The Psychological Record, 2006, 56, 83-103
EXPLORING THE DEVELOPMENT AND DISMANTLING OF
EQUIVALENCE CLASSES INVOLVING TERRORIST STIMULI
MARK R. DIXON, RUTH ANNE REHFELDT,
KIMBERLY R. ZLOMKE, and ASHTON ROBINSON
Southern Illinois University
The present paper describes 2 studies that present a
conceptual interpretation and experimental findings involving the
developing and dismantling of equivalence cl'asses consisting
of terrorist stimuli. In the first study, 8 United States citizen
participants were trained to match nonterrorist stimuli to American
and terrorist images. Afterwards, participants were tested for
derived relations between American and terrorist stimuli . Results
revealed all participants had a high probability of making predictable
responses across culturally framed stimuli during a pretest (i.e. ,
match American to American and terrorist to terrorist) , yet after
training, made fewer culturally controlled responses during the
posttest. The second study examined the acquisition rate and
resulting equivalence test performance of 7 United States citizen
participants who received training with 3 sets of visual stimuli that
consisted of (a) terrorist, (b) mixed terrorist/American, and (c)
neutral (flowers) images. Most participants acquired the relations
involving the terrorist stimuli in fewer trials and scored with higher
accuracy during testing when compared to their performance on
the other two sets (mixed terrorist/American, flowers). Implications
for various theories of stimulus equivalence are discussed.
The September 11 , 2001 terrorist attacks on the United States World
Trade Center Towers have forever changed the perception of security
in America. The psychological ramifications from these events and this
loss of national security have been devastating to all Americans, not only
those who were directly involved in the attacks (Pyszczynski, Solomon ,
& Greenberg, 2003). Significant increases in posttraumatic stress have
been reported (Meisenhelder, 2002; Thobaben , 2002) which are resulting
in more persons across the country who are now seeking psychological
services of cl inicians. Many case reports are now surfacing of persons who
have been tempted to commit suicide (Duggal , Berezkin , & John, 2002) ,
increase their need for human contact (Alper, 2002) , and seek religious
Address all correspondence to Mark R. Di xon, Behavior Analysis and Therapy
Program , Rehabilitation Institute, Southern Illinois University, Carbondale, IL 62901. (E-mail:
mdixon @siu.edu) .
84
DIXON ET AL.
affiliations as a form of coping (Meisenhelder, 2002). Recent evidence
also shows that there has been a 4.9% nationwide rise in antidepressant
prescriptions compared to the 6 months prior to the attacks (Ketti &
Bixler, 2002). Future epidemiological studies will continue to reveal the
longer-term ramifications of these terrorist attacks on the psychological
well-being of our culture at large.
With the veil of security removed from American culture, the United
States government has repeatedly instructed its citizens to have an
increased awareness for those persons who may look suspicious
of committing a terrorist act. Because the perpetrators of the 9/11
attacks were of Middle Eastern descent, a natural consequence of this
governmental instruction has been for persons of Middle Eastern heritage
to appear as suspicious-leading to an overgeneralization of this race as
potential terrorists and the false arrests of many law abiding American
citizens (msnbc.com, 2002).
What is even more unfortunate in the long run for the American society
is that members develop a prejudice, bias, or stereotype towards an entire
race of people when only small numbers of individuals within that race
correspond correctly with the perceived attitude. Social psychologists have
developed large-scale theories regarding the possession of hypothetical
constructs such as racial prejudices, stereotypes, and biases towards
members of unfamiliar groups (Brown, 1988); yet these theories fail to
account for how such "stereotypes" arise or how they can be, if they can
be, overcome. If researchers could instead conceptualize a prejudice
or stereotype as a simple stimulus class, then perhaps experimental
investigations of class development, maintenance, and disruption could
account for an otherwise immeasurable psychological construct. Behavior
analytic investigations may yield promise.
Prejudice can be viewed as a feeling or opinion created about
someone or something that does not come from direct experience or
contact with the target. In a sense it is stimulus overgeneralization.
Prejudice is built into humans through their verbal abilities and language
(Hayes, Nicolls, Masuda, & Rye, 2002). Humans use prejudice as a means
of categorizing others according to similar features. Participation in verbal
networks maintains these prejudicial views and allows them to transfer
to other targets as well. For example, ethnic origin is often perceived in
terms of the food, style of dress, and social customs of a particular group
of people: An individual who has his or her wallet stolen while traveling
through Mexico may develop a negative attitude toward persons of
Mexican nationality. This attitude may transfer to Mexican food, dress,
and artwork, as well as generalizing to persons who are not of Mexican
nationality but have similar physical features. The behavioral processes
underlying the establishment and extension of such stereotypes can be
explained by Relational Frame Theory (RFT), a behavior analytic model
of language and complex behavior (Hayes, Barnes-Holmes, & Roche,
2001). According to RFT, Mexican food, art, customs, dress, and people
of Mexican nationality are related by a frame of coordination or sameness
TERRORISM
85
for the individual in this example. Any explicit function established to one
stimulus in that network will transfer to the other stimuli in the network,
without the person having direct experiences with those stimuli (see
Dymond & Rehfeldt, 2000). This process is not unlike the negative views
towards persons of Middle Eastern descent that have come about since
the terrorist attacks on the United States.
In the laboratory, the establishment of relational networks and the
transfer of function across those networks is typically established in the
following manner: Subjects are explicitly taught conditional discriminations
between a stimulus and at least two other stimuli (i .e., A 1-B1 and A 1-C1).
