Documenti di Didattica
Documenti di Professioni
Documenti di Cultura
Summary--Data are presented from two experiments investigating the modifiability of covert
responses. We chose imagery as the target behavior because of its significant role in behavior
therapy techniques. Recent evidence on the facilitative effects of imagery in paired-associate
learning enabled us to corroborate self-reports of covert responding. In an attempt to control
the frequency of imagery, subjects were either reinforced or punished for self-reports of
specific associate methods. The data indicated predictable frequency changes in imagery
as a function of external contingency arrangements. An analysis of differential recall
performance substantiated subjects' self-reports by showing a consistent superiority on
imagined (versus non-imagined) items. These findings are interpreted as supportive of the
homogeneity assumption regarding overt and covert behavior principles. Clinical
implications are briefly discussed.
*Requests for reprints should be addressed to Michael J. Mahoney, Department of Psychology, Stanford University,
Stanford, California 94305.
tThis research was supported by the Personal Competencies Project (0506) of the Stanford Center for Research
and Development in Teaching. The authors would like to thank G. C. Davision, A. Bandura, and G. H. Bower
for their comments and suggestions.
~Now at the University of Oregon.
7
8 M I C H A E L J. M A H O N E Y , C A R L E. T H O R E S E N a n d B R I A N G. D A N A H E R
inferred from observable data. By manipulating like coin dispensing apparatus occupied the
the external consequences of that covert be- right-hand corner of the table. All experimental
havior, we obtained evidence bearing on its instructions, paired-associate lists, and recall
modifiability. Imagery (I) was chosen as an ideal tests were delivered via a tape recorder (Sony
behavior for experimental purposes both because Model TC-18). The experimenter operated
of its prominence in behavior therapy techniques both the tape recorder and the consequation
and because of recent evidence showing it to be apparatus by remote control. The initial instruc-
one of the few covert behaviors for which tions introduced the experiment as an attempt
reliability checks are possible. Although no un- to understand how people associate words. The
equivocal physical index of I has been found, a specific procedure followed that of a standard
performance index has shown itself to be reliably paired-associate learning task--i.e, two nouns
related to such behavior (Bower, in press). were associated and a later recall test presented
Numerous researchers have shown that I sig- the first member of a pair whereupon the subject
nificantly facilitates recall performance when was required to recall its associate. The learning
used as a mediator in paired-associate learning materials were four lists of 15 paired-associates
tasks (e.g. Bower, 1970, in press; Paivio, 1969; which had been employed in the previously
Paivio and Yuille, 1967, 1969; Reese, 1965). cited research of Bower and his co-workers. The
These workers have shown that when I is used pairs were presented at a 2 : 2 sec rate with an
in paired-associate learning, recall performance 8 sec interval. Each subject could use any of
is dramatically greater than when any other form four associative methods: imagery, sentence,
of mediation is employed (e.g. repetition). Self- repetition, and other. These were designated on
reports were, of course, required in these an answer sheet by the initials ' T ' , "S", " R " ,
studies. However, a highly replicable and and "O". When they had chosen to associate a
unmistakable performance superiority has particular noun pair with one of the above
consistently been associated with self-reports methods, subjects were to write the initial of the
of I in such experiments. We therefore decided associative method on their answer sheet. This
that I would serve as a target behavior for allowed the experimenter--who observed their
modification and that we would use recall performance through a one-way mirror--to
performance on a paired-associate learning task monitor and (when appropriate) consequate
as a reliability check on the self-reported their self-reported covert behavior.
occurrence of I. After the initial experimental instructions, a
list of 15 paired associates was presented. No
EXPERIMENT I: consequences were given for self-reported cog-
POSITIVE REINFORCEMENT nitive behaviors in this phase of the experiment.
