JOURNAL OF THE EXPERIMENTAL ANALYSIS OF BEHAVIOR 1999, 72, 117–137 NUMBER 1 (JULY)

A DISCRIMINATION ANALYSIS OF TRAINING-STRUCTURE

EFFECTS ON STIMULUS EQUIVALENCE OUTCOMES
R ICHARD R. S AUNDERS AND G INA G REEN
SCHIEFELBUSCH INSTITUTE FOR LIFE SPAN STUDIES,
UNIVERSIT Y OF KANSAS AND
NEW ENGLAND CENTER FOR CHILDREN,
E. K. SHRIVER CENTER FOR MENTAL RETARDATION, AND
NORTHEASTERN UNIVERSIT Y

Experiments designed to establish stimulus equivalence classes frequently produce differential out-
comes that may be attributable to training structure, defined as the order and arrangement of
baseline conditional discrimination training trials. Several possible explanations for these differences
have been suggested. Here we develop a hypothesis based on an analysis of the simple simultaneous
and successive discriminations embedded in conditional discrimination training and testing within
each of the training structures that are typically used in stimulus equivalence experiments. Our
analysis shows that only the comparison-as-node (many-to-one) structure presents all the simple
discriminations in training that are subsequently required for consistently positive outcomes on all
tests for the properties of equivalence. The sample-as-node (one-to-many) training structure does
not present all the simple discriminations required for positive outcomes on either the symmetry or
combined transitivity and symmetry (equivalence) tests. The linear-series training structure presents
all the simple discriminations required for consistently positive outcomes on tests for symmetry, but
not for symmetry and transitivity combined (equivalence) or transitivity alone. Further, the difference
in the number of simple discriminations presented in comparison-as-node training versus the other
training structures is larger when the intended class size is greater than three or the number of
classes is larger than two. We discuss the relevance of this analysis to interpretations of stimulus
equivalence research, as well as some methodological and theoretical implications.
Key words: stimulus equivalence, stimulus classes, simple discrimination, conditional discrimination,
discrimination learning, stimulus relations

Experimental analyses of stimulus equiva- Comparison Stimulus B2 if and only if Sam-

lence based on the Sidman model (1971, ple Stimulus A2 is present, and Comparison
1986, 1994) expose subjects to match-to-sam- Stimulus B3 if and only if Sample Stimulus
ple (MTS) training designed to establish con- A3 is present. In addition, subjects may learn
ditional discriminations among stimuli that to respond to Comparison Stimuli C1, C2,
are not physically similar to one another. Typ- and C3 only in the presence of Sample Stim-
ically, two or more conditional discrimina- uli B1, B2, and B3, respectively. If successful,
tions with some stimuli in common are such training establishes conditional relations
trained through differential reinforcement. between each sample (conditional stimulus)
For example, subjects may learn to respond and its corresponding correct comparison
to a comparison stimulus designated B1 if (discriminative stimulus, or S1), in this case,
and only if Sample Stimulus A1 is present, A1B1, A2B2, A3B3, B1C1, B2C2, and B3C3.
Performances indicating the development
Development of this paper was supported in part by of conditional relations are prerequisites for
National Institute of Child Health and Human Develop- testing the possibility that the trained (or
ment Grants HD25995-06 to the E. K. Shriver Center and
HD18955-13 and HD02528 to the Schiefelbusch Institute baseline) relations have the properties of
for Life Span Studies, University of Kansas, and by the equivalence, as defined in mathematics: re-
New England Center for Children. We acknowledge Joe flexivity, symmetry, and transitivity. The tests
Spradlin and Julie McEntee for their helpful comments consist of MTS trials on which subjects have
on preliminary drafts. Portions of this paper were pre-
sented at the third European meeting of the Experimen- the opportunity to demonstrate all possible
tal Analysis of Behaviour Group, July, 1997, in Dublin, untrained conditional relations among the
Ireland and at the meeting of the Association for Behav- stimuli involved in training. Symmetry would
ior Analysis, May, 1997, Chicago. be evaluated by reversing the sample and
Requests for reprints should be addressed to Richard
R. Saunders at the Parsons Research Center, 2601 Gabri-
comparison stimuli relative to training (i.e.,
el, Parsons, Kansas 67357 (E-mail: [email protected]. B1A1, B2A2, B3A3, C1B1, C2B2, C3B3 in our
ukans.edu). example). Transitivity would be evaluated by

117
118 RICHARD R. SAUNDERS and GINA GREEN

presenting sample and comparison stimuli common, or linking, sets of stimuli). Fields
that were related indirectly in training and Verhave (1987) referred to a stimulus
through a common trained relation with oth- that is related to only one other stimulus in
er stimuli (i.e., A1C1, A2C2, and A3C3). Tests training as a ‘‘single,’’ and called a stimulus
for the untrained relations C1A1, C2A2, and that is related to more than one other stim-
C3A3 would constitute simultaneous tests for ulus a ‘‘node.’’ To Fields and Verhave, the
the properties of symmetry and transitivity four-stage arrangement of Sidman et al.
(also called combined tests, or simply equiv- (1985) would be a two-node arrangement,
alence tests). Reflexivity would be evaluated with the B and C stimuli serving as nodes.
via conditional identity matching tests with These authors also suggested that the struc-
the stimuli involved in training (e.g., A1A1, ture of equivalence classes can be described
A2A2, A3A3, B1B1, etc.). If outcomes of all in terms of four parameters: the number of
tests are positive, the inference is made that stimuli in each class, the number of nodes,
the trained relations are equivalence rela- the pattern of singles relative to nodes, and
tions and that the stimuli so related constitute the pattern formed by assignment of stimuli
equivalence classes (e.g., Green & Saunders, to the roles of samples and comparisons dur-
1998; R. R. Saunders & Green, 1992; Sidman, ing training (i.e., directionality of training).
1986, 1994; Sidman et al., 1982; Sidman & The original Sidman analysis of stimulus
Tailby, 1982). equivalence did not suggest that equivalence
test outcomes should vary as a function of
training structure, order, or direction (Sid-
TRAINING STRUCTURE man & Tailby, 1982). On the contrary, it im-
The foregoing example represents the plied that if training established the intended
minimal conditions required to determine conditional relations and concurrently pre-
whether MTS training produces stimulus vented the development of extraneous stim-
equivalence: training on two arbitrary condi- ulus control, then responding on all tests for
tional discriminations with one set of stimuli untrained relations should be consistent with
in common (AB and BC in the example), fol- equivalence, regardless of the order and ar-
lowed by testing for all possible conditional rangement of the trained conditional dis-
discriminations that were not trained directly. criminations (Carrigan & Sidman, 1992;
The minimum number of stimuli in an equiv- Green & Saunders, 1998; Sidman, 1994).
alence class is three. In our example, the AB Some investigators, however, have reported
conditional discrimination was trained first, differential outcomes on equivalence tests
followed by the BC conditional discrimina- that appear to be due to training structure.
tion; the B stimuli were common to both dis- For example, one recent study with preschool
criminations. Of course, equivalence classes children found that five-member equivalence
can and often do include more than three classes were more likely following one train-
stimuli each, and the baseline conditional dis- ing sequence than another (R. R. Saunders,
criminations can and often are trained in dif- Drake, & Spradlin, 1999). All of the children
ferent sequences and with different common were exposed to two-choice MTS training de-
stimuli than in our example. The term train- signed to establish four conditional discrimi-
ing structure has been used to refer to the se- nations among 10 arbitrary visual stimuli. For
quence of conditional discriminations and 6 children, two stimuli served as the samples
the arrangements of common or ‘‘linking’’ in all four conditional discriminations; the
stimuli presented to subjects in baseline train- trained relations were designated AB, AC,
ing. Various terms have been coined to de- AD, and AE. This kind of training structure
scribe specific training structures. For exam- has been referred to as a ‘‘sample-as-node’’
ple, Sidman, Kirk, and Willson-Morris (1985) or ‘‘one-to-many’’ structure. Five other chil-
described the situation in which two stimuli dren received training with one pair of stim-
are mutually related to a third (e.g., AB, BC) uli serving as comparisons with each of four
as a ‘‘three-stage’’ training arrangement; different pairs of sample stimuli. The trained
training another linked conditional discrimi- conditional relations were designated BA,
nation such as CD created a ‘‘four-stage’’ ar- CA, DA, and EA. This training structure has
rangement (underscored letters designate been dubbed a ‘‘comparison-as-node’’ or
 DISCRIMINATION LEARNING AND STIMULUS EQUIVALENCE 119