Verbally competent subjects then typically demonstrate the emergence of
derived or untrained relations between those stimuli (i.e. , B1-A1 , C1-A1 ,
B1-C1, C1-B1) . When a particular function is established to one member
of that network (i.e., training subjects to emit one space bar press in
the presence of B1), that function will transfer to the other stimuli in the
absence of direct training (i.e., A 1 and C1 will now occasion one bar
press). Thus, all of the stimuli in the network affect responding similarly,
although the subjects did not have direct experience with all of them (see
Dymond & Rehfeldt, 2001).
The relational network paradigm has been previously used to study
prejudice in the laboratory. During the mid to latter part of the 20th
century Northern Ireland was faced with great societal "Troubles." Watt,
Keenan , Barnes, and Cairns (1991) employed a matching-to-sample
procedure to study the formation of stimulus networks in the social context
of "The Troubles" in Northern Ireland. Subjects received conditional
discrimination training between Catholic names and nonsense syllables,
and between those nonsense syllables and Protl9stant symbols. In the
testing phase of the experiment two Catholic names and one Protestant
name served as comparison stimuli and the Protestant symbols used
previously in training were presented as the sample stimuli. Five of the
twelve Northern Irish Catholic and all six of the Northern Irish Protestant
participants failed to demonstrate derived relations between the Catholic
and Protestant names or symbols (Watt et aI., 1991). These findings
suggest that a social context and preexisting stimulus relations in that
context may interfere with or override experimentally trained relations in
the laboratory. Similar research on the importance of which preexisting
relations among stimuli may have on the formation of culturally opposing
stimulus classes in the laboratory has also been investigated using
sexually related words and culturally specific slan~1 descriptions of sexual
genitalia (Plaud, Gaither, Franklin, Weller, & Barth, 1998). These authors
also found that class formation in the laboratory can be hindered by the
subject's preexperimental history: in their study the subject having a
history of experience with sexually explicit words. Together, these and
other studies (e.g., Plaud, 1995; Plaud, Gaither, Weller, Bigwood, Barth,
& von Duvillard, 1998) suggest that researchers must be aware of and
understand the impact of preexperimental experience on laboratory
attempts at stimulus class formation.
86
DIXON ET AL.
Some historical relational networks may be so strong in individuals'
repertoires that attempting to introduce new contingencies to alter the
relations may be ineffective. This may very well be the case for many
Americans who, after experiencing the terrorist attacks of September 11,
2001, have come to relate persons of Middle Eastern descent as equivalent
to terrorists. Although we are constantly reminded that "not all Middle
Eastern persons are terrorists" or that "Islam is a peaceful religion" these
verbal stimuli may be less effective in altering our relational responding
now compared to prior to September 11, 2001. Although previous research
has repeatedly shown the effects of a priori networks on experimental
preparations, none have attempted to cast the issue experimentally using
the current culturally significant stimuli of terrorism and persons of Middle
Eastern descent. Doing so would perhaps bridge the conceptual perspective
of Relational Frame Theory (Hayes, Barnes-Holmes, & Roche, 2001) , the
basic research findings of preexisting stimuli and experimental stimulus
class formation (Plaud, 1995; Plaud, Gaither, Franklin, et aI., 1998; Watt
et aI., 1991) and the growing social concern of social categorization and
terrorism (Dumont, Yzerbyt, Wigboldus, & Gordijn, 2003) that has emerged
in a post-September 11th United States of America.
Therefore , the purpose of the present paper was to further investi gate
prejudice from an experimental perspective. We examined the degree to
which American college undergraduate partiCipants would have difficulty
forming a relational network consisting of a combination of Middle Eastern
words and terrorist and American images. We also examined the ease in
which these same partiCipants could form a relational network consisting
of all terrorist images and compared it to a control or neutral set of stimuli
that had little if any extraexperimental history associated with them.
Experiment 1
Method
Participants
Eight college undergraduates participated in this institutionally
approved project for course extra-credit. All participants were born
in the United States and were either African American or Caucasian.
Participation in the entire experiment took approximately 2 hours.
Setting, Apparatus, and Stimulus Materials
All experimental conditions took place in a 10-foot by 10-foot room
equipped with a personal computer and other experimental apparatuses.
All experimental procedures took place using an IBM-compatible personal
computer that was programmed in Microsoft Visual Basic 6.0 by Kimberly
Zlomke. Stimuli used in Experiment 1 consisted of 15 graphical images
approximately 2 inches by 2 inches. A set of stimuli (i.e. , A, B, C, D, and
E) consisted of three stimuli , yielding five sets of three members each .
Figure 1 displays the 15 stimuli used in this experiment. The A stimuli
TERRORISM
Stimulus 1 Stimulus 2 Stimulus 3
A Stimuli
EJ~SJB
B Stimuli
C Stimuli
D Stimuli
E Stimuli
Figure 1. The stimuli used in during Experiment 1.
87
88
DIXON ET AL.
consisted of the words "Islam," "Afghanistan," and "Muslim." The B stimuli
consisted of images of Osama bin Laden, Mohamed Atta, and an Islamic
tank and warrior. The C stimuli consisted of images of the United States
presidential seal, the United States flag, and the Statue of Liberty. The
D stimuli consisted of images of an American Airlines airplane, a group
of camels in the desert, and the World Trade Center Twin Towers. The
E stimuli consisted of images of a piece of apple pie, a hot dog, and a
baseball diamond.
