Urcuioli 2013

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 21

JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR 2013, 99, 129–149 NUMBER 2 (MARCH)

EVIDENCE FOR RESPONSE MEMBERSHIP IN STIMULUS CLASSES BY PIGEONS


PETER J. URCUIOLI1, B. MAX JONES2, AND KAREN M. LIONELLO-DENOLF3
1
PURDUE UNIVERSITY
2
CURTIN UNIVERSITY
3
UNIVERSITY OF MASSACHUSETTS EUNICE KENNEDY SHRIVER CENTER

Response membership in pigeons’ stimulus-class formation was evaluated using associative symmetry and
class expansion tests. In Experiment 1, pigeons learned hue–hue (AA) and form–form (BB) successive
matching plus a modified hue–form (AB) task in which reinforcement was contingent upon a left versus
right side-key response after the positive AB sequences. On subsequent BA (symmetry) probe trials, pigeons
responded more often to the comparisons on the reverse of the positive than negative AB sequences and,
more importantly, preferentially pecked the side key consistent with symmetry after the reversed positive
sequences. In Experiment 2, the original three baseline tasks were supplemented by dot–white (CC)
successive matching in which reinforcement was contingent upon a left versus right side-key response after
the positive CC sequences. Class expansion was then tested by presenting nonreinforced CA and CB
successive matching probes. Comparison response rates were mostly nondifferential on CA probes but were
uniformly higher on CB probes that consisted of the C samples and B comparisons from the same,
hypothesized class. Together, these results provide evidence that responses can become members of
stimulus classes, as predicted by Urcuioli’s (2008) theory of pigeons’ stimulus-class formation and Sidman’s
(2000) theory of equivalence.
Key words: stimulus classes, response membership, associative symmetry, successive matching, conditional
discrimination, simple discrimination, pigeons, key peck

In his theoretical analysis of equivalence Consider, for example, conditional discrimi-


classes, Sidman (1994, 2000) proposed that nation training consisting of two conditional (a.
responses and reinforcers can become class k.a. sample) stimuli (A1 and A2), two discrimi-
members if the reinforcement contingencies native (a.k.a. comparison) stimuli (B1 and B2),
during training do not create a conflict between two defined responses (Resp1 and Resp2), and
establishing the n-term analytical units of two reinforcing outcomes (O1 and O2). Assume
behavior (Sidman, 1986) and class inclusion. that O1 is presented only when the subject
In other words, as long as the baseline makes Resp1 to the B1 comparison stimulus in
contingencies support discrimination learning, the context of the A1 sample, and that O2 is
responses and reinforcers can enter into presented only when the subject makes Resp2 to
equivalence relations with the stimuli with which the B2 comparison in the context of the A2
they have been associated. Sidman (1994, 2000) sample. According to Sidman (2000), these
claims that equivalence relations consist of all contingencies could generate two 4-member
pairs of events—conditional stimuli, discrimina- equivalence classes— [A1, B1, Resp1, O1] and
tive stimuli, the experimenter-defined re- [A2, B2, Resp2, O2] —and, with them, a variety
sponses, and the experimenter-defined of new or emergent relations. For example,
reinforcers—specified by the reinforcement subjects could preferentially select the A1
contingencies. comparison over the A2 comparison if now
presented with B1 as the sample and, conversely,
select the A2 comparison over the A1 compari-
This research was supported by NICHD Grant R01 son if presented with B2 as the sample (associa-
HD061322. Karen Lionello-DeNolf and Max Jones acknowl- tive symmetry). Furthermore, presentation of
edge support for research activity and manuscript prepara- one of the reinforcing outcomes (O1 or O2) or
tion from the Commonwealth Medicine Division, University
of Massachusetts Medical School. The authors thank Greg emission of one of the responses (Resp1 or
Erhardt, Lisa Macklin, Blake Polak, and Melissa Swisher for Resp2) might now occasion selection of the
their assistance in conducting this research. sample or comparison stimulus associated with
Correspondence should be addressed to Peter J. Urcuioli, that reinforcer or response in training.
Department of Psychological Sciences, 703 Third Street,
West Lafayette, IN 47907-2081 (e-mail: urcuioli@purdue.
The predicted inclusion of reinforcers in
edu). equivalence classes was confirmed by Dube,
doi: 10.1002/jeab.17 McIlvane, Mackay, and Stoddard (1987,

129
130 PETER J. URCUIOLI et al.

Experiment 1). They trained 2 intellectually data (e.g., how can Resp1 and Resp2 be
disabled adults on symbolic and identity match- “presented” as sample stimuli in testing?) and
ing-to-sample (MTS) tasks in which different to the interpretive ambiguities of other findings.
foods (O1 and O2) served as reinforcers for An example of the latter is the emergent
correctly selecting different comparisons (B1 differential sample responding effect reported
and B2) following different sample stimuli (A1 by Manabe, Kawashima, and Staddon (1995,
and A2). After baseline training to high levels of Experiment 3). They initially trained budger-
accuracy, Dube et al. conducted nonreinforced igars on hue–form MTS in which presentation
probe trials in which the samples were replaced of the form comparisons was contingent upon
by their associated foods, and found that high- versus low-frequency vocal calls (Resp1 vs.
both participants always selected the compari- Resp2) to the hue samples. Later, identity MTS
son stimulus that, in theory, belonged to trials in which the form stimuli appeared as both
the same class as the O1 or O2 reinforcer samples and comparisons were added to the
(see Dube & McIlvane, 1995; Hall, Mitchell, hue–form trials in each session. On the form–
Graham, & Lavis, 2003; and Joseph, Overmier, & form trials, either a high or a low vocal call to
Thompson, 1997 for similar findings). each form sample produced the comparisons.
Demonstrating emergent relations between Despite these nondifferential sample–response
the samples and comparisons of different MTS contingencies, the budgerigars began to vocal-
tasks that involve the same set of outcomes ize differentially to the form samples as their
provides additional, indirect evidence for rein- form–form MTS accuracy increased. Specifical-
forcer class membership. For example, Maki, ly, they made a high-frequency call (Resp1) to
Overmier, Delos, and Gutmann (1995, Experi- the form sample that occasioned the same
ment 3) reported that young children trained form–comparison choice as the hue sample to
on AB and CD MTS, each involving blue versus which a high-frequency call was required, and
red poker chips (later exchangeable for food vs. they made a low-frequency call (Resp2) to the
small toys) as reinforcers for correct choice, form sample that occasioned the same form–
subsequently matched each D comparison to comparison choice as the hue sample to which a
the A sample associated with same colored low-frequency call was required.
poker chip (i.e., AD matching). Presumably, Manabe et al. (1995) interpreted their results
stimulus classes containing these matching as evidence that the explicitly trained [vocal call
stimuli developed via the common reinforcers ! form comparison] relations on the hue–form
they shared during training (see also Dube, trials were symmetrical. In other words, that
McIlvane, Maguire, Mackay, & Stoddard, 1989; training purportedly yielded two stimulus classes,
Schenk, 1994). Indeed, Maki et al. showed that each containing a specific hue, call, and form.
the emergent AD effect did not occur in the Consequently, each form sample eventually
absence of differential outcomes training. occasioned the high- or low-frequency call in its
Data obtained from nonhuman animals also respective class (see Sidman, 1994). Saunders and
indicate that stimuli associated with the same Williams (1998), however, argued that adventi-
reinforcing outcome belong to the same stimu- tious reinforcement of high- versus low-frequency
lus class (Urcuioli, 2005; 2013). For instance, calling on the added form–form MTS trials could
sample stimuli that share a common differential- equally well account for the findings. Moreover,
outcome association during baseline training Urcuioli and Vasconcelos (2008b) adapted
are interchangeable with one another in new Manabe et al.’s procedure for use with pigeons
contexts (e.g., Astley & Wasserman, 2001; Ed- (cf. Urcuioli, Pierce, Lionello-DeNolf, Friedrich,
wards, Jagielo, Zentall, & Hogan, 1982; Honig, Fetterman, & Green, 2002) and found that
Matheson, & Dodd, 1984; Meehan, 1999; see differential responding to the added samples
also Urcuioli & Zentall, 1992), and the same is emerged only if those samples occasioned the
true for comparison stimuli as well (Urcuioli & same comparison selections as the originally
DeMarse, 1997). trained samples. Nominal identity between the
In contrast to the reinforcer–membership added samples and the comparisons from initial
findings, compelling evidence for response training was insufficient to yield the emergent
membership in stimulus classes has been effect, contrary to a symmetry account.
relatively sparse. This is due partly to methodo- Urcuioli, Lionello-DeNolf, Michalek, and
logical difficulties in obtaining directly relevant Vasconcelos (2006) sought direct evidence for
RESPONSE MEMBERSHIP IN STIMULUS CLASSES 131