‘‘many-to-one’’ structure (cf. K. J. Saunders, formances on tests for the untrained relations
Saunders, Williams, & Spradlin, 1993; Urcuio- among those stimuli were likely to be (e.g.,
li & Zentall, 1993; Urcuioli, Zentall, Jackson- Fields, Verhave, & Fath, 1984). Although the
Smith, & Steirn, 1989). Positive outcomes on results of some experiments seem to support
equivalence tests were found for all 5 chil- this hypothesis (e.g., Fields, Adams, Verhave,
dren who had comparison-as-node training, & Newman, 1990; Kennedy, 1991; Kennedy,
but positive outcomes were found for only 2 Itkonen, & Lindquist, 1994), procedural var-
of 6 children who had sample-as-node train- iations and stimulus characteristics in those
ing. experiments make them difficult to interpret.
The results obtained by R. R. Saunders et (These will be discussed later.) This account
al. (1999) replicated those of previous exper- has also been criticized because it invokes la-
iments with adolescents and adults with men- bels (nodal distance, associative distance) for
tal retardation (Drake & Saunders, 1987, cit- a structural property (the number of nodes)
ed in K. J. Saunders et al., 1993; R. R. that may imply that other hypothetical struc-
Saunders, Wachter, & Spradlin, 1988; Sprad- tural properties are at work (Sidman, 1994,
lin & Saunders, 1986). Across these studies, p. 539). Further, the account falls short of ex-
five-member equivalence classes were estab- plaining different outcomes on tests for
lished in only 1 of 7 subjects trained with the equivalence in basic behavioral terms.
sample-as-node procedure, but they were es- With respect to sample-as-node versus com-
tablished in 6 of 6 subjects trained with the parison-as-node differences, Sidman (1994,
comparison-as-node procedure. Apparent pp. 527–528) raised the possibility that the
structure-related differences have also been former might establish differential contextual
reported with normally capable adult subjects control of trained conditional relations by
(Barnes, 1992, as cited in Barnes, 1994; negative stimuli (incorrect comparisons),
Fields, Hobbie, Adams, & Reeve, in press). In which could lead to negative outcomes on
contrast, a recent study with normally capable tests for the properties of equivalence. An al-
adults suggested that sample-as-node training ternative account was postulated by Spradlin
was more likely to produce three-member and his colleagues (K. J. Saunders et al., 1993;
equivalence classes than was comparison-as- Spradlin & Saunders, 1986; also see Sidman,
node training (Arntzen & Holth, 1997). Fi- 1994, pp. 526-527). They speculated that sam-
nally, some investigators have reported nega- ple-as-node training did not consistently pro-
tive outcomes on tests for some properties of duce equivalence classes because the training
equivalence following baseline training in contingencies did not require all the simple
which several conditional discriminations discriminations subsequently called for on
were trained in sequence, with multiple nod- tests for equivalence. Spradlin and colleagues
al or linking stimuli (e.g., AB, BC, CD, DE; noted that when AB and AC are trained, for
see Arntzen & Holth, 1997; Fields, Landon- example, subjects need only discriminate the
Jimenez, Buffington, & Adams, 1995; Holth A samples from one another, the B compari-
& Arntzen, 1998). This has been described as sons from one another, and the C compari-
a linear-series training structure (e.g., Green sons from one another; the training contin-
& Saunders, 1998). gencies do not require discrimination of each
B stimulus from each C stimulus. The B ver-
sus C discriminations are called for, however,
SOME HY POTHESES ABOUT on the BC and CB trials that constitute tests
TRAINING-STRUCTURE EFFECTS for the properties of equivalence. In contrast,
Why might different training structures comparison-as-node training (e.g., BA and
yield different outcomes on equivalence CA) requires successive discrimination of all
tests? Several possibilities have been suggest- B and C stimuli across trials and the simul-
ed. With respect to linear-series training taneous discrimination of all A comparisons
structures, Fields and colleagues offered an within trials to fulfill the training contingency
account based on nodal or ‘‘associative’’ dis- requirements. That is, this training structure
tance. They suggested that the larger the potentially establishes all of the simple dis-
number of nodes potentially linking stimuli criminations required for consistently positive
indirectly in training, the less robust the per- outcomes on tests for the properties of equiv-
120 RICHARD R. SAUNDERS and GINA GREEN

alence (K. J. Saunders et al., 1993; R. R. Saun- establish that two sample stimuli (e.g., B1 and
ders et al., 1999; Spradlin & Saunders, 1986; C1) both control a response to the same com-
also see Barnes, 1994; Sidman, 1994). parison (e.g., A1), might not the subject then
treat B1 and C1 as if they are the same—that
is, fail to discriminate them? And would this
THE DISCRIMINATION not suffice to produce apparently positive
ACCOUNT: AN ELABORATION outcomes on all tests for equivalence (e.g.,
AND EXPANSION trials testing B1C1 and C1B1), as long as the
We propose that, with some additional de- subject discriminates B1 and C1 from the
velopment and elaboration, the discrimina- stimuli in the other prospective equivalence
tion analysis suggested by Spradlin and col- classes? We maintain that the answer to these
leagues provides a parsimonious account of questions is no, for the following reasons: At
the differential effects of training structures the beginning of a well-designed stimulus
on equivalence test outcomes, one that is con- equivalence experiment, the subject is ex-
sistent with basic principles of stimulus con- posed to a group of unsorted stimuli that do
trol and does not invoke constructs like as- not bear any consistent physical resemblance
sociative distance or mediation (cf. McIlvane to one another. The initial training contin-
& Dube, 1992; Sidman, 1986, 1994). Here we gencies are designed to establish a relation
undertake such an elaboration by examining between each comparison stimulus and a par-
(a) how simple simultaneous and successive ticular sample stimulus. Therefore, the con-
discriminations are necessarily embedded in tingencies require discrimination of every
conditional discriminations; (b) which and sample from every other sample presented
how many of those component simple dis- successively across trials, all samples from all
criminations are presented to subjects in each comparisons, and all comparison stimuli pre-
of the training structures commonly used in sented simultaneously within trials (Sidman,
stimulus equivalence experiments; and (c) 1986). Training procedures that enhance the
the component simple discriminations re- probability that every stimulus will be discrim-
quired for consistently positive outcomes on inated from every other stimulus—such as
tests for equivalence following training with contingencies that specify a different re-
each structure. Finally, we reanalyze the re- sponse to each conditional (sample) stimulus
sults of several published studies of stimulus and to each discriminative (comparison)
equivalence from this perspective. stimulus—should foster the development of
conditional discriminations and, therefore,
Assumptions and Definitions the development of equivalence classes (Sid-
For our analysis, we adopt a critical as- man, 1994, pp. 413–414).
sumption that has been stated or implied by Like training trials, tests for stimulus equiv-
several other authors (e.g., McIlvane & Dube, alence also present conditional discrimina-
1996; K. J. Saunders et al., 1993; Spradlin & tions that are composed of simple successive
Saunders, 1986; Sidman, 1994): For perfor- and simultaneous discriminations among ex-
mances to meet criteria for acquisition of the trained perimental stimuli. Therefore, for consistent-
baseline relations as well as criteria for positive out- ly positive outcomes on all tests for the prop-
comes on all tests for stimulus equivalence, each erties of equivalence, every stimulus must be
stimulus must be discriminated from every other discriminated from ever y other stimulus.
stimulus in the experiment. On its face, this as- More important, however, test trials may pre-
sumption might appear counterintuitive. It sent some simple successive and simultaneous
might seem that in order to ‘‘pass’’ all tests discriminations that were not presented at all
for stimulus equivalence, subjects need only in training, or that are presented differently
discriminate all stimuli in each class from all on tests than in training (e.g., successively vs.
stimuli in the other classes (between-class dis- simultaneously). If the relevant simple dis-
criminations, such as A1 vs. A2 vs. A3, B1 vs. criminations that compose the tested condi-
B2 vs. B3), and need not discriminate stimuli tional discriminations are not made, negative
within classes from one another (A1 vs. B1, outcomes on some or all tests for the prop-
B2 vs. C2, C3 vs. D3, etc.). The reasoning erties of equivalence will result (see McIlvane
might go like this: If training contingencies & Dube, 1996; K. J. Saunders et al., 1993; R.
 DISCRIMINATION LEARNING AND STIMULUS EQUIVALENCE 121

R. Saunders et al., 1999; Sidman, 1986, 1994; We describe the types and proportions of
Spradlin & Saunders, 1986). We argue here those discriminations that are presented and
that different training structures make such not presented in each of the training structures
outcomes more or less probable because they typically used in stimulus equivalence exper-
are more or less likely to establish the nec- iments, to reflect the understanding that
essary simple discriminations. whether the training contingencies actually
To bolster our contention that consistently establish those discriminations is always an
positive outcomes on tests for stimulus equiv- empirical question.
alence require discrimination of every stim- Our analysis also assumes (a) the use of si-
ulus from every other stimulus in the exper- multaneous MTS procedures; (b) that base-
iment, it is important to note that although line conditional discriminations are mixed at
such outcomes imply that the stimuli within some point before testing for equivalence be-
classes are substitutable for one another, it gins (e.g., if AB and AC relations are trained,
does not necessarily follow that they are in- all of those trial types are presented together
discriminable from one another. On the con- within one or more training session before
trary, from a practical standpoint it is vital for testing); (c) that all trial types testing for a
organisms to discriminate that stimuli that particular property of equivalence (e.g., BC
can substitute for one another in certain con- and CB) are presented together within the
texts nonetheless have distinctive features. same session; and (d) that training and test-
For example, we may observe that a child ing are conducted with balanced MTS con-
matches an apple, a picture of an apple, and ditional discrimination procedures (Green &
the printed word APPLE to one another in Saunders, 1998). For example, Comparison
every possible sample-comparison arrange- C1 is always presented with Comparisons C2
ment. That observation might lead us to con- and C3; C1 is never presented as a compari-
clude that those stimuli are substitutable for son with any stimuli from other sets (e.g., B2,
one another, and constitute an equivalence D3). A final assumption is that no differential
class. But that observation should not lead to consequences are arranged for responses on
the conclusion that those stimuli are not dis- test trials.
criminated. Indeed, for the class to be func-
tional for the child, it is necessary for discrim- Simple Discriminations Embedded in
inations among the members to be Conditional Discriminations
maintained—so that the child does not try to
eat the picture or the printed word, for ex- We turn now to a review of simple and con-
ample—at the same time as the stimuli are ditional discrimination training contingen-
treated as equivalent (substitutable) in cer- cies, to reiterate precisely why acquisition of
tain contexts. (Whether discriminations conditional discriminations necessitates ac-
among stimuli are in fact maintained during quisition of the component simple discrimi-
or after demonstrations that they are equiva- nations.
lent could be evaluated empirically via con- Simple discriminations. In the presence of a
ditional identity matching tests; cf. Dube, particular antecedent stimulus, if a particular
McIlvane, & Green, 1992.) Of course, it is response is followed by a particular conse-
also necessary for all members of the ‘‘apple’’ quence, the response may come to occur
class to be discriminated from all members of more often in the presence of the stimulus
other stimulus classes. In other words, per- (S1) than in its absence or in the presence
formances that consistently demonstrate stim- of another stimulus (S2). When repeated ap-
ulus equivalence entail discriminations plication of these contingencies leads to re-
among stimuli within as well as between clas- sponding in the presence of the S1 but not
ses. in the presence of the S2, one can infer that
For purposes of our analysis, we refer to a simple discrimination has been established.
the grand total of all possible simple simul- When the S1 and the S2 are presented con-
taneous and successive discriminations currently on each of a series of trials, the pro-
among the stimuli in a stimulus equivalence cedure is termed a simultaneous simple discrim-
experiment as required for consistently posi- ination. When the stimuli are presented one
tive outcomes on all training and test trials. at a time, on unsystematically alternating tri-
122 RICHARD R. SAUNDERS and GINA GREEN

