ISSN: Volume 7 Number 2

Transcription

1 ISSN: Volume 7 Number 2 Winter 2013

2 The Analysis of Gambling Behavior (AGB) is a peer-reviewed, refereed publication that contains original general-interest and discipline-specific articles related to the scientific study of gambling behavior Executive Editor Jeffrey N. Weatherly University of North Dakota Associate Editor Andrew Brandt Ohio Wesleyan University Editorial Board Members Jennifer Austin Lewis Bizo Mark R. Dixon University of Glamorgan The University of Waikato Southern Illinois University Simon Dymond Gordon R. Foxall Charles A. Lyons Swansea University Cardiff University Eastern Oregon University Alyssa N. Wilson Benjamin N. Witts Bryan Roche St. Louis University University of Nevada, Reno National University of Ireland, Maynooth Content of the Analysis of Gambling Behavior AGB contains general-interest and discipline-specific articles related to the scientific study of gambling behavior. Articles appropriate for the journal include a) full-length research articles, b) research reports, c) clinical demonstrations, d) technical articles, and e) book reviews. Each category is detailed below along with submission guidelines. Research Articles: A manuscript of full length (approximately double-spaced pages) that may contain multiple experiments and that makes a substantial original contribution to the published literature. Research Reports: A manuscript of reduced length (approximately 10 double-spaced pages with a single table or figure) that may be less experimentally rigorous than a Research Article, a replication of or a failure to replicate a prior published article, or pilot data that demonstrate a clear relationship between independent and dependent variables. The Result and Discussion sections of research reports should be combined. Clinical Demonstrations: A manuscript of reduced length (no more than 8 double-spaced pages with a single table or figure) that lacks the rigor of a true experimental design, but demonstrates behavior change in one or more participants gambling behavior in a clinically relevant situation. The Result and Discussion sections of clinical demonstrations should be combined. Technical Article: A manuscript of either full or reduced length, depending on necessity, that describes either a new technology available that would be of interest to researchers or a task-analysis-type description of how to utilize existing technology for the conducting of research. Examples of appropriate topics may include altering of gambling equipment to allow for scientific study, how to use computer software to simulate gambling games, or interfacings neuroimaging devices to allow for the study of gambling behavior. Book Reviews: A review of a contemporary book related to gambling not more than 3 years after the publication date of the book. The review should be no more than 15 double-spaced pages in length.

3 ANALYSIS OF GAMBLING BEHAVIOR Contents Volume 7, Number 2, Winter 2013 ISSN: Research Article Hoon, A.E., & Dymond, S. Altering Preferences for Concurrently Available Simulated Slot Machines: Nonarbitrary Contextual Control Over Gambling Choice. 35 Research Report Weatherly, J.N. Comparing the Contingencies that Maintain Gambling Behavior in an Online Sample of Younger and Older Adults. 53 Literature Review Witts, B.N. Cumulative Frequencies of Behavior Analytic Journal Publications Related to Human Research on Gambling 59 33

4 34

5 Analysis of Gambling Behavior 2013, 7, Number 2 (Winter, 2013) Altering Preferences for Concurrently Available Simulated Slot Machines: Nonarbitrary Contextual Control Over Gambling Choice Alice E. Hoon & Simon Dymond Swansea University Structural characteristics of slot machines, such as color, have been implicated in the maintenance of problem gambling. Behavior-analytic research has demonstrated that preferences for identically programmed, concurrently available simulated slot machines can be brought under contextual control, which may provide a functional account of the control exerted by structural characteristics. Specifically, when participants were trained that the color yellow was a contextual cue for more than and the color blue was a contextual cue for less than, participants showed an increased preference for the yellow slot machine, despite both machines being identical in schedule and magnitude of reinforcement. The present experiments sought to replicate and extend these findings in several ways. First, we sought to overcome limitations of pretest/posttest designs by employing a nonconcurrent multiple baseline design, counterbalancing the contextual cues, and employing problem gamblers as participants. Experiments 1 and 2 found that slot machine preferences could be altered in accordance with contextual cues in problem gamblers, and Experiment 3 reported that these preferences could be reversed. All three experiments found that extended exposure to the payout contingencies of a slot machine may weaken the control exerted by the contextual cues. Keywords: Slot machine; Contextual control; Nonarbitrary; Problem gamblers; Experiment The structural characteristics of slot machines or electronic gaming machines have been implicated in the persistence and maintenance of problem gambling (Griffiths, 1990). The term structural characteristic encompasses many features including lights, colors, sounds, and bill payment options. Adding or removing some of these features has associated effects on persistence and levels of self-reported enjoyment in slot machine play (Loba, Stewart, Klein, & Blackburn, 2001; Sharpe, Walker, Coughlan, Enerson, & The present experiments were submitted as part of the requirement for Alice E. Hoon s doctoral thesis. Address all correspondence to: Alice E. Hoon College of Medicine Swansea University Singleton Park Swansea SA2 8PP, United Kingdom a.e.hoon@swansea.ac.uk Blaszczynski, 2005). For instance, Loba et al. (2001) reported that video lottery terminals that had fast reels and produced sounds were played for longer periods of time than those that had slower reels and no sound. Behavior-analytic gambling research on the effects of structural characteristics has sought to investigate how color may influence slot machine choice. Zlomke and Dixon (2006) conducted a study in which participants were presented with two concurrently available computer simulated slot machines identical in schedule and magnitude of reinforcement and differing only in color (one slot machine was yellow and the other was blue). The slot machines paid out according to a random ratio 0.5 schedule, such that each spin was independent of the last and a winning trial occurred on 50% of trials. Magnitude of reinforcement was held constant such that one 35

6 36 CONTEXTUAL CONTROL credit was required to play a slot machine, therefore one credit was lost on a losing spin, and one credit was won for a winning spin. Participants showed no particular preference for either slot machine, allocating approximately equal responding to both machines. Participants were then given a nonarbitrary relational training task in which the color yellow was established as a contextual cue for more than and the color blue was established as a contextual cue for less than. Following this training, participants were given a further 50 slot machine trials to play under identical conditions as before. It was found that participants allocated increased responding to the yellow slot machine, despite both slot machines being identical in payout probability (Zlomke & Dixon, 2006). Researchers have replicated and extended these findings (Fredheim, Otterson, & Arntzen, 2008; Hoon, Dymond, Jackson, & Dixon, 2007, 2008; Johnson & Dixon, 2009; Nastally, Dixon, & Jackson, 2010). For instance, Fredheim et al. (2008) used an identical procedure to Zlomke and Dixon (2006) with non-problem gamblers and found that only four out of twelve participants showed an increased preference for the more than slot machine at posttest. In a second experiment, the authors altered the way in which instructions were presented so that individuals who had not reached criterion responding for nonarbitrary training and testing within one hour had the instructions repeated to them. Secondly, a brief interview was conducted following the experiment to identify whether participants had attended to the contextual cue (the Color Group) or simply responded to the comparison stimuli independent of the contextual cue (the Number Group). Results were analysed in terms of whether participants were assigned to either the Color Group or the Number Group. It was found that in the Color Group, eight out of twelve participants showed increased preference for the more than (yellow) slot machine at posttest whereas in the Number Group only two participants out of six showed increased preference towards the more than slot machine at posttest (Fredheim et al., 2008). The original nonarbitrary relational training procedure used by Zlomke and Dixon (2006) has potential limitations, which may explain the finding of Fredheim et al. (2008). Specifically, during relational training, participants were presented with the contextual cue (a yellow or blue colored screen) followed by three comparison stimuli. For example, the screen appeared yellow and then $1, $5, and $10 notes were presented. This method has been criticised because, firstly, it is ambiguous due to there being two correct responses (both $5 and $10 are more than $1), and secondly, it may lead to the more than cue being established as a cue for opposite (Hoon et al., 2007, 2008). It is possible that these factors may partly explain why Fredheim et al. (2008) only observed the predicted effect in a minority of participants because the ambiguity of the training task resulted in insufficient contextual functions having been established, which was unlikely to influence posttest performance. In order to resolve these limitations, Hoon et al. (2007, 2008) employed a training task in which only two comparison stimuli were presented. This ensured there was only one correct response per trial. During baseline, participants did not show a particular preference for either slot machine; however, following the nonarbitrary relational training and testing phase, participants did show an increased preference for the yellow more than slot machine. These results demonstrate that a preference for a structural characteristic such as color may override the direct payout contingencies of a slot machine. While there are merits to the research already conducted on the influence of contextual cues in simulated slot machine gambling, there are some methodological limitations with the research designs that have been used.

