JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR VOLUME 9, NUMBER 4 JULY, 1966

STIMULUS GENERALIZATION: THE ORDERING AND

SPACING OF TEST STIMULI'
JOAN G. STEVENSON2
BROWN UNIVERSITY

Twenty-four pigeons learned a successive discrimination between 500 m/A (S+) and 574 my
(S-). When tested in extinction, some birds received stimuli around S+, with no S- presenta-
tions. These birds showed a positive peak shift, with maximum responding not at 550 m/,
but displaced to 538 m,u and 544 m,u. Other birds were tested with stimuli around S-, with
no S+ presentations. These birds showed a negative shift, with least responding not at 574 my,
but at 586 m/t. Though the first group was tested around S+ and the second around 5-, total
responding between groups did not differ. When retested on the other half of the continuum,
however, birds that had gone from the S+ half to the S- half responded fewer times than
those that had gone from the S- half to the S+ half. In a second experiment, reducing
stimulus spacing from 6 m,u to 2 m/A produced flatter gradients and decreased the amount of
positive shift. In a third experiment, birds were tested across the whole continuum with
stimuli presented in serial order. A sequence from 538 myL to 586 m/A produced no responding
after the first part of the session; a sequence from 586 mA to 538 m,u produced responding
throughout the session.

Hanson (1959) found that successive dis- few responses have occurred to the S-, no
crimination training to a positive stimulus shift occurs in testing (Terrace, 1964). Unlike
correlated with reinforcement (S+), and a the usual successive procedure, this "errorless"
negative stimulus correlated with non-rein- procedure does not produce a change in the
forcement (S-), produced a post-discrimina- rate or latency of S+ responses opposite to the
tion gradient with a peak shifted away from change in rate or latency of S- responses
the S+, in a direction opposite that of the S-. (Terrace, 1963). These results suggest that op-
This positive peak shift has occurred only erations which fail to produce such "behav-
when training has involved a successive dis- ioral contrast" (Reynolds, 1961) during
crimination procedure (Guttman, 1959; Han- (S+,S-) training lead to the absence of a shift
son, 1959; Honig, Thomas, and Guttman, in subsequent testing. Thus, only when train-
1959; Honig, 1962; Thomas, 1962; Terrace, ing operations have involved a successive
1964; Friedman and Guttman, 1965). No shift (S+,S-) procedure in which S- responding
has occurred after training on a simultaneous and behavioral contrast occurred has a shift
discrimination (Honig, 1962), or after S+ been observed in testing.
training with blackouts followed by massed Given training operations sufficient to pro-
extinction trials to S- (Honig et al., 1959; duce a peak shift, the question arises as to
Friedman and Guttman, 1965). Even with suc- whether S- responding must also occur in
cessive discrimination training, if zero or very testing. Most experiments have varied test con-
ditions following simple S+ training without
any S-. Hiss and Thomas (1963) have shown
'This paper is based on a dissertation submitted to that if only one stimulus is presented to each
the Graduate School of Brown University in partial bird during testing, the resulting group gradi-
fulfillment of the requirements for a Ph.D. degree. The ent is similar to gradients obtained by pre-
author was supported by a U.S.P.H.S. predoctoral fel- senting every stimulus to each bird. But in
lowship. The author would especially like to thank Dr.
D. S. Blough for his guidance throughout the course of another experiment, gradients obtained using
this research, and also Dr. E. Hearst and Dr. R. A. single stimulus presentations were different
Hinde for their comments on the manuscript. from gradients obtained using multiple stimu-
2Reprints may be obtained from the author, Sub- lus presentations (Kalish and Haber, 1963).
Dept. of Animal Behaviour, Madingley, Cambridge, However, Thomas and Barker (1964) showed
England. that this difference is probably not due to the
457
 19383711, 1966, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1901/jeab.1966.9-457 by CAPES, Wiley Online Library on [11/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
458 JOAN G. STEVENSON
method of stimulus presentation, but to a on the amount of shift were determined by
steepening of the gradient as testing, or extinc- using 6 m,u spacing in one experiment, and 2