response membership in pigeons using a trans- comparisons. Shimizu found that on these RA
fer-of-control assessment following training and RC test trials, subjects also made class-
on many-to-one MTS, a conditional discrimina- consistent comparison selections.
tion known to reliably produce acquired sample Lionello-DeNolf and Braga-Kenyon (in press)
equivalence (e.g., Lowe, Horne, & Hughes, 2005; have also reported evidence for response
Urcuioli, Zentall, & DeMarse, 1995; Wasserman, membership in adult humans. Subjects were
DeVolder, & Coppage, 1992). During many-to- trained to make two different response patterns
one training, two patterns of center-key pecking (pressing a computer touch screen 10 times in
(viz., rapid vs. slow pecking generated by fixed- rapid succession or two times spaced several
ratio vs. differential-reinforcement-of-low-rates- seconds apart) to red and green stimuli. Next,
of-responding schedules, respectively) served as they were trained on form–hue MTS in which
“samples” that occasioned the same reinforced the required response to each stimulus was a
comparison selections as two other, visual sample single touch. Finally, using a procedure similar
stimuli. Pigeons then learned to match either the to Urcuioli et al. (2006), the different response
visual or the response-pattern samples to new patterns served as samples in MTS and the
comparison stimuli (reassignment training), comparisons were either the hue or form stimuli
and this was followed by a transfer test which (in separate test sessions). Two participants
assessed their ability to match the remaining made class-consistent comparison selections on
samples to the new comparisons. If acquired their first exposure to the tests, and a 3rd did so
equivalence between the response and visual after repeated tests and additional training.
samples had been established in many-to-one The positive results reported by Shimizu (2006)
training (Urcuioli, DeMarse, & Zentall, 1994), and by Lionello-DeNolf and Braga-Kenyon (in
then the new comparison choices learned to one press) are theoretically important especially in
set of samples during reassignment training view of the previous negative or ambiguous
should be immediately occasioned by the other findings from the nonhuman animal literature.
set in testing. Three separate experiments Encouraged by their results, the present study
(Urcuioli et al., 2006, Experiments 1, 2, and 4), sought to demonstrate response membership in
however, failed to confirm this acquired equiva- stimulus classes by pigeons, albeit with a different
lence prediction (although see Urcuioli & approach to the definition of responses. Given
Vasconcelos, 2008a for another interpretation the likely difficulties of establishing different
of the negative findings). temporal patterns of comparison pecking and
To date, Shimizu (2006) has reported the arranging valid emergent-relations tests (see, for
most convincing demonstration of response example, Lionello & Urcuioli, 1998), we defined
membership in stimulus classes. Normal adults the responses as moving toward and pecking keys
learned two 3-alternative MTS tasks (AB and in different locations; namely, pecking a left key
CD) in which reinforcement was contingent not versus pecking a right key. These spatial responses
only on correct comparison selections but also were incorporated into successive MTS (e.g.,
on the correct execution of one of three 4-part Wasserman, 1976, cf. Konorski, 1959), a condi-
computer-mouse movements (e.g., left, right, tional discrimination in which samples and
up, down). Thus, baseline training was actually comparisons are presented sequentially on a
ABR and CDR, where R refers to the different single manipulandum and comparison respond-
comparison–response movements. Subsequent ing is reinforced only when the comparison
testing revealed a large number of emergent matches (physically or symbolically) the preced-
relations (e.g., AC, BC, AD, and BD matching) ing sample. In our procedure, reinforcement
consistent with the idea that the common on the positive sequences in symbolic (AB)
responses across baseline tasks brought their successive matching was contingent upon differ-
respective samples and comparisons together ent side-key responses following the comparison
into the same classes. In addition, on some test stimuli on those trials. In other words, a simple
trials, different mouse-movement sequences discrimination between the comparison stimuli
were “presented” as samples. These trials began (e.g., peck left after B1 and peck right after B2)
with the appearance of a solid white square and was added to a conditional discrimination
subjects were asked to make one of the between those stimuli (i.e., peck the comparisons
previously learned mouse movements; the on A1B1 and A2B2 trials but not on A1B2 and
correct movement was then followed by the A2B1 trials).
132 PETER J. URCUIOLI et al.

Given that pigeons exhibit a variety of the-second-ordinal position (viz., R2). Third,
emergent relations like associative symmetry class members are the elements of the rein-
(Frank & Wasserman, 2005; Urcuioli, 2008, forced sample–comparison combinations. For
Experiment 3) after successive matching example, reinforced responding to a triangle
training (see also Strasser, Ehrlinger, & comparison following a red sample will yield a
Bingman, 2004; Sweeney & Urcuioli, 2010; [red sample, triangle comparison] class. Fourth,
Urcuioli, 2011; Urcuioli & Swisher, 2012), Exper- elements common to more than one stimulus
iment 1 asked whether combining this type of class produce class union (e.g., Mackay, Wilkin-
conditional discrimination training with simple son, Farrell, & Serna, 2011; Maki et al., 1995; see
(left- vs. right-response) discrimination training also Sidman, 1994). For example, if training
would provide evidence for response member- involves multiple successive matching tasks—for
ship in stimulus classes. Specifically, would BAR example, AB and BB matching—in which one
symmetry, where R refers to different side-key reinforced AB combination consists of a red
responses, emerge from ABR training? sample and a triangle comparison (R1 and T2)
and one reinforced BB combination consists of
Experiment 1 a triangle sample and a triangle comparison (T1
and T2), the common T2 element should merge
The rationale for, and the design of, Experi- the [R1, T2] and [T1, T2] classes into one [R1,
ment 1 were also based on Urcuioli’s (2008) T1, T2] class.
theory of stimulus-class formation in pigeons. For this experiment, pigeons were concur-
The important theoretical assumptions from rently trained on three tasks: hue–form (AB),
which predictions regarding response member- hue–hue (AA) and form–form (BB) successive
ship (see below) were derived are as follows. matching using red and green hues, and triangle
First, stimulus-class formation is facilitated by and horizontal-lines forms (see Table 1). AA and
continual juxtaposition of nonreinforcement BB were identity tasks: Reinforcement occurred
for certain sample–comparison combinations only after matching sample–comparison sequen-
with reinforcement for other sample–compari- ces—namely, red–red (R1–R2), green–green
son combinations. Successive MTS (cf. Nelson & (G1–G2), triangle–triangle (T1–T2), and hori-
Wasserman, 1978; Wasserman, 1976), in which zontal–horizontal (H1–H2). In the symbolic
both samples and comparisons are presented (AB) task, reinforcement also occurred only
individually, is ideal in this regard because half after certain sample–comparison sequences—
of all sample–comparison combinations end in for example, red–triangle (R1–T2) and green–
nonreinforcement independently of the level of horizontal (G1–H2) —and only if pigeons made
conditional discrimination accuracy. Second, an experimenter-designated correct side-key
the functional stimuli in successive matching are response (viz., pecking a lit left key or pecking
compounds consisting of the nominal matching a lit right key) after comparison offset. An
stimuli plus their ordinal positions within a trial. incorrect side-key response on these otherwise
Thus, a red sample stimulus is functionally red- “positive” trials ended these trials without
in-the-first-ordinal position (viz., R1), whereas a reinforcement. The remaining symbolic
red comparison stimulus is functionally red-in- sequences—for example, red–horizontal (R1–H2)

Table 1
Baseline Successive Matching Contingencies in Experiment 1

Hue–Form Symbolic Hue–Hue Identity Form–Form Identity

R ! T - FI 5 s – Leftþ R ! R - FI 5 sþ T ! T - FI 5 sþ
R ! H - EXT R ! G - EXT T ! H - EXT
G ! T - EXT G ! R - EXT H ! T - EXT
G ! H - FI 5 s – Rightþ G ! G - FI 5 sþ H ! H - FI 5 sþ
Note. R ¼ red, G ¼ green, T ¼ triangle, H ¼ horizontal, FI ¼ fixed interval schedule, EXT ¼ nonreinforced, þ ¼
reinforced. The first and second center-key stimuli in a trial sequence (sample and comparison, respectively) are shown to the
left and to the right of the arrows, respectively. Positive (i.e., potentially reinforced) hue–form matching trials ended with
food only if pigeons made the designated (left or right) side-key choice response after the comparison stimulus; otherwise,
the trial ended without food. Counterbalancing of the hue–form contingencies has been omitted.
RESPONSE MEMBERSHIP IN STIMULUS CLASSES 133

and green–triangle (G1–T2) —were defined as Urcuioli (2008) hypothesizes, then baseline
“negative”: They involved no lighting of the side training should yield the two 5-member stimulus
keys and no reinforcer delivery, ending automati- classes shown in Figure 2—namely, [R1, R2, T1,
cally with comparison offset after a fixed time. T2, Left3] and [G1, G2, H1, H2, Right3]. The
Figure 1 depicts the six stimulus classes elements of these classes predict associa-
hypothesized to result from baseline training tive symmetry (Frank & Wasserman, 2005;
(cf. Urcuioli, 2008). The bottom two rows show Urcuioli, 2008) in pigeons’ performances on
the classes arising from hue identity (AA) and novel BA test trials. Specifically, their compari-
form identity (BB) successive matching. Note son–response rates should be higher on T1–R2
that the identity classes contain the matching and H1–G2 test trials (viz., the reverse of the
sample and comparison of each positive (re- positive baseline trials) than on T1–G2 and H1–-
inforced) sequence (e.g., R1 and R2, and G1 R2 test trials (viz., the reverse of the negative
and G2). The top row shows the corresponding baseline trials). More importantly, when given a
classes arising from hue–form (AB) symbolic left versus right choice on the positive BA test
matching. These include not only the sample trials, pigeons should make more left responses
and comparison stimulus of each reinforced (Left3) than right responses on the novel T1–R2
sequence (for example, R1 and T2) but also the test sequence and more right responses (Right3)
correct side-key responses for those sequences than left responses on the novel H1–G2 test
(for example, the left response or Left3). The sequence. Such side-key preferences on the test
“3” denotes the third ordinal position because sequences would constitute evidence for re-
the reinforced side-key response on a positive sponse membership in stimulus classes by
symbolic matching trial follows the sample pigeons.
(ordinal position 1) and comparison (ordinal
position 2). Method
The bolded italics are used to highlight Subjects. Six White Carneau retired
elements common to more than one class—for breeders obtained from the Palmetto Pigeon
example, R1 (red sample) in both the symbolic Plant (Sumter, SC) began the experiment. One
and the hue identity classes. If common elements pigeon’s participation was discontinued for
cause their respective classes to merge, as health reasons and another’s because it did
not achieve the required baseline matching
accuracies after 85 training sessions; their data
are not reported. Of the remaining 4 pigeons, 1
(CHC1) was experimentally naïve and 3 (CHC3,
CHC4, and CHC6) had previously participated
in an unrelated experiment and in a successive
matching experiment with differential-
reinforcement-of-other-behavior contingencies
(Urcuioli, 2010). All were maintained at 80% of
their free-feeding body weights by restricting
feeding to the experimental sessions and

