The Psychological Record, 2011, 61, 93–112
Slot Machine Preferences of Pathological and
Recreational Gamblers Are Verbally Constructed
Mark R. Dixon, Holly L. Bihler, and Becky L. Nastally
Southern Illinois University Carbondale
The current study attempted to alter preferences for concurrently
available slot machines of equal payout through the development of
equivalence classes and subsequent transfers of functions. Participants rated stimuli consisting of words thought to be associated with
having a gambling problem (e.g., desperation and debt), words associated with overcoming a gambling problem (e.g., happiness and
wealth), and various colors according to level of “pleasantness.” Next,
participants alternated play between two simulated casinos, each containing two slot machines. Equivalence classes were formed between
pleasantly and aversively rated stimuli, stimuli associated with the slot
machines at each casino, and neutral color stimuli. Intermittent reexposure to the various casino and slot machine options demonstrated
that preferences could be altered based on derived transfer of emotive
functions via equivalence relations.
Key words: choice, gambling, equivalence, transfer of function, slot machine
With Americans spending more than $90 billion on gambling in 2006
(American Gaming Association, 2007), our country may have a new favorite
pastime. It is now possible to participate in legalized gambling in every state
except Hawaii and Utah (Ghezzi, Lyons, & Dixon, 2000), and casinos across the
country attract millions of tourists every year. However, it appears that with
the increased availability of gambling comes an unfortunate consequence.
Although casinos market their product as an innocent form of recreation, the
less publicized side of the gaming industry is one of potential gambling disorders that can result in significant financial problems, family dysfunction, and
various forms of criminal behavior (Shaffer & Kidman, 2004). Prevalence rates
for this level of gambling behavior have now reached 1–2% of the adult population worldwide (Petry, 2005). The natural agenda for researchers interested
in gambling is to determine how this behavior goes from being a harmless
leisure activity to an addiction that adversely affects one’s life.
For researchers from the behavior-analytic tradition, the task has been
to address why some gamblers develop pathology and some do not, even
though they are exposed to identical programmed contingencies when
Correspondence concerning this article should be addressed to Mark R. Dixon, Behavior
Analysis and Therapy Program, Rehabilitation Institute, Southern Illinois University Carbondale,
IL 62901. E-mail: [email protected]
94
Dixon et al.
engaging in gambling activities. However, variability of human responding
on reinforcement schedules compared to that of nonhumans was observed in
basic operant experiments early in the field’s history (Baron & Galizio, 1983;
Galizio, 1979). It is accounting for this difference with a single behavioral
mechanism that has been of great interest within the behavioral community,
and this question is at the heart of a behavioral analysis of gambling. Dixon
and Delaney (2006), among others, have suggested the differences observed
between gamblers are due to the role of verbal behavior and its impact on
subsequent responding (gambling). The majority of research in this area has
focused specifically on the effect of derived verbal rules on gambling (Hoon,
Dymond, Jackson, & Dixon, 2008; Johnson & Dixon, 2009; Nastally, Dixon, &
Jackson, 2010; Zlomke & Dixon, 2006). These types of rules may affect the
choice making of a gambler without ever being paired directly with the contingencies involved in gambling. Instead, it is possible that these rules may
form via a transfer of stimulus functions once some relations (though not
all) among a set of functionally equivalent stimuli have been directly trained.
The important point here is that stimuli need not be topographically equivalent. For example, if A is the same as B, and A is the same as C, a verbally
competent human will be able to derive that B is the same as C without any
direct reinforcement. Additionally, the functions of B will transfer to the C
stimulus.
Several studies have demonstrated that stimulus functions can frequently
transfer through members of equivalence classes (e.g., Barnes-Holmes, Keane,
Barnes-Holmes, & Smeets, 2000; Hoon et al., 2008; Johnson & Dixon, 2009;
Zlomke & Dixon, 2006). For example, a study by Barnes-Holmes et al. (2000)
examined how this transfer of functions can affect participant preferences for
brands of soft drinks. Participants initially tasted and provided “pleasantness”
ratings for two glasses of soda with identical ingredients labeled as Brand X
and Brand Y. Upon completion of the soft drink ratings, two equivalence
classes were trained (cancer
vek, vek
Brand Y; holidays
zid, zid
Brand X). Following the matching-to-sample training and testing procedures,
four out of six participant ratings dramatically increased for the soda brand
that participated in the equivalence class with the word holidays and decreased
for the brand that was a member of the class containing the word cancer. It is
important to note that preferences were created for one brand of soda over the
other even though the ingredients were identical. It may be possible that some
preferences for certain gambling activities over others are created through
similar transfers of function even though rates of payout may be the same.
To illustrate how preference may be manipulated via a transfer of function in a gambling context, Zlomke and Dixon (2006) conducted a study in
which participants were initially asked to respond to two concurrently available slot machines differing only in color (one yellow and one blue). On average, in pretesting participants demonstrated very little preference for one slot
compared to the other. However, following a matching-to-sample procedure
that differentially reinforced the selection of the comparison stimuli that was
greater than the sample in the presence of a yellow contextual cue, eight out
of nine participants preferred the yellow slot machine. It appeared that the
function of “greater than” had been transferred to the yellow contextual cue
and because the slot machine shared formal properties with this cue, participants responded differently than they had in pretesting, even though the payout rates of the slot machine remained the same. This study provided initial
ALTERING SLOT MACHINE PREFERENCES
95
evidence that recreational gamblers’ preferences may be altered through the
transfer of stimulus functions.
Hoon et al. (2008) replicated the results of Zlomke and Dixon (2006) by
utilizing gambling-related stimuli in a relational training procedure consisting of only two comparison stimuli at a time, as opposed to the traditional
matching-to-sample procedure. Five out of six participants demonstrated
a preference for the slot machine that shared contextual features with the
greater-than cue following the training procedure. The effect of function
transfer on choice making in a gambling context has even been demonstrated
with children (Johnson & Dixon, 2009). During a children’s board game, preferences for red- or blue-colored die were increased or decreased from baseline based on matching-to-sample training procedures that established relations between “greater than” or “less than” and the two colors.