tion, progresses. Furthermore, Hearst and mMu spacing in the second. Since the ordering
Poppen (1965) have shown "that the oppor- of the halves assumed importance, a third ex-
tunity for subjects to compare stimuli during periment involved a generalization test which
testing . . . is not critical in accounting for the simply began at either end of the continuum
steepening-in-extinction effect" (p. 83). These and worked through to the other.
results following simple S+ training imply
that gradients following (S+,S-) training METHOD
might also be independent of the method of
stimulus presentation. Honig (1962) varied Subjects and Apparatus
test conditions after simultaneous and succes- Twenty-four male White Carneaux pigeons,
sive (S+,S-) training as well as after simple between 1 and 2 yr of age, all experimentally
S+ training. Following any given training con- naive, were maintained at approximately 75%
dition, the peak of the gradient obtained un- of free-feeding weights.
der simultaneous test conditions occurred at The birds worked in a ventilated, light-tight
the same stimulus value as the peak obtained aluminum box. Each bird pecked a translu-
under successive conditions. The present study cent, circular, 3/4-in. response key, operated by
further alters the test conditions, to eliminate a minimum force of 10 g. Below and to the
any opportunity for S- responding or com- left of this key was an aperture for food pre-
parison of S+ and S-; it asks if a positive sentation via a solenoid-operated food mag-
peak shift will occur if the S- and stimuli azine. During each 4-sec food presentation, a
around it are not presented in testing. white, luminescent disc dimly illuminated the
The present study also asks if a negative feeding aperture. A speaker in the box pro-
shift will occur if the S+ and surrounding vided white masking noise.
stimuli are not presented. A negative shift is Stimulus system. Light from a ribbon fila-
defined as displacement of a minimum away ment lamp, run on stabilized voltage at 17
from the S-, in the direction opposite the S+. amp ac, went to a motor-driven Bausch and
In most peak shift experiments, there have Lomb 250-mm grating monochromator, with
been so few responses to the S- that a further a band width of 13.2 mu. The light beam then
decrease in responding to adjacent stimuli passed through an infra-red absorbing filter,
could not convincingly be observed. But in two achromatic lenses, and an appropriate
one experiment (Guttman, 1965), when re- Wratten neutral density filter, which was
sponding did occur to the S- in testing, due placed in the beam by a motor-driven filter
to specific pretraining over the whole con- wheel. (The experimenter had taken bright-
tinuum and a weak discrimination training ness measurements at the response key, with
criterion, a negative shift was observed. If a a MacBeth illuminometer. Because of the sim-
negative shift were to occur in the present ex- ilarity between human and pigeon spectral
periment, it would indicate that specific pre- sensitivity functions (Blough, 1957), it is as-
training, a weak discrimination criterion, and sumed that the corresponding filter correc-
the presence of the S+ in testing are not tions provided stimuli of approximately con-
necessary to produce a negative shift. stant luminance for the pigeon.) Finally, with
In short, a peak shift has occurred only after a shutter open, the beam could pass through
successive (S+,S-) training, where S- re- another lens, to be projected on the back of
sponding was observed. The first two experi- the response key.
ments of the present study ask if the absence Control and recording. From an adjacent
of successive (S+,S-) presentations during room, relays, timers, and tape programmers
testing affects the occurrence and amount of controlled presentation of stimuli and rein-
shift. The method involves training a succes- forcements. A cumulative recorder and count-
sive discrimination, and then testing over only ers recorded all key pecks.
half of the continuum: the half centered
around the S+ to investigate a positive shift, Procedure
and the half around the S- to investigate a Training. All 24 birds received the same
negative shift. Effects of spacing of test stimuli training. In session 1, each bird was magazine
 19383711, 1966, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1901/jeab.1966.9-457 by CAPES, Wiley Online Library on [11/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ORDERING AND SPACING OF TEST STIMULI 459