7 ALICE E. HOON & SIMON DYMOND 37 First, the pretest/posttest design, which has been used in all of the previous research on this topic, does not remove all threats to internal validity because it may be subject to test/re-test sensitivity. Second, in previous experiments, all participants received an identical number of baseline exposures (50 trials) to the slot machine pretest phase. Any resulting changes in slot machine preferences may not have appeared stable because of the spontaneous change following a certain number of trials. Within single case research, there are a number of alternatives to pretest/posttest designs that may overcome some of theses limitations. In particular, the non-concurrent multiple baseline design may be suitable in this regard (Harvey, May, & Kennedy, 2004; Kennedy, 2005; Watson & Workman, 1981). In a concurrent or non-concurrent multiple baseline design, participants receive differing lengths of baseline trials before the intervention is implemented. In this case, functional control is demonstrated when changes in behavior are seen only once the intervention is applied and not for any other reason. Another form of single case design is a reversal design. In a simple reversal design baseline levels of responding are recorded and once stability is achieved, treatment is implemented. These baseline data then act as a comparison condition to the treatment condition (Chambless & Hollon, 1998) to see if the behavior has changed as predicted. To date, only one published experiment has used a design other than pretest/posttest to research gambling behavior (Dixon & Holton, 2009) and only one experiment has used a reversal procedure (albeit as part of a pretest/posttest design; Nastally et al., 2010). In a study on delay discounting, Dixon and Holton (2009) presented participants with hypothetical choices involving differing amounts of money with differing lengths of delay before receiving the money. On each trial the contextual cues, color pink or color purple, were simultaneously presented. Participants then completed a nonarbitrary training procedure that was similar to that of Zlomke and Dixon (2006) except that the two contextual cues (color pink and color purple) were trained as better than and worse than. In the final phase, participants were re-presented with the delay discounting task only this time the contextual cues had acquired functions of better than or worse than, such that participants showed less frequent discounting (Dixon & Holton, 2009). Nastally et al. (2010) adopted the same procedure as described in Hoon et al. (2008), except that a second nonarbitrary training phase was presented in which the contextual cue that had initially been trained as the more than cue became the contextual cue for less than. Additionally, the color slot machine that was most preferred during initial pretest was targeted at the less than contextual cue. It was found that participants preferences for the slot machines reversed following the second nonarbitrary phase such that preference was now shown for the slot machine the same color as the cue trained as more than in the second phase, however this reversal effect was not seen in the pathological gambling group. Nastally et al. (2010) also reported that the problem gamblers took five times as long to reach criterion responding in the initial nonarbitrary task. Studies conducted to date on contextual control of slot machine preferences may have implications for furthering understanding the role of verbal behavior in gambling. Early reports (e.g., Zlomke & Dixon, 2006) described the shift in response allocation towards the more than slot machine as indicative of transformation of functions, even though the relational training intervention involved purely nonarbitrary relations. As others have outlined (e.g., Dymond & Roche, 2010; Dymond & Whelan, 2007), while these studies fall some way short of modern definitions of verbal behavior, they do emphasize

8 38 CONTEXTUAL CONTROL the central role played by relational responding in generating the resulting effects. We will return to this issue in the General Discussion. The present experiments sought to replicate and extend the findings of Hoon et al., (2007, 2008) using a non-concurrent multiple baseline design. In Experiment 1, the contextual cues were counterbalanced across participants. In Experiment 2, nonarbitrary training task was only implemented once responding became stable. In Experiment 3, a reversal design was incorporated to examine whether preferences could be shifted and reversed. EXPERIMENT 1 METHOD Participants Three British participants (2 males, 1 female), aged 20 to 22 years (M = 21; SD = 1) attending Swansea University were recruited through personal contacts. Participants completed the South Oaks Gambling Screen (SOGS; Lesieur & Blume, 1987). A score on the SOGS of 3 or 4 indicates potential problem gambling, and a score of five or above indicates probable pathological gambling (Lesieur & Blume, 1987). P2 scored zero on the SOGS, while P1 and P3 both scored three (M = 2, SD = 1.41). Apparatus The experiment took place in a small room containing a desk, a desktop computer with 16-inch display, full sized keyboard and a two-button click mouse. Stimulus presentation and the recording of responses were controlled by the computer and were programmed in Visual Basic. Design A non-concurrent multiple baseline across participants design was used and the contextual cues were counterbalanced so that two participants were trained that yellow was the more than cue and blue the less than cue, and one participant was trained that blue was the more than cue and yellow the less than cue. The color that was trained as the more than and less than cues were predetermined in Experiment 1, therefore participant preference in the baseline task did not influence the contextual cues that were to be established. Procedure There were three phases to the experiment; slot machine baseline trials, the nonarbitrary relational training intervention and slot machine trials post-intervention. Slot machine Task: Baseline The slot machine task was employed to obtain data on participants baseline choices towards two concurrently available slot machines that were identical in schedule and magnitude of reinforcement and differed only in background color, one being predominantly yellow and the other being predominantly blue. Participants were presented with the following on-screen instructions: On the following screen you will see a button in the middle of the screen. When you click on the button with your mouse two slot machines will be revealed. Click your mouse on the slot machine you would like to play and earn as many points as possible. On clicking the button on the screen, participants were presented with a grey screen which revealed a red button in the centre of the screen containing the instructions Click here. Clicking the red button presented participants with a screen containing a blue rectangular box named Slot machine 1, and a yellow rectangular box named Slot machine 2. These boxes were randomly positioned on opposite sides of the screen throughout trials to control for position bias. To play a slot machine, participants clicked the Spin button on the left hand side of the screen. All participants start-

9 ALICE E. HOON & SIMON DYMOND 39 ed with 100 credits and only one credit could be bet at a time. On clicking the spin button the reels spun for three seconds and sound effects were heard which were similar to those of actual casino slot machines. A winning spin consisted of three identical symbols on the pay off line, and resulted in one credit being awarded in the Total Credits box at the top left of the screen. A losing spin consisted of two matching symbols or no matching symbols and one credit was subtracted from the Total Credits box. After playing a slot machine, participants were taken back to the initial grey screen with Click here button and a new phase began. Each participant was presented with a different number of baseline trials, predetermined by the experimenter: P1 received 40 baseline slot machine trials, P2 received 80 baseline trials and P3 received 120 baseline trials. Nonarbitrary Relational Training This phase established more than and less than contextual functions for the background colors, yellow and blue. For P1 and P2, the color yellow was trained as the contextual cue for more than and blue was the cue for less than. This was counterbalanced for P3 so that blue was trained as the more than cue and yellow was trained as the less than cue. During nonarbitrary training the background screen would appear either yellow or blue, then after approximately three seconds, two stimuli would appear on screen, one stimulus on the left and the other on the right (see Figure 1). The stimuli presented consisted of images which represented different quantities, for example a five dollar poker chip and a twenty five dollar poker chip. Participants were required to select an image by clicking on the image with a mouse. On selecting the correct stimulus, the word Correct was displayed on the screen for one second and a chime sound effect was heard, whereas following an incorrect response the word Wrong was presented and a buzzer sound effect was heard. One point was awarded for each correct response, which was displayed at the top centre of the screen. The computer programme automatically proceeded to the next trial. There were 48 trials in the training phase and participants had to respond correctly across 43 trials in order to progress to the test phase. If criterion was not met, they were exposed to the training phase again. Overall, participants were trained with four different sets of stimuli in the relational training phase. The stimuli used were pound notes ( 5, 20, 50), dice (1, 4, 6), poker chips ($5, $25, $500), and letter grades (A+, C+, D-). Two images were presented onscreen at a time (see Table 1 for a graphical display of all the training trials that were presented). P1 and P2 learned to select the image that represented more than when the background color was yellow and the image that represented less than when the background color was blue. The reverse was true for P3 in the counterbalanced condition. Nonarbitrary Relational Testing The purpose of this phase was to test whether the more than and less than relations established during training would be applied to four novel sets of stimuli. The novel stimulus sets consisted of coins (1p, 20p, 1), playing cards (4 of spades, 9 of spades, king of spades), jackpots (5 million, 10 million, 20 million) and positions (1 st, 8 th 10 th ). Participants were required to respond correctly across all 48 trials. If a participant failed the test phase, they were re-exposed to the training phase. The format of the test phase was identical to the training phase except that no feedback was given. Slot machine Task: Post-Intervention This phase was to investigate whether the nonarbitrary relational training task would increase responding to a particular slot

10 40 CONTEXTUAL CONTROL Table 1. Graphical representation of all the different trial types that were presented to participants during nonarbitrary relational training. More 5 20 More 5 50 More Less 5 20 Less 5 50 Less More More More Less Less Less More $5 $25 Less $5 $25 More A+ C+ Less A+ C+ More $5 $500 Less $5 $500 More A+ D- Less A+ D- More $25 $500 Less $25 $500 More C+ D- Less C+ D- machine. P1 received 120 post-intervention slot machine trials, P2 received 80 slot machine trials and P3 received 40 slot machine trials. This ensured that all participants completed a total of 160 slot machine trials throughout the whole experiment. RESULTS AND DISCUSSION All participants completed the nonarbitrary relational training and testing. Criterion for the training phase required that participants obtain a score of at least 43 in order to progress to the test phase. Table 2 shows that P1 took the most number of training

11 ALICE E. HOON & SIMON DYMOND 41 Figure 1. An example of a less than (left panel) and a more than trial (right panel) where blue is trained as the less than cue and yellow is trained as the more than cue. In the presence of the less than contextual cue, the image portraying the lesser quantity is reinforced (indicated with an arrow). In the presence of the more than contextual cue, the image portraying the greater quantity is reinforced (indicated with an arrow). Table 2. The number and mean number (with standard deviation) of correct trials during nonarbitrary relational training and testing phase in Experiment 1. Participant Nonarbitrary Relational Training (/48) Nonarbitrary Relational Testing (/48) Mean Standard devtion trials to successfully progress to the test phase, whereas P2 and P3 only required three exposures to training. The mean number of trials required to each criterion in the training was (SD = 9.86). In the test phase participants were required to get all 48 trials correct to complete the task. All participants passed the test following only one exposure to the task. Figure 2 depicts the number of responses allocated towards the more than slot machine (that is, the slot machine that was the

