JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR 1979, 319 177-187 NUMBER 2 (MARCH)

CONTROL BY STIMULUS FEATURES DURING FADING1

JUDITH DORAN AND JAMES G. HOLLAND
C. W. POST COLLEGE AND UNIVERSITY OF PITTSBURGH

Sixteen children were given four successive circle-size discrimination problems with lumi-
nance as the fading stimulus. Children who were first presented with a difficult size dis-
crimination failed to acquire this discrimination. Those who first received an easy discrim-
ination learned the difficult discrimination. At the end of each 10-trial block, two probe
stimuli were presented to monitor any shift in control from luminance to size. One probe
was the same size as the positive stimulus but of different luminance; the other was the
same luminance but of different size. If, in the course of fading, size and luminance both
controlled responding, fading was successful. If luminance alone controlled responding
until the end of fading, the size discrimination was not established. Dual control, and thus
successful fading, resulted when the target stimuli were very discriminable, or when the
target stimuli were subtly different provided that previous fading series had first estab-
lished less subtle discriminations.
Key words: stimulus control, fading, errorless discrimination learning, transfer of stimu-
lus control, window press, children

In recent years there have been several procedural analysis to determine the necessary
demonstrations of discriminations acquired and sufficient conditions for the establishment
without errors. These results were not pre- of errorless discrimination learning.
dicted from the prevailing theoretical concep- The fading procedures used in establishing
tion that extinction was necessary to form a errorless discriminations are similar in some
discrimination. Interest was largely focused respects to procedures used to investigate se-
on the fact that errorless discrimination learn- lective attention to compound stimuli. Logi-
ing was possible, on how errorless learning dif- cally, before the fading cue is removed, the
fered from errorful learning in various by- fading cue and target feature form a com-
products, and on the contributions that an pound stimulus. Either feature is sufficient to
analysis of errorless learning might make to control responding; thus, attention may be
discrimination learning theories that do not shared or selective. In selective attention
include interacting gradients of excitation and studies (e.g., Chase & Heinemann, 1972; John-
inhibition (Terrace, 1972). What has received son & Cumming, 1968; vom Saal & Jenkins,
relatively less attention, however, is a detailed 1970), control by one feature of a compound
stimulus in certain instances prevents or blocks
'This research is based upon a dissertation submitted the acquisition of control by other features
by Judith Doran to the University of Pittsburgh in present in the compound stimulus; in other
partial fulfillment of the requirements for the PhD instances, however, control by both features is
degree. This research was conducted at the Learning acquired during compound stimulus training.
Research and Development Center at the University of
Pittsburgh, supported in part as a research and devel- Thus, there are interesting similarities in pro-
opment center by funds from the National Institute of cedures for fading and selective attention and
Education (NIE), United States Department of Health, in behavioral outcomes.
Education and Welfare. The opinions expressed do not Selective attention studies with redundant
necessarily reflect the position or policy of NIE and no compound stimuli have suggested certain re-
official endorsement should be inferred. An earlier ver-
sion of this paper was presented at the 1976 meetings lations among various stimulus features; for
of the Midwestern Association of Behavior Analysis in example, pigeons are more likely to respond
Chicago. We thank Jeff Corey, John Jaeger, Karl Scheu- on the basis of visual cues rather than auditory
erman, and Carol Solomon for their helpful sugges- cues (Miles & Jenkins, 1973); monkeys are
tions. Reprints may be obtained from Judith Doran,
Department of Psychology, C. W. Post College, Long more likely to respond on the basis of color
Island University, Greenvale, New York 11548. rather than form or size (Warren, 1953). But
177
178 JUDITH DORAN and JAMES G. HOLLAND
these assertions must be qualified because it effect of changes in luminance between S+
has also been shown that cue selection can be and S- upon the acquisition of control by
altered by changes in discriminability of the size was evaluated for three size discrimination
available stimulus features (Gilner, Pick, Pick, problems ranging in difficulty. Three se-
& Hales, 1969; Imai & Garner, 1965; Miles & quences of size discriminations were investi-
Jenkins, 1973) or by prior training with the gated to determine if discriminability of the
feature less likely to gain control (Chase & target feature is affected by the learner's mas-
Heinemann, 1972). tery of easier discriminations along that di-
Previous studies with sea lions (Schuster- mension.
man, 1967) and with retarded children (Tou-
chette, 1971) measured stimulus control dur-
ing successful fading series and suggested that METHOD
shifts in stimulus control occur. The target Subjects
feature, previously ineffective in controlling
differential responding, becomes effective as Sixteen children, aged 6 to 7 yr, from neigh-
demonstrated by the lack of errors when the boring schools served.
fading cue is eliminated. In a similar study
with pigeons, Fields, Bruno, and Keller (1976) Apparatus
demonstrated that, during a successful fading A 20- by 15-cm display panel containing a
series, the target feature first gains control single response key was placed on a child's
while the fading cue still controls responding. table directly in front of a child's chair. The
However, responding to one of the probes key, 2.5 cm square, was mounted in the center
used to measure the controlling feature may of the panel. The key could be illuminated
have been affected by an earlier history of by white circles with diameters ranging from
extinction in the presence of the same stimulus 5 mm to 14 mm on a black background. These
compound. Furthermore, the Fields et al. study circles were projected by a 28-V inline projec-
examined stimulus control shifts only during tor (Industrial Electronic Engineers, Inc.,
a successful fading series. Terrace (1963) with Series 1820L). The luminance and, to a slight
pigeons and Cohen (1968) with children both extent, the spectral composition of a particu-
found that a fading series which was successful lar circle were controlled by resistors inserted