als, the procedure is termed a successive simple

discrimination.
Conditional discriminations. In conditional
discrimination training, simple discrimina-
tions are brought under the control of addi-
tional antecedents, or conditional stimuli
(Sidman, 1986). A common and reliable
preparation for establishing conditional dis-
criminations is MTS training. The MTS pro-
cedures used in most stimulus equivalence
experiments involve a minimum of two dif-
ferent conditional stimuli (samples) and two
different discriminative stimuli (compari-
sons) per conditional discrimination. The
same comparison stimuli are presented on ev-
ery trial while the sample stimulus varies un-
systematically from trial to trial. In the least
complicated procedure, contingencies are ar-
ranged so that each comparison stimulus is
discriminative for reinforcement (S1) in the
presence of one and only one conditional
(sample) stimulus, and is not discriminative
for reinforcement (S2) in the presence of a
second sample stimulus. That is, the func-
tions of the discriminative stimuli change
from trial to trial, depending on which sam-
ple stimulus is present. Although three or
more samples and comparisons are generally
desirable (Carrigan & Sidman, 1992; Green
& Saunders, 1998; Johnson & Sidman, 1993;
Kelly, Green, & Sidman, 1998; Sidman, 1987), Fig. 1. Representations of three basic training struc-
we use just two samples and comparisons tures used in stimulus equivalence experiments: compar-
here and elsewhere for ease of illustration. ison as node (upper panel) with three potential classes
of five stimuli each; sample as node (middle panel) with
In a typical stimulus equivalence experi- two potential classes of four stimuli each; and linear se-
ment, multiple training trials are presented in ries (lower panel) with four potential classes of three
a session or block. A different sample is pre- stimuli each. Arrows point from stimuli used as samples
sented on each trial, and the same compari- to comparison stimuli.
sons are presented on every trial. Each sam-
ple is usually presented equally often within
a session or block of trials, in unsystematic some successive (cf. Green & Saunders, 1998;
order. The positions of the comparisons, es- K. J. Saunders & Spradlin, 1989, 1993; Sid-
pecially the S1, vary unsystematically from man, 1986).
trial to trial. Sample stimuli are presented
one at a time across trials, with some time Common Stimulus Equivalence
elapsing between presentations; thus, dis- Training Structures
criminating among them involves successive Many variations of the training structures
simple discriminations. Comparison stimuli described by Fields and Verhave (1987) ap-
are presented together on every trial, as are pear in the stimulus equivalence literature,
the samples and comparisons that constitute but there are three basic prototypes. The top
each trial type; thus, discriminating among panel of Figure 1 shows a comparison-as-node
them involves simultaneous simple discrimi- (many-to-one) structure for potentially devel-
nations. In short, performing conditional dis- oping three equivalence classes with five stim-
criminations necessarily involves a number of uli per class. This training structure usually
simple discriminations, some simultaneous, involves the use of three-choice MTS proce-
 DISCRIMINATION LEARNING AND STIMULUS EQUIVALENCE 123

dures. The middle panel of Figure 1 shows

the sample-as-node (one-to-many) structure
for potentially developing two equivalence
classes with four stimuli per class. The train-
ing represented is usually arranged using two-
choice MTS procedures. The linear-series
structure is shown in the lower panel of Fig-
ure 1. The schematic depicts four potential
classes of three stimuli each, which usually in-
volves four-choice MTS procedures. Most
stimulus equivalence experiments in the lit-
erature have used comparison-as-node, sam-
ple-as-node, or linear-series training struc-
tures, or variations thereof, to produce two or
more equivalence classes (Green & Saunders,
1998; for some examples of variations, see
Fields, Adams, & Verhave, 1993; Kennedy,
1991; Pilgrim & Galizio, 1995; Spradlin, Cot-
ter, & Baxley, 1973; Wetherby, Karlan, &
Spradlin, 1983).
Simple Discriminations Presented in
Various Training Structures
Comparison as node. Figure 2 represents
comparison-as-node training (BA and CA)
leading potentially to the development of two
equivalence classes of three stimuli each. Tri-
al types for this training, and for tests of sym-
metry and equivalence, are shown beneath
the schematic. In this structure, 15 simple dis-
criminations are presented in training, as fol-
lows: Comparison Stimuli A1 and A2 are pre-
sented simultaneously on training trials (one Fig. 2. Comparison-as-node training to produce two
simple discrimination). Each comparison three-member equivalence classes. Solid arrows indicate
trained relations, dashed-line arrows indicate tests for the
stimulus is presented simultaneously with property of symmetry, and the double-pointed dotted-
each sample stimulus; that is, A1 is presented line arrows indicate combined tests for the properties of
with B1, B2, C1, and C2 (four discrimina- symmetry and transitivity (or equivalence tests). Trial
tions), and A2 is presented with B1, B2, C1, types for training trials and tests for symmetry and equiv-
alence are shown below the schematic. Asterisks indicate
and C2 (four discriminations). The two B experimenter-designated correct comparisons (for posi-
stimuli and the two C stimuli are presented tive test outcomes).
successively across training trials as samples
when BA and CA trials are mixed (six dis-
criminations). For the training contingency between samples and comparisons (i.e., each
requirements to be fulfilled consistently, all B stimulus from each C stimulus). In this
15 of these simple discriminations must be case, all of those simple discriminations were
made. presented during baseline training. Although
Tests for stimulus equivalence—BC and CB the B and C stimuli were not presented as
trial types—also present certain simple dis- simultaneous discriminations, they were pre-
criminations as components of the tested sented as successive (sample) discriminations
conditional discriminations: the successive when the BA and CA training trials were
discrimination of all B and C sample stimuli, mixed prior to testing. Because simultaneous
two simultaneous discriminations between discriminations are generally easier than suc-
the comparison stimuli (B1 vs. B2 and C1 vs. cessive discriminations (Brady & Saunders,
C2), and four simultaneous discriminations 1991; Carter & Eckerman, 1975; Urcuioli et
124 RICHARD R. SAUNDERS and GINA GREEN

al., 1989), we would expect most subjects to

have little difficulty with the BC simultaneous
discriminations called for on the equivalence
tests in this example. Similarly, all of the sim-
ple discriminations involved in tests for sym-
metry (AB and AC) were presented during
training. Only one (A1 vs. A2) might pose a
problem for some subjects, because the dis-
crimination between those stimuli was pre-
sented simultaneously in training, but it is
presented successively on the symmetry tests.
Most of the other simple discriminations
called for on the symmetry tests are exactly
the same as in training; two (B1 vs. B2 and
C1 vs. C2) that were presented successively in
training appear as simultaneous discrimina-
tions on symmetry tests. The discriminations
between the A stimuli and the B and C stim-
uli remain the same as in training, except
that sample and comparison roles are re-
versed. In short, training with this compari-
son-as-node structure presents all of the sim-
ple discriminations involved in ever y
symmetry and equivalence test trial type.
(Tests for transitivity alone are not possible
following training with this structure.)
Sample as node. Quite a different picture
emerges from an analysis of the simple dis-
criminations presented during sample-as-
node training. Figure 3 represents sample-as-
node training (AB and AC) leading
potentially to the development of two equiv-
alence classes of three stimuli each. In train- Fig. 3. Sample-as-node training to produce two three-
ing, Comparison Stimuli B1 and B2 are pre- member equivalence classes. Solid arrows indicate
sented simultaneously, as are Comparisons C1 trained relations, dashed-line arrows indicate tests for the
property of symmetry, and the double-pointed dotted-
and C2. Samples A1 and A2 are presented line arrows indicate combined tests for the properties of
successively, and are presented simultaneous- symmetry and transitivity (or equivalence tests). Trial
ly with the B and C comparisons. Although types for training trials and tests for symmetry and equiv-
the B and C stimuli are presented in pairs alence are shown below the schematic. Asterisks indicate
successively across trials when AB and AC experimenter-designated correct comparisons (for posi-
tive test outcomes).
training trials are mixed, the B and C stimuli
are never pitted against one another within a
trial (see Barnes, 1994), nor are they pre- as samples across trials on the symmetry tests,
sented successively as samples. Thus, success- and as samples across trials as well as samples
ful training with this sample-as-node structure and comparisons within trials on the tests for
potentially results in every stimulus being dis- equivalence. On symmetry tests, the simulta-
criminated from every other stimulus except neous discrimination of A1 from A2 as com-
the B stimuli from the C stimuli. Discrimi- parisons is not likely to be problematic, be-
nations among the B and C stimuli are first cause those stimuli were presented in
called for either on tests for equivalence (BC, successive discrimination format in training.
CB) or on tests for symmetry (BA, CA), In summary, negative results on equivalence
whichever is presented first. As the lower por- and symmetry tests are expected to be more
tion of Figure 3 shows, all four B and C stim- likely following sample-as-node training than
uli must be discriminated from one another comparison-as-node training, because sam-
 DISCRIMINATION LEARNING AND STIMULUS EQUIVALENCE 125