Subjects. Subjects for the present study were Rather, it served as a baseline estimate of the
drawn from an introductory psychology course frequency of various covert mediating responses.
at Stanford University. Although all subjects Thereafter, an intervention was announced by
had the opportunity to earn money from the the following taped instructions:
experiment, its monetary potential was not
advertised and all participants were asked to During the remainder of the experiment you will be
awarded money whenever the method by which you
refrain from discussing the experiment with choose to associate a n o u n pair corresponds to a pre-
classmates. The experiment was described as one established standard. Notice that there is a large box
located on the desk in front of you. A dime will be
involving "verbal learning". dispensed from that box whenever your choice of
Procedure. Subjects in this experiment (four associative method is considered appropriate. Thus,
for example, you might receive a dime for using the
male, six female) were seated at a table in front repetition method to associate a particular n o u n pair.
of a one-way mirror through which the experi- Which method is appropriate may vary from pair to
pair or list to list. Y o u should attempt to earn as many
menter monitored their overt responses. A box- dimes as possible and to vary your choice of associative
COVERT BEHAVIOR MODIFICATION: AN EXPERIMENTAL ANALOGUE 9
method so as to maximize your earnings. You will, of a specific (covert) behavior was obtained, con-
course, be allowed to keep whatever money you have
earned at the end of the experiment. ditions were implemented which were designed
to successively increase, decrease, and again
Depending on their self-reported baseline fre- increase the targeted behavior. The four paired-
quency of I behavior, subjects were assigned to associate lists corresponded to the phases of
one of two conditions. When baseline imaging baseline, intervention, reversal, and re-interven-
was infrequent (less than 50 per cent), they were tion which characterize the empirical case study,
assigned to an Imagery-Repetition-Imagery In accordance with their assigned conditions,
(IRI) condition. Conversely, subjects with high subjects received dimes during the remainder of
frequency baseline imaging (greater than the experiment for self-reports of I or R
50 per cent) were assigned to a Repetition- behavior. A recall test for all 60 paired-associates
Imagery-Repetition (RIR) condition. In each of was then given. It was necessary to delay the
these groups (N--5=5) the targeted covert recall test until the end of experimental manipu-
behavior was reinforced, then extinguished, and lations because previous research had indicated
then reinforced again. After their baseline phase, the irreversibility of I behavior once its recall
IRI subjects received dimes for self-reported I superiority had been demonstrated (Tieman,
during the second list of paired-associates. personal communication). Following the recall
During the third list, the mediating response of test, each subject was given his earnings and
repetition (R) was reinforced. On the final list, asked (1) if he were aware of the experimental
I responses were again rewarded. The converse contingencies, and (2) if he had been consist-
was true of RIR subjects--i.e, after baseline, ently honest in reporting the covert associative
R responses, then I responses, and then R methods he employed.
responses were reinforced during lists two,
three, and four respectively. The experimental RESULTS
paradigm was intended to parallel conventional The results of the first experiment are pre-
research designs in the experimental analysis of sented in Fig. 1. The mean frequency of
behavior. After an initial frequency estimate of self-reported I behaviors for each group reflect
15
Reinforcement
Group curves e - - e I R I N=5
c---o RIR N =5
+ A
._E IC / \
o
+
5
/
7,\
V ",
0 1 . 1 t t
Baseline Intervention ReverSal Re- ~tervention
Experimental phase
FIG. 1. Mean frequency of imagery when subjects were alternately reinforced for self-reports of imagery (I) or
repetition (R).
10 MICHAEL J. MAHONEY, CARL E. THORESEN and BRIAN O. D A N A H ER
15 R ~)St6
\ ,,,, /
I0 \ / ' , 1
Y Y
"\ I\
I I ~ Q I
.g Boseline Tntervention Reversol Re-inlervenfion
g
!
h
lO
I I I ~S~k3.. I
Bosel'ine Intervention Reverse! Re-interventior~
Experimental phase
PIe. 2. Individual response curves of subjects whose performance conformed most (upper figure) or least (lower
figure) to the reinforcement contingencies,
COVERT BEHAVIOR MODIFICATION: AN EXPERIMENTAL ANALOGUE 11
was observed in the relative superiority of I as much of your $4.00 as you can and to vary your choice
of associative method in such a way as to maximize
an associative method. Over all 10 subjects, your earnings. You will, of course, be allowed to keep
I-recall averaged 40 per cent whereas non-I- whatever money you have earned at the end of the
experiment.