Procedure
The initial instructions verbally given to each participant were:
You are going to see four images presented on your screen.
Your job is to choose one of the three images on the bottom of
the screen by clicking on it with the mouse. There will be parts
of the experiment where feedback is not given. The computer
. is still keeping track of your responses so continue to do your
best. Sometimes when you are correct you will receive one point.
Incorrect responses will not result in any points. Please try to earn
as many points as you can. The more points you earn, the quicker
you will finish. Do you have any questions?
The experimenter answered any questions by repeating relevant
sections of the instructions. After the experimenter answered any
questions, she left the room.
Phase 1 (Pretest). A 36-trial pretest was conducted to serve as a control
condition to examine each participant's preexperimental history of relating
certain stimuli together. Pretest trials consisted of 18 B-C (terrorist images
to American images) and 18 C-B (American images to terrorist images)
trials where the first letter in this notation represents a single member of
the set of stimuli that was presented as a sample stimulus while the second
letter represents two of the three member set of stimuli that were presented
as comparison stimuli. The third comparison stimulus presented on a 8-C
trial was one randomly selected member of the D stimulus class (airplane,
camel, or World Trade Towers), and the third comparison stimulus
presented on a C-B trial was one randomly selected member of the E
stimulus class (apple pie, hotdog, or baseball). This test was conducted
without any reinforcement, or in other words, on extinction.
Phases 2 and 3 (A-8 / A-C Training). The A-B phase consisted of
training the A-B relationship, or matching a sample stimulus from the words
"Islam," "Afghanistan," or "Islam" to the three comparison stimuli (terrorist
images). The A-C phase consisted of training the A-C relationship, or
matching a sample stimulus from the words "Islam," "Afghanistan," or
"Islam" to the three comparison stimuli (American images). Each of
the A-B and A-C training phases consisted of blocks of 18 trials each.
To advance to the next training phase a criterion of 90% (16 correct
responses out of 18 trials within one block) needed to be met.
Phase 4 (Mixed training). After criterion was met in both the A-B
TERRORISM
89
and A-C training phases, the participant progressed into a mixed training
phase. This phase consisted of 36 intermixed A-B and A-C trials. Criterion
for completion of this phase was 90% (32 responses correct out of 36
trials within one block).
Phase 5 (Posttest). The posttest consisted of 108 intermixed trials.
Thirty-six trials were identical to those presented in the pretest, 36
additional trials of B-C and C-B relations were also presented without
the additional D or E comparison stimulus. During these 36 additional
trials all three members of the comparison class were presented as
possible response options. The remain ing 36 trials tested for symmetrical
responses with B-A and C-A trials. Prior to the first trial of the test the
participant viewed the following instructions via the computer monitor:
You will no longer receive feedback following your responses. Continue
to do the best you can. The computer is recording your score.
The experimenter thoroughly debriefed each participant after the
completion of the posttest.
Experiment 1 Results
Table 1 displays the number of blocks of training trials to reach
mastery criterion for all training phases. For A-B and A-C training, a score
of 16 correct or better out of an 18-trial block was deemed as meeting
Table 1
Number of Blocks of Training Trials to Reach Mastery Criterion in Experiment 1
Participant
Number
8
9
A-B Training
Blocks
12
A-C Training
Blocks
1
Mixed (A-B & A-C)
Training Blocks
1
5 3 2
11
12
13
2
3
5
2
2
7
1
1
1
14
15
16
2
4
4
2
4
2
1
3
1
Note . Mastery criterion was 16/18 correct for A-B and A-C and 32/36 correct for mixed.
criterion. For Mixed A-B / A-C training , a score of 32 correct or better out
of a 36-trial block was deemed as meeting criterion. All participants were
able to meet criteria for each type of training. A-B training took anywhere
from 2 to 12 blocks of trials, A-C training 1-7 blocks, and Mixed 1-3 blocks.
Figure 2 shows the average response latencies compared across A-B
training trials and A-C training trials for all participants in the experiment.
Longer response latencies were emitted for the A-C training trials which
presented the Middle Eastern words as samples and the American images
as comparisons as compared to the A-B training trials which presented
the Middle Eastern words and the terrorist images. PartiCipants 8, 11, 12,
DIXON ET AL.
90
4.5
-
4
A-8 Training Trials
1/1
"0
c:
0
3.5
""
u
GI
1/1
>
u-
c:
2.5
GI
~
..J
GI
1/1
c:
0
A-
/
A-C
~
3
n,;""
T""
2
1.5
1/1
~
0.5
0
8
9
11
12
13
14
15
16
Participant Number
Figure 2. Response latencies (seconds) for participants in Experiment 1 on A-B and A-C
training trials.
13, and 14 displayed longer response latencies during A-C training trials
than during A-B trials. The 3 participants who did not have higher average
response latency during the A-C trials (Participants 9, 15, & 16) had only
a slightly longer latency during A-B trials.
The top panel of Figure 3 illustrates that participants consistently
demonstrated extraexperimental relations during the pretest between
the C stimuli (presidential seal, flag, and Statue of Liberty) and E stimuli
(hotdog, apple pie, baseball) and images as well as between the B
stimuli (Bin Laden, Atta, and tank) and the D stimuli (airplane, camel, or
World Trade Towers). This was evident by the much higher than chance
scores during the pretest for the stereotype specific D (terrorist images)
and E (American images) stimuli as opposed to either of the other two
comparison stimuli. The top panel of Figure 2 displays the percentage of
correct responses made during the pretest, as well as the percentage of
incorrect responses that were made for either a D or E stimulus. Of the
8 participants 6 (8, 9, 11, 13, 14, & 15) responded incorrectly with D or E
stimulus choices during the pretest on at least 80% of all possible trials.