for one type of stimulus was unsuccessful with into the projector circuit. The luminance
other stimuli. Why fading should work for values ranged from approximately -1.5 log ft.
some target stimuli and not others is as yet L., the dimmest, to + 1.28 log ft. L., the bright-
an unanswered question. est. Luminance measurements were made with
To investigate this question, the present a Salford Electrical Instruments photometer.
study measured stimulus control during suc- The experimental room throughout each ses-
cessful and unsuccessful fading and analyzed sion was illuminated at -1.7 log ft. L. by
factors influential in affecting the type of means of a tungsten lamp containing a 60-W
stimulus control shown throughout fading. bulb and overhead fluorescent lights con-
One possible factor, suggested by research in trolled by a rheostat. A Grason-Stadler marble
selective attention, is discriminability of the dispenser was mounted on the left side of the
stimulus features. In a fading series, changes panel, and a door chime was mounted under
in discriminability may occur. As the fading the table.
cue becomes less discriminable, the target cue, Reinforcement consisted of a marble dis-
if highly discriminable, is more likely to be- pensed into a long tube and a ringing of the
come effective in controlling differential re- chime. After each session, marbles were ex-
sponding. But if the fading and target features changed for M&M's. At the end of the last size
are both low in discriminability, which may discrimination session, the child took home a
obtain at the final stages of a fading series, the small toy that had been selected at the begin-
fading cue controls until it is removed. Then ning of training.
errors occur. Stimulus changes, response recording, and
This experiment analyzed the acquisition reinforcement contingencies were automati-
in young children of circle-size discrimina- cally controlled by electromechanical equip-
tions, using a luminance fading series. The ment.
 STIMULUS CONTROL DURING FADING 179
sented in six 10-trial blocks, each with five S+
Procedure and five S- trials randomly presented, except
Preliminary training. The child was brought that no more than three trials of either type
to the experimental room and seated in front occurred in succession. Each S+ trial was
of the stimulus panel on which S+ ap- terminated either by a response which turned
peared. These instructions were given: "Press off the stimulus light and, with a .75 probabil-
this circle. Sometimes when you press this ity, produced a marble and sounded the chime,
circle, a bell will ring and a marble will drop or by a 5-sec interval without a response. After
down this tube. You get one M&M for every a 4-sec ITI, the next trial began unless a re-
two marbles." sponse had occurred during ITI. Each re-
After each circle press in this preliminary sponse during ITI reset a timer which delayed
session, an intertrial interval (ITI) occurred the onset of the next trial by 4 sec. Each S-
during which the key was dark. After the first trial ended after its specified duration if no
response, the ITI was .5 sec. Over the next response occurred. In the instances of a re-
seven trials, the ITI increased in .5-sec steps sponse during S-, the timer reset so that the
until ITI was 4 sec. Any response during ITI trial terminated only after the required inter-
reset a timer which delayed the onset of S+ val (See Table 1) passed without a response.
by the value of ITI during which the response Upon the termination of S-, an ITI of 4 sec
was made. followed before the onset of the next trial.
An intermittent schedule of reinforcement Probe stimuli were used to assess control by
of responses to S+ was introduced during this the size and luminance cues. Figure 1 pre-
session to minimize the disruptive effects of sents the size and luminance of S+, S-, and
unreinforced probe trials to be used in later probe stimuli for each of the three problems.
sessions. Each circle press was reinforced for One type of probe, the target probe, was a
the first 20 S+ trials. Responses on the next small bright circle with luminance equal to
12 S+ trials were reinforced with a probability S+ at +1.28 log ft. L. and size equal to the
of .92 (11 of the 12 trials); responses on the S- of the particular problem, either 5 mm,
next 12 S+ trials were reinforced with a prob- 10 mm, or 12 mm. The other type of probe,
ability of .83 (10 of the 12 trials); and, fi- the fading probe, was a circle as large as S+
nally, over the last 24 S+ trials, responses that took on the luminance value of S- at
were reinforced with a probability of .75 (18 five points along the fading series, indicated
of the 24 trials). by the superscript "1" in Table 1. Ten probe
A child who responded during each S+ and trials occurred during each session, five of
did not respond during any ITI on the last 12 each type. After every 10th trial, each type of
trials was considered to have achieved cri- probe was presented once with the order ran-
terion. If criterion was not met, the prelimi- domly chosen. A probe trial was terminated
nary session continued until 12 trials elapsed either by a response or by the end of a 5-sec
with responses to each S+ and with no re- period without a response. No reinforcement
sponses during ITI. was given on probes. Ten criterion trials, with
Size discrimination training. Size discrimi- S+ and S- at full luminance and differing
nation training began on the next session. The only in size, followed the 10th probe trial.
following instructions were given: "Do you If the subjects responded to each S+ and did
remember the circle you saw yesterday? Today not respond to each S- on criterion, the size
you will see yesterday's circle and some other discrimination had been established.
circles, but only press yesterday's circle. Today In each session, the same fading series was
you will get an M&M for every marble that used. The luminance of the S- increased
comes down. Remember some are different, over 25 S- trials from -1.5 log ft. L. to +1.28
only press yesterday's circle. Ready?" On the log ft. L. Table 1 presents the luminance
following days, the child was told: "Remem- values of S- along the fading series. At the
ber, some are different, only press yesterday's end of the fading series, both S+ and S- were
circle. Ready?" Each training session consisted at +1.28 log ft. L.; size was the only distin-
of successive presentations of 70 trials. They guishing feature between S+ and S-. As
included 30 S+ trials, 30 S- trials, and 10 shown in Table 1, within each block during
probe trials. S+ trials and S- trials were pre- which the S- was undergoing changes in lumi-
180 JUDITH DORAN and JAMES G. HOLLAND
Targ3et Fading
S- Probe Probes