However, across behavioral, and nonbehavioral, approaches to gambling,
questions have been raised about the external validity of studying gambling
under simulated conditions in artificial settings (Delfabbro, 2004). Although
the previously cited studies addressed the component of verbal behavior
that definitely applies to gambling, an experimental procedure involving a
two-choice paradigm may not accurately represent the complex, context­
dependent choices that gambling may involve. To more accurately depict that
process, experiments should expand their methodology to include multiple
(at least more than two) options for participant response patterns to occur. It
is only through these means that we may more closely represent the multifaceted networks in an individual whose history is impossible to fully know.
Furthermore, none of the prior published studies by behavioral researchers
that have appeared in peer-reviewed journals have utilized actual pathological gamblers—the clinical population of interest.
Therefore, the purpose of the current study was to extend the findings of
prior studies on the transfer of stimulus functions in a gambling context by
expanding the methodology used to assess preferential responding in gamblers—both pathological and nonpathological.
Method
Participants and Screening
Four participants (two men and two women, ages 24 to 40 years) were
asked to fill out the South Oaks Gambling Screen (SOGS; Lesieur & Blume,
1987). The SOGS is a 20-item self-report screening instrument and is the most
frequently used standardized assessment of problem gambling (Shaffer, Hall,
& Vander Bilt, 1999), with a score of 5 or more indicative of probable pathological gambling. SOGS scores for Participants 1 through 4 were 5, 7, 0, and
9, respectively, although all participants reported prior experience with some
form of gambling. The experimenter actively recruited participants through a
public posting, and the participants received a gift card and/or course extra
credit in exchange for participating.
Setting and Apparatus
The experiment took place in a small, unoccupied room in a university
laboratory setting. Participants had no contact with other participants during
96
Dixon et al.
the course of the experiment in order to reduce the possibility of communication of any kind. The experiment was presented on an HP Pavilion laptop
computer and was programmed using Visual Basic.NET.
Experimental Design and Procedure
Table 1 graphically represents the procession of each phase in the experiment. Participants were exposed to a multiple schedule of reinforcement in
which each component consisted of a concurrent schedule yielding reinforcement 33% of the time. This reinforcement percentage differed from other similar published studies on gambling (Hoon et al., 2008; Zlomke & Dixon, 2006)
that have used a 50% payback percentage and was utilized to extend the external validity of the investigation. Each component of the multiple schedule was
presented on every other trial. After the presentation of 60 trials of the multiple
schedule, participants were exposed to various types of conditional discrimination training and testing procedures and were reexposed to the multiple schedule immediately following each of these training and testing procedures.
Phase 1: Stimuli rating. Participants were presented with 22 total stimuli,
one at a time, and were asked to rate each one on a scale from 1 (I don’t like it
at all) to 10 (I like it very much). Each stimulus was randomly presented three
times. Three types of stimuli were presented. One set of stimuli consisted of
seven words that describe possible negative consequences of problem gambling:
death, AIDS, desperation, debt, divorce, hate, and pain. Another set of stimuli
consisted of seven words that may be seen as positive effects of avoiding problem gambling: wealth, health, happiness, family, love, pleasure, and hope. The
third type of stimuli consisted of eight colored squares. The colors used were
red, orange, yellow, green, blue, purple, pink, and white.
The ratings obtained during this part of the experiment allowed the experimenter to create participant-specific equivalence classes for each participant. In order to examine transfer of functions and change preference, it is
important to be sure that stimuli do in fact have the presumed pleasant or
unpleasant emotive functions. The two words rated the lowest (the most aversive), the two words rated the highest (the most pleasant), and the four colors
rated most neutrally (closest in absolute value to 5) were used to create equivalence classes. The rest of the previously rated words were not used during any
other part of the experiment. For clarity, the word rated as most aversive will
be referred to as Aversive 1, the word rated as the second most aversive will be
referred to as Aversive 2, the words rated as the most pleasant will be labeled
in the same fashion as Pleasant 1 and Pleasant 2, and the neutrally rated colors
will be referred to as Neutral 1 through Neutral 4. The instructions for this
phase were as follows:
You will be presented with a series of words and pictures. Read
each word and look at each picture carefully. After you have done
so, click on the word or picture using the cursor. After you click
on the word or picture, a rating scale will appear. Please rate the
word or picture based on how you feel about it on a scale from 1
being “I don’t like it at all” to 10 “I like it very much.” If you have
no feelings about the word/picture or feel neutral, rate the word or
picture a 5. To rate the word or picture, slide the pink bar to the
number (1–10) you wish to rate the word or picture. After you have
rated the word or picture click on the button labeled “CONTINUE.”
ALTERING SLOT MACHINE PREFERENCES
97
Table 1
A Schematic Overview of All Phases Within the Experiment
Phase 1:
Stimuli Rating
Phase 2:
Initial Casino Exposure
Phase 3:
Conditional Discrimination Training of Casino-A Stimuli
2 sets of training A–B, A–C, mixed A–B/A–C; test B–C & C–B
Two 3-member classes: Pleasant & aversive
Phase 4:
Second Casino Exposure
Phase 5:
Altered Conditional Discrimination Training of Casino-A Stimuli
1 set of mixed training A1–C2/A2–C1; test C–A
Phase 6:
Third Casino Exposure
Phase 7:
Conditional Discrimination Training of Casino-B Stimuli
2 sets of training A–B, A–C, mixed A–B/A–C; test B–C & C–B
Two 3-member classes: second-most pleasant & aversive
Phase 8:
Fourth Casino Exposure
Phase 9:
Altered Conditional Discrimination Training of Casino-B Stimuli
1 set of mixed training A3–C4/A4–C3; test C–A
Phase 10:
Final Casino Exposure
The words and pictures presented appeared in the center near the top of
the screen. The scale was positioned below the stimuli, with the numbers 1
Table
2
through 10
labeled
above the sliding scale. The numbers 1, 5, and 10 were
Equivalence
Classes
and Preference
for Stimuli
During
labeled “I don’t
like it at
all,” “neutral,”
and “IPercentages
like it very much,”
respectively.