the B and C stimuli are presented successively

as samples; they were presented simulta-
neously in training. As noted previously, this
might pose a problem for some subjects. In
addition, the CB symmetry test trials call for
the successive discrimination of C1 from C2,
which was never presented in training, where-
as the BA symmetry test trials involve succes-
sive (B1 vs. B2) as well as simultaneous (A1
vs. A2, B stimuli vs. A stimuli) discriminations
that were presented in training. Thus, differ-
ent outcomes may occur on these two types
of symmetry test trials. Note that with linear-
series training designed to establish larger
equivalence classes (e.g., AB, BC, CD, DE),
only the test for symmetry of the final con-
ditional discrimination (e.g., ED) will suffer
from the potential problem just described for
CB test trials.
Referring to the lower portion of Figure 4,
it is evident that the tests for transitivity and
equivalence following linear series training
present both simple and successive discrimi-
nations among stimuli (A and C) that are nev-
er pitted against one another in training. The
Fig. 4. Linear-series training to produce two three- only context in which those stimuli are pre-
member equivalence classes. Solid arrows indicate sented together in training is when the AB
trained relations, dashed-line arrows indicate tests for the and BC training trials are mixed; there the A
property of symmetry, and the double-pointed dotted-
line arrow indicates the test for the properties of transi- and C stimuli are presented successively
tivity (AC) and symmetry and transitivity combined, or across trials within a session, but as samples
equivalence (CA). Trial types for training trials and tests and comparisons, respectively, on different
for symmetry, transitivity, and equivalence are shown be- trial types. Thus, as with sample-as-node train-
low the schematic. Asterisks indicate experimenter-des-
ignated correct comparisons (for positive test outcomes).
ing structures, linear-series structures are ex-
pected to result in a higher probability of fail-
ure on equivalence tests than training with
comparison-as-node structures.
ple-as-node training does not present some of
the simple discriminations that make up the Simple Discriminations in Larger
conditional discriminations called for on the Training Structures
tests. Table 1 summarizes the preceding analysis.
Linear series. Examination of the simple dis- To reiterate, in any training structure de-
criminations involved in linear-series training, signed to produce two equivalence classes of
shown in Figure 4, reveals yet another pat- three stimuli each, the grand total of possible
tern. In AB and BC training to establish two simple simultaneous and successive discrimi-
classes of three stimuli each, the B stimuli are nations among all the stimuli is 15. The third
presented simultaneously as comparisons, as column of the table contrasts the proportion
are the C stimuli. The A and B stimuli are of those discriminations not presented in each
presented simultaneously as samples and of the training structures. It shows that all of
comparisons on the AB training trials, and the potential simple discriminations are pre-
the B and C stimuli are presented simulta- sented in comparison-as-node training, mean-
neously on the BC training trials. The A and ing that the subject encounters no novel dis-
B stimuli are presented successively as sam- criminations on tests for equivalence and
ples when the AB and BC training trials are symmetry. Training with the other structures,
mixed. On the symmetry tests (BA and CB), however, omits some of the simple discrimi-
126 RICHARD R. SAUNDERS and GINA GREEN

Table 1
Simple discriminations in each of three training structures designed to produce two three-
member equivalence classes.

Number not Number not

presented in presented in
Number not training, but training, but
presented in presented on presented on
Structure Grand total training/total equivalence tests symmetry tests

Comparison
as node 15 0/15 0 0
Sample as node 15 4/15 4 4
Linear series 15 4/15 4 0

nations that are presented on tests for equiv- all of those discriminations are presented on
alence, as indicated by the entries in the both equivalence and symmetry tests. Follow-
fourth column. The entries in the fifth col- ing linear-series training, all of the simple dis-
umn show that the discriminations not pre- criminations not presented in training are
sented in sample-as-node training are pre- presented on equivalence tests, but none of
sented in tests for symmetry. This is not the them are presented on symmetry tests.
case for linear-series training. Table 3 shows the effects of holding pro-
Table 2 compares the numbers of simple spective class size at four stimuli per class
discriminations presented in the three train- while increasing the number of prospective
ing structures when training is designed to classes to three and employing three-choice
produce two equivalence classes of four stim- MTS training procedures (to insure balanced
uli each. Comparing the entries in Table 3 trial types per the assumptions underpinning
with those in Table 2 reveals that increasing our analysis). The total number of possible
class size from three to four stimuli increases discriminations increases to 66. As in the pre-
the total number of simple discriminations ceding examples, comparison-as-node train-
from 15 to 28. It also changes the proportion ing presents all of them. In contrast, increas-
of those discriminations that are presented in ing the number of prospective equivalence
training, as well as the proportion of simple classes from two to three alters the propor-
discriminations not presented in training but tion of the total simple discriminations that
called for on tests for the properties of equiv- are presented in sample-as-node and linear-
alence. In comparison-as-node training, all series training structures. As Table 3 indi-
the simple discriminations among the stimuli cates, 27 of the possible simple discrimina-
are presented in training. In sample-as-node tions among the experimental stimuli are not
and linear-series training structures, 12 sim- presented in sample-as-node and linear-series
ple discriminations are not presented in training; all of those discriminations appear
training. Following sample-as-node training, on the equivalence tests that follow training

Table 2
Simple discriminations in each of three training structures designed to produce two four-
member equivalence classes.

Number not Number not

presented in presented in
Number not training, but training, but
presented in presented on presented on
Structure Grand total training/total equivalence tests symmetry tests

Comparison
as node 28 0/28 0 0
Sample as node 28 12/28 12 12
Linear series 28 12/28 12 0
 DISCRIMINATION LEARNING AND STIMULUS EQUIVALENCE 127

Table 3
Simple discriminations in each of three training structures designed to produce three four-
member equivalence classes.

Number not Number not

presented in presented in
Number not training, but training, but
presented in presented on presented on
Structure Grand total training/total equivalence tests symmetry tests

Comparison 66 0/66 0 0
as node
Sample as node 66 27/66 27 27
Linear series 66 27/66 27 0

with both of those structures. Those 27 dis- the category totals would be the same. For
criminations are also presented on symmetry example, the discrimination of A1 from C3
tests following sample-as-node, but not linear- would not be presented in linear-series train-
series, training. ing, but the discrimination of B1 from C3
Table 4 shows the complete map of the sim- would be.
ple discriminations summarized in Table 3 As the series of preceding tables suggests,
for sample-as-node training. This table distin- the differences between the proportion of
guishes the simple discriminations that are simple discriminations presented in compar-
and are not presented in training, and re- ison-as-node training and those presented in
flects the roles of the stimuli (i.e., samples, the other structures increase as prospective
comparisons) involved in each simple dis- class size and number of classes increase. Ta-
crimination. A similar map for linear-series ble 5 shows that for sample-as-node and lin-
training would be organized differently, but ear-series training, the number of simple dis-

Table 4
Simple discriminations in sample-as-node training with 12 stimuli and three conditional dis-
criminations (AB, AC, AD) to produce three four-member equivalence classes.

Discrimination Stimuli involved Simple discriminations

Presented in training
Successive Samples A1 vs. A2 A1 vs. A3 A2 vs. A3
Simultaneous Samples and A1 vs. B1 A1 vs. B2 A1 vs. B3
comparisons A1 vs. C1 A1 vs. C2 A1 vs. C3
A1 vs. D1 A1 vs. D2 A1 vs. D3
A2 vs. B1 A2 vs. B2 A2 vs. B3
A2 vs. C1 A2 vs. C2 A2 vs. C3
A2 vs. D1 A2 vs. D2 A2 vs. D3
A3 vs. B1 A3 vs. B2 A3 vs. B3
A3 vs. C1 A3 vs. C2 A3 vs. C3
A3 vs. D1 A3 vs. D2 A3 vs. D3
Comparisons B1 vs. B2 B1 vs. B3 B2 vs. B3
C1 vs. C2 C1 vs. C3 C2 vs. C3
D1 vs. D2 D1 vs. D3 D2 vs. D3
Not presented in training Comparisons B1 vs. C1 B1 vs. C2 B1 vs. C3
B1 vs. D1 B1 vs. D2 B1 vs. D3
B2 vs. C1 B2 vs. C2 B2 vs. C3
B2 vs. D1 B2 vs. D2 B2 vs. D3
B3 vs. C1 B3 vs. C2 B3 vs. C3
B3 vs. D1 B3 vs. D2 B3 vs. D3
C1 vs. D1 C1 vs. D2 C1 vs. D3
C2 vs. D1 C2 vs. D2 C2 vs. D3
C3 vs. D1 C3 vs. D2 C3 vs. D3
128 RICHARD R. SAUNDERS and GINA GREEN