0'
0
DO '0
Luminance +1.28 Progresses +1.28 -1.5 -1.35 -1.06 +0.11 +1.28
(log ft. L.) from
-1.5 to
+1.28

Size (mm) 14 12 12 14 14 14 14 14
10 10
5 5
Figure 1. Size and luminance values of S+, S- and probes for each of three successive discrimination problems.
The rows describe the different problems; columns show stimuli used within each problem.

nance, the duration of S- increased over the Three sequences of problems were investi-
five trials from 3 to 5 sec. This gradual length- gated. Some subjects were given the difficult
ening of the S- within a block was used as problem first, then the easy and intermediate
an additional means of preventing errors. problems, and finally a second session on the
Each subject received four size discrimina- difficult problem. This is called the Difficult/
tion sessions. In each session, S+ was a circle Gradual Progression Condition. Other sub-
14 mm in diameter with a luminance of +1.28 jects received the easy problem and the inter-
log ft. L. In different problems, S- was a mediate problem before two sessions on the
circle 5 mm in diameter (the easy problem), difficult problem. This condition is called the
10 mm in diameter (the intermediate prob- Gradual Progression Condition. Other sub-
lem), or 12 mm in diameter (the difficult jects were given the difficult problem in each
problem). of the four sessions. This is called the Difficult
Condition. On the fourth session, subjects in
Table 1 all conditions received the difficult problem;
Duration (sec) and luminance values (log ft L) of S- their performance was expected to reflect their
during fading in successive 5-trial blocks. experimental histories, with only subjects in
Dura- Blocks2 the Difficult/Gradual Progression and Gradual
tion., Progression Conditions responding differen-
(sec) 1 2 3 4 5 tially on the target stimuli at criterion when
3.0 bulb off -1.50 -1.35 -1.06 + .11 the fading cue was removed.
3.5 -1.50 -1.35 -1.21 - .77 + .40
4.0 -1.50 -1.35 -1.21 - .48 + .69
4.5 -1.50 -1.35 -1.06 - .18 + .99 RESULTS
5.0 -1.501 -1.351 l.1061 + .111 +1.281
Preliminary Session
'Value used in fading probe.
2In Criterion testing, the duration of S- was 5.0 and The preliminary session presented only S+
luminance was +1.28. trials followed by ITI. After a nonreinforced
 STIMULUS CONTROL DURING FADING 181
response on the intermittent schedule, some the number and location of errors within the
subjects responded during the following ITI fading series on criterion trials.
when the key was dark. The preliminary ses- Between each block of 10 trials (indicated
sion was then extended until 12 trials oc- by a triangle on the abscissa) a set of two
curred with a response to each S+ and no probe trials occurred in random order: a
responding during ITI. The mean number of target probe as bright as S+ but only as large
intertrial responses was 3.4, 2.5, and 2.6 for as the S- of the particular problem, and a
the Gradual Progression, Difficult/Gradual fading probe as large as S+ but only as bright
Progression, and Difficult Condition, respec- as the preceding S- in the fading series (see
tively. The difference in group means was Figure 1). Performance on the two probes,
not reliable at the .05 level of confidence considered as a pair or set, indicated the
(Kruskal-Wallis one-way analysis of variance). nature of the stimulus control at five points
along the fading series for each size discrimi-
Size Discrimination Sessions nation problem. By the fifth probe set at the
Individual data for the four size discrimina- end of the fading series, the fading probe
tion problems given to each subject are shown was also as bright as S+. Thus the fifth probe
in Figure 2. The individual curves present set contained criterial probes in that the tar-
performance during the programmed series, get probe was the same as the criterion S-
on criterion trials, and on probe trials. The and the fading probe was the same as the cri-
number of trials on which an error occurred terion S+. Because the luminance cue at this
is plotted along the ordinate for each size dis- point in the series had been completely faded,
crimination problem. An error is defined as no control by luminance was possible.
a response to S- or the lack of a response to Stimulus control was measured through use
S+ within 5 sec of its onset. Because S- was of the probe data. Control by luminance alone
not terminated until a required number of was indicated when the subject responded on
sec had elapsed without a response, multiple the target probe and did not respond on the
errors on an S- trial were possible. (Multiple fading probe. Control by size alone was indi-
errors occurred relatively infrequently, how- cated when the subject did not respond on
ever, and will be presented later in Table 2.) the target probe but responded on the fading
To simplify the discussion in the text that fol- probe. Dual control by luminance and size
lows, a trial with errors will be referred to was indicated when the subject did not re-
simply as an "error." spond on either probe. Control by neither
The subjects were randomly assigned to one luminance nor size was indicated when the
of three groups, each with a different se- subject responded on both probes. The probe
quence of size problems. The sequence given data were only meaningful, however, if re-
each subject is indicated by the numbers in sponding on S+ and S- was essentially error-
parentheses. Problems are identified by the less.
difference in mm of S+ and S- circle diame- Difficult condition. The records of five sub-
ters; thus, the easy problem is identified on jects given four sessions on the most difficult
the figures as "9-mm difference," the inter- problem (2 mm difference in S+ and S- di-
mediate problem as "4-mm difference," and ameter) are in the upper left panel of Figure 2.
the difficult problem as "2-mm difference." All five subjects performed nearly errorlessly
The abscissa presents trial-by-trial perfor- until the fifth trial block of each session when
mance, labeled according to trial blocks of 10 the luminance cue became very subtle. Three
trials each. Each block contained five S+ and of the five subjects, Carol, Jean, and Arletha,
S- trials. Criterion trials occurred in Block 6. made extensive errors on all four presentations
The same sequence of S+ and S- trials was of the criterion trials in Block 6. Two sub-
given to each subject on each of the four size jects, Matt H. and Susie, also made errors on
problems. For each trial, the data line moved criterion trials, but their performance showed
one step horizontally. If no errors occurred, some evidence of control by size.
the data points formed a straight horizontal The probe data for the three nonlearners,
line. Each error, however, moved the data Carol, Jean, and Arletha, revealed almost ex-
line one step upward. The amount and loca- clusive control by luminance through the
tion of deviation from the horizontal indicate fourth probe set. Then, as the luminance cue
182 JUDITH DORAN and JAMES G. HOLLAND
DlIFFICULT PROBE SET = v
(2222) mm CONTROL BY:
LUMINANCE = 0--
SIZE = -0-
BOTH(DUAL) = -E-
Matt H. NEITHER =

