Conditional discriminations, symmetry, and semantic priming☆
Introduction
Stimulus equivalence refers to the observation that, after having learned a few overlapping conditional relations among stimuli, human subjects will demonstrate a number of other conditional relations among those stimuli without direct training or reinforcement (Sidman et al., 1989, Sidman and Tailby, 1982). For example, having learned to match the stimulus A1 to stimulus B1 (and not B2 or B3) and having learned to match stimulus B1 to stimulus C1 (and not C2 or C3), human participants will readily match B1 to A1, C1 to B1, A1 to C1, and C1 to A1 – without training or reinforcement.
Sets of stimuli for which the above description holds true are called stimulus equivalence classes because they satisfy the requirements of equivalence relations as described in mathematical set theory which states that a relation of equivalence obtains among members of a set if it can be shown that the elements are related via reflexivity, symmetry, and transitivity (cf. Sidman, 1994). Laboratory generated equivalence classes have a number of interesting characteristics that have led researchers to suggest that they may serve as effective analogues of linguistic and other complex human performances. Notice, first, that replacing the non-representative forms that are typically used in equivalence studies with everyday stimuli (the spoken word ‘cat’, a picture of a cat, the written word [cat], etc.) immediately transforms the conditional relations that define equivalence classes into linguistically relevant performances. For example, trials in which pictures serve as sample stimuli and the written or spoken word serve as comparison stimuli are good analogues of picture comprehension and trials in which the written word serves as the sample stimulus and the picture and spoken word serve as comparisons are good analogues of word comprehension and reading, respectively.
Furthermore, research has shown that stimulus functions established for one member of an equivalence class will extend to other members of the equivalence class without any training or contingencies supporting such extension (e.g.; Catania et al., 1989, Dougher et al., 1994). For example, Dougher et al. first directly trained two overlapping conditional discriminations and documented the existence of 2, four-member equivalence classes. The experimenters then established a conditioned-startle reflex with one element (B1) of the existing equivalence class and tested for the extension of that stimulus function to other class members (C1 and D1). They found that the conditioned-stimulus functions of B1 readily transferred to other members (C1 and D1) of the equivalence classes. This observed extension of function also mimics an important feature of natural languages, namely, the ability of a word (or a collection of words) to serve as a symbol for object or events (generically, referents) in the world. Laboratory-generated equivalence classes, then, may have properties that mimic certain aspects of linguistic functioning and may, therefore, serve as useful analogs for the laboratory study of such complex behavior.
These definitional and extra-definitional properties of laboratory-generated equivalence classes have led behavioral researchers to suggest that laboratory-generated stimulus equivalence relations and naturally developing and expanding semantic networks may be related (Barnes and Hampson, 1993, Cullinan et al., 1994, Fields, 1987, Hayes and Hayes, 1992, Reese, 1991, Sidman, 1986). It is important, however, to ascertain the extent to which laboratory-generated equivalence relations have the properties of more naturally occurring linguistic phenomena. One such phenomenon, of great interest to linguists and cognitive psychologists, is the semantic-priming effect.
In general, priming refers to a behavioral change in responding to a stimulus as a function of previous exposure to the same or related stimuli (Voss et al., 2010, Voss et al., 2010Voss et al., 2010; Schacter, 1987; Richardson-Klavehn and Bjork, 1988). “Semantic priming” refers to the observation of priming effects with words as stimuli in human participants. For instance, participants are faster in reporting that two words are related to each other than they are to report that two words are not related to each other (e.g., Meyer and Schvaneveldt, 1971). The task typically used to assess semantic priming is called the Lexical Decision Task (LDT). A typical trial begins with the presentation of a prime stimulus (usually a spoken or printed word) followed quickly by a target stimulus (also usually a spoken or printed word). The participant is then required to report whether the words are related or not by pressing one of two buttons to indicate a ‘yes’ or ‘no’ response. Two main measures fall out of this task. One is a measure of the participant’s choices following the presentation of the target stimulus (e.g., whether the participant says “yes” or “no”). The second is a measure of the time the participant takes to make either response.
A common finding in these procedures is that participant's reaction times are systematically faster when they are accurately reporting “Yes” than when they are accurately reporting “No”. For example, if asked whether two words go together (say ‘YES’) or not (say ‘NO’), participants are faster to say “Yes” (or select Yes) when the words are “animal” and “tiger” than they are to say “No” (or select No) when the words are “animal” and “coffee”. The faster reaction times for related words are said to be the result of more efficient retrieval dynamics resulting from generalized activation of the semantic network by the presentation of the prime stimulus (e.g., Meyer and Schvaneveldt, 1971).
If equivalence relations are to serve as viable analogues or models of naturally occurring semantic networks (Hayes and Hayes, 1992, Sidman, 1986), it is important to ascertain whether stimuli within equivalence classes are effective as primes and targets relative to stimuli not related via equivalence. The results of two studies (Hayes and Bisset, 1998, Barnes-Holmes et al., 2005) suggest that equivalence relations among otherwise unrelated stimuli are sufficient to produce the semantic priming effect. For example, Barnes-Holmes et al. trained the prerequisites for and documented the existence of two 4-member equivalence classes using nonsense words. After training and testing was complete, participants completed a LDT utilizing the stimuli comprising the equivalence classes as well as novel stimuli. Across the three experiments the researchers found priming effects for stimuli in equivalence classes. An important contribution of the Barnes-Holmes et al. study was their use of procedures that were typical of the conventional research on semantic priming which makes the task of comparing the effects of other variables of interest easier.