During the posttest, all participants increased their percentage of
correct responses over pretest percentages. Of the 8 participants, 4 (8,
11, 14, & 16) demonstrated equivalence responding with a 90% correct
score on the combined trials posttest. Three additional participants (9, 12,
& 15) increased their percentage correct by a minimum of 40%. Only 1
participant (13) responded at a chance percentage level (33%). The lower
91
TERRORISM
Inoorrect RespOnsES
to aD orE
100
-
Pretest
,
stmu'l
,
,
Correct
Responses
,
J~
I
o
8
9
11
12
13
14
15
[l
16
Posttest (Matched Pretest Trials)
100
l
o
8
9
to
11
n
12
13
14
n
15
16
Posttest (aliI trials cormined)
100
80
60
f-
40
20
n
o
8
9
h
11
11
12
13
14
1
15
16
Participant Numbelr
Figure 3. Percentage of correct responses for participants in Experiment 1 during the pretest and
posttest. Incorrect responses to a D or E (culturally consistent) stimulus are plotted for comparison.
92
DIXON ET AL.
two panels of Figure 3 display the percentage of correct responses during the
posttest for just the 36 matched posttest trials that were identical to the pretest
trials in their inclusion of a 0 or E comparison, as well as all 108 test trials
combined. Although 5 of the 8 participants did continue to make responses
errors to the 0 and E stimuli, the frequency of such responses dramatically
decreased from pretest to posttest. Figure 4 displays the percentage of
correct responses on symmetry (8-A or C-A) and equivalence trials (8-C or
C-8). Again, all participants' errors during both test trial types were due to
responses for the 0 and E comparison stimulus.
Posttest (5ym & Equiv Trials)
III
GI
III
100
C
0
c-
..
III
GI
80
a:
u
GI
t:
60
0
()
....0
..
GI
40
0)
«$
c
GI
u
20
io..
GI
Q.
0
8
9
11
12
13
14
15
16
Participant Number
Figure 4. Percentage of correct symmetry relations (B-A & C-A) and equivalence relations
(C-B & B-C) made by participants in Experiment 1 during the postlest.
Discussion of Experiment 1
The results of Experiment 1 illustrate that during the pretest,
participants often responded in accordance with preexisting frames of
coordination between American symbols and other American images,
and Middle Eastern symbols and terrorist images. The majority of our
participants opted to demonstrate the culturally consistent relations of
coordination by selecting the "E" stimuli when given the "C" stimuli as
samples, and the "0" stimuli when given the "8" stimuli as samples.
These results are consistent with previous research, which has shown
that subjects are most likely to demonstrate preexisting relations between
culturally relevant stimuli rather than demonstrating a novel relationship
involving those same stimuli (e.g., Plaud, Gaither, Franklin, et aI., 1998).
However, all participants in Experiment 1 emitted fewer errors during
TERRORISM
93
the posttest relative to the pretest. The prese~nt results suggest that
programmed contingencies can to some degree override the preexisting
relationships between stimuli and have important practical ramifications.
Persons who possess racial or religious stemotypes often tend to
inaccurately attribute personal characteristics such as "good or bad,"
"right or wrong," "enemy or friend" to members oJ a certain race, religion,
or ethnic background. It may be possible to reduce the effects of such
prejudices by establishing relational networks bE3tween stimuli from that
network and novel stimuli. For example, the individual who had a wallet
stolen in Mexico might learn to relate stimuli from that verbal network to
novel, positive stimuli, for example, coming into contact with a Peace
Corps worker who is of Mexican heritage.
At the same time, the experimental contin!~encies did not appear
to completely override subjects' preexisting relations. Most participants
still made errors during the posttest, and those errors were often a
selection for a novel ("D" or "E") comparison stimulus. Thus, subjects'
extraexperimental histories could not be completE3ly dismantled. This was
particularly the case for Participant 13, who required seven blocks of trials
to demonstrate mastery criterion for A-C training, suggesting that she may
have had the strongest extraexperimental framEl of opposition between
the Middle Eastern words and the American ima~~es.
The longer response latencies during A-C training trials relative to A-B
training trials that were observed with many participants may suggest that
the Middle Eastern words (Islam, Muslim, and Arab) were more strongly
related in extraexperimental frames of coordination with terrorist images
than in frames of opposition with American images. In fact, pitting these two
types of relational frames against each other could have possibly led to the
low posttest scores that were observed for 4 of the 8 participants. Future
research may wish to unravel this possibility more systematically by examining
multiple stimulus sets containing stimuli that participate in extraexperimental
frames of coordination and stimuli that are unlikE!ly to participate in such
extraexperimental frames of coordination. During Experiment 2 we attempted
to further investigate the relative ease and difficultness of equivalence class
formation using stimuli that were either (a) novel and neutral, (b) presumably
extraexperimentally nonequivlant (i.e., those used iin Experiment 1), and (c)
presumably extraexperimentally equivalent.
Experiment 2
Method
Participants
Seven college undergraduates participated in this institutionally
approved project for course extra-credit. All participants were born
in the United States and were either African American or Caucasian.
Participation in the entire experiment took approximately 2 hours.
DI XON ET AL.