recall averaged 22 per cent. Most subjects
reported awareness o f the experimental contin- Thereafter, the consequation apparatus was
gencies. Two individuals in the R I R condition operated whenever the targeted decelerative
indicated that they had occasionally reported I behavior was reported. IRI subjects were pun-
when they had in fact used some other asso- ished for imaging on the second list, for
ciative method. A within-subject analysis of repeating on the third, and for imaging on the
their recall difference (per cent I-recall minus fourth. RIR subjects were exposed to the con-
per cent non-I,recall) revealed it to be con- verse of these conditions. The recall test and
siderably less than that o f subjects who said they debriefing procedures o f Experiment I were
had been consistent in their self-reports of covert employed.
behavior (10"5 per cent versus 24-7 per cent
respectively). This, o f course, would be expected RESULTS
if non-I-recall were being inflated by surreptitious The results o f the second experiment are
imaging. presented in Fig. 3, Again, group curves are
employed both for simplicity and because of the
E X P E R I M E N T II: P U N I S H M E N T remarkable correspondence between individual
Subjects. Subjects were again recruited from and group performances. The degree o f inter-
an introductory psychology course at Stanford subject variability is illustrated in Fig. 4 which
University. depicts the two most extreme individual per-
Procedure. The general paradigm o f Experi- formances in each group. The least conforming
ment I was employed except that subjects (six subject (S No. 9) indicated during debriefing
male, six female) were alternatively punished for that he had intentionally responded contrary to
self-reports of I and R behavior in paired- the contingencies to show that "he wasn't so
associate learning. The optional associative easily managed". It is interesting to note the
methods were restricted to I and R in order to consistency with which all other punishment
enhance the informative feedback o f negative subjects performed. S No. 20, the least con-
consequation. The paired-associate lists o f forming R I R subject, responded very consist-
Experiment I were employed and subjects were ently with experimental conditions. Likewise,
assigned to an IRI or an R I R punishment con- the correspondence between the reinforcement
dition depending on the self-reported baseline and punishment data are striking (compare
frequency of the covert target behaviors. After Figs. 1 and 3).
baseline, the following taped instructions were The punishment contingencies were found to
presented: be extremely effective in altering response fre-
quency. Of the 36 predicted I transitions for
During the remainder of the experiment you will hear individual subjects, 34 were in the expected
a noise whenever the method by which you choose to
associate a noun pair corresponds to a pre-established direction and one showed no change (sign test
standard. Notice that there is a large box located on P<0-001, two-tailed). A within-subject com-
the desk in front of you. The noise will come from that
box whenever your choice of associative method is parison o f I-recall showed the former to be
considered inappropriate. Thus, for example, you superior in 11 o f 12 cases (sign test P < 0 . 0 0 1 ,
might hear the noise for using the repetition method
to associate a particular n o u n pair. You are presently two-tailed). The recall difference for the 12th
the recipient of $4.00 cash for signing up and attending subject was zero. Averaged across individuals,
today's experiment. Each time you hear the noise
indicating you have chosen the inappropriate associative I-recall was 42 per cent as compared to 25-5 per
method, a quarter will be subtracted from your $4.00. cent for non-I items. As in Experiment I, most
Which method is inappropriate may vary from pair to
pair or list to list. You should attempt to maintain as subjects were able to describe the experimental
12 MICHAEL Jo MAHONEY, CARL E. THORESEN and BRIAN G. D A N A H E R
Punishment
*--e IRI N=7
Group curves o---o RIR N=5
/
P
/
o ;. /
o~
E
10
'k/" .,-"\ \ / /\ i
o
JN '/ / \o
N / ' ,"
/ / ,'
S
:E
V '\ i \ I
0 ] I I I i
Baseline Intervention Reversal Re-intervention
Experimental phase
FIG. 3. Mean frequency of imagery when subjects were alternately punished for salt-reports of imagery (I) or
repetition (R).
Punishment
e----e IRI
Individual curves
o---o RIR
/e 75#12
10
", A / \ "
V V Y
5
,,'\ /,,, / \
==
.E 0
' ,v V \,...