Table 5 are acquired at all. This may account for the

Simple discriminations by class size and number of gradual emergence of equivalence that has
classes for linear series and sample-as-node training struc- been reported by a number of investigators,
tures.
a point to which we will return later.
Number of
discriminations
Stimuli Number of Grand total not presented APPLICATION OF THE
per class classes discriminations in training DISCRIMINATION ANALY SIS TO
3 2 15a 4a EQUIVALENCE RESEARCH
3 36 9
4 66 16 The analysis presented here suggests some
5 105 25 testable hypotheses about the results of stim-
4 2 28b 12b ulus equivalence experiments conducted with
3 66c 27c various training structures. The most general
4 120 48 one is that, all other things being equal, sam-
5 190 75 ple-as-node and linear-series training are less
5 2 45 24 likely to yield positive outcomes on tests for
3 105 54 equivalence than is comparison-as-node train-
4 190 96
5 300 150 ing. In the following sections we examine rel-
evant published studies for evidence bearing
6 2 66 40
3 153 90 on this and related hypotheses.
4 276 160
5 435 250 Studies Involving a Small Number of
a From
Small Classes
Table 1.
b From Table 2.
c From
Investigators have reported mixed out-
Table 3.
comes on tests for equivalence in young chil-
dren following either sample-as-node training
criminations not presented in training (Barnes, Browne, Smeets, & Roche, 1995;
increases as a direct function of class size and Barnes, McCullagh, & Keenan, 1990; Devany,
class number. Moreover, given a particular Hayes, & Nelson, 1986; Pilgrim, Chambers, &
number of classes, the proportion of discrim- Galizio, 1995; Sidman et al., 1985; Sidman,
inations not presented to the total number of Willson-Morris, & Kirk, 1986) or linear-series
discriminations also increases as class size is training (Lazar, Davis-Lang, & Sanchez, 1984;
increased. The proportion is relatively unaf- Michael & Bernstein, 1991) to establish two
fected by increases in number of classes, how- three-member classes. In an experiment with
ever, when class size is held constant. Whether 3 children aged 6 to 7 years, a sample-as-node
increases in the proportion of discriminations structure was used to establish three three-
not presented will have different effects than member classes of tactile stimuli. Equivalence
increases in number alone is a question for classes were eventually established, but only
future research. after extensive retesting (O’Leary & Bush,
In sum, sample-as-node or linear-series 1996). Another study, using comparison-as-
training structures for large equivalence clas- node training to establish three three-mem-
ses present only a fraction of the simple dis- ber classes of arbitrary visual stimuli with chil-
criminations that make up the conditional dren aged 7 to 12 years, produced positive
discriminations encountered on tests for the test outcomes immediately following training
properties of equivalence. It is possible that (Williams, Saunders, Saunders, & Spradlin,
some new discriminations will be acquired 1995). In contrast, Eikeseth and Smith (1992)
over the course of repeated testing, which is reported that two three-member classes were
often conducted when positive outcomes are established in none of 4 young children with
not seen on initial tests (see R. R. Saunders autism following sample-as-node training. It is
& Green, 1992). Because these additional dis- important to note that the objective of the
criminations must be acquired in the absence initial testing and training in all of these ex-
of trial-by-trial differential consequences, they periments was to establish just two or three
are not likely to be acquired rapidly, if they three-member equivalence classes, so that the
 DISCRIMINATION LEARNING AND STIMULUS EQUIVALENCE 129

total number of simple discriminations in- plicitly presented in baseline training. We

volved was quite small. noted previously that although sample-as-
Most experiments using sample-as-node node training (e.g., AB, AC) does not explic-
and comparison-as-node training to establish itly require discrimination of comparison
small equivalence classes with adults have stimuli from different trial types (e.g., B stim-
produced positive outcomes. Green (1990) uli from C stimuli) because those stimuli are
reported development of two auditory-visual never pitted directly against one another, they
as well as two all-visual equivalence classes of are presented successively across trials as pairs
three members each in adults with mild men- of comparisons when training trials are
tal retardation following sample-as-node mixed. With small potential classes, this ex-
training. In another study using a sample-as- posure may suffice for acquisition of the B
node structure, three three-member classes versus C discriminations by some subjects.
were established in 15 of 16 normally capable The same possibility applies to linear-series
adults (Innis, Lane, Miller, & Critchfield, training for a small number of small classes.
1998). Pilgrim and Galizio (1990) also re- For example, the A stimuli may be discrimi-
ported positive outcomes in normally capable nated from the C stimuli when AB and BC
adults following sample-as-node training. trials are mixed during training. If so, positive
More recently, the same investigators report- test outcomes would seem equally likely fol-
ed establishing two four-member equivalence lowing training with either structure when
classes in normally capable adults using a the total number of simple discriminations
mixed comparison-as-node/sample-as-node involved is relatively small. Alternatively, pos-
training structure (Pilgrim & Galizio, 1995). itive outcomes might result from interspersing
Investigators using linear-series training test trials with training trials in test sessions.
structures designed to establish small equiv- Previously untrained simple discriminations
alence classes have reported mixed results. could develop over the course of testing due
Fields, Adams, Newman, and Verhave (1992) to (a) the additional exposure to training tri-
reported that immediately after linear-series als, (b) the juxtaposition of test trials that in-
training, positive outcomes on tests for all clude those discriminations with training tri-
properties of equivalence were evident for als, or (c) both. This might explain the
only 3 of 14 adults; repeated or specially de- gradual emergence of equivalence-consistent
signed testing was necessary to yield positive test performances that has been documented
outcomes in the other subjects. Similar re- in a number of studies (e.g., Lazar et al., 1984;
sults were obtained in the same laboratory Sidman et al., 1986; Spradlin et al., 1973).
with other adult subjects, and successful lin-
ear-series training and testing with one set of Studies Involving Larger Classes
stimuli was not always followed by positive With sample-as-node or linear-series train-
outcomes following linear-series training with ing designed to produce more than two po-
a second set of stimuli (Adams, Fields, & Ver- tential classes or classes with more than three
have, 1993; Buffington, Fields, & Adams, members, the number of new discriminations
1997; Fields et al., 1997). The studies by presented to subjects on tests for equivalence
Fields and colleagues employed consonant- is greater than with smaller classes (refer to
vowel-consonant trigrams as stimuli. In con- Table 5). Therefore, we would expect struc-
trast, three-choice linear-series training with ture-related differences in test performances
olfactor y stimuli, trigrams, and arbitrar y to be more likely or more marked when train-
forms (Annett & Leslie, 1995) and a mixture ing is designed to produce larger classes. Re-
of trigrams and haptic stimuli (Tierney, de sults of a series of studies by R. R. Saunders
Largy, & Bracken, 1995) produced positive and colleagues, discussed previously, are con-
results in most adult subjects. sistent with this prediction: In 23 of 28 young
In summary, the observation that sample- children and individuals with mental retar-
as-node and linear-series training established dation, four- and five-member equivalence
equivalence classes in some subjects suggests classes were established following compari-
that the successive discriminations that were son-as-node training; in only 3 of 13 subjects
called for on the tests for equivalence were were such classes established following sam-
established, even though they were not ex- ple-as-node training (Drake & Saunders,
130 RICHARD R. SAUNDERS and GINA GREEN

1987, cited in K. J. Saunders et al., 1993; K. presented in the one-node tests, the A versus
J. Saunders et al., 1993; R. R. Saunders et al., D and B versus E discriminations in the two-
1999; R. R. Saunders, Saunders, Kirby, & node tests, and the A versus E discriminations
Spradlin, 1988; R. R. Saunders, Wachter, & in the three-node tests. A typical test session
Spradlin, 1988; Spradlin & Saunders, 1986). intersperses test trials (AC, BD, CA, EC, etc.)
A study with adult subjects, in which five- among baseline trials (AB, BC, CD, DE) with
member equivalence classes were established each trial type appearing equally often. This
in only 2 of 12 cases following linear-series means that the B, C, and D stimuli will appear
training, also corroborates this prediction three times more often than either A stimulus
(Fields et al., 1995). and 50% more often than either E stimulus
Results of other studies, however, appear to on baseline trials during each test session. If
be inconsistent with our prediction. For ex- previously untrained discriminations develop
ample, Spradlin, Saunders, and Saunders over the course of testing, as we speculated
(1992) reported successful development of previously, then we hypothesize that the or-
two five-member classes in normally capable der in which those untrained discriminations
children immediately following linear-series are acquired will correspond to the frequency
training. Kennedy (1991) employed a of reexposure to particular stimuli on base-
‘‘branching’’ linear-series structure, or what line trials during testing. That is, one-node
might be considered a combination of linear- tests (BD and DB) should produce positive
series and sample-as-node training, to estab- results first. Other tests involving B, C, and D
lish three seven-member classes with normal- stimuli (i.e., one-node tests for AC, CA, CE,
ly capable adults. Gradual emergence of and EC and two-node tests for AD, DA, BE,
equivalence was reported for most subjects, and EB) should produce positive results next,
with multinode relations emerging last. Sim- and tests involving the A and E stimuli (i.e.,
ilar results were reported for typically devel- the three-node AE and EA tests) should be
oping children following sample-as-node the last to produce positive results. In other
training (K. J. Saunders et al., 1993). As we words, patterns of gradual emergence that
noted above and elsewhere (R. R. Saunders have been attributed to nodal or associative
& Green, 1992), simple discriminations that distance (e.g., Fields et al., 1990; Kennedy,
were not presented in training may be ac- 1991; Kennedy et al., 1994) may instead re-
quired over the course of testing, even in the flect gradual acquisition of simple discrimi-
absence of differential trial-by-trial conse- nations as a function of frequency of stimulus
quences. When there are a large number of presentation during testing.
these, acquisition may be more likely to occur To evaluate this possibility, we looked for
in normally capable adults and older children studies reporting nodal distance effects that
than in individuals with severe learning dif- conformed to the procedural assumptions
ficulties or young children. This could ac- underlying our analysis (simultaneous MTS
count for the results reported by Kennedy procedures, all baseline trials mixed before
(1991), K. J. Saunders et al. (1993), and testing, test-trial types for each property of
Spradlin et al. (1992). equivalence presented together within a ses-
sion, balanced MTS trials in training and test-
Gradual Emergence of Equivalence ing, and no differential consequences on test
Next, we pick up the thread of our earlier trials). Unfortunately, we found none. The
discussion about gradual emergence of equiv- seminal experiment on nodal distance (Fields
alence, this time in the context of linear-se- et al., 1990), for example, employed unbal-
ries training structures. As Table 5 shows, lin- anced test trials on which comparisons from
ear-series training (e.g., AB, BC, CD, DE) and different stimulus sets were mixed (e.g., CA
testing for two five-member equivalence clas- test trials had A and C stimuli as compari-
ses yield 24 simple discriminations that are sons). Kennedy (1991) also employed unbal-
not presented in training but that are called anced trial types. Other studies purporting to
for on test trials (e.g., A1 vs. C1, B2 vs. D2, show nodal distance effects had other meth-
C1 vs. E2, etc.). In the terminology of the odological features that made them unsuit-
nodal distance literature, the A versus C, B able for our discrimination analysis. For in-
versus D, and C versus E discriminations are stance, extensive pretesting and the use of
 DISCRIMINATION LEARNING AND STIMULUS EQUIVALENCE 131