Slot Machine
Phasesthis
for stage,
Participant
1
Phase Equal-Probability
2: Initial casino exposure.
During
participants
selected
Phasecomputer-simulated
2
Phase 4
Phase
6
Phase
8
Phase 10
between and played two different
slot machines
at two
different “casinos”
Casino 1(Casino 1 and Casino 2). Therefore, between the two casinos
A1 (Win)
90
80
27
93
100
B1 (Pleasure)
100
98
Dixon et al.
there were four total slot machines. At each casino, participants were presented with a choice screen with two buttons. Each of these buttons represented a different slot machine. The buttons at Casino 1 were labeled “Win”
and “Lose.” The buttons at Casino 2 were labeled “Jackpot” and “Bankrupt.”
These words were considered relative opposite gambling-related stimuli, and
while it may be assumed that a participant would prefer slot machines labeled
“Win” and “Jackpot” over “Lose” and “Bankrupt,” our concurrent-operants
procedure allowed this to be assessed experimentally. Selecting one of the
buttons resulted in the presentation of the corresponding slot machine. After
one spin, or trial, participants switched to the other casino, where they were
presented with the choice screen for the other casino.
The casinos looked almost identical in form. The only differences were that
each was labeled (Casino 1 or Casino 2), the background color was different
(one was light orange and the other was bright red), and the buttons for each
slot machine were labeled. The slot machines were identical in form, that is,
there were no physical differences among the slot machines. All the computersimulated slot machines were programmed on a variable ratio (p = .33) schedule. In addition to the visual indication of a winning spin (the three pictures
lined up on the payoff line and the addition of credits), the participant also
heard an auditory sound while the word “Winner” flashed on the screen.
Participants completed a total of 80 trials during this initial phase. The
first 10 trials at each casino were forced (five trials on each slot machine).
During the forced trials, only one button was present on each casino’s choice
screen (instead of two buttons). Upon selecting that button, the participant
played the corresponding slot machine for one trial. Upon completion of the
trial, play switched to the other casino where only a single slot machine was
available. Play alternated between casinos and slot machines until the five
forced trials were completed at each casino.
After the forced trials, both of the buttons representing each of the casinos were present on the choice screens and participants were permitted
to select between the two slot machines at each casino. After completion of
the forced trials, the participant started out at Casino 1 and played one trial
on his/her slot machine of choice. Once the trial was over, the participant
moved to Casino 2 and was given the choice between two slot machines
at that casino. Participants continued to alternate between Casino 1 and
Casino 2 and continued to choose which slot machine they preferred to play
at each casino until they completed 60 choice trials (80 total trials).
The following directions were presented to participants during this phase:
During the following part of the experiment, you will be presented with a screen with one or two buttons. The buttons will
bring you to the corresponding slot machine. Therefore, after selecting one of the slot machines by clicking on one of the buttons,
you must play the slot machine by clicking on the button labeled
“SPIN.” After pressing the spin button, the slot machine will proceed to spin and stop, resulting in either a win or loss of tokens.
If all three of the pictures land on the payoff line, you will be
awarded with credits; any other combination of stimuli will result
in a loss of credits. Your current total number of credits will be
displayed in the upper right-hand corner. After you play the slot
machine, you will be presented with a button that reads “PRESS
ALTERING SLOT MACHINE PREFERENCES
99
HERE TO CONTINUE.” Press the button and you will be taken to
the other casino. This casino and your credits earned are totally
separate from the other casino. Therefore, credits you earn at one
of the casinos do not transfer or mix with the credits at the other
casino. This casino will look identical to the other casino and will
operate in the same manner. After choosing which slot machine
to play and playing that slot machine, you will be returned to the
first casino you played. At this casino you will begin each time at
the choice screen and will get to choose which slot machine you
would like to play next. You will continue to alternate between
casinos and get to choose which slot machines you want to play at
each of the casinos until you are instructed to stop. TRY TO WIN
AS MUCH AS POSSIBLE.
For ease of communication, the casino containing the most preferred
slot machine, where the experimenter attempted to alter preferences
first, will be referred to as Casino A. The other casino will be referred to
as Casino B. These names will be used because either Casino 1 or Casino 2
could have contained the most highly preferred slot machine.
Phase 3: Conditional discrimination training of Casino A stimuli. After
initial preference for each slot machine was assessed, play was discontinued
and a conditional discrimination program immediately began. The matchingto-sample program consisted of the following phases: A–B training, A–C
training, mixed A–B/A–C training, and testing for equivalence. This resulted
in the formation of two 3-member equivalence classes. The purpose of creating these equivalence classes was to strengthen the existing preference
for the preferred slot machine at only one of the casinos via the transfer of
stimulus functions.
Since preferences for either of the slot machines and emotional ratings
of stimuli may differ, the stimuli used to establish equivalence classes and
alter preferences were participant specific. The stimuli that were used and
equivalence classes that were formed were dependent on several different
factors including: (a) which slot machine at which of the two casinos the participant showed the strongest preference for during the second phase, (b)
which words the participant rated as most aversive (Aversive 1 and Aversive
2), (c) which words the participant rated as the most pleasant (Pleasant 1 and
Pleasant 2), and (d) which colors were rated the closest to neutral (Neutral 1,
Neutral 2, Neutral 3, and Neutral 4). One class consisted of A1, the word associated with the most preferred slot machine, B1, the most pleasantly rated
word (Pleasant 1), and C1, one of the neutrally rated colors (Neutral 1). The
other class consisted of A2, the word associated with the less preferred slot
machine (at the same casino as the slot machine in Class 1), B2, the most
aversively rated word (Aversive 1), and C2, one of the neutrally rated colors
(Neutral 2). The following instructions were presented:
In a moment some words and pictures will appear on this
screen. Your task is to look at the word or picture at the top of
the screen and choose one of the two words or pictures at the
bottom of the screen by selecting one of them using the touchpad or mouse. During the beginning of this stage, the computer will provide you with feedback on your performance. If
you make a correct response you will hear a brief high-pitched
100
Dixon et al.
sound and will see the word “CORRECT” flash across the screen.