trained, shaped to key peck, and given 25 rein- training session, both groups received all nine
forcements on a continuous reinforcement stimuli. For every test day, the order of presen-
(CRF) schedule. The response key was contin- tation was determined by reading across a
uously illuminated by 550 mu, and a house stimulus array containing 10 rows, with each
light was on during this session. In every sub- stimulus appearing only once in each row.
sequent session, the only light besides the The array was further arranged so that each
response key was a dim feeder light which stimulus followed itself and every other stim-
accompanied each 4-sec reinforcement. In ses- ulus at least once.
sion 2, every bird first produced 25 reinforce- Experiment II. The same rules of order of
ments on a CRF schedule in the presence of presentation held as in Exp I. Four more
550 m,u. From this point on, a variable inter- stimuli were added to each half, so that 2 mu
val 1-min schedule (VI 1) was in effect when- spacing instead of 6 mpt spacing occurred
ever the key was illuminated by 550 m/A. This around the S+ and S-. Group I received the
last part of session 2 consisted of 60 30-sec positive half on day 1, and Group II the nega-
presentations of 550 m,u, and session 3 con- tive. The halves were reversed on day 2.
sisted of 120 presentations, all alternated with Experiment III. The order of stimulus pre-
10 sec of black-out. For all S+ presentations, sentations was not mixed, but arranged in a
the 30-sec timer stopped during each reinforce- successively increasing or decreasing sequence.
ment. Each stimulus was presented 10 times in a row.
Successive discrimination training began in Starting at the positive end of the continuum,
session 4. In each discrimination session, 30- Group I rcteived 538 mMA for 10 successive pre-
sec stimulus presentations again alternated sentations, then 544 mMA for 10 presentations,
with 10 sec of black-out but the stimulus was then 550 m,u, and so on in 6 m,u steps to 586
either 550 mu (S+) or 574 mu (S-). Each oc- m,u. Group II began at the negative end, with
curred 60 times. The S+ and S- were pre- 586 m,u, and worked toward the positive. On
sented in a mixed order, with the restriction day 2, the conditions were reversed, so that
that within every block of 10 presentations, Group I began with 586 mu, and Group II
each occurred five times. Furthermore, neither with 538 m,u.
stimulus could occur more than three succes-
sive times. Discrimination training stopped RESULTS and DISCUSSION
after four discrimination sessions (sessions 4-7)
only if the following criterion had been met: Discrimination training
no responses to S- over any two successive By the end of session 7, 21 birds had met the
blocks (a total of 10 S- presentations). If the discrimination criterion. Three birds, #10,
criterion had not been met by the end of ses- #17, and #21, required 24, 10, and 59 extra
sion 7, training continued until it was met. blocks of stimuli before meeting criterion.
Considering all 24 birds over the last 60 S+
Testing presentations (30 sec each), total S+ respond-
For all three experiments, conditions of ex- ing ranged from 776 to 3683; and over the last
tinction held during testing. As in training, 60 S- presentations, total S- responding
30-sec stimulus presentations alternated with ranged from 0 to 283. High S+ responding did
10-sec black-outs. Each experiment involved not imply high S- responding, for the Spear-
eight birds, placed into two groups of four, man rank correlation coefficient equalled
each group having given the same total num- -0.01. The S+/(S+ and S-) response ratios,
ber of responses on the last day of training. calculated over the last 60 S+ and the last 60
Experiment I. On test day 1, Group I re- S- presentations, ranged from 0.85 to 1.00.
ceived five different stimuli, ranging from 538 When birds were placed into test groups, the
m,u to 562 m,u in 6 mu steps. This range is between-group ratios did not differ. The re-
called the "positive half," since it is centered training ratios for all birds ranged from 0.94
around the S+, 550 m,u. Group II received the to 1.00.
"negative half," 562 m,u to 586 mju. On test day
2, the halves were reversed, so that Group I Experiment I
now received the negative half, and Group II Positive shift. On day 1 of testing, Group I
the positive half. On test day 3, after one re- received the positive half of the continuum.
 19383711, 1966, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1901/jeab.1966.9-457 by CAPES, Wiley Online Library on [11/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
460 JOAN G. STEVENSON
The mean gradient shows a peak at 544 m/A, experiment. His S+ was also 550 mju, and one
followed by a decrease in responding through of his training groups had an S- of 570 mu,
550 m,, and out to 562 m, (Fig. 1). A positive only 4 mu from the present S- of 574 mu. The
shift is shown by all four birds in Group I. main differences occurred during testing.
The individual gradients are higher at both Whereas the present groups received five
538 mu and 544 mp than at 550 m/A. Two stimuli spaced 6 m,u apart over only a 24 mMu
gradients peak at 538 m,, and two at 544 mu. range which did not include the S-, Hanson's
Although Group II did not receive the posi- group received 13 stimuli spaced 10 mu apart
tive half of the continuum until day 2, its over a 140 mu range, which did include the
mean gradient is very similar to that of Group S-. The gradient of Hanson's experiment
I (Fig. 2). All four birds in Group II show a peaked at 540 mM, a value not given in the
positive peak shift; they responded more to present experiment, in which four peaks oc-
538 mu and 544 m, than to 550 mM. As with curred at 538 m,u and four at 544 mML. Averag-
the individual gradients of Group I, two peaks ing these values gives a peak at 541 mM, very
occur at 538 m, and two at 544 m,. close to Hanson's 540 mM peak. Hanson mea-
The above gradients obtained over only the sured the amount of displacement of his gra-
positive half of the continuum may be com- dients by determining for each subject the per-
pared with those of Hanson (1959), who ob- cent of the total number of responses emitted
tained gradients over the whole continuum. to stimuli to the left of the S+. These values
The present training procedure and criterion ranged from about 480% to about 820%. In the
of discrimination were adapted from Hanson's present experiment, values ranged from 63%