12 42 CONTEXTUAL CONTROL same color as the more than contextual cue) during baseline and post-intervention. All participants showed relatively stable levels of responding during baseline, suggesting no marked preferences for either colored slot machine, however it should be noted that P2 initially showed a small preference for the slot machine that was to be targeted as the more than contextual cue. Following the relational training and testing intervention, two of the three participants (P1 & P2) showed an increase in the number of responses allocated to the more than slot machine. This increase remained stable for the remainder of the post-intervention phase. The participant (P3), who received the shortest post-intervention phase, showed a smaller increase in response allocation to the more than slot machine. Overall, the percentage difference in responding allocated towards the more than slot machine from baseline to post-intervention was 32.5% for P1, 30% for P2, and 5% for P3. The findings of Experiment 1 support those of Hoon et al. (2007, 2008) and Zlomke and Dixon (2006) that preferences for concurrently available slot machines may be altered in accordance with contextual cues. Although there was a clear shift in slot machine preferences for P1 and P2, this effect was less evident for P3, who had received the lowest number of postintervention trials. Thus, a limitation of Experiment 1 was that the participant who received the shortest exposure to the postintervention trials also produced unstable responding. Accordingly, this made it difficult to assess the effects of the relational training intervention on slot machine preferences. If more trials had been given following the intervention then this may have resulted in stable responding towards one slot machine allowing any slot machine preference to be assessed. In order to overcome this limitation, the number of baseline trials should be determined on the basis of visual analysis (level and trend) and the intervention should be employed only once responding is stable. The same stability criteria could then be adopted during the postintervention phase. An additional limitation of Experiment 1 was that the color to be targeted as the more than contextual cue was predetermined. It is possible that when presenting participants with a concurrent choice between slot machines, a preference for a particular colored machine may be seen during baseline. If that color is then targeted as the more than cue, a smaller effect size may be seen when comparing baseline and postintervention slot machine preferences. To overcome this issue, Nastally et al. (2010) targeted the color slot machine that was the least preferred following baseline as the more than contextual cue, rather than predetermining the contextual cues for more than and less than. Experiment 2 was conducted to explore this issue in participants who met a SOGS classification of problem gambler. Additionally, in Experiment 2, similar to Nastally et al. (2010) the color of the least preferred slot machine during baseline was targeted as the contextual cue for more than in the nonarbitrary training task in order to rule out any resulting shift in preferences towards the more than slot machine occurring on the basis of pre-existing color preferences. EXPERIMENT 2 METHOD Participants Three male participants of British and Asian descent aged 20 to 24 years (M = 22.67; SD = 2.31) were recruited through campus-wide advertising the study. Only participants with a minimum SOGS score of 3 were recruited. Participants SOGS scores were 7, 6 and 3 (M = 5.33, SD = 1.70).

13 ALICE E. HOON & SIMON DYMOND 43 Figure 2. Response allocation to the more than slot machine in baseline and post-intervention in Experiment 1.

14 44 CONTEXTUAL CONTROL Procedure The procedure of Experiment 2 was identical to that of Experiment 1 with the following exceptions. Firstly, participants were not given a pre-determined number of baseline trials. Instead, baseline slot preferences were monitored every 30 trials and once responding appeared stable the nonarbitrary relational training task was given. Responding was said to be stable when slot machine preferences fell within a range of two trials across three consecutive data points. This was assessed by the experimenter. Secondly, whereas in Experiment 1 the contextual cues had been counterbalanced across participants, in Experiment 2, the color of the least preferred slot machine during baseline was targeted as the more than contextual cue in the relational training intervention. Following the relational training intervention, participants were represented with the concurrent slot machine task. In the same way as the baseline task, responses were monitored every 30 trials and the experiment only ended once responses appeared stable. Finally, only individuals with a minimum SOGS score of 3, indicating a potential problem gambler, were recruited. RESULTS AND DISCUSSION All participants completed the nonarbitrary relational training and testing phase (see Table 3). P4 and P5 required two exposures to the training task before progressing to the test phase, whereas P required three exposures. The mean number of trials required to meet criterion for the training phase was (SD = 8.09). P4 and P5 passed the nonarbitrary test after only one exposure to the task, whereas P6 required two exposures. Table 3. The number and mean number (with standard deviation) of correct trials during nonarbitrary relational training and testing phase in Experiment 2. Participant Nonarbitrary Relational Training (/48) Nonarbitrary Relational Test (/48) Mean Standard deviation Figure 3 shows that participants preferences for concurrently available slot machines were altered, such that participants allocated most responding to the slot machine that was the same color as the more than contextual cue. The extent of this increase in preference varied across participants, with responding by P6 showing the greatest post-intervention increase. Postintervention responses from P4 were initially variable but increased in level and trend by the fifth block of trials where an increase in preference is seen towards the more than slot machine. Responses then stabilised

15 ALICE E. HOON & SIMON DYMOND 45 Figure 3. Response allocation to the more than slot machine in baseline and post-intervention in Experiment 2.

16 46 CONTEXTUAL CONTROL whilst showing a fairly clear preference for the more than slot machine. The results for P5 are perhaps not as clear. Although P5 initially showed a slight preference for the more than slot machine, in his second block of trials P5 showed equal responding to either slot machine, and then in the last block of trials P5 showed a slight preference for the more than slot machine again. Unfortunately, P5 terminated his participation before additional post-intervention trials could be administered. For this reason, it is difficult to draw many clear conclusions regarding the preferences for P5. P6 required the highest number of baseline trials before the intervention was implemented, and then showed the most marked increase towards the more than slot machine during the final phase. It is important to note that in the first block of trials following intervention, all participants allocated increased responding to color slot machine that had been established as more than, despite this slot machine being least preferred during baseline. This suggests that the contextual cues that were established in the intervention are influencing responding. Overall, the findings of Experiment 2 demonstrated that problem gamblers preferences for one of two concurrently available slot machines can be altered in accordance with a relational training intervention that targets the color of the least preferred slot machine, and rules out competing explanations in terms of pre-existing color preferences. It is possible therefore that in real world gambling, an individual may show a preference for a particular slot machine due to a feature of that machine being related in some way to something else in the gamblers behavioral repertoire. This suggests that the factors influencing a gambler s slot machine preferences may be highly complex. In Experiment 3, an additional measure was incorporated: given that the presence of a nonarbitrary relational training task can alter participants preferences for concurrently available slot machines, if the original task were then reversed and the contextual cue originally established as the more than cue now trained as the less than cue, it would be expected that preferences for the slot machines would shift accordingly. Therefore, Experiment 3 was a further modification of Experiment 2 with the addition of a reversal design. EXPERIMENT 3 Experiment 3 employed a reversal design in which, following initial baseline, intervention and post-intervention trials, the color previously established as the more than cue was trained as the less than cue and vice versa in a second relational training intervention. METHOD Participants In Experiment 3, there were two male British participants aged 25 and 27 (M = 26; SD = 1). Only individuals with a minimum SOGS score of 3 were recruited. One participant scored 6 on the SOGS (P7) and the other scored 5 (P8) indicating that both participants were potential pathological gamblers (M = 5.5, SD = 0.5). Procedure The procedure for Experiment 3 was identical for that of Experiment 2 with the exception of additional nonarbitrary training tasks that reversed the contextual cues. P7 was given 2 reversals (therefore, 3 training tasks in total) whereas P8 was only given one reversal. The number of reversal interventions given to the participants was predetermined. Participants were presented with a different number of reversals to examine the extent to which presenting multiple reversal training interventions would still exert control over responding. As in Experiment 2, the least preferred slot machine was

17 ALICE E. HOON & SIMON DYMOND 47 targeted as the more than contextual cue for the first nonarbitrary training intervention, and this was only implemented once responding appeared stable (that is, preference feel within a range of two trials). The second (and third, in the case of P7) training intervention was also only implemented once stability had been achieved. RESULTS AND DISCUSSION Both participants completed the nonarbitrary relational training and testing phase (see Table 4). In the first nonarbitrary training task P7 completed the training phase after just one exposure to the task (i.e., 48 trials), whereas P8 required 2 exposures to the training task to the first nonarbitrary training task. Both participants passed the nonarbitrary test following just one exposure to the task. When presented with nonarbitrary training during the reversal interventions, both participants were able to respond accurately to the new relations that were established requiring only one exposure to pass both the nonarbitrary training and nonarbitrary testing tasks. Figure 4 shows participants preferences towards the more than slot machine. The upper panel depicts the responding made by P7, whereas the lower panel depicts responding made by P8. During baseline, both participants showed no clear preferences for either slot machine and responding stabilized after 60 trials. Following the first relational training intervention, both participants preferences for the more than slot machine increased as predicted. Following the second relational training intervention, P7 showed equal response allocation to the more than slot machine, whilst responding by P8 approximated to that of the earlier post-intervention phase. P7 then received a third relational training intervention and subsequently showed a decreased preference for the more than slot machine. This is perhaps due to the interaction of the direct contingencies of the slot machines. From the data of P7 it appears that as more trials are undertaken, the effect of the contextual cue weakens and the payout probability exerts more control. In a similar way, P8 initially showed a slight increase for the more than slot machine following the first intervention, then following the second intervention showed a fairly equal preference for more than and the less than slot machine. Overall, the findings of Experiment 3 demonstrated that problem gamblers preferences for one or two concurrently available slot machines may be altered in accordance with a relational training intervention, Table 4. The number and mean number (with standard deviation) of correct trials during the nonarbitrary relational training and testing phase in Experiment 3. Participant Nonarbitrary training (/48) Nonarbitrary testing (/48) Reversal 1 Nonarbitrary training (/48) Reversal 1 Nonarbitrary testing (/48) Reversal 2 Nonarbitrary training (/48) Reversal 2 Nonarbitrary testing (/48) n/a n/a Mean Standard deviation

18 48 CONTEXTUAL CONTROL Figure 4. Response allocation to the more than slot machine in baseline and post-intervention in Experiment 3. P7 is shown in the upper panel, and P8 in the lower panel.