Fig. 1. The six stimulus classes hypothesized to result


from hue–form symbolic, hue identity, and form identity
successive matching training in Experiment 1. Bolded italics Fig. 2. The two 5-member stimulus classes hypothesized
denote elements that appear in more than one stimulus to result from the merging of individual classes containing
class. common elements (cf. Figure 1) in Experiment 1.
134 PETER J. URCUIOLI et al.

providing a reduced amount of Purina Pro- rated by a 15-s intertrial interval (ITI) with the
Grains in the home cages on the days they were house light remaining on throughout each
not run. Water and grit were freely available in session.
their stainless steel, wire-mesh home cages. All Next, pecking the center-key stimuli was
were housed in a colony room that was on a 14h- reinforced on fixed-interval (FI) schedules.
10h light-dark cycle (lights on at 07:00). The first seven FI sessions contained 30
Apparatus. Pigeons were run in BRS/LVE randomized presentations each of red and
(Laurel, MD) pigeon chambers (Model PIP-016 green, and the following seven sessions con-
three-key panel inside a Model SEC-002 enclo- tained 30 randomized presentations each of the
sure). The three response keys were 2.5 cm in triangle and horizontal lines. Within a seven-
diameter, 5.7 cm apart center to center, and session block, the FI value was increased from 2
aligned horizontally in a row 7.5 cm from the to 5 s over the initial four sessions. For the last
top of the panel. Mounted behind each key was a three sessions, completion of a FI 5-s require-
BRS/LVE Model IC-901-IDD stimulus projector ment ended with food on a random 50% of
that was equipped with films and filters for occasions. Successive stimulus presentations
displaying red (R), green (G), and white (W) (trials) within a session were again separated
homogeneous fields, and three white horizontal by a 15-s ITI, the first 14 s of which were spent in
lines (H), a solid white inverted triangle (T), darkness. The house light came on for the last
and three small white circular dots (D) oriented 1 s of the ITI and remained on until the end of a
45o counter-clockwise from vertical, all on black trial. The duration of food access per delivery
backgrounds (BRS/LVE Pattern No. 692). The was constant within a session but varied
rear-mounted food hopper could be accessed individually from 1.8–6.0 s across sessions to
via a 5.8-cm-square opening located 13 cm maintain a pigeon’s 80% body weight.
below the center key. When the hopper was Successive matching acquisition. Table 1 sum-
raised, it was illuminated by a small miniature marizes the reinforcement contingencies for
bulb (ESB-28) in the surrounding metal hous- the baseline successive matching tasks. All
ing. A GE #1829 bulb located 7.6 cm above the pigeons were concurrently trained on hue–
center key provided general chamber illumina- form symbolic, hue–hue identity, and form–
tion. The opening in the metal shield surround- form identity matching. The hue–hue and
ing the bulb directed its light toward the ceiling. form–form identity tasks involved typical succes-
Ventilation and masking noise was provided by a sive matching contingencies (Nelson &
constantly running blower fan attached to the Wasserman, 1978; Wasserman, 1976). Here, a
chambers. All experimental events were con- 5-s sample stimulus presented on the center key
trolled by customized programs written in GW- was followed, after a 1-s dark interval, by a 5-s
Basic Version 3.20 and running on an IBM- comparison stimulus presented on the same
compatible computer. key. The trial ended in reinforcement if the
comparison physically matched the preceding
Procedure sample (positive trial), but without reinforce-
Preliminary training. All pigeons were ini- ment if the comparison did not match the
tially trained to peck a key displaying the preceding sample (negative trial). More specifi-
stimuli that would later appear in the succes- cally, the first key peck to the sample on a trial
sive matching tasks. In the first and second initiated a fixed-interval (FI) 5-s schedule that
sessions, 30 presentations each of the triangle ended with the sample turned off, a 1-s dark
and horizontal lines, and red and green, interval and the onset of a comparison stimulus.
respectively, appeared in randomized order On positive trials, the first comparison key peck
on the center key, and a single key peck to the after a 5-s interval timed from the first peck
lit key was followed immediately by food. In turned off the comparison and produced food.
the third and fourth sessions, there were 10 On negative trials, the comparison stimulus and
center-key presentations of triangle and hori- the house light went off after 5 s independent of
zontal, and red and green, respectively, plus any key pecking.
20 presentations each of white on the left and The sequence of events on negative hue–form
right side keys, again in random order and symbolic matching trials (viz., R1–H2 and G1–
with single key pecks to the lit keys reinforced. T2) was the same as the sequence described
Successive stimulus presentations were sepa- above for negative identity matching trials.
RESPONSE MEMBERSHIP IN STIMULUS CLASSES 135

However, on positive hue–form trials (viz., R1– 16 positive hue–form trials (81.3% correct), for
T2 and G1–H2), comparison offset was accom- five of six consecutive sessions (“criterion”).
panied by the simultaneous illumination of This was followed by a minimum of 10 over-
white on the left and right side keys. A single training sessions, the last six of which had to
peck to either side key within 3 s turned off both meet the aforementioned performance levels.
side-key stimuli and produced either food if the Symmetry testing. Following overtraining, eight
side-key response was designated “correct” or a test sessions, run in four 2-session blocks
short blackout if the response was designated separated by at least five baseline sessions at
“incorrect”. For 3 pigeons (CHC1, CHC4, and criterion, were conducted to determine wheth-
CHC6), a left response was correct after the R1– er the hue–form baseline relations were
T2 sequence and a right response was correct symmetrical. Each test session contained 96
after the G1–H2 sequence. For the remaining baseline trials divided equally among the three
pigeon (CHC3), a right response was correct baseline tasks (hue–form, hue–hue and form–
after the R1–T2 sequence and a left response form) plus eight nonreinforced probe trials in
was correct after the G1–H2 sequence. For all which the sample was either the triangle or
pigeons, incorrect side-key responses turned off horizontal lines and the comparison was either
the house light for a period equal to the red or green. The four possible form–hue
reinforcement duration for that session, as did probes (T1–R2, T1–G2, H1–R2, and H1–G2)
the absence of any side-key peck within 3 s. Due were each presented twice in each test session.
to experimenter error, the left- and right- The first probe trial did not occur until at least
response contingencies were also in effect on one of each possible baseline trial was pre-
the positive identity trials for the first 10 sessions sented, and successive probes were separated by
for 2 pigeons (CHC4 and CHC6), but were at least six baseline trials. Probe trials that were
removed thereafter. the reverse of the negative baseline trials (viz.,
Each training session consisted of an equal T1–G2 and H1–R2) ended automatically (com-
number (32) of hue–form, hue–hue and form– parison stimulus and the house light off) 5 s
form matching trials presented in random order after comparison onset. For probe trials that
except that no sample–comparison sequence were the reverse of the positive baseline trials
could occur on more than two consecutive trials. (viz., T1–R2 and H1–G2), comparison offset
Successive trials were separated by a 15-s ITI, the after 5 s was accompanied by the simultaneous
first 14 s of which were spent in darkness. The illumination of white on the left and right side
house light came on for the last 1 s of the ITI keys. A single peck to either side key turned
and remained on until the end of the food- both off without food delivery, although the
hopper cycle on reinforced trials or comparison house light remained on for a period equal to
offset on nonreinforced trials. Reinforcement the session’s reinforcement duration. All other
duration (1.8–6.0 s) was again adjusted as procedural details were identical to those
needed on a session-by-session basis for each previously described. These probe trials per-
pigeon in order to maintain its 80% body weight mitted two assessments of associative symmetry:
as closely as possible. 1) a DR based on the number of comparison
Conditional discrimination learning was as- pecks on “positive” probe trials (viz., on the
sessed by calculating a discrimination ratio (DR) reverse of the positive baseline trials) divided by
in which the total number of comparison pecks the number of comparison pecks on both
on the positive trials of each baseline task was “positive” and “negative” probe trials (viz., on
divided by the total number of comparison the reverse of the positive and negative baseline
pecks on both positive and negative trials of that trials), and 2) the percentage of side-key
task. Only pecks occurring within the first 5 s of responses to the same (left or right) side key
comparison–stimulus onset entered into these as that designated correct on the positive hue–
computations. Simple discrimination learning form baseline trials.
on the hue–form trials was assessed by calculat- Because all 4 pigeons continued to respond at
ing the percentage of all side-key responses appreciable rates to the comparisons on the
designated as “correct”. Baseline training for nonreinforced probe trials after eight test
each pigeon continued until it achieved a DR of sessions, eight additional test sessions were run
0.80 or higher on each successive matching task, after baseline recovery to criterion performance
and at least 13 correct side-key responses on the levels. These latter test sessions were also run in
136 PETER J. URCUIOLI et al.