EXPERIMENT I, GROUP I
100%4
83%
Day I 79% Day 2
Bird I 0
0 20
0 / \ 3 a
0
c 4 A
0 Group mean-
a
h.

0
0
4-
4..
4..
C)
Uw
it

538 544 550 556 562562 568 574 580 586

Wavelength (mp)
Fig. 1. Percent of total responses during extinction as a function of the positive half of the test continuum on
day 1 and the negative half on day 2. On day 1, response totals for Birds 1-4 were: 1564, 503, 1564, and 1087 re-
spectively; and on day 2: 94, 1, 42, and 6.
 19383711, 1966, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1901/jeab.1966.9-457 by CAPES, Wiley Online Library on [11/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ORDERING AND SPACING OF TEST STIMULI 461

EXPERIMENT I, GROUP 3
73%
Day 2 Day I
Bird 5 0
S 6 0
0 7 a
S SA
c
0 Group mean-
a,
06

-
0
0
0

0
UL 1I

538 544 550 556 562 562 568 574 580 586
Wavelength (mp)
Fig. 2. Percent of total responses during extinction as a function of the negative half of the test continuum on
day 1 and the positive half on day 2. On day 1, response totals for Birds 5-8 were: 1265, 783, 1430, and 296 re-
spectively; and on day 2: 1185, 1268, 643, and 397.

to 87% for all eight birds, showing consider- test day, the overall responding to it was much
able overlap with Hanson's values. That is, higher than if the whole continuum had been
the combined operations of reducing the presented (e.g., Hanson, 1959). This allows
range, number, and spacing of test stimuli, investigation of a negative shift following the
and omitting the S- did not noticeably affect usual (S+,S-) training. The gradient of
the amount of shift. Group II on day 1 shows an orderly decrease
Thus, not only do simultaneous vs succes- in percent total responses, from 562 mu to
sive presentations of test stimuli have little 586 m,u (Fig. 2). Thus, although in training
effect on the post-discrimination gradient every bird had met the criterion of no re-
(Honig, 1962), but so also does presence vs sponding over 10 S- presentations, in testing
absence of the S- in testing. Just as "active when the S- was also presented 10 times, min-
comparison" of stimuli is not necessary to pro- imum responding was displaced away from the
duce a simple generalization gradient (Hiss S-, in the direction opposite the S+. All four
and Thomas, 1963), neither is "active compar- birds in Group II show this negative shift, as
ison" of S+ and S- necessary to produce well as an orderly decrease from 562 mp to
a post-discrimination gradient. It therefore 586 m,u. When Group I received the negative
seems that production of a peak shift depends half on day 2, three birds show a maximum
largely on training, rather than testing, opera- at 562 mp (Fig. 1). However, nothing can be
tions. said of a shift, since few or no responses were
Negative shift. When only the negative half given to each stimulus beyond 562 mM.
of the continuum was presented on the first Responding to 562 mp. Figures 1 and 2
 19383711, 1966, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1901/jeab.1966.9-457 by CAPES, Wiley Online Library on [11/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
462 JOAN G. STEVENSON

show that when 562 m/A occurred in the posi- ing, both groups were tested with all test stim-
tive half it received fewest responses; and uli. The gradients of the groups are very sim-
when it occurred in the negative half, it re- ilar (Fig. 3); both show a peak at 544 mMu, and
ceived most responses (with one exception in an orderly decrease to 586 m,u. A comparison
Group I, day 2). If a typical gradient over the with days 1 and 2 shows that the four birds
whole continuum were taken and split at 562 which had previously peaked at 538 mu now
m,u, the above would also hold. Thus, results peaked at either 544 mp or 550 m,. The four
regarding both the positive shift and respond- which had previously peaked at 544 mu still
ing to 562 mu indicate that relative respond- peak here, but the second highest number of
ing over a given half of the continuum is pre- responses occurs atV550 mu rather than at 538
dictable from relative responding over the mM. It would be important to determine how
whole continuum. much of this decrease in amount of shift is
What is not directly predictable from a gra- due to retesting, over the whole continuum
dient obtained over the whole continuum is after only half, and how much to retraining.
the absolute number of responses to 562 m/A The amount of shift increases with amount of
when it appeared in each half. On day 1, the training given before all test stimuli are pre-
total number of responses to 562 mp by each sented (Thomas, 1962); and the amount of
bird in the positive group (Group I) was 1, 32, shift decreases with repeated testing in a "mul-
39, and 46; on the same day, the responses to tivalued S- situation" where the S+ is still
562 m,u by each bird in the negative group correlated with reinforcement (Pierrel and
(Group II) were 216, 355, 503, and 659. That is, Sherman, 1962). It might be that the training
more absolute responses were given to 562 m,u before testing (Thomas, 1962) affects the shift
when it was presented with the negative half one way, and training during testing (Pierrel
than with the positive half. and Sherman, 1962) or after testing (the pres-
Positive shift after retraining. After retrain- ent experiment) affects it another way.