94
Set 1
Stimulus 1
Stimulus 2
Stimulus 3
Stimulus 2
Stimulus 3
Stimulus 2
Stimulus 3
A Stimuli
B Stimuli
C Stimuli
Set 2
A Stimuli
B Stimuli
C Stimuli
Stimulus 1
A Stimuli
B Stimuli
C Stimuli
Figure 5. The stimuli used in during Experiment 2.
TERRORISM
95
Setting, Apparatus, and Stimulus Materials
All experimental conditions took place in the same room and used the
same apparatus as Experiment 1. Stimuli used in Experiment 2 consisted
of 27 graphical images approximately 2 inches by 2 inches. A set of
stimuli (i.e., A, B, C) consisted of three stimuli, yielding three sets of three
members each. Figure 5 displays the 27 stimuli used in this experiment.
Set 1 consisted of nine stimuli of flowers. This "flowers" set depicted
close up images of flowers that varied in color and shape. Set 2 consisted
of nine stimuli that were a mix of terrorist/American images including
a wheat field and combine, the World Trade Center on fire, Yosemite
National Park's Half-Dome, Bin Laden, the United States flag, Islamic
tank and warrior, the United States presidential seal, Mohamed Atta, and
the Statue of Liberty. Set 3 consisted of nine stimuli that were terrorist
images including Saddam Hussein, a Middle Eastern tank and soldiers,
Yasser Arafat, the World Trade Center on fire, an airplane used in the
September 11 th terrorist attacks, Osama bin Laden and other terrorist
persons standing in a group , suspected American terrorist John Walker,
an unnamed male terrorist, and a camel in a pen.
Research Design
All participants were exposed to the three sets of stimuli with order
of presentation counterbalanced across participants. The following
procedure was identical for all three sets of stimuli.
Procedure
Phase 1 (Pretest). A 36-trial pretest was conducted to serve as a
control condition to examine each participant's preexperimental history
of relating certain stimuli together. Pretest trials consisted of 18 B-C and
18 C-B trials where the first letter in this notation represents a single
member of the set of stimuli that was presented as a sample stimulus
while the second letter represents the three member set of stimuli that
were presented as comparison stimuli. This test was conducted without
any reinforcement, or in other words, on extinction.
Phases 2 and 3 (A-8 / A-C training). The A-B phase consisted of
training the A-B relationship, or matching a sample stimulus from the
images of flowers (Set 1), terrorist (Set 2), or mixed terrorist/American
(Set 3) to the three comparison stimuli. The A-C phase consisted of
training the A-C relationship. Each of the A-B and A-C training phases
consisted of blocks of 18 trials each. To advance to the next training
phase a criterion of 90% (16 correct responses out of 18 trials within one
block) needed to be met.
Phase 4 (Mixed training). After criterion was met in both the A-B
and A-C training phases, the participant progressed into a mixed training
phase. This phase consisted of 36 intermixed A-B and A-C trials. Criterion
for completion of this phase was 90% (32 responses correct out of 36
trials within one block).
Phase 5 (Posttest). The posttest consisted of 72 intermixed trials.
Thirty-six trials were identical to those presented in the pretest, or trials
96
DIXON ET AL.
of B-C and C-B relations. The remaining 36 trials tested for symmetrical
responses with B-A and C-A trials. Prior to the first trial of the test the
participant viewed the following instructions via the computer monitor:
You will no longer receive feedback following your responses. Continue
to do the best you can. The computer is recording your score.
All participants repeated this identical procedure for the remaining
two sets of stimuli. After completion of all three sets of stimuli, the
experimenter thoroughly debriefed each participant.
Experiment 2 Results
Table 2 displays the number of blocks of training trials to reach
mastery criterion for all training phases used in Experiment 2. For A-B
and A-C training, a score of 16 correct or better out of an 18-trial block
Table 2
Number of Blocks of Training Trials to Reach Mastery Criterion in Experiment 2
Participant
Number
20
28
29
30
46
47
48
A-B
Flowers
A-C Mixed
1
2
2
3
6
6
1
4
2
5
2
3
1
2
1
1
1
1
3
2
2
Terrorl American
A-B A-C Mixed
6
2
3
2
2
2
7
5
3
2
2
2
2
2
4
1
1
1
1
1
2
A-B
2
2
2
1
1
2
2
Terrorism
A-C Mixed
2
2
2
2
2
2
1
1
3
1
1
1
1
1
Note. Mastery criterion was 16/18 correct for A-B and A-C and 32/36 correct for mixed.
was deemed as meeting criterion. For mixed A-B / A-C training , a score
of 32 or better out of a 36-trial block was deemed as meeting criterion.
All participants were able to meet criteria for each type of training. A-B
training took anywhere from 1 to 7 blocks of trials across the various
stimulus sets, A-C training 1-5 blocks, and Mixed 1-3 blocks. There
were an overall fewer numbers of trial blocks to criterion exhibited by
participants during their exposure to the terrorism stimulus set when
compared to exposure to the flowers or mixed terrorism/American sets.
Specifically, 4 participants (29, 30, 46, & 48) had the least blocks of trials
to criterion for the terrorist set, 1 participant (20) had only fewer blocks
when exposed to the flowers set, 1 participant (47) had equal numbers of
blocks across all sets, and 1 participant had fewer blocks when exposed
to the other two sets.
Figure 6 displays mean response latencies emitted on training trials
for all partiCipants across the three stimulus sets used in Experiment 2.