stimuli (words) with which subjects had Positive outcomes on equivalence tests were
preexperimental histories likely confounded seen for 5 of the 6 instructed subjects but for
the results (Kennedy et al., 1994). Thus, new only 1 of 5 uninstructed subjects. A follow-up
experiments will be necessary to evaluate our experiment showed that when previously un-
hypothesis about gradual emergence. successful uninstructed subjects were given
instructed training with new stimuli, they too
Differential Responding passed equivalence tests. These results sug-
Exposure to test trials that involve novel gest that differential naming of stimuli by the
discriminations may give rise to behavior that experimenter fostered the establishment of
was not the direct product of training contin- equivalence classes in subjects with mental re-
gencies, but that may nonetheless foster ac- tardation, including those who received com-
quisition of the new discriminations. For parison-as-node training (K. J. Saunders et al.,
some subjects, preexperimental repertoires 1993). Demonstrations that equivalence clas-
may emerge in the presence of novel trial ses emerged more readily following sample-
types or arrangements, generating differen- as-node training with auditory samples than
tial responding. Among verbally sophisticated with visual samples may reflect similar pro-
humans, stimulus naming is a common form cesses (Green, 1990; Sidman et al., 1986).
of such behavior. Although perhaps not nec- That is, presenting auditory samples (names)
essary for untrained simple discriminations to in training may set the occasion for subjects
emerge, differential responding such as nam- to produce names for the other experimental
ing could foster the acquisition of those dis- stimuli, thereby fostering simple discrimina-
criminations and, therefore, the emergence tions among them. These possibilities warrant
of equivalence-consistent performances over direct empirical testing.
the course of testing (see Dugdale & Lowe, In a related recent study, R. R. Saunders et
1990; Eikeseth & Smith, 1992; Lowe & Beasty, al. (1999) provided preschool children with
1987; McIlvane & Dube, 1996; Sidman, ‘‘instructed’’ training like that provided in
1994). previous experiments in the same laboratory
McIlvane and Dube (1996) suggested that (i.e., differential experimenter-provided oral
naming or other differential responding may names for each stimulus on the first four
occur even when experimental procedures training trials). Some subjects had compari-
do not explicitly require it. Several proce- son-as-node training, and others had sample-
dures may promote naming. For instance, K. as-node training. The former required more
J. Saunders et al. (1993) noted that in prior trials for performance to reach the training
studies comparing sample-as-node with com- criterion than did the latter. Similar differ-
parison-as-node training structures (R. R. ences were reported by R. R. Saunders, Wach-
Saunders, Wachter, & Spradlin, 1988; Sprad- ter, and Spradlin (1988) and Fields et al. (in
lin & Saunders, 1986), subjects with mental press). Thus, across comparable experiments,
retardation were given unique names for preschool children, normal adults, and ado-
each stimulus in four ‘‘instructed’’ trials at lescents with mental retardation acquired
the very beginning of conditional discrimi- baseline conditional discriminations more
nation training. On those trials, the experi- rapidly in sample-as-node training than in
menter named each of the stimuli twice, but comparison-as-node training, but were less
did not do so thereafter, and the subjects likely to produce positive outcomes on equiv-
were never required to repeat the names. Pos- alence tests immediately after training. On
itive outcomes on equivalence tests were seen the other hand, as noted above, equivalence
for all subjects given comparison-as-node classes were not established in all the subjects
training but for only 1 subject given sample- with mental retardation who had comparison-
as-node training. In a partial replication, K. J. as-node training without instructions that in-
Saunders et al. (1993) exposed 11 subjects cluded differential names for the stimuli in
with mental retardation to comparison-as- the study by K. J. Saunders et al. (1993). Tak-
node training to establish two four-member en together, these results suggested to R. R.
equivalence classes. Six subjects received four Saunders et al. (1999) that instructions may
initial instructed trials with stimulus names enhance training-structure differences. Be-
provided by the experimenter, and 5 did not. cause comparison-as-node training presents a
132 RICHARD R. SAUNDERS and GINA GREEN

larger number of successive (sample) discrim- presented simultaneously (as comparisons)

inations than does sample-as-node training, it during training. The A, B, C, D, E, and F stim-
may be more likely to set the occasion for uli, on the other hand, were all presented
differential sample naming (at least by ver- successively (as samples) during training, so
bally skilled subjects). This might foster the symmetry test trials with those stimuli likely
development of simple discriminations, not posed little difficulty for the subjects. Because
just among the samples but also among sam- the GF symmetry tests constituted one sixth
ples and comparisons presented simulta- of all symmetry test trials in the study, de-
neously, which is then likely to produce pos- creases in speed on those trial types alone
itive outcomes on tests for the properties of could account for the slight overall differenc-
equivalence (cf. Sidman, 1994, pp. 413–414). es in speed of responding between baseline
In contrast, sample-as-node conditional dis- and symmetry trials. Indeed, it seems plausi-
criminations require fewer successive discrim- ble that the decrease in response speed on
inations than does comparison-as-node train- equivalence test trials that Spencer and Chase
ing, which might decrease the likelihood that attributed to nodal distance was instead a
subjects who are capable of naming stimuli function of the numbers and types of new,
will do so during training. This may in turn untrained discriminations presented on the
make them ill-prepared to perform the new various test trials. Further, it is possible that
discriminations called for on tests. Either or some of the simple discriminations that were
both of these possibilities may account for not explicitly presented in the conditional
training-structure differences reported in discriminations trained early in the linear se-
some stimulus equivalence studies (R. R. ries were nonetheless acquired over the
Saunders et al., 1999). course of subsequent training, whereas those
trained late in the series were not. The order
Response Speed and Verbal Self-Reports in which conditional discriminations were
A recent study used linear-series training trained was AB, BC, CD, DE, EF, FG. As we
(AB, BC, CD, DE, EF, FG) in an attempt to suggested earlier, simple discriminations
establish three seven-member equivalence among some stimuli that were never present-
classes with 12 college students (Spencer & ed together as samples within a session or as
Chase, 1996). The investigators reported that comparisons within trials (e.g., the B and D
accuracy of responding on tests for the prop- stimuli) might develop when those stimuli
erties of equivalence decreased with increas- are presented successively across trials (e.g.,
ing nodal distance for most subjects, but the in sessions mixing AB, BC, and CD training
effect was small and transient. In contrast, trials, as in the Spencer and Chase study).
speed of correct responding on tests of tran- This possibility could not arise, however, for
sitivity and combined tests of symmetry and stimuli introduced late in the linear series
transitivity was inversely related to nodal dis- (e.g., F and G stimuli). Our analysis would
tance for nearly all subjects, a relation that predict, therefore, that subjects might re-
was maintained with repeated testing. Sub- spond faster on two-node tests involving stim-
jects responded considerably faster on base- uli introduced early in the series (e.g., BD,
line and symmetry trials than on transitivity DB) than on two-node tests involving stimuli
and combined trials during test sessions, and introduced later (e.g., DG, GD), because the
somewhat faster on baseline trials than on former involve simple discriminations that
symmetry test trials. There was no overall dif- were more likely to have been acquired than
ference in speed of responding to transitivity the latter. Response speed on the trials with
and combined trials. stimuli from the middle of the training series
It is interesting to speculate whether the (BE, EB, CF, FC) should fall in between. All
differential response speeds reported by three-, four-, and five-node tests would involve
Spencer and Chase (1996) might have been some of the stimuli introduced late in the
a function of differential acquisition during training series, so they should be excluded
training of the simple discriminations re- from such an analysis.
quired on the tests. The GF symmetry tests, The differential response speeds reported
for example, required successive discrimina- by Spencer and Chase (1996) seem to paral-
tions among G stimuli, which had only been lel verbal self-reports of accuracy on symme-
 DISCRIMINATION LEARNING AND STIMULUS EQUIVALENCE 133