EXPERIMENT I, Day3
Group I Groupl
Bird I O Bird 5 o
20 6 0
0 3 a 7 a
0 4A 8 A
0
c Group mean Group mean -
0
0c
0

40
0
CL
0
0
0~

538 550 562 574 586 538 550 562

Wavelength (mp)
Fig. 3. Percent of total responses during extinction following one day of retraining. Response totals for Birds
1-4 were: 1427, 602, 377, and 702 respectively; and for Birds 5-8: 2184, 1115, 1251, and 403.
 19383711, 1966, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1901/jeab.1966.9-457 by CAPES, Wiley Online Library on [11/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ORDERING AND SPACING OF TEST STIMULI 463

placement, as measured by percent of total re-

Experiment II sponses to stimuli to the left of the S+, is
Positive shift. With only 2 m,u spacing, the smaller for all birds in Group I of Exp II (a
mean gradient for Group I on day 1 shows a range of 55%-61%) than for all birds in
positive shift, with a peak at 544 mju, and an Group I of Exp I (63%-77%). The mean gra-
orderly decrease to a minimum at 562 mu (Fig. dient for Group II when it received the posi-
4). Similarly, the individual curves for Group tive half shows an orderly increase to 546 mM,
I show fairly orderly gradients and a peak followed by a decrease out to 562 mu, with one
shift, in spite of difficult conditions for a dis- deviant point at 552 mpt (Fig. 5). Individual
crimination. That is, stimuli were spaced only gradients do not differ from those of Group I
2 mu apart, equated for brightness, presented with respect to peak location or amount of
successively, and accompanied by no differen- displacement. A comparison between Fig. 1
tial reinforcement. and 4 indicates that the gradient flattens when
The additional stimuli did affect the more stimuli are added around the S+. In-
amount of shift. In Exp I, two birds in Group deed, relative responding to the left of the S+
I peaked at 538 mM, and two at 544 m,u. If the is less in Exp II than in Exp I. That is, for
amount of shift is independent of the test Group I on day 1 of each experiment, re-
series, then when 546 mu and 548 mu are sponding to 538 m,u divided by responding to
added, some birds should still peak at 538 mMu. 550 mu was less for all four birds in Exp II
Instead, two birds peaked at 544 mju and two than for all four in Exp I. This also holds for
at 546 mM. Furthermore, the amount of dis- relative responding to 544 m,A. On the other

EXPERIMENT r, GROUP I

Day I Day 2
Bird 9 0
' 10 i
0 4. 11 a
C
c 12A
0
cL
Group mean-
0 3C

0
0
-
4- 2t
4-
c
0

538 544 550 556 562 562 568 574 580 586
Wavelength (mp)
Fig. 4. Percenit of total responses during extinction as a function of the positive half of the test continuum on
day 1 and the negative half on day 2. On day 1, response totals for Birds 9-12 were: 465, 1598, 718, and 985 re-
spectively; and on day 2: 19, 9, 6, and 31.
 19383711, 1966, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1901/jeab.1966.9-457 by CAPES, Wiley Online Library on [11/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
464 JOAN G. STEVENSON

EXPERIMENT I, GROUP 31

Day 2 Day I
Bird 13 0
14 x
15 A
0 16 A
n
U)
c Group mean
S
co
0
a'