1,, " I I , i , I
Baseline Intervention Reversal Re-intervention
g
u2
IO
,/ eS~9
\ /
\ /
\ /
\ /
I 1 I I I
Boleline In'tervent'ion Reversal Re- in'tervenfion
Experimental iohose
FIG. 4. Individual response curves of subjects whose performance conformed most (upper figure) or least (lower
figure) to the punishment contingencies.
COVERT BEHAVIOR MODIFICATION: A N EXPERIMENTAL ANALOGUE 13
contingencies as debriefing. Two of the IRI (R) can become a boring mode of associative
punishment subjects reported that they had learning. This speculation is suggested by the
occasionally marked I when they were in fact research of Paivio and Yuille (1969) who found
employing R as an associative method. Four that subjects soon tired of R as a mediating
others admitted having frequently misrepre- response. It is possible that misrepresentation is
sented their covert behavior. An analysis of their less frequent in clinical situations involving
recall difference revealed that highly misrepre- motivated volunteers or clients. Nevertheless,
sentative subjects showed the least difference in the present data suggest that therapists working
recall performance (20-5 per cent). Subjects with covert response patterns would do well to
reporting occasional misrepresentations showed remember that at least t w o behaviors are being
moderate superiority (25.0 per cent), and the manipulated---one overt (e.g. verbal reports)
single IRI subject who said he had been con- and one covert (e.g. imagery). Their correspond-
sistent in his self-reports showed the greatest ence can only be cautiously presumed until
difference (27.0 per cent). related overt indices provide substantiation.
The implications of the present study for
DISCUSSION covert behavior modifiers are twofold: (1) some
The foregoing results lend support to the covert (cognitive) behaviors can be modified by
contention that covert behaviors are functionally their consequences, and (2) overt behavioral
related to their consequences. The covert re- indices may be very helpful in assessing the
sponse class of imagery was shown to be both magnitude and accuracy of self-reported changes
acceleratively and deceleratively manipulable via in covert responding. These findings substantiate
external contingency arrangements. A counter- previously reported attempts to modify covert
balanced order of manipulations across experi- behaviors. They also lend support to the
ments controlled for such variables as primacy, homogeneity assumption regarding overt and
recency, interference, and specific pair difficulty. covert behavior principles.
Although the act of imaging was inferred Given the clinical utility of imagery responses
from subjects' self reports, concurrent changes (Cautela, 1969) and the functional relationship
in recall performance support the notion that between covert behaviors and overt performance,
some consistent change in cognitive behavior the present evidence would seem to have direct
was taking place. A peripheralist interpretation bearing on therapeutic techniques in behavior
of the present results--i.e, that the behavior modification. Since covert responses play a
modified was that of writing the letter ' T ' on a significant role as reinforcers, punishers, and
piece of paper--leaves one with the dilemma of discriminative cues for other (often overt)
explaining why the peripheral response of responses, their modifiability presents a major
writing ' T ' was consistently associated with concern for adaptive behavior change. More-
superior recall performance. However, the fact over, cognitive forms of self-control have shown
that 36 per cent of the subjects in this study themselves to be therapeutically promising (cf.
reported that they had misrepresented their Mahoney, 1972; Thoresen and Mahoney, in
covert responses at least once suggests the need press). Finally, the fact that imagery responses
for further investigation. Although the superi- were demonstrated to be significantly manipu-
ority of I-recall allowed the present investigators lable in this study suggests the possibility that
to substantiate self reports, the tendency to mis- cases of imagery deficit may be amenable to
represent covert responses in situations of less treatment. For example, planned learning ex-
experimental control could prove troublesome periences for the strengthening of vivid multi-
to covert behavior modifiers. One possible sensory (e.g. visual, auditory, olfactory) imagery
explanation for the degree of misrepresentation may be possible (Danaher and Thoresen, in
evidenced in the present study is that repetition press; Phillips, 1971). What with the important
14 MICHAEL J. MAHONEY, CARL E. THORESEN and BRIAN G. DANAHER