try and equivalence tests demonstrated in an- a large number of classes. The discrimination
other study. College students who were given analysis also suggests an alternative interpre-
sample-as-node training and testing to poten- tation of some findings in the stimulus equiv-
tially establish two four-member equivalence alence literature that have been attributed to
classes were asked to report verbally whether such variables as nodal distance. For exam-
they believed their test-trial responses were ple, we have suggested that differential re-
correct or incorrect (Lane & Critchfield, sponse speeds or latencies on various types of
1996). Symmetry and equivalence test results test trials may reflect differential acquisition
(i.e., MTS performances) were highly accu- of the component simple discriminations that
rate. Self-reported evaluations of accuracy are due to training structure. Unlike nodal
ranged from 96% to 100% on symmetry tests distance, this is a characteristic of experimen-
and 79% to 100% on equivalence tests. That tal procedures that relates to familiar behav-
is, despite nonverbal responses on equiva- ioral principles.
lence test trials that were highly consistent Further, some specific, testable predictions
with equivalence, verbal reports seemed to re- follow from our main hypothesis:
flect some uncertainty about the accuracy of 1. Sample-as-node training should yield less
those responses, more so than on symmetry accurate performances on tests for symmetry
tests. Lane and Critchfield’s results are diffi- than does linear-series training, provided
cult to interpret, however, because of the class size and number of classes are equated.
composition of the training trials. Although 2. Following sample-as-node training, sym-
training was designed to establish only two metry tests present successive discriminations
classes, three-choice MTS procedures were among former comparison stimuli serving as
employed instead of balanced two-choice pro- samples, and simultaneous discriminations
cedures. This meant that stimuli from differ- among former samples serving as compari-
ent stimulus sets were mixed as comparisons: sons. Positive outcomes on these symmetry
AB trials had the C stimuli as the third com- tests entail demonstration of the simple suc-
parisons, AC trials included D comparisons, cessive and simultaneous discriminations nec-
and AD trials had B comparisons. Thus, un- essary for positive results on equivalence tests.
like sample-as-node training with standard Thus, following sample-as-node training, if
procedures in which the B, C, and D stimuli subjects are given equivalence tests only after
are never juxtaposed within trials, Lane and producing positive outcomes on the symme-
Critchfield’s training procedures might have try tests, all test outcomes are likely to be pos-
established simple discriminations among all itive. If equivalence tests are given first, how-
the experimental stimuli prior to testing. Fur- ever, positive outcomes are less likely.
ther research will be necessary to determine 3. Any procedure that leads to differential
how such unbalanced trial configurations responding to each of the stimuli in the ex-
might influence the numbers and types of periment, whether explicitly arranged by the
simple discriminations established in train- experimenter or arising from subjects’ preex-
ing. perimental histories, should establish simple
discriminations among all the stimuli, there-
by mitigating the differential effects of train-
SUMMARY AND ing structures.
CONCLUSION Although our review suggests that pub-
We have suggested, based on an analysis of lished research on stimulus equivalence pro-
the simple discriminations that make up the vides some support for the discrimination
trained and tested conditional discrimina- analysis of training-structure effects, the evi-
tions in typical stimulus equivalence experi- dence is neither overwhelmingly confirma-
ments, that sample-as-node and linear-series tory nor discomfirmatory. This may reflect
training are less likely to produce positive the complexities of equivalence research at
outcomes on all tests for the properties of least as much as the validity of our analysis.
equivalence than is comparison-as-node train- In this now sizable body of research, proce-
ing. These differential outcomes should be dural and subject variability is so great that it
more likely or more marked when training is is very difficult to make comparisons across
designed to establish relatively large classes or studies or to draw general conclusions. For
134 RICHARD R. SAUNDERS and GINA GREEN

example, we suggested that experiments de- more difficult discriminations (i.e., more suc-
signed to produce small numbers of small cessive discriminations among samples) than
equivalence classes should produce positive sample-as-node training, it seems plausible
outcomes regardless of training structure, be- that the stability of the baseline performances
cause of their relatively low discrimination de- engendered by the two training structures dif-
mands in terms of both the total number and fered at the point at which the isolated test
the types of component simple discrimina- trials were presented. This might account for
tions required. Yet Fields and his colleagues the smaller number of subjects in whom
have consistently reported negative outcomes equivalence classes were established following
from such experiments, even though their comparison-as-node training than following
subjects were normally capable adults who sample-as-node training.
should not have had difficulty learning the A more recent study by the same investi-
requisite discriminations (e.g., Fields et al., gators poses some interesting challenges to
1992). One striking difference between those our discrimination analysis. Holth and Arntz-
experiments and many others in the basic en (1998) reported that different mixtures of
stimulus equivalence literature is that the arbitrary stimuli and familiar stimuli (e.g.,
stimuli were printed trigrams rather than ar- pictures of common objects) in linear-series
bitrary forms or sounds. This suggests that training structures (AB, BC) produced differ-
specific characteristics of the stimuli might ac- ent results with normally capable adult sub-
count for the high rate of equivalence test jects. Recall that this structure yields test trials
failures in the Fields laboratory. In particular, for transitivity (AC) and equivalence (CA)
the presence of some identical or physically that require both simultaneous and succes-
similar letters among trigrams might make sive discriminations among the A and C stim-
for especially difficult simple discriminations uli not presented in training (see Table 1 and
among experimental stimuli that we contend Figure 4). When familiar pictures constituted
are necessary for positive equivalence out- the A and C stimulus sets, the B set, or all
comes. We have not reanalyzed the experi- three sets, positive equivalence test outcomes
ments from the Fields laboratory for this pos- were produced more often than when only
sibility, but our analysis suggests it may be a the A or the C stimuli were familiar pictures,
plausible explanation for the reported equiv- or when all the stimuli were arbitrary. En-
alence failures (and see K. J. Saunders et al., hanced outcomes with familiar pictures in
1993). both the A and C positions or in all three
Another study that appears to contradict positions in the linear series are entirely con-
our analysis was reported recently by Arntzen sistent with our analysis: Subjects presumably
and Holth (1997). Their data are consistent could discriminate among all those stimuli
with our contention that linear-series training before the experiment began, so they should
is not very likely to produce positive equiva- have had no difficulty making the simple dis-
lence outcomes, but are inconsistent with our criminations called for on the AC and CA
predictions about the outcomes of sample-as- tests. When just the A stimuli or the C stimuli
node and comparison-as-node training struc- were familiar to subjects prior to the experi-
tures. Following training to establish two ment, there may not have been enough pre-
three-member classes, Arntzen and Holth re- existing simple discriminations to carry them
ported positive test outcomes in fewer sub- through the AC and CA tests. It is not readily
jects following comparison-as-node training apparent to us, however, why equivalence test
than sample-as-node training, and in even outcomes were also enhanced when just the
fewer following linear-series training. As R. R. B stimuli, but not the A and C stimuli, were
Saunders et al. (1999) noted, however, these familiar to the subjects (i.e., discriminated)
investigators conducted their tests in isola- when they entered the experiment, because
tion, that is, in blocks of test trials rather than the B stimuli did not appear at all on the tests
with test trials interspersed among training for transitivity and equivalence. Whatever the
trials. This raises a question as to how well the explanation, the Holth and Arntzen experi-
trained conditional relations were main- ment suggests an interesting approach to ex-
tained during testing. Given the observation amining simple discrimination effects within
that comparison-as-node training arranges various training structures.
 DISCRIMINATION LEARNING AND STIMULUS EQUIVALENCE 135

We hope the analysis presented here will Devany, J. M., Hayes, S. C., & Nelson, R. O. (1986).
prompt research to resolve some of the ap- Equivalence class formation in language-able and lan-
guage-disabled children. Journal of the Experimental
parent inconsistencies in the stimulus equiv- Analysis of Behavior, 46, 243–257.
alence literature, particularly inconsistencies Dube, W. V., McIlvane, W. J., & Green, G. (1992). An
that appear to be the bases for ongoing de- analysis of generalized identity matching-to-sample
bates about such issues as naming, gradual test procedures. The Psychological Record, 42, 17–28.
Dugdale, N., & Lowe, C. F. (1990). Naming and stimulus
emergence, nodal or associative distance, and equivalence. In D. E. Blackman & H. Lejeune (Eds.),
of course, training-structure effects. We do Behavior analysis in theory and practice: Contributions and
not mean to suggest that all of the reported controversies (pp. 115–138). Hillsdale, NJ: Erlbaum.
differences in stimulus equivalence outcomes Eikeseth, S., & Smith, T. (1992). The development of
can be accounted for by differences in the functional and equivalence classes in high-function-
ing autistic children: The role of naming. Journal of
simple discriminations presented in the vari- the Experimental Analysis of Behavior, 58, 123–133.
ous training structures; other potential sourc- Fields, L., Adams, B. J., Newman, S., & Verhave, T.
es of extraneous stimulus control should be (1992). Interactions among emergent relations dur-
examined as well. We do mean to suggest that ing equivalence class formation. The Quarterly Journal
the role of simple discriminations as inherent of Experimental Psychology, 45B, 125–138.
Fields, L., Adams, B. J., & Verhave, T. (1993). The effects
components of conditional discriminations of equivalence class structure on test performances.
has been underappreciated, and that careful The Psychological Record, 43, 697–712.
attention to this factor can lead to improved Fields, L., Adams, B. J., Verhave, T., & Newman, S.
experimental designs for stimulus equiva- (1990). The effects of nodality on the formation of
lence work. equivalence classes. Journal of the Experimental Analysis
of Behavior, 3, 345–358.
Fields, L., Hobbie, S. A., Adams, B. J., & Reeve, K. F. (in
press). Effects of training directionality and class size
REFERENCES on equivalence class formation by adults. The Psycho-
logical Record.
Adams, B. J., Fields, L., & Verhave, T. (1993). Formation Fields, L., Landon-Jimenez, D. V., Buffington, D. M., &
of generalized equivalence classes. The Psychological Adams, B. J. (1995). Maintained nodal distance ef-
Record, 43, 553–566. fects in equivalence classes. Journal of the Experimental
Annett, J. M., & Leslie, J. C. (1995). Stimulus equiva- Analysis of Behavior, 64, 129–145.
lence classes involving olfactory stimuli. The Psycholog- Fields, L., Reeve, K. F., Rosen, D., Varelas, A., Adams, B.
ical Record, 45, 439–450. J., Belanich, J., & Hobbie, S. A. (1997). Using the
Arntzen, E., & Holth, P. (1997). Probability of stimulus simultaneous protocol to study equivalence class for-
equivalence as a function of training design. The Psy- mation: The facilitating effects of nodal number and
chological Record, 47, 309–320. size of previously established equivalence classes. Jour-
Barnes, D. (1994). Stimulus equivalence and relational nal of the Experimental Analysis of Behavior, 67, 367–389.
frame theory. The Psychological Record, 44, 91–124.
Fields, L., & Verhave, T. (1987). The structure of equiv-
Barnes, D., Browne, M., Smeets, P., & Roche, B. (1995).
alence classes. Journal of the Experimental Analysis of Be-
A transfer of functions and a conditional transfer of
havior, 48, 317–332.
functions through equivalence relations in three- to
Fields, L., Verhave, T., & Fath, S. (1984). Stimulus equiv-
six-year-old children. The Psychological Record, 45, 405–
alence and transitive associations: A methodological
430.
Barnes, D., McCullagh, P. D., & Keenan, M. (1990). analysis. Journal of the Experimental Analysis of Behavior,
Equivalence class formation in non-hearing impaired 42, 143–157.
children and hearing impaired children. The Analysis Green, G. (1990). Differences in development of visual
of Verbal Behavior, 8, 19–30. and auditory-visual equivalence relations. American
Brady, N. C., & Saunders, K. J. (1991). Considerations Journal on Mental Retardation, 95, 260–270.
in effective teaching of object-to-symbol matching. Green, G., & Saunders, R. R. (1998). Stimulus equiva-
Augmentative and Alternative Communication, 7, 112– lence. In K. A. Lattal & M. Perone (Eds.), Handbook
116. of research methods in human operant behavior (pp. 229–
Buffington, D. M., Fields, L., & Adams, B. J. (1997). En- 262). New York: Plenum Press.
hancing equivalence class formation by pretraining of Holth, P., & Arntzen, E. (1998). Stimulus familiarity and
other equivalence classes. The Psychological Record, 47, the delayed emergence of stimulus equivalence or
69–96. consistent nonequivalence. The Psychological Record, 48,
Carrigan, P. F., & Sidman, M. (1992). Conditional dis- 81–110.
crimination and equivalence relations: A theoretical Innis, A., Lane, S. D., Miller, E. R., & Critchfield, T. S.
analysis of control by negative stimuli. Journal of the (1998). Stimulus equivalence: Effects of a default-re-
Experimental Analysis of Behavior, 58, 183–204. sponse option on emergence of untrained stimulus
Carter, D. E., & Eckerman, D. A. (1975). Symbolic relations. Journal of the Experimental Analysis of Behavior,
matching by pigeons: Rate of learning complex dis- 70, 87–102.
criminations predicted from simple discriminations. Johnson, C., & Sidman, M. (1993). Conditional discrim-
Science, 187, 662–664. ination on equivalence relations: Control by negative
136 RICHARD R. SAUNDERS and GINA GREEN