The top panel of this figure shows the comparison between response
latencies on A-B trials and A-C trials for each of the three stimulus sets.
No differences between these two trial types were found. The bottom
panel of this figure shows the comparison between correct and incorrect
TERRORISM
97
4.5
4
3.5
3
2.5
2
1.5
'U'
Q)
0.5
(/)
,..,
'-'"
0
Flowers
0
Terror/American
Terrorism
s::
Q)
tU
....J
Q)
4.5
Incorrect Responses
./
4
(/)
s::
0
a.
(/)
3.5
Q)
2.5
0=:
3
2
1.5
0.5
0
Flowers
Terror/American
Terrorism
Stimulus Set
Figure 6. Response latencies (seconds) for participants in Experiment 2 on A-B and A-C
training trials (top panel) and on correct and incorrect responses (bottom panel).
responses across both A-B and A-C training trials combined for each of
the three stimulus sets. The average response latency was higher prior
to an incorrect response than prior to a correct response for each of the
three stimulus sets.
Figure 7 shows the percentage of correct responses made by each
participant during the testing conditions of Experiment 2. The top panel of
this figure depicts pretest scores of each participant for all three stimulus
sets. All 7 participants responded at chance levels (33%) for most sets.
The bottom panel of this figure depicts posttest scores of each participant
for all three stimulus sets. Of the 7 participants, 6 (20, 28, 29, 30, 46, &
47) were able to demonstrate equivalence, as measured by 90% correct
DIXON ET AL.
98
Pre-Test
100
80
Terrorist
Fbwers
60
1
40
.....u
(J)
!o-
"-
o
1
20
o
20
U
(J)
0)
('U
1
Terror/Pmerican
.....
c
100
~
80
28
29
30
46
47
48
Post-Test
(J)
(J)
0...
60
40
20
o
20
28
29
30
46
47
48
Participant Number
Figure 7. Percentage of correct responses for participants in Experiment 2 during the pretest
and posttest for each of the three stimu lus sets.
on any posttest, for at least one stimulus set. No participant was able
to demonstrate equivalence for all three stimulus sets. The terrorist set
equivalence test was passed by 6 participants (20, 28, 29, 30, 46, & 47).
The flower set equivalence test was passed by 4 participants (20, 28, 46,
& 47). The mixed terrorism/American set was passed by only 1 participant
(30). Comparisons between sets revealed that all 7 participants had a
higher posttest score on the terrorist set than on the mixed terrorism/
American set. The mixed terrorism/American set was the lowest posttest
score for 5 of the 7 participants (20, 28, 46, 47, & 48).
TERRORISM
99
Discussion of Experiment 2
The findings of Experiment 2 illustrate that although none of the
participants were able to derive the B-C and C-8 relations for all three
sets of stimuli during the posttest with 90% test trials correct or higher, 6 of
the? did so for at least one set. In every case, each participant was able to
make more correct derived relational responses when exposed to the set
of stimuli that consist of all terrorist images than when exposed to the set
of stimuli that consist of a mixture of terrorist/American images. It is thus
possible that the terrorist images existed in an extraexperimental frame
of coordination for the participants. As a result, it was much easier for
participants to form relations with the terrorist set relative to the other two
stimulus sets. Previous research has shown that preexisting relations are
more readily demonstrated in laboratory experiments than are relations
involving novel stimuli (Plaud, Gaither, Franklin, et aI., 1998). Our current
data again appear to support those findings and extend them into the
context of terrorism.
The other sets of stimuli consisted of images for which participants
had no previous history of relating. The flowers set, while attempting
to serve as a control set of stimuli, may have been too arbitrary for
participants to respond with a high degree of accu racy. It may have been
more desirable to have used a control set of stimul i that included a variety
of everyday objects. Also, the mixed terrorist/AmHrican set required the
participant to make responses that may have conflicted with previously
established relations; in fact, terrorist and American stimuli may have
existed in an extraexperimental frame of opposition as suggested by the
results in Experiment 1.
In the present experiment participants were reinforced merely with
points redeemable for nothing for making correct responses during
training. The relative reinforcing value of the "points" may have not been
enough to outweigh the relative aversiveness of selecting a comparison
stimulus that was not consistent with an extraexperimental relational
network during exposure to the mixed terrorist/American stimulus set.
When faced with the choice between the small reinforcing value of
a point or the larger reinforcing value of responding in accordance
with an extraexperimental network, our participants' often selected to
respond congruent with the culturally preestablished relational network
rather than respond against that network and gain a mere point. Future
research might wish to investigate the relative eas,e of deconstructing an
extraexperimental relational frame by using varying sizes of reinforcers
for correct responding. Using the mixed terrorist/American images
stimulus set, one might instate experimental conditions whereby points
are redeemable for (a) nothing, (b) course extra-credit, or (c) money and
assess participants' relative training and testing performances.
Most of the participants of Experiment 2 demonstrated higher
response latencies for incorrect responses during tlraining than for correct
responses. This finding is consistent with previous research which has
100
DIXON ET AL.
used response latency as a supplemental measure of the strength of
emergent performances (see Dymond & Rehfeldt, 2001). Furthermore the
longer latencies may suggest that participants were attempting to generate
self-rules or strategies to aid in solving the conditional discrimination
problem at hand (Rehfeldt & Dixon, 2000). A future study might attempt to
have participants talk aloud throughout the experiment (see Hayes, 1986
for a description of this procedure). The amount of emitted verbal behavior
that was produced on correct and incorrect trials could be compared to
see if increases in frequency occurred on trials that the participant was
unsure about the correct response.