S u s Sie
DIFFICULT/GRADUAL PROGRESSION
(2942)mm
I
oCa

cc Rosanna
_pp ~ Jean
0
UJ rel l E
[.0
David
Arletha
cn GRADUAL PROGRESSION
-J (9422)mm E
El O 0 0E ~~EEEl S
CI
Kirsten
Ara
N._ _
El =w E M. 0 El _El Ml

Amy Charlotte

El El ~~~~-E 0 OLj IM."r,

El 0 El -3 *Bess SQO

1El9 0 0 5 lo El."E ElS. *

David G.
Eml El l E 0
o 0 El El~ ~~~
S
Matt B.
El
ID
~~0-~~~~ennifenier
E w-
l 0

1 2 3 4 5 6 1 2 3 4 5 6
TRIAL BLOCKS
Figure 2. Performance of sixteen children given four sessions of successive circle-size discrimination problems.
For each child, Session One is the top data line. Each problem had a 50-trial fading series on luminance (Blocks
1-5), a 10-trial criterion (Block 6), and 5 probe sets (indicated by filled triangles on the abscissa). Each trial moves
the data line one unit to the right; each errorful trial moves the line one step upward. Each child received one
of three conditions. The Difficult Condition, indicated by (2222)mm, presented four sessions on the "Difficult" dis-
crimination (2 mm difference between S+ and S-). The Gradual Progression Condition, (9422)mm, presented size
discrimination problems in the sequence: 9 mm, 4 mm, 2 mm, 2 mm. In the Difficult/Gradual Progression Con-
dition, (2942)mm, the sequence was 2 mm, 9 mm, 4 mm, and 2 mm.
 STIMULUS CONTROL DURING FADING 183