If you make an incorrect response, you will hear a brief lowpitched sound and see the word “INCORRECT” flash across the
screen. Later on, the task will become more difficult and feedback will no longer be presented. It is important that you try
to make as many correct choices as possible during this entire
stage of the experiment. The more correct responses you make,
the faster you will complete this segment of the experiment.
During each trial, the sample stimulus was centered near the top of the
screen, while the two comparison stimuli were in a single row near the bottom of the screen. Selection of the correct stimulus immediately resulted
in the presentation of a high-pitched sound and the word “Correct” on the
screen. Incorrect selections resulted in the presentation of a low-pitched
sound and the word “Incorrect” on the screen. The next trial began 2 s after
“Correct” or “Incorrect” was presented.
The A–B training consisted of blocks of 18 trials. During each trial block,
the sample stimuli (A1 and A2) were displayed nine times each in a random
sequence. The positions of comparison stimuli (B1 and B2) were randomized.
The criterion for completion of this phase was one block with at least 16 correct responses. If less than 16 correct responses were made, participants began another 18-trial block of A–B training and continued to complete 18-trial
blocks until the criterion had been met.
Upon meeting the criterion for A–B training, the participants immediately began A–C training. This training was identical, except the comparison
stimuli now consisted of C stimuli (C1 and C2). Once the criterion was met,
the participant advanced to mixed A–B/A–C training. If the criterion was not
met, the participant began another 18-trial block of A–C training. They did
not return to A–B training.
Mixed A–B/A–C training consisted of a combination of the first two parts
of training. The order of the trials was randomized. The same auditory and
visual stimuli that were used in the first two parts of the training were presented contingent on correct and incorrect responses. Each trial block consisted of 36 trials. Criterion for completion was 32 of 36 correct responses.
If criterion was not met, only mixed A–B/A–C training was repeated. This
concluded the training segment of Phase 3.
The remainder of Phase 3 consisted of testing trial blocks, during which
the discriminant auditory and visual stimuli were no longer provided contingent on responding. During testing, the emergence of derived relations
was assessed. For one 18-trial block, the sample stimuli consisted of B stimuli and the comparison stimuli consisted of C stimuli. For the other 18-trial
block, the sample stimuli consisted of C stimuli and the comparison stimuli
consisted of B stimuli. These blocks were not repeated. Criterion for equivalence was set at 16 of 18 correct responses. If this criterion was not met, the
participant repeated mixed A–B/A–C training until the criterion was met.
Phase 4: Second casino exposure. After participants met criterion for
equivalence, they were reexposed to the slot machines at Casino 1 and
Casino 2. This phase was identical to Phase 2 except for two important differences: (a) There were no forced trials, and (b) the buttons for the slot machines at Casino A were labeled with any two of the six stimuli from the
newly formed equivalence classes. The choice screens at Casino A randomly
alternated between presentations of A1 and A2, B1 and B2, or C1 and C2. At
ALTERING SLOT MACHINE PREFERENCES
101
Casino B, only A3 and A4 were presented at every choice screen. This phase
consisted of 60 choice trials (30 at each casino).
Phase 5: Altered conditional discrimination training of Casino A
stimuli. The physical layout and the consequences for correct and incorrect
responses were the same as Phase 3; however, this phase consisted of an
18-trial block of mixed A1–C2/A2–C1 training. The sample stimulus consisted of either A1 or A2, and both C1 and C2 served as comparisons during
every trial. The positions of the comparisons were randomized. The criterion
for completion of this phase was one block with at least 16 of 18 correct
responses. If the criterion was not met, the phase was repeated until at least
16 correct responses were made.
Next, the emergence of untrained relations was tested. Discriminant auditory and visual stimuli were no longer provided contingent on responding.
The sample stimuli consisted of either C1 or C2, and A1 and A2 served as
the comparison stimuli. The criterion for this phase was 16 of 18 correct
responses. If the criterion was not met, A1–C2/A2–C1 training was repeated.
Phase 6: Third casino exposure. This phase was identical to Phase 4
with one exception: The buttons at Casino A were labeled with either A1
and A2 or C1 and C2. For example, at Casino 1 the button for Slot 1 might
have been labeled “Win” (A1) and Slot 2 would have been labeled “Lose” (A2).
During the following trial at Casino 1, the buttons for the slots might have
been blue (C2) and purple (C1) boxes.
Up to this phase, only variables related to Casino A had been manipulated, whereas variables at Casino B remained constant. The same types of
conditional discrimination training sessions that took place during Phases 3
and 5 that were intended to alter responding at the first casino were subsequently implemented in an attempt to alter preferences at Casino B.
Phase 7: Conditional discrimination training of Casino B stimuli. This
phase was identical to Phase 3 with the exception of the stimuli used and the
equivalence classes formed. In this phase only stimuli from Casino B were
used. No changes were made at Casino A. One of the classes consisted of A3,
the word label of the most preferred slot machine at Casino B (as assessed by
preference percentage during Phase 6), B3, the second most pleasantly rated
word (Pleasant 2), and C3, one of the neutrally rated colors (Neutral 3). The
other class consisted of A4, the word label of the less preferred slot machine
at Casino B, B4, the second most aversively rated word (Aversive 2), and C4,
one of the neutrally rated colors (Neutral 4).