General Discussion
Together, the present two experiments provide a possible empirical
foundation of relational frame theory's application to prejudice as
suggested possible by (Hayes, Barnes-Holmes, & Roche, 2001). During
the pretest in Experiment 1, it was shown that common words describing
Middle Eastern culture were related in a frame of coordination with
terrorism-type stimuli for all our participants. While it is debatable if the
events that took place on September 11, 2001 caused this relationship
to occur, we suggest that at a minimum it participated in altering the
previously existing relational networks these words were in. In other words,
there was possibly a transfer of function between words that describe an
entire race of people and images that describe very few members of this
population (Dymond & Rehfeldt, 2001). What results is a new relational
network that has been constructed based on incorrect information. This
finding alone is not necessarily novel and supports previous research on
the transfer of function (Dougher, 1998), but it does tend to suggest how
easily flexible stimulus functions are in a rather naturalistic and socially
relevant context of terrorism. Furthermore in Experiment 1, we utilized a
rather novel pretest and posttest that consisted of stimuli never presented
in training. These D and E stimuli of Experiment 1 served as a continued
baseline with which to measure the relative success or failure of the
experimental training on overriding preexisting relational networks. Future
research in the area of transfer of function may wish to incorporate similar
pretest and posttest stimuli as a supplementary measure of the potential
strength that preexisting stimuli networks may influence experimental
training protocols.
While the pretest results of Experiment 1 suggest that Middle Eastern
words are readily related to terrorist stimuli over American stimuli,
the training and testing performance of the partiCipants suggests that
programmed reinforcement contingencies can override the preexisting
relational networks to some degree. These data are promising in that
they provide empirical support towards a dismantling of an existing frame
of coordination between stimuli and a reorganization of a new frame of
coordination based on the derived transformation of functions between
stimuli presented during training. However, it may be critical to identify
TERRORISM
101
salient enough reinforcers to produce the desired effect. We believe future
research on this topic is necessary and may wish to incorporate more
substantial reinforcers than points on a computer screen. Perhaps verbal
praise from a member of the prejudiced racial group could be an option.
Our procedure was very artificial in the sense that we attempted to
have participants relate stimuli such as George Bush to Osama Bin Laden.
Relating these stimuli together does not necessarily teach an American
person to deconstruct a frame of terrorism. However, by attempting to
create very impractical frames of coordination between known terrorists
and Americans, and being able to do so somewhat successfully, it would
suggest that such relations should be able to devl910p between unknown
persons of Middle Eastern descent and pro-American type stimuli. In
other words, our study attempted to demonstrate an effect with the most
improbable extreme stimuli. Better success experimentally and practically
might be made using less intense stimuli and also have greater external
validity when attempting to understand how transfers of function occur
that result in nonterrorist persons of Middle Eastern descent being
labeled, named, or scorned as terrorists.
The ease in which mastery criterion was met and derived relations
emerged during Experiment 2 for the set of all terrorist stimuli suggests
how quickly it is for stimuli sharing formal or functional similarities to be
constructed into a relational frame of coordination. Reinforcement from
the verbal community may strengthen these relat~ons even more so. As
our culture becomes increasingly concerned about the threat of terrorism,
we believe that prejudice against Middle Eastern persons may become
increasingly common. Awareness of prejudice from an empirically based
behavior analytic perspective is the first step towards identifying ways to
overcome it.
The present two experiments have attempted to establish an
experimental groundwork for a Relational FramEl Theory interpretation
of the formation and the possible dismantlingl of relational frames
of terrorism and prejudice. While the present data suggest such an
interpretation is possible, much more research needs to be conducted
that addresses the development of stimulus networks involving prejudice.
For example, unknown persons of Middle Eastern descent could be used
as stimuli, rather than known terrorists, and participants could be taught
conditional discriminations between them and pro-American comparison
stimuli. Other persons of non-Middle Eastern descent, perhaps even
white American males who have committed terrorist-type crimes could
be used as stimuli and participants would be reinforced for correctly
relating them to anti-American comparison stimuli. This type of procedure
would hopefully teach participants that not all persons of Middle Eastern
descent are terrorists, and not all white males are anti-terrorists. As a
result, prejudice will be shown to not be a by-product of our culture, but
rather a behavior, like all others, that can be shaped and altered through
a manipulation of environmental variables.
Although we have limited our outline of the formation of stimulus
102
DIXON ET AL.
classes involving images of terrorism and the discussion of our obtained
results to Relational Frame Theory, additional conceptualizations and data
interpretations could be made via other theories of relational responding .
For example, the "Naming Theory" (Horne & Lowe, 1996) suggests that
when one names a stimulus, it may be concurrently serving a discriminative
function for relating matching and equivalence tasks. In the context of
our experiment, a subject may have for example named the stimulus of
Osama bin Laden, which supplementally functioned as an occasion to
match that stimulus to other terrorism related stimuli. The naming theory
also suggests that when the naming of a specific stimulus occurs, this
process may be a mediating response that aids in generalization from
the learning environment to other environments. In our experiment, the
participants had a history of media exposure that certain persons are
called, or named, terrorists. Those persons called terrorists were of Middle
Eastern descent and that feature may be subsequently named in novel
environments with novel stimuli-nan-terrorists that are of Middle Eastern
descent. Sidman's (1994) conceptualization of stimulus equivalence could
also successfully describe the formation of stimulus classes and the
transfer of stimulus functions involving images of terrorism , and may also
explain our findings regarding the difficulty of class formation for many of
our subjects when we used preexperimentally class inconsistent stimuli as
we did in the mixed terrorism / American stimulus set. One advantage we
see Relational Frame Theory having as a conceptual foundation for the
present data is the theory's extensive description and growing empirical
reports of the formation of other derived relations beyond equivalence or
sameness which include, opposition , comparison, and distinction (Hayes,
Barnes-Homes, & Roche, 2001). Future research interested in exploring
the development and dismantling of stimulus classes with terrorist stimuli
should construct relational networks other than equivalence as we have
done in the present set of experiments.