19 ALICE E. HOON & SIMON DYMOND 49 however, where the contextual cues are reversed with an additional relational training and testing procedure the level of altered preferences decreases as further interventions are presented. This suggests that the effects of the intervention targeting the background colors interacted with the concurrent, matched schedule of programmed reinforcement, leading to diminished control by the background colors. Nastally et al. (2010) found that whilst nonproblems gamblers showed shifting preferences in slot machine preference when the contextual cues were reversed; this effect was diminished in the problem gambling group. Nastally et al (2010) suggested that these differences between the nonproblem gambling group and the gambling group could be the result of maladaptive rule formations (Delfabbro, 2004) and self governed rule adherence which has been reported in clinical populations (Wulfert, Greenway, Farkas, Hayes, & Dougher, 1994). Given that the present study only employed potential problem gamblers, it is not possible to compare whether presenting additional reversals in which the contextual cues are switched, exerts more control in nonproblem gamblers compared to problem gamblers. GENERAL DISCUSSION The current experiments replicated the findings of Hoon et al. (2007, 2008) and extended them by using a design that allowed for a better demonstration of experimental control. In Experiment 1, a clear shift was seen in P1 and P2 who allocated the majority of trials to the more than slot machine following the nonarbitrary intervention. The effect was less clear in P3. Unfortunately, as P3 was given the shortest number of postintervention trials, this did not allow for his responding to become stable and, therefore, preference cannot clearly be determined from his data. This participant also received the highest number of baseline trials and had therefore most experience of the direct contingencies of the slot machines prior to nonarbitrary training. For this reason, it is possible that the payout probabilities may have exerted greater control than the contextual cue. In Experiment 2, the number of baseline and post-intervention trials given to each participant were not predetermined, but instead responses were monitored until responding appeared stable. All three participants showed an increase in response allocation towards the more than slot machine following the intervention. Experiment 3 incorporated a reversal design and the results were particularly interesting, as the data show that with extended exposure to the contingencies of reinforcement and additional nonarbitrary training tasks, the control exerted by the nonarbitrary training intervention begins to diminish and the schedules of reinforcement appear to influence responding. Across all experiments, every participant, except P3, showed an increase in preferences towards the more than slot machine in the first ten trials following the first intervention. This increased response allocation was not, however, always maintained during all post-intervention trials. The findings of the present experiments, particularly Experiment 3 are perhaps not quite as clear as the initial studies by Zlomke and Dixon (2006) and Hoon et al. (2007, 2008). The finding that participants do not always show a consistent preference for the more than slot machine following relational training is likely due to the payout probability of each slot machine: With the probability of the slot machines being 0.5, it is highly plausible that a participant may experience a string of losses on what has been trained as the more than slot machine resulting in switching over to the slot machine that was the same color as the less than cue. A contextual cue can be trained to represent more than,

20 50 CONTEXTUAL CONTROL however when that cue is paired with a random ratio schedule such as that of a slot machine, the direct contingencies of the schedule of reinforcement of that machine are also going to influence responding and may conflict with the individual s understanding of the properties of the contextual cue. Whilst it has long been understood that contingency-shaped behavior and the schedules of reinforcement are an important factor in gambling behavior (Skinner, 1974), the extent to which contextual cues may interact with or override direct contingencies of reinforcement are not clearly understood. The interaction between the contextual cues and the contingencies of the slot machine highlight the need for research on the analysis of gambling behavior to present slot machine tasks under extinction (or, more accurately, non-reinforcement). Had participants only been able to play the slot machines but not actually experience any wins or losses (see Dymond, McCann, Griffiths, Cox & Crocker, 2012, for a related example), then the contextual cue may well have continued to control behavior. However, an obvious limitation of presenting trials under non-reinforcement is the challenge it presents to ecological validity: in a casino environment it is always possible that an individual will experience winning trials or variants of winning with conditioned reinforcement properties ( losses disguised as wins ; Dixon et al in Addiction). Thus, it remains an important empirical issue to ascertain the conditions under which partial and nonreinforcement interact with structural characteristics such as contextual cue in initiating and maintaining gambling choice.. An alternative to presenting slot machines under non-reinforcement would be to vary the payout probability of the slot machine (Dymond et al., 2012). The payout probability of the slot machines in the present experiments was 0.5 with five credits being awarded for a winning spin; therefore participants were in credit at the end of the experiment. This is fairly generous compared to those of casino slot machines in which the payout probabilities favor the casino (Parke & Griffiths, 2006). It would be interesting to see to what extent the contextual cues exert control when the payout probability was set to 0.3 as wins would occur less frequently therefore the contextual cue may function as a more salient rule and continue to exert control over behavior. Although the current findings may supplement the existing literature on the role of nonarbitrary contextual control of gambling behavior, such an explanation of gambling behavior is may not be complete. Electronic gaming machines are rarely, if ever, controlled solely by the formal properties of the stimuli and the nature of stimulus functions are beyond such formal characteristics. It has been suggested that for a more complete account of gambling behavior, the role of verbal behavior as defined by relational frame theory (RFT; Hayes, Barnes-Holmes, & Roche, 2001) must be addressed (Dymond & Roche, 2010). Although the studies by Zlomke and Dixon (2006), Hoon et al. (2008) and Nastally et al. (2010) have provided preliminary insight regarding how gambling may not be controlled solely by schedules of reinforcement alone, these experiments do not supplement a strictly verbal account of gambling as defined by RFT (Hayes et al., 2001). According to RFT, for an event to be considered a verbal event responding must be arbitrarily applicable. The training tasks in both of these experiments consisted of nonarbitrary relational responding, therefore non-humans should, in principle, be able to complete such tasks (see Reese, 1968) as the organism receives reinforcement for selecting the larger or smaller stimulus, and consequently, the organism is then able to respond to the relation between the stimuli (Hayes et al., 2001).

21 ALICE E. HOON & SIMON DYMOND 51 For these reasons, a nonarbitrary model of gambling cannot be considered a verbal account of gambling and experiments that aim to provide such an account must include the arbitrarily applicable nature of verbal behavior. Humans however, are able to respond to arbitrary relations in which there is no physical relation between the stimuli. For example, a human participant can be trained that stimulus B is more than stimulus A, even though the physical properties of stimulus B are not any greater than stimulus A. The second reason why the aforementioned studies are not true verbal experiments is that an integral component of the RFT account of verbal behavior is that of derived relational responding. In the present study, participants were directly trained that the color yellow was a contextual cue for more than and color blue was a contextual cue for less than. This training intervention is not dissimilar to non-human literature on identity matching to sample and oddity from sample, in which animals are trained to select the comparison stimulus that is the same (identity) in the presence of one sample stimulus, or the stimulus that is different (oddity) in the presence of another stimulus, through differential reinforcement (e.g., Cumming & Berryman, 1965). It is possible for humans, however, to derive relations between arbitrary stimuli that have not been directly trained. For example, if an individual is trained that stimulus B is more than A, and stimulus C is more than B; the individual is then able derive that A is less than B, and B is less than C. Furthermore, they can also derive that C is more than A, and A is less than C, therefore, from just two trained relations, the human participant is able to derive a further four untrained relations. Such principles have only been robustly demonstrated in human participants and may explain highly complex human behaviour where a direct contingency of reinforcement account falls short (Dymond et al., 2012; Dymond & Roche, 2010; Hayes et al., 2001). Given that it is unlikely that gambling behaviour can ever be wholly accounted for by schedules of reinforcement, it is vital that for a more complete account that includes the fundamental components of RFT outlined above be incorporated into further empirical research. REFERENCES Chambless, D. L., & Hollon, S. D. (1998). Defining empirically supported theories. Journal of Consulting and Clinical Psychology, 66, 1, Cooper, J. O., Heron, T. E., & Heward W. E. (2007). Applied Behavior Analysis. New Jersey: Pearson/Prentice Hall. Cumming, W. W., & Berryman, R. (1965). The complex discriminated operant: studies of matching-to-sample and related problems. In D. I. Mostofsky (Ed.), Stimulus generalization (pp ). Stanford: University Press. Dixon, M. R., & Holton, B. (2009). Altering the magnitude of delay discounting by pathological gamblers. Journal of Applied Behavior Analysis, 42, Dixon, M. J., Harrigan, K. A., Sandhu, R., Collins, K., Fugelsang, J. A. (2010). Losses disguised as wins in modern multi-line video slot machines. Addiction, 105, Dymond, S., McCann, K., Griffiths, J., Cox, A., & Crocker, V. (2012). Emergent response allocation and outcome ratings in slot machine gambling. Psychology of Addictive Behaviors, 26(1), Dymond, S., & Roche, B. (2010). The impact of derived relational responding on gambling behavior. Analysis of Gambling Behavior, 4, Eckerman, D. A. (1970). Generalization and response mediation of a conditional discrimination. Journal of the Experimental Analysis of Behavior, 13,