was removed in the fifth trial block, errors oc- errors because of their mastery of easier size
curred: subjects responded on nearly all S+ discriminations. A comparison of each sub-
and S- trials. Probe data demonstrated a gen- ject's performance on the third and fourth
eral lack of control by the target feature- session with the performance of subjects in
size-with only Carol showing control by size the Difficult Condition on the third and fourth
on a single probe set in the first session when session supports this expectation. Four of the
the luminance cue was no longer available. On five subjects in the Gradual Progression Con-
the criterion trials of the first session, she dition performed errorlessly on all criterion
made two errors. On her next three sessions trials of the difficult problem (2 mm differ-
of the difficult problem, however, there was ence); the fifth subject, Ara, made only one
no indication of control by size on criterion error on criterion in the third session.
trials or on probes. None of the probes for In general, probe data showed dual control
Carol, Jean, or Arletha revealed dual control by luminance and size well before the lumi-
by luminance and size. nance cue was completely faded on the easy,
The other two subjects in this group, Matt intermediate, and difficult size discriminations.
H. and Susie, acquired limited and unreliable Some individual differences in stimulus con-
control by size after errors were made, but trol were revealed by the probes. For example,
this control did not persist in subsequent ses- on the easy problem (9 mm difference), three
sions. For Susie, even after errorless perfor- subjects shbwed dual control from the first
mance on criterion on the second and third through the fourth probe set (Ara, Bess, and
sessions, responding on subsequent sessions be- Matt B.); one subject showed dual control
gan under exclusive control of luminance. through the third probe set (Amy); and one
Size control developed late in each session, subject showed dual control only on the first
only after errors occurred as luminance was probe set, and control by size alone on the fol-
completely faded in the fifth trial block. Matt lowing four probe sets (David G.). All sub-
H.'s errors decreased consistently from the jects showed control by size alone on the fifth
first through the fourth session, but on each probe set and performed errorlessly on cri-
of the first three sessions, errors occurred in terion for the two easier problems. On the
the fifth block as the luminance cue became first administration of the difficult problem,
more and more subtle. Unlike the probe data however, two subjects showed control by lu-
for the other four subjects, Matt's probe data minance alone (Ara and Matt B.). With Matt
did indicate extensive periods of dual control B., dual control emerged by the third probe
by luminance and size. set. Ara had some difficulty as the luminance
In general, for three of the five subjects cue was removed, indicated by two errors in
given four sessions on the difficult problem, the fifth trial block of the third session. The
there was no indication of control of respond- other three subjects showed dual control on
ing by size on any of the four sessions. Re- the first four probe sets of the third session,
sponding was nearly errorless and under the followed by control by size alone on the fifth,
sole control of luminance until the luminance criterion probe set.
differences became very subtle. When lumi- Difficult/Gradual progression condition. Six
nance was eliminated, responding was no subjects were given the most difficult problem
longer differential. The other two subjects as their first size discrimination problem, fol-
in this group acquired limited and unreliable lowed by a progression in problem difficulty
control by size after errors were made, but in the sequence: easy, intermediate, and diffi-
this control did not persist in subsequent ses- cult. When the most difficult problem was
sions. In general, on each session, errors oc- presented first, the fading series uniformly
curred as the luminance cue was removed. failed to establish the size discrimination. As
Gradual progression condition. Data from shown in the right side of Figure 2, the sub-
these five subjects are shown in the lower left jects performed reasonably well until quite
panel of Figure 2. This group received the near the end of the luminance fading series.
easy problem and intermediate problem be- Two of the subjects (Kirsten and David R.)
fore two sessions on the difficult size discrimi-- made no errors in the first four blocks, the
nation problem. It was expected that they others made a few scattered errors early in
would learn the difficult size problem without the fading sequence. However, by the time the