In summary, Relational Frame Theory has claimed to provide a
behavior analytic account of prejudice, racism, and the formation of
stereotypes (Hayes et aI. , 2002). To date, however, the relational frame
conceptualization has lacked the empirical groundwork to support the
claim. Our data lend evidence to the formulation of Hayes et al. (2002)
and also provide for a starting place to investigate terrorism and prejudice
from a behavioral perspective. As the threats to national security in the
United States continue to grow, the time is now for behavior analysts to
gain a greater understanding of the psychological ramifications for the
individual society member and the homeland at large that is exposed to
a new culture of terror.
References
ALPER, G. (2002). Up close and personal: September 11, through the lens of a
psychotherapist. Journal of Loss and Trauma, 7, 251-261.
TERRORISM
103
BROWN , R. J. (1988). Group processes: Dynamics within and between groups.
Oxford: Blackwell.
DOUGHER, M. J. (1998) . Stimulus equivalence and the untrained acquisition of
stimulus functions. Behavior Therapy, 29,577-591.
DUGGAL, H. S., BEREZKIN, G., & JOHN, V. (2002) . PTSD and TV viewing of
the World Trade Center. Journal of the American Academy of Child and
Adolescent Psychiatry, 41 , 494-495.
DUMONT, M., YZERBYT, V., WIGBOLDUS , D., & GORDIJN, E. H. (2003) . Social
categorization and fear reactions to the September 11 th terrorist attacks.
Personality and Social Psychology Bulletin, 29, 1509-1520.
DYMOND, S., & REHFELDT, R. A. (2000). Understanding complex behavior: The
transformation of stimulus functions. The Behavior Analyst, 23, 239-254.
DYMOND, S. , & REHFELDT, R. A. (2001). Supplemental measures of derived
stimulus relations. Experimental Analysis of Human Behavior Bulletin, 19.
HAYES, S. C. (1986) . The case of the silent dog - verbal reports and the analysis
of rules. A review of Ericsson and Simon's Protocol analysis: Verbal reports
as data. Journal of the Experimental Analysis of Behavior, 45, 351-363.
HAYES, S. C., BARNES-HOLMES, D., & ROCHE , B. (2001). Relational frame
theory: A post-Skinnerian account of human language and cognition. New
York: Knapp .
HAYES, S. C. , NICOLLS, R., MASUDA, A. , & RYE, A. K. (2002). Prejudice, terrorism,
and behavior therapy. Cognitive and Behavioral Practice, 9, 296-301.
HORNE, P. J., & LOWE, C. F. (1996) . On the origins of naming and other symbolic
behavior. Journal of the Experimental Analysis of Behavior, 65, 185-241 .
KETTL, P., & BIXLER, E. (2002). Changes in psychotropic drug use after
September 11 . Psychiatric Services , 53, 1475-1476.
MEISENHELDER, J. B. (2002). Terrorism , posttraumatic stress, and religious
coping. Issues in Mental Health Nursing, 23, 771-782.
MSNBC, America At War (n .d.). US Muslims opt for home schooling: Parent cite
post-Sept. 11 fears, cultural issues. Retrieved April 1, 2003 from http://
msnbc.com/news/840205.asp
PLAUD, J. J. (1995). The formation of stimulus equivalences: Fear-relevant versus
fear-irrelevant stimulus classes. The Psychological Record , 45, 207-222.
PLAUD, J. J., GAITHER, G. A. , FRANKLIN , M., WELLER, L. A. , & BARTH, J.
(1998). The effects of sexually explicit words on the formation of stimulus
equivalence classes. The Psychological Record, 48, 63-79.
PLAUD, J. J., GAITHER, G. A. , WELLER, L. A. , BIGWOOD, S. J. , BARTH, J. , &
VON DUVILLARD, S. P. (1998). Rational-emotive behavior therapy and the
formation of stimulus equivalence classes. Journal of Clinical Psychology,
54,597-610.
PYSZCZVNSKI, T., SOLOMON, S., & GREENBERG, J. (2003). In the wake of9/11: The
psychology of terror. Washington, DC: American Psychological Association.
REHFELDT, R. A., & DIXON , M. R. (2000). Investigating the relation between
self-talk and emergent stimulus relations. Experimental AnalYSis of Human
Behavior Bulletin, 18, 28-30.
SIDMAN, M. (1994). Equivalence relations and behavior: A research story.
Boston, MA: Authors Cooperative.
THOBABEN, M. (2002). The aftermath of the terrorist attacks on September 11 ,
2001 : Posttraumatic stress disorder. Home Health Care Management and
Practice, 45, 398-399.
WATT, A., KEENAN , M. , BARNES, D. , & CAIRNS, E. (1991). Social categorization
and stimulus equivalence. The Psychological Record, 41, 33-50.