22 52 CONTEXTUAL CONTROL Fredheim, T., Ottersen, K., & Arntzen E. (2008). Slot-machine preferences and self-rules. Analysis of Gambling Behavior, 2, Griffiths, M. D. (1990). The acquisition, development and maintenance of fruit machine gambling in adolescents. Journal of Gambling Studies, 6, Harvey, M. T., May, M. E., & Kennedy, C. H. (2004). Nonconcurrent multiple baseline designs and the evaluation of educational systems. Journal of Behavioral Education, 13, Hayes, S. C., Barnes-Holmes, D., & Roche, B. (2001). Relational Frame Theory; A Post-Skinnerian Account of Human Language and Cognition. New York: Kluwer Academic/Plenum Publishers Hoon, A., Dymond, S., Jackson, J. W., & Dixon, M. R. (2007). Manipulating contextual control over simulated slot machine gambling. Analysis of Gambling Behavior, 1, Hoon, A., Dymond, S., Jackson, J. W., & Dixon, M. R. (2008). Contextual control of slot machine gambling: Replication and extension. Journal of Applied Behavior Analysis, 41, Johnson, T. E. & Dixon, M. R. (2009). Influencing children s pregambling game playing Via conditional discrimination training. Journal of Applied Behavior Analysis, 42, Kennedy, C. H. (2005). Single-Case Designs for Educational Research. Boston: Pearson Education/Allyn and Bacon,. Lesieur, H. R., & Blume, S. B. (1987). The South Oaks gambling screen (SOGS): A new instrument for the identification of pathological gamblers. American Journal of Psychiatry, 144, Loba, P., Stewart, S. H., Klein, R. M. & Blackburn, J. R. (2001). Manipulations of the features of standard video lottery terminal (VLT) games: Effects in pathological and non-pathological gamblers. Journal of Gambling Studies, 17, 4, Nastally, B. L., Dixon, M. R., & Jackson, J. W. (2010). Manipulating slot machine preference in problem gamblers through contextual control. Journal of Applied Behavior Analysis, 43, Parke, J., & Griffiths, M. (2006). The psychology of the fruit machine: The role of structural characteristics (revisited). International Journal of Mental Health Addiction, 4, Sharpe, L., Walker, M., Coughlan, M., Enerson, K., & Blaszczynski, A. (2005). Structural changes to electronic gaming machines as effective harm minimization strategies for non-problem and problem gamblers. Journal of Gambling Studies, 21, 4, Skinner, B. F. (1974). About behaviorism, New York: Vintage Books. Watson, P. J., & Workman, F. A. (1981). The non-concurrent multiple baseline across-individuals design: An extension of the traditional multiple baseline design. Journal of Behavior Therapy and Experimental Psychiatry, 12, 3, Zlomke, K.R., & Dixon, M.R. (2006) Modification of slot-machine preferences through the use of a conditional discrimination paradigm. Journal of Applied Behavior Analysis, 39, Action Editor: Mark R. Dixon

23 Analysis of Gambling Behavior 2013, 7, Number 2 (Winter, 2013) Comparing the Contingencies that Maintain Gambling Behavior in an Online Sample of Younger and Older Adults Jeffrey N. Weatherly University of North Dakota Age is a risk factor for developing pathological gambling. Endorsing gambling as an escape has also been associated with the disorder. The present study recruited 120 people either years of age or 45 years of age or older so as to determine how age was related to the contingencies that maintain gambling behavior. Results showed that younger adults displayed more gambling problems, as measured by the Problem Gambling Severity Index, than did older adults. Younger adults also endorsed gambling for positive reinforcement to a lesser extent and gambling as an escape to a greater extent than older adults, as measured by the Gambling Functional Assessment Revised. For both groups, gambling behavior was primarily maintained by positive reinforcement. However, gambling as an escape, but not for positive reinforcement, was a significant predictor of gambling problems. The present results replicate previous research linking age and gambling problems. The results also provide a potential reason for why that link may exist. Keywords: Gambling; Age; Escape; Positive Reinforcement According to Petry (2005), there are several risk factors for pathological gambling. The strongest factor is substance use and abuse, with substance abusers being at a much higher risk than nonusers. Another factor is ethnicity, with ethnic minorities being at greater risk than those in the majority population. The next factor is sex, with males displaying pathological gambling at a significantly higher frequency than females. Marital status is another risk factor, with individuals who are single or divorced being at greater risk than those who are married. Socioeconomic status is the fifth factor, with the less affluent displaying pathology at a greater frequency than the more affluent. The final risk factor is age. Young adults display higher rates of pathological gambling than older adults. These factors are not the only ones Address all correspondence to: Jeffrey N. Weatherly Department of Psychology University of North Dakota Grand Forks, ND jeffrey.weatherly@und.edu associated with pathological gambling, however. In fact, a specific contingency of reinforcement, escape, has been linked to the disorder. Gambling as an escape is recognized as a symptom of pathological gambling (American Psychiatric Association, 2000). Endorsing gambling as an escape has been shown to be strongly correlated with scoring high on measures designed to screen for pathological gambling such as the South Oaks Gambling Screen (SOGS; Lesieur & Blume, 1987; see Weatherly & Derenne, 2012; Weatherly, Dymond, Samuels, Austin, & Terrell, in press b) and the Problem Gambling Severity Index (PGSI; Ferris & Wynne, 2001; see Weatherly, 2013). Endorsing gambling as an escape has shown to be related to how individuals gamble in a laboratory situation (e.g., Martner, Montes, & Weatherly, 2012; Weatherly, Montes, & Christopher, 2010). Specifically, participants who score high in gambling as an escape gamble differently (e.g., play more times) than participants who score low in gambling as an escape. Finally, escape contingencies have been integrated into theoretical accounts for why people 53

24 54 YOUNGER VS. OLDER might display pathological gambling (e.g., Blaszczynski & Nower, 2002). Blaszczynski and Nower suggested that emotionally vulnerable individuals were most prone to turn to gambling as an escape. What is not presently known is how the contingency of escape might be related to the risk factors for pathological gambling as outlined by Petry (2005). For instance, younger adults are more apt to display pathological gambling than are older adults. One could hypothesize that therefore younger adults would be more likely than older adults to endorse gambling as an escape. However, that is not the only possible outcome. Younger adults are less likely to be financially secure than older adults, so that may make them more likely to gambling for positive reinforcement (e.g., winning money). Then again, younger adults might gamble to escape economic hardship. Likewise, one could potentially justify why older individuals, such as retirees, might turn to gambling as an escape (e.g., from boredom). The present study was designed to determine how the contingencies of positive reinforcement and escape might be related to age. A sample of younger and older adults was recruited. These individuals then completed the Gambling Functional Assessment Revised (GFA-R; Weatherly, Miller, & Terrell, 2011), which is designed to determine if the respondent s gambling behavior is maintained by positive reinforcement and/or by escape. The participants also completed the SOGS and the PGSI. The hypotheses were that younger adults would be more likely than older adults to display gambling problems, as would be expected from the research literature (Petry, 2005), and that they would also be more likely than older adults to endorse gambling as an escape, which would be consistent with their heightened rates of gambling problems. METHOD Participants The participants were 120 (65 males; 55 females) individuals from the United States. Sixty-six of the participants were between years of age and 54 participants were 45 years of age or older (with six reporting to be 65 years of age or older). One hundred one (84.2%) were Caucasian. Fifty-two participants (43.3%) reported being single, divorced, or widowed, while 32 (26.7%) reported being in a relationship and 36 (30.0%) reported being married. Ninety-eight participants (81.7%) reported an annual income of less than $50,000. Thirty-four of the participants (28.3%) indicated that they regularly smoked cigarettes. Younger (21-24 years of age) and older adults (45 years of age or older) were recruited to participate in this online study. These age ranges were chosen because they ensured that the participants in the different groups differed in age by at least 20 years. Participants completed the study on Amazon s Mechanical Turk (MTurk; and were paid $0.25 in Amazon.com credit for their participation. Materials and Procedure Participants were first presented with informed-consent information about the study as approved by the Institutional Review Board at the University of North Dakota. The second item they were presented with was a demographic questionnaire that asked about the data reported in the participants section. The third item presented to the participants was the SOGS (Lesieur & Blume, 1987), which is a self-report measure pertaining to the respondent s gambling history. Lesieur and Blume (1987) proposed that a score of five or more on the SOGS is indicative of the potential presence of pathology. They also reported that the internal consistency of the SOGS was high ( = 0.97), although subsequent researchers have reported lower lev-