stimuli. Journal of the Experimental Analysis of Behavior, lence of behavioral and mathematical equivalence.
59, 333–347. Journal of the Experimental Analysis of Behavior, 57, 227–
Kelly, S., Green, G., & Sidman, M. (1998). Visual identity 241.
matching and auditory-visual matching: A procedural Saunders, R. R., Saunders, K. J., Kirby, K. C., & Spradlin,
note. Journal of Applied Behavior Analysis, 31, 237–243. J. E. (1988). The merger of equivalence classes by
Kennedy, C. H. (1991). Equivalence class formation in- unreinforced conditional selection of comparison
fluenced by the number of nodes separating stimuli. stimuli. Journal of the Experimental Analysis of Behavior,
Behavioural Processes, 24, 219–245. 50, 145–162.
Kennedy, C. H., Itkonen, T., & Lindquist, K. (1994). No- Saunders, R. R., Wachter, J. A., & Spradlin, J. E. (1988).
dality effects during equivalence class formation: An Establishing auditory stimulus control over an eight-
extension to sight-word reading and concept devel- member equivalence class via conditional discrimina-
opment. Journal of Applied Behavior Analysis, 27, 673– tion procedures. Journal of the Experimental Analysis of
683. Behavior, 49, 95–115.
Lane, S. D., & Critchfield, T. S. (1996). Verbal self-re- Sidman, M. (1971). Reading and auditory-visual equiv-
ports of emergent relations in a stimulus equivalence alences. Journal of Speech and Hearing Research, 14, 5–
procedure. Journal of the Experimental Analysis of Behav- 13.
ior, 65, 355–374. Sidman, M. (1986). Functional analysis of emergent ver-
Lazar, R. M., Davis-Lang, D., & Sanchez, L. (1984). The bal classes. In T. Thompson & M. D. Zeiler (Eds.),
formation of visual stimulus equivalences in children. Analysis and integration of behavioral units (pp. 213–
Journal of the Experimental Analysis of Behavior, 41, 251– 245). Hillsdale, NJ: Erlbaum.
266. Sidman, M. (1987). Two choices are not enough. Behav-
Lowe, C. F., & Beasty, A. (1987). Language and the ior Analysis, 22, 11–18.
emergence of equivalence relations: A developmental Sidman, M. (1994). Equivalence relations and behavior: A
study. Bulletin of the British Psychological Society, 40, A42. research story. Boston: Authors Cooperative.
McIlvane, W. J., & Dube, W. V. (1992). Stimulus control Sidman, M., Kirk, B., & Willson-Morris, M. (1985). Six-
shaping and stimulus control topographies. Behavior member stimulus classes generated by conditional-dis-
Analysis, 15, 89–94. crimination procedures. Journal of the Experimental
McIlvane, W. J., & Dube, W. V. (1996). Naming as a fa- Analysis of Behavior, 43, 21–42.
cilitator of discrimination. Journal of the Experimental Sidman, M., Rauzin, R., Lazar, R., Cunningham, S., Tail-
Analysis of Behavior, 65, 267–272. by, W., & Carrigan, P. (1982). A search for symmetry
Michael, R. L., & Bernstein, D. J. (1991). Transient ef- in the conditional discriminations of rhesus monkeys,
fects of acquisition history on generalization in a baboons, and children. Journal of the Experimental Anal-
matching-to-sample task. Journal of the Experimental ysis of Behavior, 37, 23–44.
Analysis of Behavior, 56, 155–166. Sidman, M., & Tailby, W. (1982). Conditional discrimi-
O’Leary, C. A., & Bush, K. M. (1996). Stimulus equiva- nations vs. matching-to-sample: An expansion of the
lence in the tactile modality. The Psychological Record, testing paradigm. Journal of the Experimental Analysis of
46, 509–517. Behavior, 37, 5–22.
Pilgrim, C., Chambers, L., & Galizio, M. (1995). Reversal Sidman, M., Willson-Morris, M., & Kirk, B. (1986).
of baseline relations and stimulus equivalence: II. Matching-to-sample procedures and the development
Children. Journal of the Experimental Analysis of Behavior, of equivalence relations: The role of naming. Analysis
63, 239–254. and Intervention in Developmental Disabilities, 6, 1–19.
Pilgrim, C., & Galizio, M. (1990). Relations between Spencer, T. J., & Chase, P. N. (1996). Speed analyses of
baseline contingencies and equivalence probe perfor- stimulus equivalence. Journal of the Experimental Anal-
mances. Journal of the Experimental Analysis of Behavior, ysis of Behavior, 65, 643–659.
54, 213–224. Spradlin, J. E., Cotter, V. W., & Baxley, N. (1973). Estab-
Pilgrim, C., & Galizio, M. (1995). Reversal of baseline lishing a conditional discrimination without direct
relations and stimulus equivalence: I. Adults. Journal training: A study of transfer with retarded adolescents.
of the Experimental Analysis of Behavior, 63, 225–238. American Journal of Mental Deficiency, 77, 556–566.
Saunders, K. J., Saunders, R. R., Williams, D. C., & Sprad- Spradlin, J. E., Saunders, K. J., & Saunders, R. R. (1992).
lin, J. E. (1993). An interaction of instructions and The stability of equivalence classes. In S. C. Hayes &
training design on stimulus class formation: Extend- L. J. Hayes (Eds.), Understanding verbal relations (pp.
ing the analysis of equivalence. The Psychological Record, 29–42). Reno, NV: Context Press.
43, 725–744. Spradlin, J. E., & Saunders, R. R. (1986). The develop-
Saunders, K. J., & Spradlin, J. E. (1989). Conditional ment of stimulus classes using match-to-sample pro-
discrimination in mentally retarded adults: The effect cedures: Sample classification versus comparison clas-
of training the component simple discriminations. sification. Analysis and Intervention in Developmental
Journal of the Experimental Analysis of Behavior, 52, 1–12. Disabilities, 6, 41–58.
Saunders, K. J., & Spradlin, J. E. (1993). Conditional Tierney, K. J., de Largy, P., & Bracken, M. (1995). For-
discrimination in mentally retarded subjects: Pro- mation of equivalence classes incorporating haptic
gramming acquisition and learning set. Journal of the stimuli. The Psychological Record, 45, 431–437.
Experimental Analysis of Behavior, 60, 571–585. Urcuioli, P. J., & Zentall, T. R. (1993). A test of compar-
Saunders, R. R., Drake, K. M., & Spradlin, J. E. (1999). ison-stimulus substitutability following one-to-many
Equivalence class establishment, expansion, and mod- matching by pigeons. The Psychological Record, 43, 745–
ification in preschool children. Journal of the Experi- 759.
mental Analysis of Behavior, 71, 195–214. Urcuioli, P. J., Zentall, T. R., Jackson-Smith, P., & Steirn,
Saunders, R. R., & Green, G. (1992). The nonequiva- J. N. (1989). Evidence for common coding in many-
 DISCRIMINATION LEARNING AND STIMULUS EQUIVALENCE 137

to-one matching: Retention, intertrial interference, Williams, D. C., Saunders, K. J., Saunders, R. R., & Sprad-
and transfer. Journal of Experimental Psychology: Animal lin, J. E. (1995). Unreinforced conditional selection
Behavior Processes, 15, 264–273. within three-choice conditional discriminations. The
Wetherby, B., Karlan, G. R., & Spradlin, J. E. (1983). The Psychological Record, 45, 613–627.
development of derived stimulus relations through
training in arbitrary-matching sequences. Journal of the Received January 23, 1998
Experimental Analysis of Behavior, 40, 69–78. Final acceptance March 29, 1999
 ERRATUM

Weiss, S. J., & Panlilio, L. V. (1999). Blocking a selective association in pigeons.

JEAB, 71, 13–24. On p. 15, the frequency of the tone is given as 400 Hz when
it should be 440 Hz.

138