four blocks of two sessions with a minimum of 8-10 sessions, stabilizing at nearly 100% by the end
five baseline sessions at criterion separating of training.
each block. The average hue–identity DR over the last five
For all statistical analyses reported in this overtraining sessions preceding testing was 0.94
paper, Type I error rate was set at 0.05 using the (range: 0.91–0.97). The corresponding DRs for
tabled F values reported by Rodger (1975) for form identity and hue–form matching were 0.94
controlling error rates on a per decision basis. (range: 0.91–0.95) and 0.93 (range: 0.86–0.98),
respectively. The average percentage of correct
Results left versus right responses on the hue–form
Successive matching acquisition. Figure 3 (symbolic) task over these five sessions was
plots discrimination ratios (DRs) for each pigeon 98.5% (range: 96.3–100%).
on each baseline task (solid symbols) for the first Symmetry testing. The panels shown in the
40 training sessions (in blocks of two sessions) and top two rows of Figure 4 plot the number of
the percentages of correct choice on the symbolic comparison pecks/s averaged over the first
(hue–form) task (open triangles). Discrimination eight test sessions on the hue–form baseline
ratios (DRs) reached criterion levels most rapidly trials (open circles) and on the nonreinforced
on hue identity matching for 3 of the 4 pigeons form–hue symmetry probe trials (filled circles).
(CHC1, CHC3, and CHC6), but most slowly for The corresponding panels of Figure 5 plot the
the remaining pigeon (CHC4). Nevertheless, all corresponding data averaged over the second
pigeons were performing at criterion levels (DR set of eight test sessions. For the baseline trials,
at or above .80) on all tasks within 40 training “positive” and “negative” refer to the (potential-
sessions. Choice accuracy on the hue–form ly) reinforced (R1–T2 and G1–H2) and the
(symbolic) trials was around chance (50% nonreinforced (R1–H2 and G1–T2) sample–
correct) at the beginning of training, but quickly comparison sequences, respectively. For the
rose to its criterion level (81.2% or above) within symmetry probes, “positive” and “negative” refer

% Correct Symbolic Choices


1.00 100 1.00 100
Discrimination Ratio

0.80 80 0.80 80

0.60 60 0.60 60

Hue identity
0.40 40 0.40 40
Form identity
Hue-form (symb.)
0.20 20 0.20 20
CHC1 Symb. choice CHC4

0.00 0 0.00 0

1.00 100 1.00 100


% Correct Symbolic Choices
Discrimination Ratio

0.80 80 0.80 80

0.60 60 0.60 60

0.40 40 0.40 40

0.20 20 0.20 20
CHC3 CHC6

0.00 0 0.00 0
0 2 4 6 8 10 12 14 16 18 20 0 2 4 6 8 10 12 14 16 18 20
Blocks of 2 Sessions Blocks of 2 Sessions

Fig. 3. Acquisition of the three concurrently trained successive matching tasks over the first 20 two-session blocks of
training for each pigeon in Experiment 1. Discrimination ratios (filled symbols) are the proportions of all comparison
responses occurring on the positive sample–comparison trials of each task. The percentage of correct symbolic choices (open
triangles) is the percentage of correct side-key responses on positive hue–form matching trials.
RESPONSE MEMBERSHIP IN STIMULUS CLASSES 137

First 8 Test Sessions

4.0 4.0

3.0 3.0

Pecks/s
2.0 2.0

1.0 1.0
CHC1 CHC3

0.0 0.0
Positive Negative Positive Negative

5.0 3.0
Baseline
4.0 Probe
Pecks/s

2.0
3.0

2.0
1.0
1.0
CHC4 CHC6
0.0 0.0
Positive Negative Positive Negative
Trial Type Trial Type

100
% Symmetrical Side-key Rs

80
* *

60

40

20

0
1 3 4 6
Pigeon (CHC)

Fig. 4. Top panels: Comparison–response rates in pecks/s (1 SEM) on the symbolic matching baseline trials (open
circles) and the nonreinforced symmetry probe trials (filled circles) averaged over the first eight test sessions for each pigeon
in Experiment 1. Positive ¼ potentially reinforced symbolic baseline trials and test trials in which the samples and
comparisons of the potentially reinforced baseline trials were presented in reverse order. Negative ¼ nonreinforced
symbolic baseline trials and test trials on which the samples and comparisons of the nonreinforced baseline trials were
presented in reverse order. Bottom panel: The percentage of symmetry-consistent side-key responses for each pigeon on the
positive symmetry probe trials averaged over its first eight test sessions in Experiment 1. Error bars ¼ one SEM; stars denote
significant deviation from chance (50%).

to the reverse of the potentially reinforced and positive hue–form combination) from each test
the nonreinforced baseline sequences (viz., T1– session within an eight-session block. Similarly,
R2 and H1–G2, and H1–R2 and T1–G2), each negative baseline datum point is the
respectively. Each positive baseline datum point average number of pecks on four randomly
represents the average number of pecks on four selected negative trials (two of each negative
randomly selected positive trials (two of each hue–form combination) from each of these test
138 PETER J. URCUIOLI et al.

Second 8 Test Sessions

4.0 4.0

3.0 3.0
Pecks/s
2.0 2.0

1.0 1.0
CHC1 CHC3
0.0 0.0
Positive Negative Positive Negative

4.0 3.0
Baseline
Probe
3.0
Pecks/s

2.0
2.0
1.0
1.0
CHC4 CHC6
0.0 0.0
Positive Negative Positive Negative
Trial Type Trial Type

*
% Symmetrical Side-key Rs

100

* * *
80

60

40

20

0
1 3 4 6
Pigeon (CHC)

Fig. 5. Top panels: Comparison-response rates in pecks/s (1 SEM) on the symbolic matching baseline trials (open
circles) and the nonreinforced symmetry probe trials (filled circles) averaged over the second eight test sessions for each
pigeon in Experiment 1. Positive ¼ potentially reinforced symbolic baseline trials and test trials in which the samples and
comparisons of the potentially reinforced baseline trials were presented in reverse order. Negative ¼ nonreinforced
symbolic baseline trials and test trials on which the samples and comparisons of the nonreinforced baseline trials were
presented in reverse order. Bottom panel: The percentage of symmetry-consistent side-key responses for each pigeon on the
positive symmetry probe trials averaged over its second eight test sessions in Experiment 1. Error bars ¼ one SEM; stars
denote significant deviation from chance (50%).

sessions. Each probe-trial datum point is the at the end of the acquisition phase. In short, the
average of the four positive, or four negative, test hue–form baseline conditional discriminations
trials from each test session across an eight- remained intact during testing. More impor-
session block. tantly, probe-trial comparison-response rates
Throughout testing, pigeons pecked the were numerically higher on the symmetrical
comparisons at much higher rates on positive versions of the positive than on the negative
than on negative baseline trials, just as they did baseline trials for all pigeons. For the first eight
RESPONSE MEMBERSHIP IN STIMULUS CLASSES 139

test sessions, this difference in probe-trial comparison sequences in the hue–form task.
response rates was statistically significant in The remaining sequences never ended in
analysis of variance (ANOVA) only for pigeons reinforcement and these negative sequences
CHC1 and CHC3, Fs(1, 62) ¼ 6.14 and 18.96, occurred as frequently as the positive sequences,
respectively. For the second set of eight test a condition Urcuioli (2008) hypothesized may
sessions, the difference was statistically signifi- be crucial for pigeons’ stimulus-class formation.
cant for all 4 pigeons, Fs(1, 62) ¼ 3.95, 15.80, The present experiment shows that requiring an
4.44, and 3.90 for pigeons CHC1, CHC3, CHC4, additional, simple discrimination between com-
and CHC6, respectively. parisons on the potentially reinforced (positive)
The bar graphs shown in the bottom panels of AB trials did not preclude the emergence of BA
Figures 4 and 5 show the average percentages of symmetry.
symmetry-consistent (“symmetrical”) left versus Second, the test results show that associative
right side-key responses on the form–hue probe symmetry was evident in the pigeons’ side-key
trials for the first and second set of eight test preferences as well as in the rate at which they
sessions, respectively. For the first eight sessions, pecked the comparisons on the probe test trials.
these percentages were around 50% correct for Specifically, after learning to peck the left key on
the 2 pigeons (CHC1 and CHC3) whose probe- R1–T2 baseline trials, pigeons then pecked the
trial comparison response rates showed evi- left key more often than the right key on the
dence of symmetry, whereas they were signifi- symmetrical T1–R2 probe trials. Likewise, after
cantly above chance for the 2 pigeons (CHC4 learning to peck the right key on G1–H2
and CHC6) whose corresponding response baseline trials, pigeons pecked the right key
rates did not show evidence of symmetry, x2 more often than the left key on the symmetrical
(1) ¼ 0.80, 0.12, 5.83, and 7.26 for pigeons H1–G2 probe trials. For 2 of the 4 pigeons, this
CHC1, CHC3, CHC4, and CHC6, respectively. effect did not materialize during the first eight
For the second set of test sessions, however, the test sessions, but it was clearly apparent in the
percentages of symmetrical side-key responses second set of eight test sessions. It is important to
were significantly above that expected by chance note that this delayed emergence of symmetry-
for all pigeons, x2(1) ¼ 6.37, 3.86, 5.76, and consistent side-key preferences cannot be
8.60, for pigeons CHC1, CHC3, CHC4, and dismissed as further learning on these test
CHC6, respectively. trials because all test-trial responses were
nonreinforced.
Discussion It is curious that the pigeons that did not show
The results of this experiment are noteworthy symmetrical side-key preferences in the first set
for two reasons. First, they replicate Urcuioli’s of test sessions nonetheless showed clear
(2008, Experiment 3) and Frank and Wasser- evidence of symmetry in their probe-trial
man’s (2005, Experiment 1) findings of associa- comparison–response rates during these ses-
tive symmetry in pigeons. Specifically, after sions. Likewise, the other 2 pigeons exhibiting
concurrently training AB (hue–form), AA significant symmetry-consistent side-key prefer-
(hue–hue) and BB (form–form) successive ences in the first set of test sessions did not show
matching, pigeons responded more frequently a significant symmetry effect in their compari-
to the comparisons on nonreinforced BA son–response rates. We have no explanation for
(form–hue) probe trials when those trials this dissociation. In any event, the results from
reversed the order of the positive (reinforced) the second set of eight test sessions showed that
AB baseline sequences than when they reversed all pigeons performed in a manner consistent
the order of the negative (nonreinforced) AB with associative symmetry, both in the rates at
baseline sequences. This occurred despite the which they pecked the comparisons on the
fact that, unlike those previous experiments, positive versus negative BA probe trials and in
reinforcement on the positive AB baseline trials their left versus right side-key preferences on the
was also contingent upon a correct left versus positive BA probes.
right side-key response following the compar- The choice results from BA testing suggest
isons on those trials. This meant that not all strongly that the left versus right responses were
positive AB trials ended in food reinforcement, members of the same class as the samples and
especially early in training. Nevertheless, rein- comparisons preceding those responses. None-
forcement occurred only after certain sample– theless, we need to consider another possible
140 PETER J. URCUIOLI et al.