25 JEFFREY N. WEATHERLY 55 els of internal consistency ( = 0.69, Stinchfield, 2002; = 0.81, Stinchfield, 2003). Lastly, the SOGS has shown to have good temporal reliability (r = 0.89 at four weeks and r = 0.67 at 12 weeks; Weatherly, Miller, Montes, & Rost, 2012). The fourth item was the PGSI (Ferris & Wynne, 2001), which is a self-report measure designed for use in the general population. It measures negative consequences associated with the one s gambling. Increasing scores are indicative of an increasing number of negative consequences experienced because of one s gambling. The PGSI has shown to measure one construct (McMillen & Wenzel, 2006), have good internal consistency ( = 0.84; Ferris & Wynne, 2001), and good temporal reliability (r = 0.78; Ferris & Wynne, 2001). Lastly, the participants completed the GFA-R (Weatherly et al., 2011). The GFA-R is a self-report functional assessment tool designed to measure whether the individual s gambling behavior is reinforced by positive reinforcement and/or escape. High scores on the positive reinforcement and escape subscales suggest that the individual s gambling is maintained by those contingencies. The GFA-R has good internal consistency ( = 0.91; Weatherly et al., 2012) and temporal reliability (r = 0.80 at four weeks and r = 0.81 at 12 weeks; Weatherly et al., 2012). The psychometric properties of the GFA-R have also been replicated when used in a sample from Japan (Weatherly, Aoyama, Terrell, & Berry, in press a) and the United Kingdom (Weatherly et al., in press b). Participants completed the study on MTurk, which has shown to be a valuable way to collect data from a broad population (e.g., Shapiro, Chandler, & Mueller, 2013; Shapiro & Stewart, 2011). The participants were presented the materials in the order described above and were compensated within three days of completing the materials. RESULTS AND DISCUSSION An initial set of analyses was conducted using the participants sex, ethnicity, marital status, annual income, and smoking of cigarettes as covariates because these are potentially measures of the other risk factors of gambling and could potentially help account for any effect of participants age. However, the results did not indicate that these covariates were significant, and therefore they were excluded from the results of the analyses reported below. For all statistical analyses, significance was met at p <.05. The first analysis was an analysis of variance (ANOVA) on the SOGS scores of the younger (M = 2.56, SD = 3.5) and older participants (M = 1.63, SD = 2.94). Although the younger adults scored higher on the SOGS than the older adults, this difference was not statistically significant, F(1, 118) = 2.42, p =.122, η 2 =.020. A similar ANOVA was then conducted using participants PGSI scores as the dependent measure. The results of this analysis were significant, F(1, 118) = 5.08, p =.026, η 2 =.041, indicating that the younger participants had experienced more negative outcomes because of their gambling (M = 3.00, SD = 5.14) than had the older participants (M = 1.22; SD = 2.96). The next analysis was a two-way mixed model ANOVA, with group serving as the between-subjects variable and GFA-R subscale score serving as the repeated measure. This analysis did not yield a significant main effect of age group, F(1, 118) = 0.17, p =.681, η 2 =.001, indicating that the total scores on the GFA-R did not differ as a function of age. The main effect of subscale was significant, F(1, 118) = , p <.001, η 2 =.673, indicating that participants scored significantly higher on gambling for positive reinforcement (M = 19.03, SD = 10.92) than for gambling as an escape (M = 4.86, SD = 7.26). Finally, the interaction between group and GFA-R subscale was also significant, F(1,

26 56 YOUNGER VS. OLDER 118) = 4.93, p =.028, η 2 =.040. The younger adults scored lower on gambling for positive reinforcement (M = 17.85, SD = 11.11) than the older adults (M = 20.48, SD = 10.61), but higher on gambling as an escape (M = 5.52, SD = 8.14) than the older adults (M = 4.06, SD = 6.00). Despite the difference in the extent to which participants in each group endorsed gambling as an escape, endorsing gambling as an escape was a significant predictor of gambling problems for participants in both groups. This conclusion is supported by the results of several simultaneous linear regressions that were conducted separately for both groups, with their SOGS and PGSI scores as the dependent measure and their GFA-R subscale scores as the predictor variables. When the SOGS scores of the younger adults were subjected to this analysis, the model was significant, F(2, 63) = 36.01, p <.001, R 2 =.533. GFA-R escape subscale scores, β =.706, p <.001, but not positive reinforcement subscale scores, β =.047, p =.638, were significant predictors of SOGS scores. The analysis on PGSI scores of the younger adults again found that the model was significant, F(2, 63) = 59.68, p <.001, R 2 =.655. Again, GFA-R escape subscale scores, β =.842, p <.001, but not positive reinforcement subscale scores, β = -.072, p =.402, were significant predictors of PGSI scores. The same analyses were repeated on the data from the older adults. When the SOGS scores were analyzed, the regression model was significant, F(2, 51) = 26.28, p <.001, R 2 =.507 and the GFA-R escape subscale scores, β =.694, p <.001, but not positive reinforcement subscale scores, β =.047, p =.446, were significant predictors of SOGS scores. The analysis on PGSI scores showed that the model was significant, F(2, 63) = 25.86, p <.001, R 2 =.504. Again, GFA-R escape subscale scores, β =.701, p <.001, but not positive reinforcement subscale scores, β =.023, p =.402, were significant predictors of PGSI scores. Age is a risk factor for pathological gambling, with one s risk of the disorder decreasing with age. Gambling problems have also been associated with gambling as an escape. The goal of the present study was to determine if endorsing gambling as an escape would also be associated with age. The results indicated that younger adults displayed higher scores on the GFA-R escape subscale than did the older adults. However, both groups displayed significantly higher scores on the GFA-R positive reinforcement subscale than on the escape subscale. Likewise, GFA-R escape scores were significant predictors of gambling problems, as measured by the SOGS and PGSI, for both groups. The present results therefore offer a potential reason for why age is risk factor for pathological gambling. Whereas older adults endorse gambling for positive reinforcement to a greater extent than younger adults, the younger adults endorse gambling as an escape to a greater extent than older adults. This outcome suggests that the younger adults may be less capable of dealing with aversive situations than are older adults and thus more likely to turn to gambling to escape those situations. Of course, the difference in endorsing gambling as an escape between the age groups may have occurred because the older adults have developed alternative means to deal with aversive situations. Interestingly, the present study attempted to control for the other risk factors of gambling (e.g., socioeconomic status, ethnicity) and including those factors into the statistical analyses would not have changed the results. Thus, whatever is driving the younger adults to endorse gambling as an escape does not appear to be statistically related to the other risk factors for pathological gambling. With that said, all participants gambling behavior was maintained to a significantly greater extent by positive reinforcement than

27 JEFFREY N. WEATHERLY 57 by escape. This result is consistent with other research that has utilized the GFA-R (e.g., Weatherly & Derenne, 2012; Weatherly et al., in press a, b). It might also help explain why the vast majority of people who gamble do not develop problems with gambling; their gambling is primarily maintained by positive reinforcement. The present study is not without its limitations. All of the data collected represent self-report measures, and therefore one cannot verify their accuracy or be confident that the responses correspond to actual behavior. In terms of race and annual income, the sample was relatively homogenous and different results might have been observed had a more diverse sample been recruited. The present sample also does not qualify as a clinical sample, with only 18 of the 120 participants scoring five or more on the SOGS. Therefore, one cannot conclude that similar relationships would be identified in a sample of clinically diagnosed pathological gamblers. Finally, the data were collected from individuals who were motivated to complete the materials for the equivalent of $0.25. It is possible that the present results represent some special characteristics of such a sample. Despite these potential limitations, the present results do support previous research that indicated that younger adults are more likely to suffer from gambling problems than older adults. They also add to those findings that younger adults appear more likely to endorse gambling as an escape, and less like to endorse gambling for positive reinforcement, than older adults. Despite this, the gambling behavior of both younger and older adults is maintained to a greater degree by positive reinforcement than by escape. Likewise, the presence of gambling problems, at least as measured by the SOGS and PGSI, are predicted by endorsing gambling as an escape, but not by gambling for positive reinforcement, for both younger and older adults. Future studies that employ longitudinal research designs would be helpful in determining how these factors related to gambling change with age. REFERENCES American Psychiatric Association (2000). Diagnostic and Statistical Manual of Mental Disorders (4 th ed., text revision). Washington, D.C.: Author. Blaszczynski, A., & Nower, L. (2002). A pathways model of problem and pathological gambling. Addiction, 97, Ferris, J., & Wynne, H. (2001). The Canadian Problem Gambling Index: Final report. Ottawa, ON: Canadian Center on Substance Abuse. Lesieur, H.R., & Blume, S.B. (1987). The South Oaks Gambling Screen (SOGS): a new instrument for the identification of pathological gamblers. American Journal of Psychiatry, 144, Martner, S.G., Montes, K.S., & Weatherly, J.N. (2012). Using unsolvable anagrams to induce escape: Will it increase gambling behavior? Analysis of Gambling Behavior, 6, McMillen, J., & Wenzel, M. (2006). Measuring problem gambling: Assessment of three prevalence screens. International Gambling Studies, 6, Petry, N.M. (2005). Pathological Gambling: Etiology, Comorbidity, and Treatment. Washington, D.C.: American Psychological Association. Shapiro, D.N., Chandler, J., & Mueller, P.A. (2013). Using Mechanical Turk to study clinical populations. Clinical Psychological Science, 1, Shapiro, D.N., & Stewart, A.J. (2011). Parenting stress, perceived child regard, and depression in stepmothers and biological mothers. Family Relations, 60,