4-
0
20-

0
CL
b.-
10-
a.

538 544 550 556 562 562 568 574 580 586
Wavelength (rmn)
Fig. 5. Percenit of total responises during extinction as a function of the negative half of the test continuum oni
day 1 and the positive half on day 2. On day 1, response totals for Birds 13-16 were: 692, 2355, 1025, and 1395 re-
spectively; and on day 2: 1149, 1442, 22, and 1945.
side of the S+, relative responding to 556 mp, ker, 1964; Hearst and Poppin, 1965). A series
and 562 mu did not differ between experi- of experiments would be required to deter-
ments. mine whether adding more stimuli or de-
Thus, with a 6 mju to 2 mj reduction in creasing spacing is the important variable in
spacing, the amount of positive shift decreased changing relative responding.
and the gradient flattened. Comparing the 6 Negative shift. Group II received the nega-
m,u gradient taken over a small range with tive half on day I. The resulting gradient was
Hanson's 10 m,u gradient taken over a large also less steep than in Exp I (Fig. 2 and 5). Re-
range showed no such change in amount of sponding to both 562 m,u and 568 m,u, relative
shift. Furthermore, when Friedman (1961) to S-, was less in Exp II than in Exp I. No
tested over 5, 10, and 20 m,u spacing, with the change in relative responding occurred to
smallest range occurring with 5 mju spacing, stimuli to the right of S-. The gradient for
he found no differences in slopes of resulting Group II on day 1 shows a decrease in re-
gradients. It is therefore possible that a change sponding from 562 mM, through the former
in relative responding occurs only with ex- S-, and out to 586 mu (Fig. 5). Gradients for
tremely small spacing, and/or with the range three of the birds are similar to the mean gra-
held constant. When spacing was decreased dient. The other bird, #16, shows a general
with the range held constant, it was necessary decline, but it is not orderly. With Group I on
to add more stimuli, and the amount of ex- day 2, two birds show a maximum at 562 mp,
tinction increased. However, it is unlikely that but the other two had stopped responding be-
simply increasing the amount of extinction fore 562 m, was presented (Fig. 4).
would cause a flattening (Thomas and Bar- Responding to 562 mu. Responding to 562
 19383711, 1966, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1901/jeab.1966.9-457 by CAPES, Wiley Online Library on [11/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ORDERING AND SPACING OF TEST STIMULI 465

m,u relative to a given half is again predictable from the positive to the negative end on day
from gradients taken over the whole con- I. As shown in Fig. 6, three birds peaked at
tinuum. When 562 mju appeared in the posi- 538 mu, and one at 544 mju. This apparent
tive half, seven birds gave the lowest number peak shift may be an extinction effect, since
of responses to this stimulus. When 562 m,u 550 m,u was presented later in the session than
appeared in the negative half, the two birds in 538 mju and 544 m/u. All birds showed an or-
Group I that responded over more than the derly decrease to the right of the peak. How-
first few stimulus presentations, and all birds ever, responding did not continue past the first
in Group II responded most to it (Fig. 4 and half of the session. Two birds did not respond
5). As in Exp I, the absolute number of re- after 556 m,u, and two did not after 562 m,u.
sponses to 562 mu was less when 562 mu Similarly, an abrupt cut-off in responding oc-
appeared in the positive half than in the nega- curred with Group II on day 2, with no re-
tive half. On day 1, the total number of re- sponses after 544 mu for one bird, and none
sponses to 562 mu when in the positive half after 556 mju for three birds (Fig. 7).
was only 3, 4, 4, and 48 for each bird; on This abrupt stop did not occur with Group
day 1 the total number to 562 m,u when in the II on day 1, when it went from the negative to
negative half was 187, 233, 295, and 579. the positive end of the continuum (Fig. 7). For
all four birds, responding continued through-
Experiment III out day 1, and even increased from 586 mu
In this experiment, the test stimuli were through the former S-. One gradient peaks at
presented in either a successively increasing or 574 m,u (the third stimulus presented), two
decreasing order, so that each group started at peak at 568 mu (the fourth), and one peaks at
one end of the continuum and then proceeded 556 mu (the fifth). Later in the session, ef-
in 6 mu steps to the other end. Group I went fects of extinction take over; the group curve

EXPERIMENT X, GROUP I
Day I Day 2
Bird 17 0
18 *
0 19 a
0 20 A
40
c Group mean
0
cL
0
0
b-
0
4.-
0
4)-

C.)

538 550 562 574 586 538 550 562 574

_
586
+ - +
Wavelength (mp)
Fig. 6. Percent of total responses during extinction as a function of positive to negative ordering on day 1 and
negative to positive ordering on day 2. On day 1, response totals for Birds 17-20 were: 983, 1278, 992, and 316 re-
spectively; and on day 2: 15, 518, 316, and 104.
 19383711, 1966, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1901/jeab.1966.9-457 by CAPES, Wiley Online Library on [11/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
466 JOAN G. STEVENSON

