/avoidance tendency, adding stronger evidence to support the assertion that drug-related cues can elicit overt behavioural approach tendency in substance users (Field et al. 2008).

The AAT used in Weirs et al. (2009, 2010) studies was grounded on the same approach/avoidance response (i.e. arm flexion versus arm extension) to the PPT. However, AAT differs from PPT in that AAT instructs participants to respond in accordance with an irrelevant feature (e.g. picture orientation) rather than the drug relatedness of stimuli. In this way, AAT implicitly measures automatic tendencies behind participants' conscious reactions. In contrast, PPT that requires participants to explicitly categorise stimulus pictures as drug related or neutral may involve certain conscious deliberation into the process. However, the present study yielded coherent result with Wiers et al. (2009) finding, suggesting that drug users' behavioural approach tendency for drug-related stimuli may not be profoundly influenced by conscious control. Besides, AAT uses the zooming feature to solidify the respective sensation of approach and avoidance, which is not included in the PPT. Although the relation between arm movements and evaluative valence has been strongly supported (i.e. arm flexion links with positive valence and arm extension links with negative valence; see review of Neumann et al. 2003), there could be alternative interpretations of the arm movements. For instance, arm flexion could be interpreted as either approach (i.e. pulling the object toward oneself) or avoidance behaviour (i.e. withdrawing the hand from the object); similarly, arm extension could also be interpreted as either avoidance (i.e. pushing the object away) or approach behaviour (i.e. extending the hand to grasp the object). With the zooming feature, participants' resistance against the ambiguity of approach versus avoidance behaviours can be enhanced (Rinck and Becker 2007; Wiers et al. 2009). In this point, the lack of the zooming feature was a potential methodology limitation of the PPT used in the present study. However, on the other hand, in a control experiment using fix-sized pictures as stimuli (i.e. no zooming feature), Rinck and Becker (2007; see at the footnote 3) replicated the result of the experiment deploying the zooming feature when the instruction was body reference (e.g. “pulling/pushing the joystick towards/away from yourself”, which was the same instruction used in the present study); whereas the effect vanished when the instruction was stimulus reference (e.g. “pulling/pushing the joystick away from/towards the picture”). Hence we might expect that the influence of the lack of the zooming feature on the current result was not significant.

The present study separated AH participants' approach and avoidance tendencies and yielded dissociative results. AH individuals had a much stronger bias to approach heroin-related stimuli than NC individuals did; meanwhile, they were more resistant to avoid the same stimuli. Whereas notably, the latter bias was less profound compared to the former one. It was an extension of the previous findings, which only inspected relative behavioural approach tendencies (i.e. the RT difference between approach responses and avoidance responses) (Field et al. 2006). According to incentive motivational models, chronic drug use turns the drug or the drug-related cue into a powerful incentive (Robinson and Berridge 1993). In line, it is easy to understand drug users' conditioned approach response toward the stimuli that are considered relevant to the drug intake ritual (Tomie 1996). In addition, the high incentive value may also cause those stimuli to become spontaneously more irresistible (i.e. difficult to avoid), whereas the effect is relatively indirect and less profound. However, at this point the notion remains rather speculative and awaits further validation. More research is encouraged to clarify this issue.

Although the study yielded interesting results, it should be considered in the context of methodological limitations. First of all, AH participants responded generally slower in all experimental conditions compared with NC participants (see Tables 1 and 2). Such an overall group difference may possibly covariate with the differences observed on behavioural tendencies. Secondly, only male abstinent heroin inpatients were included in the current study. It would be worthy trying to include more representative groups (e.g. female heroin abusers and current heroin abusers) in the future study and determine whether the current finding could be widely generalised. In addition, the severity of addiction to heroin (i.e. quantity and duration of heroin abuse) was not deliberately controlled in the present study, which might influence AH individuals' behavioural approach tendencies. Furthermore, the current AH sample did not distinguish intravenous-injecting heroin abusers from smoking/snorting abusers, who may have differed perception of heroin intake ritual. Probably, they may react to various drug-related stimuli (e.g. syringe versus tinfoil) differently. However, stimulus pictures used in the present study had been aforehand rated by an extra group of AH inpatients including both types of abusers. They did not rate the stimulus pictures differently. At any events, future studies may advance the research by emphasising on these variables and reveal whether heroin abusers' behavioural approach tendencies are mediated by means of drug intake and the severity of dependence.