184 JUDITH DORAN and JAMES G. HOLLAND
fading cue was eliminated or almost elimi- David R., performed perfectly on the criterion
nated, all subjects showed a lack of control trials and on the preceding block when the
by the target feature. During Block 6 with luminance cue was quite subtle. These two
criterion trials, four subjects responded on all subjects showed the characteristic performance
five S- trials, and the remaining two re- of the earlier group; others made a few errors.
sponded on four of the five S- trials. Two made errors only in the fifth block of
The probe technique revealed the type of trials in the final stage of the luminance fad-
stimulus control during what superficially ap- ing. The remaining two, Jennifer and Levi,
peared to be a successful fading sequence. showed errors in the fifth block and on the
While few errors were made, only one of the criterion trials as well, with Jennifer respond-
subjects showed dual control at any point in ing to two of the five S- trials. Levi's per-
the first fading session, and that was a tempo- formance was characteristic of a nonlearner
rary dual control at the third probe set for (cf. Arletha's data in the Difficult Condition).
Jennifer. With this single exception, the first He responded errorlessly through almost all
three sets of probes all showed luminance of the fading series, but the probes showed
control alone. On the fourth probe, the last that he continued under the control of lumi-
one in which luminance remained as even a nance alone until, on the fourth probe set, he
subtle cue, the subjects differed in whether showed control by neither feature. Shortly
they showed a continued reliance on lumi- thereafter he began responding to S- both
nance alone or whether they showed control by just before and just after the final criterion
neither feature. And finally, in the fifth cri- probe which, paradoxically, indicated size
terial probe, no control was demonstrated; control.
subjects responded on both probes and on For four of the six subjects, the fading series
S+'s and S-'s alike. and the progression in difficulty in the size
In their second and third sessions when discrimination worked to establish more or less
these subjects were given the fading series with errorless learning. For a fifth subject, Jennifer,
the easy and intermediate size discriminations, there was considerable evidence of control by
their performance along the fading series was the target feature, but that control was not
quite'similar to the performance of the chil- completely reliable; and for the sixth, Levi,
dren in the Gradual Progression Condition. one would hesitate to claim successful control.
Dual control by both properties was shown Overall, the performance of these subjects
quite early, usually by the first probe set in on the difficult discrimination given after the
both of these problems. Occasionally, the early two easier discriminations compares somewhat
probes showed control by luminance alone unfavorably with the performance of sub-
(David R. and Rosanna), followed by dual con- jects in the Gradual Progression Condition.
trol. By the fifth criterial probe set, all subjects The exposure on the first session to the diffi-
were controlled by size alone, as luminance cult problem before the progressive series
was no longer available as a cue. In two in- seemed to hinder, somewhat, performance on
stances (David R. and Jennifer), size alone the fourth session despite intervening error-
controlled during the fourth probe set when less performance on the easy and intermediate
the luminance cue became more subtle. problems. That the easier problems did have
When, on the fourth session, these subjects a facilitating effect, however, is revealed when
were again given the difficult discrimination this group's performance is compared with
(2 mm difference), luminance alone typically that of subjects in the Difficult Condition
controlled responding according to the early who had no experimental history with easier
probes, and dual control emerged at least be- size problems.
fore the fourth probe set. In one instance
(Jennifer), control had shifted from dual con- Multiple Errors on S- Trials
trol to control by size alone on the fourth Multiple errors on S- trials were possible
probe set. Overall, the subjects in this group because S- was not terminated until a re-
on their second exposure to the difficult prob- quired number of seconds without a response
lem made more errors than subjects on their had elapsed. (The duration of S- increased
first or second exposure in the Gradual Pro- from 3 to 5 sec within each trial block.) In
gression Condition. Two subjects, Kirsten and Figure 2, trials with errors were plotted with-
 STIMULUS CONTROL DURING FADING 185