explanation of the pigeons’ side-key preferences response on these positive trials (see Table 2).
on the BA probes. Note that on the baseline AB Nonmatching (viz., negative) trials were always
trials (see Table 1), a left side-key response was nonreinforced. Trials of this CC identity match-
reinforced only after a red sample and only after ing task were later intermixed with those of the
a triangle comparison (viz., only on R1–T2 three previously learned tasks, and training
trials). Likewise, a right side-key response was continued until criterion levels of performance
reinforced only after a green sample and only were achieved on all four tasks.
after a horizontal comparison (viz., only on G1– Figure 6 shows the stimulus classes hypothe-
H2 trials). On the probe trials during testing, sized to arise from all four baseline tasks (cf.
pigeons had the opportunity to choose between Urcuioli, 2008). The stimulus and response
left and right responses after these same notations are the same as before and, once
nominal stimuli, albeit in reversed order (e.g., again, elements common to more than one class
on T1–R2 and H1–G2 probes). Perhaps, then, are bolded and italicized. The two new stimulus
pigeons preferentially responded left after the classes associated with DW identity matching are
triangle sample and the red comparison be- shown at the bottom. The two elements in the
cause the triangle and red stimuli, individually DW identity classes common to other classes
or in combination, had become discriminative (specifically, to the symbolic classes) are the left
for a left-key response during baseline training. and right side-key responses (Left3 and Right3).
Similarly, perhaps pigeons preferentially re- If these and other common, across-class ele-
sponded right after the horizontal-lines sample ments cause their respective classes to merge,
and the green comparison because both hori- the net result will be the two 7-member classes
zontal and green, individually or in combina- depicted in Figure 7. Given these enlarged
tion, had become discriminative for a right-key classes, new sample–comparison relations in-
response during training1. volving the D and W samples should emerge in
Obviously, this simple discrimination expla- testing. Specifically, pigeons should respond
nation ignores the potentially influential ordi- more to the comparisons on novel D1–R2 and
nal positions of these stimuli in testing vis-á-vis W1–G2 trials than to the comparisons on novel
training. Nevertheless, it is a plausible alterna- D1–G2 and W1–R2 trials because only the
tive to an account which claims that the former trials consist of elements from the
reinforced left versus right side-key responses same stimulus class. Likewise, they should
(Left3 and Right3) are members of the same respond more to the comparisons on novel
stimulus classes as the sample and compa- D1–T2 and W1–H2 trials than to the compar-
rison stimuli preceding them in AB baseline isons on novel D1–H2 and W1–T2 trials for the
training (cf. Figure 2). The purpose of same reason. Experiment 2 tested these
Experiment 2, then, was to conduct a follow- predictions.
up response membership test whose results
would not be amenable to this alternative Method
explanation. Subjects and Apparatus. The 4 pigeons that
completed Experiment 1 initially participated in
Experiment 2 this experiment. One pigeon (CHC6) was
dropped prior to testing because it failed to
In Experiment 2, pigeons learned a fourth maintain its baseline performances.
successive matching task using three small white The apparatus was the same as those used in
dots (D) and a homogeneous white (W) field as Experiment 1. However, two new center-key
stimuli. Reinforcement was arranged only on stimuli were introduced: three small white
trials in which the comparison physically circular dots (D) oriented 45o counter-clockwise
matched the preceding sample and only if from vertical on black background and a white
pigeons made a “correct” left or right side-key (W) homogeneous field (BRS/LVE Pattern No.
692).
1
Procedure. This experiment consisted of
These points are particularly apropos for pigeons CHC4 four phases of baseline training and recovery,
and CHC6 because they were erroneously run with side-key
choice (viz., simple discrimination) contingencies on the which began approximately 2-3 months after the
matching hue–hue and form–form trials during their first 10 completion of Experiment 1, and two test
acquisition sessions. phases. Each is briefly described below.
RESPONSE MEMBERSHIP IN STIMULUS CLASSES 141

Table 2
Expanded Baseline Successive Matching and Choice Contingencies in Experiment 2

Hue–Form Symbolic Hue–Hue Identity Form–Form Identity

R ! T - FI 5 s – Leftþ R ! R - FI 5 sþ T ! T - FI 5 sþ
R ! H - EXT R ! G - EXT T ! H - EXT
G ! T - EXT G ! R - EXT H ! T - EXT
G ! H - FI 5 s – Rightþ G ! G - FI 5 sþ H ! H - FI sþ
Dot–White Identity
D ! D - FI 5 s – Leftþ
D ! W - EXT
W ! D - EXT
W ! W - FI 5 s – Rightþ
Note. R ¼ red, G ¼ green, T ¼ triangle, H ¼ horizontal, D ¼ dots, W ¼ white, FI ¼ fixed interval schedule, EXT ¼
nonreinforced, þ ¼ reinforced. The first and second center-key stimuli in a trial sequence (sample and comparison,
respectively) are shown to the left and to the right of the arrows, respectively. Positive (i.e., potentially reinforced) hue–form
and dots–white matching trials ended with food only if pigeons made the designated (left or right) side-key choice response
after the comparison stimulus; otherwise, the trial ended without food. Counterbalancing of the hue–form contingencies and
of the dots–white identity choice contingencies have been omitted.

Baseline recovery. Each pigeon (CHC1, for each pigeon continued until its DR was .80 or
CHC3, and CHC4) was re-trained on the three higher on all three tasks, and it made a
baseline successive matching tasks used in minimum of 13 (of 16) correct side-key
Experiment 1 (see Table 1). Baseline re-training responses (81.3% correct) on the hue–form
task, for five of six consecutive sessions. This was
followed by at least 10 overtraining sessions with
performances at or above these criteria for five
of the last six sessions.
Successive matching acquisition with D and W
stimuli. After baseline recovery, each pigeon
was trained on successive matching using D
and W as center-key (sample and comparison)
stimuli. Identity matching contingencies were in
effect for all pigeons (see Table 2): Food
reinforcement was scheduled on trials in which
the sample and comparison were nominally
identical (viz., D1–D2 and W1–W2); trials on
which they differed (viz., D1–W2 and W1–D2)
ended without food. In addition, reinforcement

Fig. 6. The eight stimulus classes hypothesized to result


from hue–form symbolic, hue identity, form identity, and
dot–white (DW) identity successive matching training in Fig. 7. The two 7-member stimulus classes hypothesized
Experiment 2. Bolded italics denote elements that appear in to result from the merging of individual classes containing
more than one stimulus class. common elements (cf. Figure 6) in Experiment 2.
142 PETER J. URCUIOLI et al.