reaches a maximum at 568 mju, and then de- on day 2, while Group II received the reverse.
creases. On day 2, when Group I began at Experiment III differed in that stimulus pre-
586 m,u and went to 538 mju, the peak is even sentations were ordered along the continuum,
further to the left with a group peak at 550 and each group received the two halves within
mju (Fig. 6). Taken individually, two birds the first session.
peak here, one at 544 mMf, and one at 538 mni, Response totals in Exp I, where stimuli
the last stimulus presented. In addition to this were mixed, did not differ from those in Exp
change in peak location during a second test III, where stimuli were ordered. That is, a
session, the percent of total responses for comparison of Groups I over their first-pre-
stimuli over the last part of the session indi- sented positive half (538 mu - 562 m,u) shows
cates more relative responding later in the 503, 1087, 1564, and 1564 responses for each
session for Group I, day 2 than for Group II, bird in Exp I; and 316, 983, 992, and 1278 re-
day 1. That is, the combined percents of total sponses for each bird in Exp III. A comparison
responses to the last three stimuli presented of Groups II over their first-presented negative
(550 mu, 544 mu, and 538 mu) are greater for half (586 mu - 562 m,) shows 296, 783, 1265,
each bird in Group I than for each bird in and 1430 responses for each bird in Exp I; and
Group II. 393, 1146, 1189, and 1957 responses for each
bird in Exp III. Thus, over the restricted
Response totals, Exp I, II, and III range of 24 m,u (i.e., 538 m,u - 562 m,u or 586
Within each experiment, birds went either mp - 562 mu), ordered vs mixed stimulus pre-
from the positive to the negative half of the sentations did not differentially affect total
continuum, or from the negative to the posi- responding.
tive. In Exp I and II, Group I received the Within each experiment, one might ask if
positive half on day 1 and then the negative the group getting the positive half responded

EXPERIMENT m, GROUP 3
Day I Day 2
Bird 21 O
22 0
a@ 40- 23 a
0 24 A
Group mean
0 3

°
c
0
0
0
.0- 20.
4-
c
U
h.
0

a.-io

550 562 574 586 538 s$o1* 562

-
574
-
586
Wavelength (mp)
Fig. 7. Percent of total responses during extinction as a function of negative to positive ordering on day 1 and
positive to negative ordering on day 2. On day 1, response totals for Birds 21-24 were: 1842, 2903, 840, and 1633
respectively; and on day 2: 140, 164, 24, and 88.
 19383711, 1966, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1901/jeab.1966.9-457 by CAPES, Wiley Online Library on [11/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
ORDERING AND SPACING OF TEST STIMULI 467

more than the group getting the negative half, It is puzzling that the positive half of the
and furthermore if this difference held when continuum did not produce more responding
the halves were reversed for each group. In or- on the first test day than did the negative half.
der to decrease individual differences in re- On day 1, the range of absolute responses in
sponse totals, each bird's total over a given Exp I for the positive half was 503 for the low-
half was weighted by its total (or rate, since est bird to 1564 for the highest bird vs 296 to
time was constant) for the final training ses- 1430 for birds getting the negative half; in
sion. (Hearst, Koresko, and Poppin [1964], Exp II for the positive half (465-1598) vs (692-
have shown that there is a positive correlation 2355) for the negative half; and in Exp III for
between response rate during VI training and the ordered positive half (316-1278) vs (393-
total responses during generalization testing.) 1957) for the ordered negative half. Yet on
Figure 8 presents these weighted totals taken day 2, the range of absolute responses in Exp
over each half, and over the halves combined. I for the positive half was (397-1268) vs only
Points are connected simply to join the data (1-94) for the negative half; in Exp II for the
for any given bird. The first set of points in positive half (22-1945) vs (6-31) for the nega-
each graph represents the first half presented; tive half; and in Exp III for the ordered posi-
for Group I, it was the positive half, and for tive half following the ordered negative half
Group II the negative. Within any of the on day 1 (654-1391) vs (0-6) for the ordered
three experiments, the two groups do not dif- negative half following the ordered positive
fer. The second set of points represents the half on day 1.
second half presented. In each experiment, the The concept of behavioral contrast (Rey-
groups do differ, with each bird in the group nolds, 1961) comes closest to handling this
now receiving the positive half (Group II) lack of difference in response totals, followed
giving more weighted responses than the by a difference. Contrast experiments suggest
group (Group I) now receiving the negative that responding to stimuli associated with dif-
half. This difference is reflected in the totals ferent frequencies of reinforcement would be-
over both halves, with Group II, which went come more different if these stimuli were pre-
from negative to positive giving more than sented to the same subject, rather than to dif-
Group I, which went from positive to nega- ferent subjects. That is, if responses to the
tive. positive and negative half did not differ when
EXPERIMENT I EXPERIMENT I EXPERIMENT m
I.bU- WIW r-^-
o Groupl (-,t)-4-