Phase 8: Fourth casino exposure. After participants met criterion for
equivalence during Phase 7, they were reexposed to the slot machines at
Casino 1 and Casino 2. This phase was identical to Phases 4 and 6 with the
exceptions of the buttons presented. At Casino A the only buttons at each
choice trial consisted of the A1 and A2 stimuli (for example, “Win” and
“Lose”). At Casino B, the choice screen randomly alternated between presentations of A3 and A4, B3 and B4, or C3 and C4. Each choice screen consisted
of one stimulus from each of the newly formed equivalence classes.
Phase 9: Altered conditional discrimination training of Casino B stimuli.
This phase was identical to Phase 5 except that the purpose was to alter baseline
conditional discriminations at Casino B; therefore the training included different stimuli. This phase consisted of an 18-trial block of mixed A3–C4/
A4–C3 training. The sample stimulus consisted of either A3 or A4, and both
C3 and C4 served as comparisons during every trial. During testing for the
Conditional Discrimination Training of Casino-A Stimuli
2 sets of training A–B, A–C, mixed A–B/A–C; test B–C & C–B
Two 3-member classes: Pleasant & aversive
102
Dixon et al.
Phase 4:
Second
Casino Exposure
emergence of untrained
relations,
the sample stimuli consisted of either C3
or C4, while the comparisons consisted of A3 and A4.
Phase 10: Final casino exposure. During this phase only A1 and A2
Phase 5:
were presented at the choice
screen for Casino A. The choice screen at
Conditionalalternated
Discrimination
Training ofA3
Casino-A
Stimuli
CasinoAltered
B randomly
between
and A4
or C3 and C4. All the slot
1 set of mixed training A1–C2/A2–C1; test C–A
machine characteristics were identical to the equal probability payout rates
in Phases 4, 6, and 8.
Phase 6:
Results
Third Casino Exposure
Across Phases 3, 5, 7, and 9, which were comprised of conditional discrimination training, only one participant (Participant 2) required repeated exposure
7:
to any of the training blocksPhase
in order
to meet criterion and advance to Phases 4,
Conditional Discrimination Training of Casino-B Stimuli
6, 8, and
10.
This
occurred
during
A–B training
in &Phase
2 sets of training A–B, A–C, mixed A–B/A–C;
test B–C
C–B 3. Participants 1, 3, and
4 all metTwo
criterion
within
each
phase after
one& trial
block.
3-member
classes:
second-most
pleasant
aversive
The stimuli used during the conditional discrimination training were
dependent on the ratings of the words and colored squares during Phase 1.
Therefore, the classes created
were
Phase
8: participant specific. During the initial conFourth
Casino Exposure
ditional discrimination
training
(Phase 3), two classes were formed. The stimuli for Class 1 were as follows: A1 = word label from the most highly preferred
slot (during Phase 2), B1 = Pleasant 1, and C1 = Neutral 1. The stimuli for Class
Phase
9: from the less preferred slot machine (at the
2 were as follows: A2 = word
label
Altered Conditional
Discrimination
Training
of Casino-B
Stimuli 2. The classes formed
same casino
as A1), B2
= Aversive
1, and
C2 = Neutral
1 set of mixed training A3–C4/A4–C3; test C–A
during Phase 7 were created in the same manner using the word labels from
the slot machines at the remaining casino: Pleasant 2, Aversive 2, Neutral 3,
and Neutral 4. The stimuli that comprised Class 1, Class 2, Class 3, and Class
Phase 10:
4 are shown for each participant
in Tables 2, 3, 4, and 5.
Final Casino Exposure
Percentage of responses allocated during Phase 2 and Phase 6 was
used to identify the preferred slot machines, around which the equivalence
Table 2
Equivalence Classes and Preference Percentages for Stimuli During
Equal-Probability Slot Machine Phases for Participant 1
Phase 2
Phase 4
Phase 6
Phase 8
Casino 1
A1 (Win)
90
80
27
93
B1 (Pleasure)
100
C1 (Orange)
90
67
A2 (Lose)
10
20
73
7
B2 (Death)
0
C2 (Red)
10
33
Casino 2
A3 (Jackpot)
90
97
100
100
B3 (Happiness)
100
C3 (Purple)
100
A4 (Bankrupt)
10
3
0
0
B4 (AIDS)
0
C4 (Yellow)
0
Phase 10
100
0
93
7
7
93
ALTERING SLOT MACHINE PREFERENCES
Table 3
Equivalence Classes and Preference Percentages for Stimuli During
Equal-Probability Slot Machine Phases for Participant 2
Phase 2
Phase 4
Phase 6
Phase 8
Casino 1
A1 (Lose)
67
10
47
50
B1 (Hope)
0
C1 (Pink)
10
53
A2 (Win)
33
90
53
50
B2 (Death)
100
C2 (Blue)
90
47
Casino 2
A3 (Jackpot)
53
50
50
70
B3 (Pleasure)
60
C3 (Green)
40
A4 (Bankrupt)
47
50
50
30
B4 (AIDS)
40
C4 (Purple)
60
103
Phase 10
50
50
73
33
27
67
Table 4
Equivalence Classes and Preference Percentages for Stimuli During EqualProbability Slot Machine Phases for Participant 3
Phase 2
Phase 4
Phase 6
Phase 8
Phase 10
Casino 2
A1 (Jackpot)
63
100
87
97
93
B1 (Hope)
100
C1 (Red)
100
7
A2 (Bankrupt)
37
0
13
3
7
B2 (Debt)
0
C2 (White)
0
93
Casino 1
A3 (Win)
43
90
97
100
100
B3 (Pleasure)
100
C3 (Green)
100
0
A4 (Lose)
57
10
3
0
0
B4 (Divorce)
0
C4 (Orange)
0
100
classes would be built. On some occasions, this resulted in equivalence
classes with stimuli that likely held different emotive functions. For example, as shown in Table 3, Participant 2 allocated a larger number of
responses to the slot labeled “Lose” at Casino 1, resulting in the following
equivalence
classes: A1 = Lose, B1 = Hope, and C1 = Pink for Class 1
Table
5
Equivalence
Classes
and Preference
Percentages
Stimuli
During
Equal- that
and
A2 = Win,
B2 = Death,
and C2 = Pink
for for
Class
2. Table
5 shows
Probability Slot Machine Phases for Participant 4
Phase 2
Phase 4
Phase 6
Phase 8
Phase 10
Casino 1
104
Dixon et al.