on matching (positive) trials was contingent Class expansion testing. These 104-trial test
upon a “correct” left versus right side-key sessions consisted of 96 baseline trials, divided
response. In other words, immediately after equally among the four baseline tasks (cf.
comparison offset on the D1–D2 and W1–W2 Table 2) plus eight nonreinforced probe trials.
sequences, the left and right side keys were each As in Experiment 1, test sessions were run in
lit by a white stimulus, and a single peck to either blocks of two separated by five baseline sessions
side key within 3 s immediately turned both off at criterion levels of performance. There were
and yielded food (if the choice response was two types of test sessions, one in which the
designated “correct”) or turned off the house nonreinforced probe trials consisted of D or W
light (if the choice response was designated samples followed by R or G comparisons (viz.,
“incorrect”). If neither side key was pecked D1–R2, D1–G2, W1–R2, and W1–G2) and the
within 3 s (a very infrequent event), the side keys other in which the probe trials consisted of D
and house light went off automatically. For or W samples followed by T or H comparisons
pigeons CHC1 and CHC4, a left response was (viz., D1–T2, D1–H2, W1–T2, and W1–H2).
reinforced after the D1–D2 sample–comparison Each test block was always of one type (CA or
sequence and a right response was reinforced CB, respectively), and two blocks of one type
after the W1–W2 sequence. For pigeon CHC3, were alternated with two blocks of the other
the opposite contingencies were in effect. type until a total of eight test sessions of each
Each DW successive matching session con- type were completed. For pigeon CHC1,
sisted of 60 trials divided equally among the four the order of testing was T/H comparisons
possible trial types. All other procedural details (CB) – R/G comparisons (CA) – T/H compar-
were identical to those described for successive isons (CB) – R/G comparisons (CA); for
matching acquisition in Experiment 1. Training pigeons CHC3 and CHC4, testing occurred in
for each pigeon continued until it achieved a DR the reverse order. There was no left–right
(total number of comparison pecks on positive choice component on any test trial in this
trials  total number of comparison pecks on experiment. Otherwise, all other details were
both positive and negative trials) of at least .80 identical to the corresponding test sessions in
and made at least 25 out of 30 correct side-key Experiment 1.
responses (83.3%) for five of six consecutive
sessions. At this point, sessions of DW identity Results
matching were alternated across days with Baseline performances. Pigeons learned the
training on the three concurrent successive DW identity task quickly, reaching criterion levels
matching tasks from Experiment 1, and this of performance within 6-10 sessions, and these
continued until criterion-level performances accuracies were maintained when this task was
were met for both session types for five sessions later combined with the other three baseline
each. tasks. The average DRs for the five baseline
Combined training with all baseline tasks. Prior sessions prior to the first R/G comparison test
to testing, pigeons received additional training session were .94, .96, .96, and .92 for hue–form
with trials from all four baseline tasks inter- symbolic, hue identity, form identity, and DW
mixed. These 96-trial sessions consisted of 24 identity matching, respectively, and the percen-
trials of each baseline task, hue–form symbolic tages of correct side-key responses for hue–form
(AB), hue identity (AA), form identity (BB), and symbolic and DW identity matching were 99.4%
dot–white identity (CC), divided equally among and 99.4%, respectively. The average DRs for the
the four sample–comparison sequences com- five baseline sessions prior to the first T/H
prising each task. Combined training continued comparison test session were .97, .95, .95, and
for each pigeon until its DR was at least .80 for all .94, for hue–form symbolic, hue identity, form
four tasks and it made at least 10 out of 12 identity, and DW identity matching, respectively,
correct side-key responses (83.3%) on both and the corresponding percentages of correct
hue–form symbolic and dot–white identity side-key responses for hue–form symbolic and
matching for five of six consecutive sessions. DW identity matching were 99.4% and 98.9%,
At least 10 overtraining sessions then followed. respectively.
Testing commenced when criterion levels of Performances on the baseline trials during
performances were met for the last five of six testing with each set of comparisons were well
overtraining sessions. maintained, with only scattered instances in
RESPONSE MEMBERSHIP IN STIMULUS CLASSES 143

which a DR fell below .80. The average DRs for task (which had been extensively trained in
hue–form symbolic, hue identity, form identity, Experiment 1).
and DW identity were .94, .94, .94, and .92, Testperformances. The top panel of Figure 8
respectively, across the eight test sessions with shows the results from each pigeon averaged
the R and G comparisons, and .95, .94, .97, and over the eight test sessions in which probe trials
.95, respectively, across the eight test sessions consisted of D and W samples followed by R and
with the T and H comparisons. The correspond- G comparisons (viz., CA tests). The bottom
ing percentages of correct side-key responses on panel of the figure shows the corresponding
the baseline hue–form and DW identity tasks data from the test sessions in which probe trials
were 98.6% and 84.8%, respectively, for the consisted of D and W samples followed by T and
former tests, and 99.3% and 84.5%, respectively, H comparisons (viz., CB tests). “Positive” and
for the latter tests. These accuracies never fell “negative” baseline trials in both panels refer to
below 80% correct on the baseline hue–form the potentially reinforced versus nonreinforced
symbolic task during either type of test, but did symbolic training sequences (R1–T2 and G1–H2
occasionally on the baseline DW identity task versus R1–H2 and G1–T2), respectively. (Each
(although never below 70% correct). The lower baseline datum point was computed in the same
overall accuracy on DW identity than on hue– fashion as that described in Experiment 1.) For
form symbolic matching is probably attributable the probe trials, “positive” and “negative” refer
to less training on the former than on the latter to test-trial sequences in which the sample (D1

R-G Comparisons Test

4.0 4.0 4.0


Baseline
Probe
3.0 3.0 3.0
Pecks/s

2.0 2.0 2.0

1.0 1.0 1.0


CHC1 CHC3 CHC4
0.0 0.0 0.0
Positive Negative Positive Negative Positive Negative
Trial Type Trial Type Trial Type

T-H Comparisons Test

4.0 3.0 4.0


Baseline
Probe
3.0 3.0
Pecks/s

2.0
2.0 2.0
1.0
1.0 1.0
CHC1 CHC3 CHC4
0.0 0.0 0.0
Positive Negative Positive Negative Positive Negative
Trial Type Trial Type Trial Type
Fig. 8. Top panels: Comparison-response rates in pecks/s (1 SEM) on the symbolic matching baseline trials (open
circles) and the class expansion probe trials (filled circles) using red and green comparisons averaged over eight test sessions
for each pigeon in Experiment 2. Bottom panels: Comparison-response rates in pecks/s (1 SEM) on the symbolic matching
baseline trials (open circles) and the class expansion probe trials (filled circles) using triangle and horizontal-lines
comparisons averaged over eight test sessions for each pigeon in Experiment 2.
144 PETER J. URCUIOLI et al.

or W1) and comparison (R2 or G2, or T2 or H2) hue–form (AB) symbolic and dot–white identity
were drawn from the same hypothesized stimu- (CC) baseline tasks were members of the
lus class or different stimulus classes, respective- same hypothetical classes containing the sam-
ly (see Figure 7). Thus, the D1–R2 and W1–G2 ples and comparisons of those tasks, and 2)
sample-comparison sequences are “positive”, classes containing common response elements
and the D1–G2 and W1–R2 sequences are (see Figure 6) merged in the same manner as
“negative”, for the R/G comparison tests. classes containing a common sample or com-
Likewise, the D1–T2 and W1–H2 sample– parison element (see Figure 7). Put simply, we
comparison sequences are “positive”, and the assessed whether association with the same
D1–H2 and W1–T2 sequences are “negative”, response would render two stimuli equivalent.
for the T/H comparison tests. The probe-trial data from Experiment 2
Figure 8 shows that all pigeons maintained generally confirmed this prediction. Although
their baseline hue–form symbolic performances the differences between comparison-response
in both tests as evidenced by much higher rates on positive versus negative trials involving
comparison response rates on positive than on R and G comparisons were not statistically
negative baseline trials. Probe-trial response rates significant when those rates were averaged
averaged across the eight R/G comparison tests over all eight tests, 1 of the 3 pigeons showed
(top panel), on the other hand, were mostly significantly higher rates on the positive probe
nondifferential: The slightly higher rates on the trials during its initial two R/G test sessions.
positive than on the negative probe trials on these Moreover, all 3 pigeons showed significantly
tests were not significantly different for pigeons higher comparison-response rates on positive
CHC1, CHC3, and CHC4, Fs(1, 62) ¼ 1.05, 0.48, than on negative probe trials averaged over all
and 0.30, respectively. Moreover, overall probe- test sessions in which the D and W samples were
trial response rates were considerably lower than followed by the T and H comparisons. These
on the baseline trials, and were considerably results are consistent with the notion that
lower than the overall probe-trial rates with these responses can become members of stimulus
same (R and G) comparisons on the first eight test classes (Sidman, 2000) given that “positive”
sessions in Experiment 1 (see Figure 4). These could only be defined by assuming that common
differences might reflect the cumulative effects of left and right side-key responses merged smaller
extinction on the probe trials (they were always classes into larger ones that contained the D
nonreinforced) which, in turn, could have and W samples and particular hue and form
diminished the chances of obtaining a significant comparisons (see Figure 7).
difference between positive and negative probe- Because comparison response rates on the
trial response rates with these comparisons. probe trials served as the sole dependent
Interestingly, CHC3’s positive versus negative variable in Experiment 2, this experiment
probe-trial rates over just the first two R–G test effectively avoided the interpretation issue that
sessions (1.38 versus 0.12 pecks/s, respectively) we faced in Experiment 1. Recall that in
were significantly different, F(1, 14) ¼ 5.68. Experiment 1, pigeons showed symmetry-con-
By contrast, the results from eight test sessions sistent side-key preferences on BA (form–hue)
involving the T/H comparisons (bottom panel probe trials that were the reverse of the explicitly
of Figure 8) showed significantly higher com- trained AB (hue–form) trials. One interpreta-
parison-response rates on positive than on tion of this result is that the left and right
negative probe trials for all 3 pigeons, Fs(1, responses were also members of stimulus classes
62) ¼ 166.24, 17.72, and 77.86 for pigeons involving the A and B elements of the reinforced
CHC1, CHC3, and CHC4, respectively. Indeed, combinations of the symbolic (AB) and identity
these response-rate differentials were compara- (AA and BB) tasks (cf. Figure 2). However,
ble to those observed on the hue–form baseline because the nominal stimuli on the BA probe
trials. trials were the same as those on the AB baseline
trials, it is possible that symmetry-consistent
Discussion left versus right responding occurred because
Experiment 2 tested the prediction that new each side-key response had been explicitly
stimulus relations involving D and W samples reinforced following a specific A stimulus and
and R, G, T and H comparisons would emerge a specific B stimulus on the positive AB trials
if 1) the side-key responses reinforced on the (e.g., left after red and after triangle—cf.
RESPONSE MEMBERSHIP IN STIMULUS CLASSES 145