28 58 YOUNGER VS. OLDER Stinchfield, R. (2002). Reliability, validity, and classification accuracy of the South Oaks Gambling Screen (SOGS). Addictive Behaviors, 27, Stinchfield, R. (2003). Reliability, validity, and classification accuracy of a measure of DSM-IV diagnostic criteria for pathological gambling. American Journal of Psychiatry, 160, Weatherly, J.N. (2013). The relationship between endorsing gambling as an escape and the display of gambling problems. Journal of Addiction, 2013, 7 pages. Weatherly, J.N., Aoyama, K., Terrell, H.K., & Berry, J.C. (in press a). Comparing the Japanese version of the Gambling Functional Assessment Revised to an American sample. Journal of Gambling Issues. Weatherly, J.N., & Derenne, A. (2012). Investigating the relationship between the contingencies that maintain gambling and probability discounting of gains and losses. European Journal of Behavior Analysis, 13, Weatherly, J.N., Dymond, S., Samuels, L., Austin, J.L., & Terrell, H.K. (in press b). Validating the Gambling Functional Assessment Revised in a United Kingdom sample. Journal of Gambling Studies. Weatherly, J.N., Miller, J.C., Montes, K.S., & Rost, C. (2012). Assessing the reliability of the Gambling Functional Assessment Revised. Journal of Gambling Studies, 28, Weatherly, J.N., Miller, J.C., & Terrell, H.K. (2011). Testing the construct validity of the Gambling Functional Assessment Revised (GFA-R). Behavior Modification, 35, Weatherly, J. N., Montes, K. S., & Christopher, D. M. (2010). Investigating the relationship between escape and gambling behavior. Analysis of Gambling Behavior, 4, Action Editor: Andrew Brandt

29 Analysis of Gambling Behavior 2013, 7, Number 2 (Winter, 2013) Cumulative Frequencies of Behavior Analytic Journal Publications Related to Human Research on Gambling Benjamin N. Witts St. Cloud State University This brief review investigates the frequency with which gambling-related articles in behavior-analytic journals are being published. Results of this investigation yield important information regarding the recent growth of gambling-related publications within the behavioral community. Specifically, cumulative frequencies show an increasing trend in gambling-related publications between 1979 and 2012, with substantial growth occurring after These data are considered in light of Dixon s (2007) editorial in the inaugural issue of the Analysis of Gambling Behavior. Keywords: Gambling; Gaming; Cumulative record; Publication trends Behavior analysis has been relegated as being out of touch with what gamblers face. For example, Griffiths (1999) stated that no simple parsimonious explanation of gambling maintenance [i.e., operant conditioning] will ever be sufficient to explain all cases (p. 444), and went on to say that reinforcement contingencies are simple (Griffiths & Parke, 2003; p. 12). The potential for behavior analysts to be viewed as novices in the gambling research arena or to be accused of being unable to adequately explain gambling behavior with behavioral principles is an issue Dixon (2007) addressed as a possible outcome of behavior analysts relative inaction toward gambling research. Dixon pointed out that in the years from 1997 to 2006 the rate of PsycINFO returns for the keywords Pathological Gambling increased from 29 to 142. This search was recently conducted again, and as of July, 2013, the 1997 and 2006 results increased to Address all correspondence to: Benjamin N. Witts Community Psychology, Counseling, & Marital Therapy St. Cloud State University St. Cloud, MN benjamin.witts@gmail.com 34 and 200, respectively 1, with an additional 264 being published in Of course, broader terms used in PsycINFO like Gambling yield bigger returns (e.g., 572 for 2012). Thus, in line with Dixon s original argument, while there are growing trends in gambling publications, the possibility that behavior analysts are not devoting the attention toward gambling that other branches of the behavioral sciences are certainly exists. In the inaugural issue of the Analysis of Gambling Behavior, Dixon (2007) offered two solutions to deal with the paucity of behavior-analytic research on gambling. First, Dixon argued for the establishment of a peerreviewed behavior-analytic journal on the subject of gambling, which the publication of the Analysis of Gambling Behavior accomplished. The second solution was to convince the behavioral community to turn their attention toward the issue of problem gambling, with the understanding that our field is best suited to address the needs of those identified as problem gamblers. There is evidence of increased interest in gambling research among behavior analysts (e.g., Weatherly, 2011); however, answering the question of where 1 This change may be due to increased journal coverage in PsycINFO and potential changes in search algorithms, though this is only speculative. 59

30 60 CUMULATIVE FREQUENCIES these interests lay and how they have changed with time may serve to identify growing trends and areas that need to be addressed further. While Dixon (2007) attended to problem gambling in particular, a broader treatment of the topic of gambling is necessary to build a foundation upon which behavior analysts can better respond to the need of effective treatments for that population. It is in this light that the current investigation is presented, which focuses on articles published in behavior-analytic journals that deal directly with issues related to human gambling. Results are discussed in terms of trends across journals and article types. The review concludes with a brief discussion of the future of gambling research in behavior analysis. METHOD Article Inclusion Articles were selected from seven behavior-analytic journals that have published articles on gambling up to and including the year Other journals were considered, but they either had not published gambling research (e.g., Behavior Analysis in Practice) or were not exclusively behavioral in their focus (e.g., Journal of Gambling Studies), despite having published some behavioral articles (e.g., Weatherly, Miller, Montes, & Rost, 2012). The journals included in the analyses were the Journal of Applied Behavior Analysis (JABA), the Journal of the Experimental Analysis of Behavior (JEAB), Behavior Modification (Bx Mod), The Psychological Record (TPR), The Behavior Analyst (TBA), Behavior Research and Methods (BRM), and the Analysis of Gambling Behavior (AGB). Searches were conducted through the journal s abstract search option (JABA, JEAB), through PsycINFO (Bx Mod, TPR, TBA, BRM), or through reading each hardcopy of the journal (AGB 2 ). The following terms were used for abstract and PsycINFO searches; gambling, gaming, casino, slot machine, poker. Articles were removed from the final list if the article did not limit its content to issues of human subjects. Variable Selection Journal name and year of publication were recorded for each article and served as a focal point of analysis in terms of both total articles per journal and cumulative publications over time. Article types were categorized as experimental, descriptive, methodological, theoretical, treatment, book review, technical, or other. Experimental articles manipulated some variable and measure its effect on at least one dependent variable. Descriptive articles investigated differences in measures between populations (e.g., delay discounting between pathological and nonpathological gamblers). Methodological articles provided guidelines on approaches to research topics. Treatment articles focused on the improvement of problem gambling. Book review articles, of course, provide a review of a gambling-related book. Technical articles investigated instrument development. Lastly, any article that did not meet criteria for the above mentioned article types were relegated to the category of other, and included articles such as commentaries and author responses. Three overarching populations of participants were identified; problem gamblers, nonproblem gamblers, and non-gamblers/other. Problem gamblers consisted of any identified pathological or problem gamblers. Nonproblem gamblers were comprised of recreational gamblers and those identified as historically gambling without identified pathology. The last category, non-gambler/other, was inclusive of studies in which no classificatory 2 Alternatively, sources like Google Scholar may make these searches possible for journals not listed on databases like PsycINFO.

31 BENJAMIN N. WITTS 61 tool was used to identify gambling severity, when the population consisted of undifferentiated individuals (e.g., university students), or those who reportedly never gambled. With these divisions, it was possible for one study to have two or more population categories endorsed in the final analysis. The type of gambling was codified and included slot machines, horse racing, poker (both video and live), blackjack, dice/craps, lottery, sports betting, dog racing, roulette, other (e.g., casino games ), and modified board games for children. Again, it was possible for more than one category to be included in any given study. Lastly, the incentive strategy was recorded. Five divisions of incentives were identified, and included extra credit for coursework, remaining money won after participating in the game (e.g., having $15 left after playing a slot machine would result in $15 granted to the participant), a percent chance at some monetary prize (e.g., drawing one participant name after the study to award them $25), a gift card upon completion of the study, and an additional category for those studies that reportedly gave no incentive for participation. It was possible for a study to endorse more than one type of incentive for participation. RESULTS AND DISCUSSION A total of 119 articles were identified across the seven journals subjected to analysis, beginning in 1979, and a cumulative graph of these publications by year is presented in Figure 1. Once the articles were selected, inter-observer agreement was calculated with a second observer for 28.57% (n = 34) of the articles across all variables of interest and achieved 100% agreement for occurrence and nonoccurrence of each variable. Publications by journal, article type, population, game type, and incentive are shown in Table 1. Data show that AGB is the most frequent publisher of gambling articles of the seven (n = 82; 68.91% of all articles), experimental studies are the most frequently published study type (42.02% of all articles), slot machines are the most cited gaming type investigated (24.37% of all articles), and extra credit is the most frequently reported type of incentive for gambling research (46.00% of all experimental articles). Data regarding population type show more non-problem gamblers (48.00% of experimental articles) in research studies than problem gamblers (32.00%) and nongamblers (14.00%) combined. However, when considering all factors analyzed, there Articles Published Publication Year Figure 1. Cumulative gambling-related publications across all analyzed journals.