Table 5
Equivalence Classes and Preference Percentages for Stimuli During EqualProbability Slot Machine Phases for Participant 4
Phase 2
Phase 4
Phase 6
Phase 8
Phase 10
Casino 1
A1 (Win)
73
100
60
90
57
B1 (Wealth)
80
C1 (Blue)
90
33
A2 (Lose)
27
0
40
10
43
B2 (Debt)
20
C2 (Green)
10
67
Casino 2
A3 (Bankrupt)
37
10
53
70
53
B3 (Hope)
70
C3 (Red)
60
40
A4 (Jackpot)
63
90
47
30
47
B4 (Desperation)
30
C4 (Pink)
40
60
Participant 4 was also trained to form at least two classes containing stimuli
with potentially inconsistent emotive functions (Class 3 and Class 4).
Figures 1, 2, 3, and 4 show a visual representation of response allocation (by preference percentage) across the various phases. The preference
percentages of Participants 1, 3, and 4 were consistent with the equivalence
classes formed or altered during the immediately preceding phase throughout the course of the experiment; that is, their response allocation was similar to optimal responding. See Figure 3 for an example of optimal responding and Figure 1 for a less robust example of optimal responding. These participants allocated a higher percentage of responses to A1 when presented
with the choice of A1 or A2 during Phase 4, just as they had during Phase 2.
They also allocated a higher percentage of responses to the other Class-1
stimuli (B1 and C1) than the Class-2 (B2 and C2) stimuli during Phase 4.
After classes were altered through A1–C2/A2–C1 training during Phase
5, the participants allocated a higher percentage of responses to A1 and C2
or to A2 and C1, suggesting that stimuli took on new emotive functions. For
Participant 1, the negative emotive function of C2 (that was derived during
the initial conditional discrimination training) was transferred to A1. The
emotive functions were transferred in the opposite direction for Participants
3 and 4. The positive emotive function of A1 transferred to C2, while the
negative emotive function of A2 transferred to C1, resulting in more responses allocated to A1 and C2.
After Class 3 and Class 4 were formed during Phase 7, Participants 1, 3,
and 4 allocated a larger percentage of responses to A3, B3, and C3. Following
A3–C4/A4–C3 training during Phase 9, all of the participants allocated more
responses to A3 and C4 than to A4 and C3. Participants 3 and 4 continued
to allocate a higher percentage of responses to A1 in Phase 8 and 10. This is
to be expected since the emotive function of A1 had not been altered during Phase 5 (i.e., the function of A1 transferred to C2 instead of vice versa).
ALTERING SLOT MACHINE PREFERENCES
Preference Percentages for Slot Machines
Class
construction
at Casino 1
100
A1
C1
80
Class
construction
at Casino 2
Alter classes
at Casino 1
B1
A1
105
Alter classes
at Casino 2
A1 Casino 1
A1
C1
60
40
A1
20
0
100
A3
A3
A3
C3 A3 B3
A3 Casino 2
80
60
40
20
C3
0
Phase 2
Phase 4
Phase 6
Phase 8
Phase 10
Figure 1. Preference percentages for slot machines during equal-probability slot
machine phases for Participant 1.
Class
construction
at Casino 1
Class
construction
at Casino 2
Alter classes
at Casino 1
Alter classes
at Casino 2
Casino 1
Preference Percentages for Slot Machines
100
80
60
A1
C1
40
20
C1 A1
0
A1
A1
A1
B1
Casino 2
100
80
60
A3
A3
A3
A3
40
A3
B3
C3
C3
20
0
Phase 2
Phase 4
Phase 6
Phase 8
Phase 10
Figure 2. Preference percentages for slot machines during equal-probability slot
machine phases for Participant 2.
Participant 1, however, who chose A1 on only 27% of trials during Phase 6, allocated 93% and 100% of responses to A1 in Phase 8 and Phase 10, respectively.
Participant 2’s preferences for slot machines were more variable.
Response allocation was altered at Casino 1 during Phases 4 and 6 and at
Dixon et al.
106
Preference Percentages for Slot Machines
Class
construction
at Casino 2
C1 A1 B1
100
80
60
Class
construction
at Casino 1
Alter classes
at Casino 2
Alter classes
at Casino 1
A1
A1
A1 Casino 2
A1
40
20
C1
0
100
C3 A3 B3
A3
A3
A3
Casino 1
80
60
40
A3
20
C3
0
Phase 2
Phase 4
Phase 6
Phase 8
Phase 10
Figure 3. Preference percentages for slot machines during equal-probability slot
machine phases for Participant 3.
Preference Percentages for Slot Machines
Class
construction
at Casino 1
100
80
C1
Class
construction
at Casino 2
Alter classes
at Casino 1
A1
A1 Casino 1
A1
B1
A1
Alter classes
at Casino 2
A1
60
40
C1
20
0
Casino 2
100
80
60
40
A3
C3
A3 B3
C3
A3
20
A3
A3
0
Phase 2
Phase 4
Phase 6
Phase 8
Phase 10
Figure 4. Preference percentages for slot machines during equal-probability slot
machine phases for Participant 4.
Casino 2 during Phases 8 and 10. Preference percentages did not change
at Casino 2 during Phases 4 and 6 or at Casino 1 during Phases 8 and 10
(i.e., response allocations only differed at each casino following conditional
ALTERING SLOT MACHINE PREFERENCES
107
discrimination training intended to alter responding at that casino).
However, the response allocation was not always consistent with the previously trained conditional discriminations. For example, during Phase 2 a
higher percentage of responses was allocated toward A1; therefore, an equivalence class consisting of A1 (Lose), B1 (Hope), and C1 (Pink) was created.