Table 1). Consequently, if these nominal stimuli reinforcement contingency becomes untena-
had acquired discriminative control over left ble.” (Sidman, 2000, p. 133). Second, they
versus right responding, our BA probe trials address the processes that Urcuioli (2008)
would not have constituted a test for derived hypothesized are crucial for pigeons’ stimulus-
(emergent) relations. class formation.
In Experiment 2, no such interpretive prob- The two experiments reported here tested
lem arises because the CA and CB test trials did predictions regarding response membership by
not include a left- versus right-key choice phase. training pigeons in a modified successive
Instead, response membership was indirectly matching paradigm (Wasserman, 1976; cf.
assessed by asking whether two stimuli associat- Konorski, 1959): Rather than simply reinforcing
ed with the same side-key response generate responses to the comparison stimuli on the
successive matching behavior consistent with positive symbolic (AB) baseline sequences, left
stimulus class formation. Specifically, higher and right side-key pecks were differentially
comparison-response rates on positive CA and reinforced following these sequences. By sup-
CB probes than on negative CA and CB probes plementing the conditional discrimination con-
implies that the left and right responses tingencies of successive matching with a simple
required during symbolic and DW identity discrimination involving two side-key responses
baseline training produced class merger and, following the stimuli comprising those contin-
with it, enlarged classes whose elements com- gencies, we were able to obtain independent
prised the positive test sequences. The pattern measures of conditional and simple discrimina-
of results we obtained was largely consistent with tion accuracies with which to assess emergent
the pattern predicted by response membership relations between the stimuli on positive
in those classes. sequences and the side-key responses they
occasioned (cf. Figures 1 and 6). This permitted
General Discussion a straightforward test for emergent relations
that involved those responses (cf. Figures 2
The present study was motivated in part to and 7).
test Sidman’s (1994, 2000) proposal that the In our estimation, the data we report here
reinforcement contingency itself generates provide good evidence that pigeons’ stimulus
equivalence classes and, more specifically, that classes can include their differential responses.
under certain baseline conditions, responses This evidence took two forms. First, Experiment
will become members of those classes. Shimizu 1 showed that symmetrical BAR relations
(2006) and Lionello-DeNolf and Braga-Kenyon emerged from ABR baseline training (viz.,
(in press) studied n-alternative MTS perform- symbolic – AB – matching with an added
ances in humans and reported data consistent choice-response – R – component) combined
with response membership in stimulus classes. with AA and BB identity training. In other
Here, we sought similar evidence in pigeons words, after learning which hue–form (AB)
performing successive matching tasks because sample–comparison sequences were potentially
they have previously demonstrated emergent reinforced and which side-key response (R) to
stimulus–stimulus relations indicative of class make after each positive AB combination,
formation in these tasks (e.g., Sweeney & pigeons made that same left or right response
Urcuioli, 2010; Urcuioli, 2008; Urcuioli & more often than the alternative response on test
Swisher, 2012). Another motivation for the trials that reversed the positive baseline sequen-
present study, then, was to assess whether ces. Thus, if one component of ABR training
training with successive matching contingencies involved positive trials consisting of a red
that include a differential response component sample, a triangle comparison, and a left
would generate classes containing those differ- response, then in testing, pigeons pecked left
ent responses (cf. Urcuioli, 2008). Thus, our more often than right on trials beginning with a
experiments are theoretically important for two triangle sample and a red comparison. This
reasons. First, they bear directly on Sidman’s associative symmetry effect was complemented
(1994, 2000) position regarding the origins of by the additional finding that pigeons pecked
equivalence. Indeed, if his prediction of re- more to the comparison stimuli themselves on
sponse membership was disconfirmed “…the positive test trials (viz., the reverse of the
theory that equivalence arises directly from the reinforced AB baseline trials) than on negative
146 PETER J. URCUIOLI et al.

test trials (viz., the reverse of the nonreinforced pecked the left or right key on the red–red
AB baseline trials). In short, conditional dis- and green–green test trials in accordance with
crimination performance on the BA probe trials which side-key response was associated with
coincided with simple discrimination (left key each color comparison in training (see also
peck vs. right key peck) performance on those Vasconcelos & Urcuioli, 2011). For example, if a
same trials. red comparison choice had been reinforced
Second, Experiment 2 showed that after following a left-response sample in training,
CCR training (viz., dot–white identity match- then pigeons pecked the left key more often
ing with a side-key response requirement) was than the right key on test trials where red
added to the ABR, AA, and BB baseline tasks, appeared on both keys.
pigeons responded in a class-consistent fashion This symmetry effect suggests that baseline
on CB and, to a lesser extent, on CA test trials. training had produced two classes, each of
It is important to note that the samples and which contained a response member—namely,
comparisons of the CCR task did not overlap [left response, red] and [right response, green].
with the samples and comparisons of the ABR, García and Benjumea (2006, Experiment 3) also
AA, and BB tasks; only the left and right showed that pigeons did not exhibit side-key-
responses in the CCR and ABR matching tasks response preferences in testing if they did not
did. Without reference to those common peck the left- and right-key samples during
responses, it would not be possible to classify baseline training. In other words, simply pre-
the dot and white stimuli in the CCR task as senting (without the requirement to peck) a lit
belonging to, or not belonging to, the same left key or a lit right key prior to the color
stimulus class as the red, green, triangle and comparisons during training was not sufficient
horizontal stimuli appearing on the other to produce a preference for pecking left versus
baseline tasks. In other words, “class consis- pecking right on subsequent red–red and
tent” has no meaning on the CA and CB probe green–green test trials. These findings suggest
trials unless an expanded class can be derived that the differential proprioceptive stimulation
via overlapping elements in the individual generated by emitting left- versus right-key
stimulus classes proposed to arise from the responses, rather than the differential extero-
four baseline tasks. As Figure 6 shows, two of ceptive stimulation provided by different lit-key
the required overlapping elements were the locations, was responsible for the reported
left and right responses (Left3 and Right3) symmetrical relations.
that were components of the positive symbolic Although our results support Sidman’s (2000)
(AB) and positive dot–white identity (CC) response membership proposal, they suggest an
training trials. alternative to his definition of responses in
Emergent relations involving left and right conditional discriminations. Sidman (1986,
side-key responses have also been reported in 1994, 2000) assumed that the subject’s re-
pigeons by García and Benjumea (2006, Experi- sponses in these procedures were the actions
ment 1). They devised a novel conditional required to operate whichever manipulandum
discrimination training procedure in which displayed a specific stimulus. Thus, he assumed
five consecutive pecks to either the left or right that conventional, n-alternative MTS tasks
key (the “sample”) of a white two-key display involve a single response such as a button press
replaced that display with red and green (see Table 3 in Sidman, 1986, and Figure 2 in
comparison stimuli randomized across loca- Sidman, 2000). In contrast, we defined re-
tions. Pecking the red comparison was rein- sponses in terms of which specific manipula-
forced after one side-key-response sample and ndum was operated (e.g., left-key pecks and
pecking the green comparison was reinforced right-key pecks). Although the two side-key
after the other side-key-response sample. After responses featured in only simple discrimina-
these [side-key-response – color comparison] tions (i.e., two 3-term contingencies) in our
baseline relations were learned to high levels of experiments, evidence for these responses
accuracy, the possibility that those relations were becoming members of stimulus classes suggests
symmetrical was tested by occasionally present- that the responses in conventional MTS might
ing nonreinforced probe trials where both keys also be usefully defined in terms of their spatial
were lit either red or green. García and location (see Jones, 2003). MTS can still be
Benjumea found that pigeons preferentially conceptualized as four-term contingencies
RESPONSE MEMBERSHIP IN STIMULUS CLASSES 147

providing that the various comparison–stimulus the emergence of red sample – dot compari-
arrays are viewed as the discriminative stimuli son (R1–D2) and dot sample – triangle
(see Figure 9 in Jones, 2003). Future research comparison (D1–T2) symbolic relations. In
should explore the implications of this ap- other words, these novel sequences should
proach to response definition for research yield relatively high comparison response rates
into equivalence relations. in testing.
If the assumptions of Urcuioli’s (2008) theory Ideally, evidence for response membership in
that prompted the present experiments are stimulus classes obtained from tests like those
sound, they make other predictions that can and just described would be complemented by
should be tested in future research. First, the results from tests of the sort reported by Shimizu
ostensible class-expansion effect reported in (2006) in which the differential responses
Experiment 2 should occur only if baseline DW themselves serve as probe-trial samples and
identity training requires differential side-key are found to control class-consistent comparison
responding on the two positive sample–compar- responding. For instance, after training pigeons
ison sequences. In other words, pigeons trained on tasks similar to those used in Experiment 1
in the same manner as those in Experiment 2 (viz., ABR, AA, and BB), perhaps they would
but without the left–right choice phase in the respond more often to certain comparisons
DW identity task should respond nondifferen- than to others on test trials that begin with either
tially on subsequent CA and CB probe trials. a left- or a right-key response (the “sample”) to
Nevertheless, such training (viz. concurrent produce the center-key comparison (i.e., on RA
ABR, AA, BB, and CC successive matching) and RB successive matching probes). Such an
would still be expected to yield symmetrical BA outcome would provide especially compelling
comparison-response effects and symmetrical support for the view that stimulus-class forma-
BAR spatial-response effects, like those observed tion in pigeons can, as in humans, incorporate
in Experiment 1. their different responses.
Second, Urcuioli’s (2008) theoretical as-
sumptions predict that concurrent training References
on AB, AA, and BBR successive matching will
Astley, S. L., & Wasserman, E. A. (2001). Superordinate
also yield the same two five-member classes categorization via learned stimulus equivalence: Quan-
shown in Figure 2. Consequently, if later given tity of reinforcement, hedonic value, and the nature of
a left versus right choice following the positive the mediator. Journal of Experimental Psychology: Animal
AA trials, pigeons trained in this fashion Behavior Processes, 27, 252–268.
should respond differentially to the side keys. Dube, W. V., & McIlvane, W. J. (1995). Stimulus–reinforcer
relations and emergent matching-to-sample. The Psycho-
Specifically, they should preferentially peck the logical Record, 45, 591–612.
left key after a R1–R2 probe trial and Dube, W. V., McIlvane, W. J., Mackay, H. A., & Stoddard, L. T.
preferentially peck the right key after a G1– (1987). Stimulus class membership established via
G2 probe trial. If obtained, this result could stimulus–reinforcer relations. Journal of the Experimental
Analysis of Behavior, 47, 159–175.
not be explained in terms of explicit training Dube, W. V., McIlvane, W. J., Maguire, R. W., Mackay, H. A.,
to peck left versus peck right after the red & Stoddard, L. T. (1989). Stimulus class formation and
versus green (A) nominal stimuli, respectively, stimulus–reinforcer relations. Journal of the Experimental
because no such training is provided in any of Analysis of Behavior, 51, 65–76.
the AB, AA, and BBR baseline tasks. Edwards, C. A., Jagielo, J. A., Zentall, T. R., & Hogan, D. E.
(1982). Acquired equivalence and distinctiveness
Third, new successive matching relations in matching to sample by pigeons: Mediation by
should emerge following just hue–form sym- reinforcer-specific expectancies. Journal of Experimental
bolic and DW identity training so long as both Psychology: Animal Behavior Processes, 8, 244–259.
baseline tasks include a left versus right side- Frank, A. J., & Wasserman, E. A. (2005). Associative
symmetry in the pigeon after successive matching-to-
key response requirement (i.e., ABR and CCR, sample training. Journal of the Experimental Analysis of
respectively). Thus, if a red sample – triangle Behavior, 84, 147–165.
comparison – left response (R1–T2–Left3) García, A., & Benjumea, S. (2006). The emergence of
sequence is reinforced as part of hue–form symmetry in a conditional discrimination task using
symbolic matching, and a dot sample – dot different responses as proprioceptive samples in
pigeons. Journal of the Experimental Analysis of Behavior,
comparison – left response (D1–D2–Left3) 86, 65–80.
sequence is reinforced as part of DW identity Hall, G., Mitchell, C., Graham, S., & Lavis, Y. (2003).
matching, Urcuioli’s (2008) theory predicts Acquired equivalence and distinctiveness in human
148 PETER J. URCUIOLI et al.