out distinguishing between single-error and make errors as the luminance cue was removed
multiple-error S- trials to make visible com- did not persist in multiple responding dur-
mon patterns of responding that plotting each ing S-.
error would obscure.
Table 2 presents, first, the total number of
errors that occurred on each problem for each DISCUSSION
subject including both single-error and multi- A successful fading sequence produces error-
ple-error trials; second, the number of multi- less responding throughout acquisition. In
ple-error trials; and third, the mean number order to do this, the target feature must con-
of multiple errors on multiple-error trials trol responding before the fading cue is com-
only. As can be seen in Table 2, some subjects pletely removed. In the present study, consid-
never responded to S- and some subjects never ering only the performance on S+ and S-
responded more than once on any S- trial. trials, responding in successful and unsuccess-
Those subjects who did respond repeatedly ful sequences was indistinguishable over the
during S- did so primarily in their first ses- early part of all sequences. Performance on
sion. Levi and Arletha were the only subjects probes, however, revealed striking differences
who showed extensive multiple errors. In the in the features which actually controlled er-
first session of the difficult problem as lumi- rorless responding. Exclusive control by the
nance was faded, Levi and Arletha responded fading cue was characteristic of unsuccessful
twice on nearly each S- trial through the end series; dual control by both the fading cue
of criterion. and the target features was characteristic of
In general, after the first session, multiple successful series. In general, only those sub-
responses during S- occurred infrequently. jects who showed dual control on probe stim-
The subjects who did not learn the target size uli continued to respond errorlessly on target
discrimination and who thus continued to discrimination.

le 2
Total number of errors, number of trials with multiple errors, and mean number of mul-
tiple errors for each size problem.
Mean number of
multiple errors
Total number of Number of trials on multiple
Subject errors with multiple errors error trials
Problem Problem
(mm difference between (mm difference between
S+ and S- diameters) S+ and S- diameters)
9 4 22 22 9 4 2, 22
David G. 0 0 0 0 0 0 0 0
Amy 1 0 2 0 0 0 0 0 -
Ara 0 0 4 0 0 0 1 0 2.00
Bess 0 0 0 0 0 0 0 0 -
Matt B. 0 2 0 0 0 0 0 0 -