32 62 CUMULATIVE FREQUENCIES Category Variable n Journal Article Type Population Gaming Type Incentive AGB 82 JABA 13 TPR 10 Bx Mod 5 BRM 4 TBA 3 JEAB 2 Experimental 50 Descriptive 17 Theoretical 9 Treatment 8 Technical 6 Book Review 2 Methodological 1 Other 26 Problem 16 Non-Problem 24 Non- Gambler/Other 7 Slot Machine 29 Poker 14 Horse Racing 5 Roulette 5 Blackjack 4 Dice/Craps 3 Sports Betting 3 Board Game 2 Lottery 2 Dog Racing 2 Other 1 Extra Credit 34 Money Won 15 Gift Card 11 % Chance at $ 10 None 2 Table 1. Number of articles by journal, article type, population, gaming type, and incentive. appear to be a mixture of article types, populations, foci of games studied, and incentives, demonstrating a diversity of interests and strategies amongst researchers. Figure 2 is a cumulative frequency of five article types. Book reviews and methodological types were removed from this search as their frequencies were too low to determine any specific trends. The category other was also removed as this category was too wide-ranging and may have produce a false perception of unity among those articles that potentially unrelated articles are trending in the same direction when they may not. The remaining types, descriptive, experimental, theoretical, treatment, and technical, show similar frequencies prior to However, from 2006 to 2012, the experimental type showed the greatest growth in terms of numbers of publications, followed by the descriptive type. The remaining three types are relatively similar in their trends from 2006 to One issue arose during coding. Specifically, gaming type proved difficult to code as several studies were conducted within a specific gaming venue (e.g., off-track betting parlor), which may have been unrelated to the study s focus. For example, Dixon, Marley, and Jacobs (2003) sample for studying delay discounting by pathological gamblers consisted of individuals betting on horse races at an off-track location. In this example, the gaming type was recorded as horse racing, which may or may not be related to the purpose of the study, which was to examine delay discounting rates. Additional articles that are arguably related to gambling were not included in this review. For example, articles on risky choice, which can be found in the discounting literature, were not included unless they focused on some aspect of gambling (e.g., problem gamblers). Additionally, this review is reported in light of Dixon s (2007) concern regarding behavior analysis pursuit

33 BENJAMIN N. WITTS 63 Articles Published Experimental Descriptive Theoretical Treatment Technical Publication Year Figure 2. Cumulative publications across article type. of gambling research. There exist, however, non-behavioral journals which may be of interest in other investigations regarding the impact behavior analysts might be having in the larger domain of gambling research. Such an investigation is beyond the scope of this article, and may require additional time before any discernible trend or impact emerges. In sum, the current review shows that the pursuit of gambling research is growing within our field. Specifically, cumulative records indicate an increasing frequency of gambling-related publications in seven behavior-analytic journals with a diverse range of topics and populations. With respect to Dixon s (2007) second solution to the inactivity of behavior analysts pursuit of gambling research, that of increased interest, we can conclude that as a field we are certainly moving in the right direction. These findings are promising for those starting lines of research at university settings. Specifically, the faculty member may be asked to supervise or produce research in basic, applied, and/or theoretical domains, and potentially under budgetary restrictions preventing the use of expensive equipment. Research in gambling addresses each of these concerns (e.g., Dixon & Holton, 2009; Arntzen & Stensvold, 2007; Porter & Ghezzi, 2006; Dixon & Schreiber, 2002, respectively). Additionally, the National Centers for Responsible Gaming provides grant funding specifically for gaming research, and has financially supported behavioral approaches to treatment. Both awards listed on NCRG s grant site ( were for research investigating the utility of Acceptance and Commitment Therapy in the treatment of near miss magnitude effects in underage pathological gamblers. This suggests that not only is gambling research in behavior analysis expanding, it is also capable of being financially supported from outside sources. While the increase in rate of publications in gambling by behavior analysts is promising, there is still much work to do in the refinement of current investigations and the identification and pursuit of new avenues of research. For example, research in the role of money in gaming laboratories has yielded interesting results, though more work is certainly needed. Weatherly and Meier (2007) and Peterson and Weatherly (2011) conclude that participants gamble similarly for a chance at a gift card as they do for real money. Additionally, participants economic statuses may play a signifi-

34 64 CUMULATIVE FREQUENCIES cant role in how they gamble in these laboratory conditions (e.g., Peterson & Weatherly, 2011; Weatherly, Mari, & Montes, 2012). What remains to be seen, however, is how the outcomes of these studies compare to gambling in the natural environment. Indeed, that risk of previouslyearned conditioned reinforcers (i.e., money) is at stake for the gambler is a point of concern in the development of gambling laboratories interested in external validity. Caution must be taken, however, as wagering with one s own or someone else s money in non-casino environments may be illegal in the researcher s state (see Dodds, 2008). That is, researchers may be legally restricted from laboratory work in which the participant gambles with their own money or where the credits earned in the lab are exchangeable for money. This last point, of legal restrictions, will require creative solutions in the pursuit of external validation of current methodologies, and efforts are underway which may help us answer these questions (cf. Brandt, Sztykiel, & Pietras, 2013). The role of money in gambling laboratories is but one of myriad issues to be resolved by the behavior-analytic community. For example, investigations into slot machine bonus rounds has remained untouched, and may be of interest in investigations looking at the formation or maintenance of problem gambling. Gambling research may also be the foundation from which newer theoretical approaches might be pursued, such as with relational frame theory. Indeed, the relational framing of poker rules (e.g., Hoch, Witts, & Weil, 2012; 2013; cf. Dymond & Roche, 2010), and the internet poker player (e.g., Costello & Fuqua, 2012; Witts & Lyons, 2013) may help to elucidate how to prevent or treat problem gambling, as those understanding the mathematical rules may be less inclined to gamble in a risky fashion (e.g., Costello & Fuqua, 2012) though rigorous peer-review is currently limited. As of today we have more questions than we have answers, making this a truly exciting time to be a researcher interested in gambling. REFERENCES Arntzen, E., & Stensvold, J. (2007). Treatment of compulsive gambling. Analysis of Gambling Behavior, 1, Brandt, A. E., Sztykiel, H., & Pietras, C. J. (2013). Laboratory simulated gambling: Risk varies across participant-stake procedure. The Journal of General Psychology, 140, Costello, M. S., & Fuqua, R. W. (2012). A brief behavioral intervention of harm reduction for online poker players. Analysis of Gambling Behavior, 6, Dixon, M. R. (2007). Why behavior analysts should study gambling behavior. Analysis of Gambling Behavior, 1, 1-3. Dixon, M. R., & Holton, B. (2009). Altering the magnitude of delay discounting by pathological gamblers. Journal of Applied Behavior Analysis, 42, Dixon, M. R., Marley, J., & Jacobs, E. A. (2003). Delay discounting by pathological gamblers. Journal of Applied Behavior Analysis, 36, Dixon, M. R., & Schreiber, J. B. (2002). Utilizing computerized video poker simulation for the collection of data on gambling behavior. The Psychological Record, 52, Dodds, D. (2008, Autum). Beating the odds: UND gets a law changed to create a casino laboratory. UND Discovery. Retrieved from discoveryaryarchive/fall_2008/gamblingbehaviors.htm l

35 BENJAMIN N. WITTS 65 Dymond, S., & Roche, B. (2010). The impact of derived relational responding on gambling behavior. Analysis of Gambling Behavior, 4, Fantino, E. (2008). Behavior analysis: Thriving, but how about its future? Journal of the Experimental Analysis of Behavior, 89, Griffiths, M. D. (1999). The psychology of the near-miss (revisited): A comment on Delfabbro & Winefield (1999). British Journal of Psychology, 90, Griffiths, M. D., & Parke, J. (2003). The psychology of the fruit machine. Psychology Review, 9, Hoch, V. A., Witts, B. N., & Weil, T. (2012, September). Cards, Chips, & Science: An Analysis of Rule-Governance in Gambling Behavior. Paper presented at the Florida Association for Behavior Analysis annual convention, Riverfront, FL. Hoch, V., Witts, B. N., & Weil, T. (2013, March). Derived Stimulus Relations in the Analysis of Poker Play. Poster presented at the Gambling Special Interest Group annual Conference, Reno, NV. Peterson, J. M., & Weatherly, J. N. (2011). Comparing three strategies of motivating gambling behavior in the laboratory environment. Analysis of Gambling Behavior, 5, Porter, J. C., & Ghezzi, P. M. (2006). Theories of pathological gambling. In P. M. Ghezzi, C. A. Lyons, M. R. Dixon, & G. R. Wilson (Eds.), Gambling: Behavior Theory, Research, and Application (pp ). Reno, NV: Context Press. Weatherly, J. N. (2011). Editorial comment: Turning the corner at Analysis of Gambling Behavior. Analysis of Gambling Behavior, 5, 3-4. Weatherly, J. N., & Meier, E. (2007). Studying gambling experimentally: The value of money. Analysis of Gambling Behavior, 1, Weatherly, J. N., Mari, K., & Montes, K. S. (2012). Gambling in a laboratory setting: A comparison of gambling for positive reinforcement versus as a potential escape. Analysis of Gambling Behavior, 6, Weatherly, J. N., Miller, J. C., Montes, K. S., & Rost, C. (2012). Assessing the reliability of the Gambling Functional Assessment: Revised. Journal of Gambling Studies, 28, Witts, B. N., & Lyons, C. A. (2013). Factors correlated with persistence in online Texas Hold em poker play. Analysis of Gambling Behavior, 7, Action Editor: Jeffrey N. Weatherly

36 This journal is Copyright 2013 by Jeffrey N. Weatherly, publisher, Analysis of Gambling Behavior. All rights are reserved. All information contained within is provided as is. The AGB journal, its publisher, authors, and agents, cannot be held responsible for the way this information is used or applied. The AGB journal is not responsible for typographical errors. Analysis of Gambling Behavior (AGB) is published twice per year (summer and winter). AGB is a print journal and back issues are available online. AGB is an independent publication and is in no way affiliated with any other publications. The materials, articles, and information provided in this journal have been prepared by the staff of the AGB journal for informational purposes only. The information contained in this journal is not intended to create any kind of patient-therapist relationship or representation whatsoever. Make sure to subscribe to the newest and most scientifically advanced journal for understanding gambling behavior. Back Issues are Available 1 year subscription (2 issues published per year) $50 USD Professionals and Institutions $20 USD Students (with documented student Status) Send check or money order to: Analysis of Gambling Behavior Department of Psychology University of North Dakota Grand Forks, ND Phone: (701) Fax: (701)