During Phase 4, however, Class-1 stimuli were only chosen on 10%, 0%, and
10% of trials. During Phase 8, A3 was selected on 70% of trials; however, the
preference was lower for B3 (60%) and only 40% of responses were allocated
to C3. Interestingly, during Phases 6 and 10, response allocations were consistent with the A1–C2/A2–C1 and A3–C4/A4–C3 training intended to alter
the previously established equivalence classes.
Discussion
In the current study, preferences for computer-simulated slot machines
were altered via the transfer of stimulus functions. For these participants,
response allocation was higher toward the slot machines labeled with any
three members of an established equivalence class, and the response allocations were subsequently changed following class-altering conditional
discrimination training. The findings from the current study support existing research on the transfer of functions through equivalence relations
(Barnes-Holmes et al., 2000) and, more specifically, research on altering
gambling preferences via the transfer of stimulus functions (Hoon et al.,
2008; Johnson & Dixon, 2009; Zlomke & Dixon, 2006). These results also
support previous research on derived stimulus relations utilizing reversal
of baseline conditional discriminations to strengthen experimental control
of pre- and post-test design methodology (Murphy, Barnes-Holmes, & BarnesHolmes, 2005; Smeets & Barnes-Holmes, 2005; Smeets, Barnes-Holmes,
Akpinar, & Barnes-Holmes, 2003).
For conceptual clarity it should be noted that only those trials in Testing
Phases 4 and 8 that allowed the participant to choose between slot machines
labeled with C-1 and C-2 stimuli (colors) were adequate tests of derived transfer of stimulus functions. Because the conditional discrimination procedure
involved direct training of A–B stimuli (slot name to pleasant/aversive word)
and A–C stimuli (slot name to neutral color), response allocation to slot machines labeled “Win” or “Jackpot” or pleasantly rated words among B stimuli
would only indicate a transfer of function based on symmetrical relations.
Thus, this could be explained through second-order conditioning and would
not be an example of verbal behavior as defined in the derived stimulus–
response literature (Dymond & Rehfeldt, 2000). Likewise, although the evidence of transfer of stimulus function is not robust for each participant in
all phases, the data serves as an initial demonstration of complex verbal processes in gambling participants.
It is possible that during the various phases of conditional discrimination training throughout the experiment, each equivalence class formed
could be looked at as an externally provided rule. For example, Participant 1
may have had an internal rule such as “I am more likely to win while playing the slot machine labeled ‘Win.’” Following the formation of the equivalence class, the participant would have two new external rules including,
“I am more likely to win while playing the ‘Pleasure’ slot machine” and “I
am more likely to win while playing the orange slot machine.” However, it
108
Dixon et al.
is worth noting that this interpretation would assume that the trials involving color stimuli did actually serve as pure instances of function transfer,
and as pointed out earlier, the sequence of discrimination training in the
present experiment may not support this. Further research is necessary to
determine the extent to which preexisting functions, or internally generated rules, can and will transfer to stimuli that most likely carry preexisting
functions themselves, such as emotive words and colors.
Across participants, some variability in responding can be observed.
For example, Participant 2’s preference for slot machines appeared to be affected by the conditional discrimination training during some phases but
not during others. The cause of this variability could be further examined
in an extension of the current study. By videotaping participants and asking
them to explain the “strategy” or “rule” they are following during every trial,
the experimenter may be able to assess which types of rules the participant
is following and thus whether functions of stimuli actually transferred.
Furthermore, the current study could be altered in order to separate the effects of reinforcement contingencies from the effects of rules. Providing immediate feedback (in the form of credits added or removed) for trials at one
casino but not at the other casino would allow for examination of response
allocation with and without reinforcement.
Verbal rules may also explain an interesting change in the response allocation of Participant 3. His response allocation to A3 at Casino 1 increased
at the same time that response allocation increased for A1 at Casino 2 (following conditional discrimination training) during Phase 4 (see Figure 3).
During Phase 3, equivalence classes were formed: Class 1 = A1 (Jackpot), B1
(Hope), and C1 (Red) and Class 2 = A2 (Bankrupt), B2 (Debt), and C2 (White);
these classes were intended to affect responding at Casino 2 only. However,
preference percentage for the available Class-3 stimulus A3 (Pleasure) increased from 43% in Phase 2 to 90% in Phase 4. This participant may have
formed a general rule such as “Good or positive stimuli are better” and applied this rule across all trials.
During Phase 3, the conditional discrimination training was intended
to strengthen or maintain responding at A1 while also creating a preference for B1 and C1 stimuli. However, if the classes formed were in fact
inconsistent, it is possible that the functions of the stimuli may have transferred in either direction, resulting in classes with positive or negative
emotive functions and not necessarily the emotive functions intended by
the experimenter. Participants 2 and 3 formed at least one presumably inconsistent class during the experiment; however, the experimenters could
not be sure that these classes contained stimuli with originally inconsistent functions since the A stimuli were not included in the initial wordrating task.
The data from Participant 1 was notable for various reasons. Participant 1
showed a strong preference for A1 (Win) during Phases 2 and 4. Following
A1–C2/A2–C1 conditional discrimination training he allocated a higher percentage of responses to A2 and C1 than A1 and C2. The negative emotive
functions of the previously neutral C2 (Red) stimulus appeared to transfer to
A1 (Win) instead of vice versa (the functions of A1 transferring to C2). This
is particularly interesting considering that C2 was a neutral stimulus prior
to the formation of the equivalence classes. For Participants 3 and 4, functions transferred in the opposite direction.
ALTERING SLOT MACHINE PREFERENCES
109
To address the limitations (and perhaps the variability in individual responding) in the current study, future investigations could employ several
methodological variations. First, if similar nonarbitrary stimuli are utilized
in future research, experimenters may want to consider training the reversal
of baseline preferences in initial conditional discrimination training phases.