21 9 4 22 21 9 4 22
Jennifer 13 0 0 4 1 0 0 0 2.00
Charlotte 12 0 0 2 0 0 0 0 -
Rosanna 15 0 0 1 2 0 0 0 2.00
David R. 7 1 1 0 0 0 0 0 -
Kirsten 9 0 0 0 2 0 0 0 2.00
Levi 16 0 0 7 6 0 0 1 2.14
21 2, 2s 24 2, 2, 2s 24
Carol 3 12 10 8 0 3 0 0 2.33
Jean 5 8 7 7 0 1 0 0 2.00
Arletha 27 7 7 8 8 0 1 1 2.20
Susie 3 2 2 5 0 0 0 0 -
Matt H. 5 3 2 1 1 0 0 0 2.00
186 JUDITH DORAN and JAMES G. HOLLAND
Several factors influenced dual control and 1969), graphemes were color-coded according
thus the likelihood of successful fading. When to particular sounds to be produced. Color
the same fading sequence on luminance was was the fading cue and letter shape was the
used with size discriminations of varying diffi- target feature. Observation of the target fea-
culty, errorless learning was accomplished if ture may be facilitated if the effect of gradu-
the size differences were prominent, but not if ally withdrawing or fading the original cue
the size differences were small. With the prom- is to make the target feature more prominent.
inent size cue, dual control by luminance and The present study has shown that this effect
size was usually indicated early; with the dif- is most likely if the target feature is not ex-
ficult size cue on the first session, luminance tremely subtle.
control continued until this cue was no longer The transfer method of stimulus control is
available and thereafter errors were extensive. risky, however, because the desired precursory
Errorless learning of this difficult discrimina- behavior (observing the target stimulus and
tion was accomplished in later sessions if it responding to it) may never be established. In
had been preceded by easier discrimination shoddy instructional materials, the private
tasks along the same dimension. The perfor- precursory behaviors that lead to errorless
mance of subjects who began with the difficult public behaviors may be considerably differ-
problem before the gradual progression was ent from the behavior desired by the designer;
somewhat inferior to the performance of those yet a correct response leads to reinforcement
who began with the gradual progression, even no matter what private behavior preceded it.
though both groups performed errorlessly on Nevertheless, the transfer of stimulus control
the easy and intermediate problems. Sidman method merits continued investigation and
and Stoddard (1967) and Touchette (1968) analysis to clarify further how control is ac-
also found that giving subjects an initial dif- quired by the target stimulus property at the
ficult task interfered with the subsequent ac- early stages of training.
quisition and retention of the particular dis-
crimination on which errors were made.
An important consequence of a history of REFERENCES
errors is that not only is the appropriate Chase, S., & Heinemann, E. G. Choices based on re-
behavior not learned, but also inappropriate dundant information: An analysis of two-dimen-
behavior may become established. In this study sional stimulus control. Journal of Experimental
Psychology, 1972, 92, 161-175.
the appropriate behavior required a careful Cohen, M. The role of S- responding in discrimina-
observation of the size of the circles. With tion learning (Doctoral dissertation, University of
prominent differences in the stimuli on the Pittsburgh, 1967). Dissertation Abstracts, 1968, 28,
target feature, subjects may be more likely 4308B. (University Microfilms No. 68-4429).
Fields, L., Bruno, V., & Keller, K. The stages of acqui-
to observe these differences, thereby increasing sition in stimulus fading. Journal of the Experimen-
the likelihood of the target stimulus acquir- tal Analysis of Behavior, 1976, 26, 295-300.
ing control. If these differences are gradually Gilner, C. R., Pick, A. D., Pick, H. L., & Hales, J. J.
made more subtle, this observing behavior, A developmental investigation of visual and haptic
preferences for shape and texture. Monographs of
once established, is likely to be maintained. the Society for Research in Child Development,
Inappropriate behavior included responding 1969, 34 (6, Serial No. 130).
to each circle without regard to its size. Such Imai, S., & Garner, W. R. Discriminability and prefer-
inappropriate behavior would be intermit- ence for attributes in free and constrained classifi-
tently reinforced and, especially for these cation. Journal of Experimental Psychology, 1965,
69, 596-608.
children who were already working on an Johnson, D. F., & Cumming, W. W. Some determiners
intermittent schedule, it might be expected of attention. Journal of the Experimental Analysis
to persist on the particular discrimination of Behavior, 1968, 11, 157-166.
once it was established. Kjeldergaard, P. M., Frankenstein, R., & Glaser, R.
Stepping stones to reading: Experimental edition.
Many instructional tasks begin with stimulus New York: New Century, 1969.
control already within the learner's repertoire Miles, C. G., & Jenkins, H. M. Overshadowing in oper-
and gradually change this control until the ant conditioning as a function of discriminability.
target stimulus control is established. For ex- Learning and Motivation, 1973, 4, 11-27.
Schusterman, R. J. Attention shift and errorless re-
ample, in lessons designed to teach beginning versal learning by the California sea lion. Science,
reading (Kjeldergaard, Frankenstein, & Glaser, 1967, 156, 833-835.
 STIMULUS CONTROL DURING FADING 187
Sidman, M., & Stoddard, L. T. The effectiveness of Touchette, P. E. Transfer of stimulus control: Mea-
fading in programming a simultaneous form dis- suring the moment of transfer. Journal of the Ex-
crimination for retarded children. Journal of the perimental Analysis of Behavior, 1971, 15, 347-354.
Experimental Analysis of Behavior, 1967, 10, 3-15. vom Saal, W., & Jenkins, H. M. Blocking the develop-
Terrace, H. S. Errorless transfer of a discrimination ment of stimulus control. Learning and Motivation,
across two continua. Journal of the Experimental 1970, 1, 52-64.
Analysis of Behavior, 1963, 6, 223-232. Warren, J. M. Additivity of cues in visual pattern dis-
Terrace, H. S. By-products of discrimination learning. crimination learning in monkeys. Journal of Com-
In G. H. Bower & J. T. Spence (Eds.), Learning and parative and Physiological Psychology, 1953, 46,
motivation (Vol. 5). New York: Academic Press, 1972. 484-488.
Touchette, P. E. The effects of graduated stimulus
change on the acquisition of a simple discrimination
in severely retarded boys. Journal of the Experi- Received May 2, 1977
mental Analysis of Behavior, 1968, 11, 39-48. Final acceptance August 24, 1978