L2
a c /
40- I0%0
0 0.80- V
-, --
,, pal Dy2 Dp9 ol Dy2 Dyl2 Ihl ehl fDy

JZOAO-A-
c

CL ~Day I Day 2 Days 812 Day I Day 2 Days 812 Iholf 2half of Day I

Fig. 8. Weighted response totals for each bird. Open circles represent birds tested on the positive half during
day 1 (Exp I and II) or during the first half of the session (Exp III), and on the negative half during day 2 or
during the second half of the session. Filled circles represent birds tested on the negative half on day 1 and the
positive half on day 2.
 19383711, 1966, 4, Downloaded from https://onlinelibrary.wiley.com/doi/10.1901/jeab.1966.9-457 by CAPES, Wiley Online Library on [11/11/2024]. See the Terms and Conditions (https://onlinelibrary.wiley.com/terms-and-conditions) on Wiley Online Library for rules of use; OA articles are governed by the applicable Creative Commons License
468 JOAN G. STEVENSON

each group received only one half, responses Hanson, H. M. Effects of discrimination training on
would be expected to differ when each group stimulus generalization. J. exp. Psychol., 1959, 58,
had received both halves. The present results 321-334.
Hearst, E., Koresko, M. B. and Poppin, R. Stimulus
do show that following no difference between generalization and the response-reinforcement con-
birds tested with the positive half and birds tingency. J. exp. Anal. Behav., 1964, 7, 369-380.
tested with the negative half, when the second Hearst, E. and Poppin, R. Steepened generalization
half was presented, birds going from the posi- gradients after massed extinction to the CS. Psy-
chon. Sci., 1965, 2, 83-84.
tive to negative half responded far less often Hiss, R. H. and Thomas, D. R. Stimulus generaliza-
than birds going from negative to positive. tion as a function of testing procedure and response
However, this is an extended application of measure. J. exp. Psychol., 1963, 65, 587-592.
behavioral contrast. In previous contrast ex- Honig, W. K., Thomas, D. R., and Guttman, N. Dif-
ferential effects of continuous extinction and dis-
periments, both positive and negative stimu- crimination training on the generalization gradient.
lus presentations alternated within a session; J. exp. Psychol., 1959, 58, 145-152.
here they did not. Usually, at least one of the Honig, W. K. Prediction of preference, transposition,
stimuli was correlated with a reinforcement and transposition-reversal from the generalization
gradient. J. exp. Psychol., 1962, 64, 239-248.
schedule; here, one of the stimuli had been Kalish, H. I. and Haber, A. Generalization: I. Gener-
correlated with a reinforcement schedule. alization gradients from single and multiple stimu-
Hopefully, further investigation will show lus points. II. Generalization of inhibition. J. exp.
how useful it is to extend contrast to the pres- Psychol., 1963, 65, 176-181.
Pierrel, Rosemary and Sherman, J. G. Generalization
ent context. and discrimination as a function of the SD-SA inten-
sity difference. J. exp. Anal. Behav., 1962, 5, 67-71.
Reynolds, G. S. Contrast, generalization, and the
process of discrimination. J. exp. Anal. Behav., 1961,
REFERENCES 4, 289-294. 1

Blough, 1). S. Spectral sensitivity in the pigeon. J. Stevenson, Joan G. Effects of ordering and spacing of
opt. soc. Amer., 1957, 47, 827-833. test stimuli on the post-discrimination gradient of
Friedman, H. Wavelength generalization as a func- generalization. (Doctoral dissertation, Brown Uni-
tion of spacing of test stimuli. J. exp. Psychol., versity). Ann Arbor, Mich.: University Microfilms,
1963, 65, 334-338. 1964.
Friedman, H. and Guttman, N. A further analysis of Terrace, H. S. Discrimination learning with and with-
the various effects of discrimination training on out "errors." J. exp. Anal. Behav., 1963, 6, 1-27.
stimulus generalization gradients. In D. Mostofsky Terrace, H. S. Wavelength generalization after dis-
(Ed.), Stimulus generalization. Palo Alto: Stanford crimination learning with and without errors.
Univ. Pr., 1965. Pp. 255-267. Science, 1964, 144, 78-80.
Guttman, N. Generalization gradients around stimuli Thomas, D. R. The effects of drive and discrimination
associated with different reinforcement schedules. training on stimulus generalization. J. exp. Psychol.,
J. exp. Psychol., 1959, 58, 335-340. 1962, 64, 24-28.
Guttman, N. Effects of discrimination formation on Thomas, D. R. and Barker, E. G. The effects of ex-
generalization measured from the positive-rate tinction and "central tendency" on stimulus gen-
baseline. In D. Mostofsky (Ed.), Stimulus generali- eralization in pigeons. Psychon. Sci., 1964, 1, 119-120.
zation. Palo Alto: Stanford Univ. Pr., 1965. Pp.
210-217. Received Decemiber 16, 1965