Second, to address similar limitations in the current study, a pretest of slot
machine responding could also be incorporated to detect a more pronounced
shift in preference. And lastly, to avoid or minimize the effect of preexisting
idiosyncratic functions, future researchers might also consider incorporating all arbitrary stimuli, thus removing the current words and colors that
were used in the present investigation. However, doing so might not allow
for the type of evaluation of preexperimental history with various words as
they apply to gambling and risk taking.
Additionally, a future study might consider recruiting a larger number of pathological gamblers or using a between-subjects design to directly
compare the responding of pathological to nonpathological gamblers so that
greater external validity might be obtained. It has been suggested elsewhere
that clinical populations may be less sensitive to experimental contingencies
during human operant investigations (Wulfert, Greenway, Farkas, Hayes, &
Dougher, 1994). In the current study, the question of why some gamblers’ preferences can be altered through the establishment of equivalence classes but
other gamblers’ preferences are unaffected remains unanswered. Researchers
should also expand beyond the frequently used slot machine apparatus and
explore other casino games such as roulette, video poker, and other table
games found in casinos. This will ensure that the findings that are generated
are not an artifact of slot machines but truly generalize to gambling behavior
broadly. Lastly, future research may also look at applying therapeutic treatments to alter existing gambling-related equivalence classes in order to alter
gambling-related behavior (Wilson, Hayes, Gregg, & Zettle, 2001).
In closing, the current study demonstrated that it is possible to alter gambler behavior based on derived transfer of emotive functions. This study lends
support to the existing research in this area of gambling (Hoon et al., 2008;
Johnson & Dixon, 2009; Zlomke & Dixon, 2006) and provides some insight into
idiosyncratic choice-making within this population. It is now evident that programmed contingencies only in part control the behavior of complex, verbal
organisms and studies that utilize them as participants need to take this into
account. By doing so, we can move closer to approximating ways in which to
help those who have been adversely affected by this cultural epidemic.
References
American Gaming Association. (2007). Industry information fact sheets:
Statistics. Washington, DC: American Gaming Association. Retrieved,
February 12, 2009 from http://www.americangaming.org/Industry/
factsheets/statistics_detail.cfv?id=8
Barnes-Holmes, D., Keane, J., Barnes-Holmes, Y., & Smeets, P. M. (2000).
A derived transfer of emotive functions as a means of establishing
differential preferences for soft drinks. The Psychological Record, 50,
493–511.
Baron, A., & Galizio, M. (1983). Instructional control of human operant
behavior. The Psychological Record, 33, 495–520.
110
Dixon et al.
Delfabbro, P. (2004). The stubborn logic of regular gamblers: Obstacles and
dilemmas in cognitive gambling research. Journal of Gambling Studies,
20(1), 1–21.
Dixon, M. R., & Delaney, J. (2006). The impact of verbal behavior to
gambling. In P. M. Ghezzi, C. Lyons, M. Dixon, & G. Wilson (Eds.),
Gambling: Behavior theory, research, and application (pp. 171–191). Reno,
NV: Context Press.
Dymond, S., & REHFELDT, R. A. (2000). Understanding complex behavior:
The transformation of stimulus functions. The Behavior Analyst, 23,
239–254.
Galizio, M. (1979). Contingency-shaped and rule-governed behavior:
Instructional control of human loss avoidance. Journal of the
Experimental Analysis of Behavior, 31(1), 53–70.
Garotti, M., & De Rose, J. C. (2007). Reorganization of equivalence classes:
Evidence for contextual control by baseline reviews before probes. The
Psychological Record, 57, 87–102.
Ghezzi, P., Lyons, C. M., & Dixon, M. R. (2000). Gambling from a
socioeconomic perspective. In W. K. Bickel & R. E. Vuchinich (Eds.),
Reframing health behavior change with behavioral economics (pp.
313–341). New York: Erlbaum.
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(3), 467–470.
Johnson, T. E., & Dixon, M. R. (2009). Influencing children’s pregambling
game playing via conditional discrimination training. Journal of Applied
Behavior Analysis, 42, 73–81.
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(9), 1184–1188.
Murphy, C., Barnes-Holmes, D., & Barnes-Holmes, Y. (2005). Derived
manding in children with autism: Synthesizing Skinner’s verbal
behavior with relational frame theory. Journal of Applied Behavior
Analysis, 38, 445–462.
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, 125–129.
Petry, N. M. (2005). Pathological gambling: Etiology, comorbidity, and
treatment. Washington, DC: American Psychological Association.
Shaffer, H. J., Hall, M. N., & Vander Bilt, J. (1999). Estimating the
prevalence of disordered gambling behavior in the United States and
Canada: A research synthesis. American Journal of Public Health, 89,
1369–1376.
Shaffer, H. J., & Kidman, R. (2004). Gambling and the public health. In J. E.
Grant & M. N. Potenza (Eds.), Pathological gambling: A clinical guide to
treatment (pp. 3–23). Washington, DC: American Psychiatric Publishing.
Smeets, P. M., & Barnes-Holmes, D. (2005). Establishing equivalence
classes in preschool children with one-to-many and many-to-one
training protocols. Behavioural Processes, 69, 281–293.
Smeets, P. M., Barnes-Holmes, Y., Akpinar, D., & Barnes-Holmes, D.
(2003). Reversal of equivalence relations. The Psychological Record, 53,
91–119.
ALTERING SLOT MACHINE PREFERENCES
111
Wilson, K. G., Hayes, S. C., Gregg, J., & Zettle, R. D. (2001).
Psychopathology and psychotherapy. In S. C. Hayes, D. Barnes-Holmes, &
B. Roche (Eds.), Relational frame theory: A post-Skinnerian account of
human language and cognition (pp. 211–239). New York: Kluwer
Academic/Plenum Publishers.
Wulfert, E., Greenway, D. E., Farkas, P., Hayes, S. C., & Dougher, M. J.
(1994). Correlation between self-reported rigidity and rule-governed
insensitivity to operant contingencies. Journal of Applied Behavior
Analysis, 24(4), 659–671.
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(3), 351–361.
112