discrimination learning: Evidence for associative Sidman, M. (1994). Equivalence relations and behavior: A
mediation. Journal of Experimental Psychology: General, research story. Boston: Authors Cooperative.
132, 266–276. Sidman, M. (2000). Equivalence relations and the reinforce-
Honig, W. K., Matheson, W. R., & Dodd, P. W. D. (1984). ment contingency. Journal of the Experimental Analysis of
Outcome expectancies as mediators for discriminative Behavior, 74, 127–146.
responding. Canadian Journal of Psychology, 38, 196– Strasser, R., Ehrlinger, J. M., & Bingman, V. P. (2004).
217. Transitive behavior in hippocampal-lesioned pigeons.
Jones, B. M. (2003). Quantitative analyses of matching to Brain, Behavior, and Evolution, 63, 181–188.
sample performance. Journal of the Experimental Analysis Sweeney, M. M., & Urcuioli, P. J. (2010). A reflexivity effect in
of Behavior, 79, 323–350. pigeons. Journal of the Experimental Analysis of Behavior, 94,
Joseph, B., Overmier, J. B., & Thompson, T. (1997). Food- 267–282.
and nonfood-related differential outcomes in equiva- Urcuioli, P. J. (2005). Behavioral and associative effects
lence learning by adults with Prader-Willi syndrome. of differential outcomes in discrimination learning.
American Journal on Mental Retardation, 101, 374–386. Learning & Behavior, 33, 1–21.
Konorski, J. (1959). A new method of physiological Urcuioli, P. J. (2008). Associative symmetry, anti-symmetry,
investigation of recent memory in animals. Bulletin De and a theory of pigeons’ equivalence-class formation.
L’Académie Polonaise Des Sciences, 7, 115–117. Journal of the Experimental Analysis of Behavior, 90, 257–
Lionello-DeNolf, K. M., & Braga-Kenyon, P. (in press). 282.
Membership of defined responses in equivalence Urcuioli, P. J. (2010). Associative symmetry and stimulus-
classes. The Psychological Record. class formation by pigeons: The role of non-reinforced
Lionello, K. M., & Urcuioli, P. J. (1998). Control by sample baseline relations. Behavioural Processes, 85, 226–235.
location in pigeons’ matching-to-sample. Journal of the Urcuioli, P. J. (2011). Emergent identity matching after
Experimental Analysis of Behavior, 70, 235–251. successive matching training: Reflexivity or generalized
Lowe, C. F., Horne, P. J., & Hughes, J. C. (2005). Naming and identity? Journal of the Experimental Analysis of Behavior, 96,
categorization in young children: III. Vocal tact training 329–341.
and transfer of function. Journal of the Experimental Urcuioli, P. J. (2013). Stimulus control and stimulus-class
Analysis of Behavior, 83, 47–65. formation. In G. J. Madden, W. V. Dube, T.
Mackay, H. A., Wilkinson, K. M., Farrell, C., & Serna, R. W. D. Hackenberg, G. P. Hanley, & K. A. Lattal (Eds.),
(2011). Evaluating merger and intersection of equiva- APA Handbook of Behavior Analysis, Vol. 1 (pp. 361–386).
lence classes with one member in common. Journal of the Washington, DC: American Psychological Association.
Experimental Analysis of Behavior, 96, 87–105. Urcuioli, P. J., & DeMarse, T. (1997). Further tests of
Maki, P., Overmier, J. B., Delos, S., & Gutmann, A. J. (1995). response–outcome associations in differential-outcome
Expectancies as factors influencing the conditional discriminations. Journal of Experimental Psychology: Animal
discrimination performance of children. The Psychologi- Behavior Processes, 23, 171–182.
cal Record, 45, 45–71. Urcuioli, P. J., DeMarse, T., & Zentall, T. R. (1994). Some
Manabe, K., Kawashima, T., & Staddon, J. E. R. (1995). properties of many-to-one matching with hue, response,
Differential vocalization in budgerigars: Towards an and food samples: Retention and mediated transfer.
experimental analysis of naming. Journal of the Experi- Learning and Motivation, 25, 175–200.
mental Analysis of Behavior, 63, 111–126. Urcuioli, P. J., Lionello-DeNolf, K., Michalek, S., &
Meehan, E. F. (1999). Class-consistent differential reinforce- Vasconcelos, M. (2006). Some tests of response
ment and stimulus class formation in pigeons. Journal of membership in acquired equivalence classes. Journal of
the Experimental Analysis of Behavior, 72, 97–115. the Experimental Analysis of Behavior, 86, 81–107.
Nelson, K. R., & Wasserman, E. A. (1978). Temporal factors Urcuioli, P. J., Pierce, J. N., Lionello-DeNolf, K. M., Friedrich,
influencing the pigeon’s successive matching-to-sample A., Fetterman, J. G., & Green, C. (2002). The development
performance: Sample duration, intertrial interval, and of emergent differential sample behavior in pigeons.
retention interval. Journal of the Experimental Analysis of Journal of the Experimental Analysis of Behavior, 78, 409–
Behavior, 30, 153–162. 432.
Rodger, R. S. (1975). The number of non-zero, post hoc Urcuioli, P. J., & Swisher, M. (2012). Emergent identity
contrasts from ANOVA and error rate. I. British Journal of matching after successive Matching training II: Reflex-
Mathematical and Statistical Psychology, 28, 71–78. ivity or transitivity? Journal of the Experimental Analysis of
Saunders, K. J., & Williams, D. C. (1998). Do parakeets Behavior, 97, 5–27.
exhibit derived stimulus control? Some thoughts on Urcuioli, P. J., & Vasconcelos, M. (2008a). The effects of
experimental control procedures. Journal of the Experi- within-class differences in sample responding on
mental Analysis of Behavior, 70, 321–324. acquired sample equivalence. Journal of the Experimental
Schenk, J. J. (1994). Emergent relations of equivalence Analysis of Behavior, 89, 341–358.
generated by outcome-specific consequences in condi- Urcuioli, P. J., & Vasconcelos, M. (2008b). On the origins of
tional discrimination. The Psychological Record, 44, 537– emergent differential sample behavior. Journal of the
558. Experimental Analysis of Behavior, 90, 61–80.
Shimizu, H. (2006). Testing response–stimulus equivalence Urcuioli, P. J., & Zentall, T. R. (1992). Transfer across
relations using differential responses as a sample. Journal delayed discriminations: Evidence regarding the nature
of the Experimental Analysis of Behavior, 86, 239–251. of prospective working memory. Journal of Experimental
Sidman, M. (1986). Functional analysis of emergent Psychology: Animal Behavior Processes, 18, 154–173.
verbal classes. In T. Thompson & M. D. Zeiler (Eds.), Urcuioli, P. J., Zentall, T. R., & DeMarse, T. (1995). Transfer
Analysis and integration of behavioral units (pp. 213–245). to derived sample-comparison relations following
Hillsdale, NJ: Erlbaum. many-to-one and one-to-many matching with identical
RESPONSE MEMBERSHIP IN STIMULUS CLASSES 149

training relations. Quarterly Journal of Experimental Wasserman, E. A., DeVolder, C. L., & Coppage, D. J. (1992).
Psychology, 48B, 158–178. Nonsimilarity-based conceptualization in pigeons.
Vasconcelos, M., & Urcuioli, P. J. (2011). Associative Psychological Science, 3, 374–379.
symmetry in a spatial-response paradigm. Behavioural
Processes, 86, 305–315.
Wasserman, E. A. (1976). Successive matching-to-sample
in the pigeon: Variation on a theme by Konorski. Received: June 28, 2012
Behavior Research Methods & Instrumentation, 8, 278–282. Final Acceptance: January 4, 2013

You might also like