The current study sought to expand the conditions under which the relation between equivalence classes and semantic priming are investigated. Toward that end, the first change, relative to earlier studies, was the use of non-representative forms as stimuli. The use of nonsense words as stimuli leaves open the possibility that factors like stimulus generalization (from real words) may contribute to the observed effects in unknown ways. By contrast, a priming effect observed with nonverbal and non-representational stimuli would permit a stronger test of equivalence relations as being sufficient for the priming effect. Voss et al. (2010) have partially demonstrated a semantic priming effect with geometrical shapes, but only with stimuli rated as ‘highly meaningful’ by participants, and without intraexperimentally establishing the semantic relations of interest.
Along similar lines, this study attempted to more clearly isolate the role of conditional discrimination training and testing on semantic priming by conducting the LDT both before and after conditional discrimination training and prior to testing for emergent symmetry. Presenting the LDT prior to any exposure to the stimuli and training conditions allowed us to see if there was any naturally-occurring priming in the stimulus relations of interest. It was assumed that these data would allow for a cleaner interpretation of the role of the programmed training and testing contingencies on any observed priming.
A third change in the training conditions was also designed to reduce unknown sources of variability in the priming effects of interest in the study. Several studies have shown that linear training structures can potentially introduce associative distance between members of an equivalence class (Fields et al., 1984, Fields et al., 1995; cf. Imam, 2001, Imam, 2006) and the effects of associative distances may involve reaction times in addition to the accuracy of the response (Bentall et al., 1998, Spencer and Chase, 1996; cf. Imam, 2006). In the current study we avoided this potential confound by using a one-to-many training structure and a many-to-one testing structure in which only symmetry (associative distance = 0) was assayed.
Section snippets
Participants
Six young adults (4 women, 2 men) were recruited from the University of North Texas to participate. The participants were recruited via flyers posted around campus and were selected on the basis of their availability and naïveté with respect to the terms and concepts of the experimental analysis of behavior. Participants were instructed to “do as well as they could”. All data collection for a participant occurred in a single meeting and each participant was given $10 for their involvement,
Results
All participants completed the 96-trial LDT-familiarization task without any problems. No data were collected during the LDT-familiarization task and nothing further will be said about the subjects’ performances during this task.
Discussion
Our primary question in this study was to see if laboratory-generated stimulus classes consisting of non-representational stimuli would show a priming effect. The semantic priming effect is a robust and reliable phenomenon mainly documented with words from natural languages used as stimuli. If the development of laboratory-generated stimulus classes and the development and expansion of naturally-occurring semantic networks are allied phenomena, it should be possible to relate the
References (26)
- et al.
Is the N400 effect a neurophysiological index of associative relationships
Neuropsychologia
(2013) - et al.
Finding meaning in novel geometric shapes influences electrophysiological correlates of repetition and dissociates perceptual and conceptual priming
Neuroimage
(2010) - et al.
Stimulus equivalence and connectionism: Implications for behavior analysis and cognitive science
Psychol. Rec.
(1993) - et al.
Derived stimulus relations, semantic priming, and event-related potentials: testing a behavioral theory of semantic networks
J. Exp. Anal. Behav.
(2005) - et al.
Errors and response latencies as a function of nodal number in five-member equivalence classes
Psychol. Rec.
(1998) - et al.
Transfer of function across members of an equivalence class
Anal. Verb. Behav.
(1989) - et al.
A transfer of explicitly and non-explicitly trained sequence responses through equivalence relations: An experimental demonstration and connectionist model
Psychol. Rec.
(1994) - et al.
The transfer of respondent eliciting and extinction functions through stimulus equivalence classes
J. Exp. Anal. Behavi.
(1994) - et al.
Match to Sample Program (Version 11.08a67) [Computer Software]
(1991) The structure of equivalence classes
J. Exp. Anal. Behav.
(1987)
Maintained nodal-distance effects in equivalence classes
J. Exp. Anal. Behav.
Stimulus equivalence and transitive associations: a methodological analysis
J. Exp. Anal. Behav.
Derived stimulus relations produce mediated and episodic priming
Psychol. Rec.
Cited by (4)
Intraexperimental development of priming and event related potentials effects with abstract geometrical stimuli
2020, Behavioural ProcessesCitation Excerpt :These results are consistent with the effects reported earlier. Vaidya, Hudgins, and Ortu (2015) found that conditional relations training was enough to produce a priming effect. The design of the study also allows a direct, within-subject comparison of the effect with words as stimuli and with non-words and finds only a difference in the degree of the effects reported.
Effects of different types of training models on matching-to-sample tasks
2016, Revista Latinoamericana de PsicologiaDo common elements predict class merger: A test of Sidman's theory of equivalence
2021, Journal of the Experimental Analysis of BehaviorSocial Categorization and Stimulus Equivalence: A Systematic Replication
2020, Psychological Record
- ☆
These data were collected in partial fulfillment of the requirements of a master’s thesis by the second author while at The University of North Texas. We would like to thank reviewers of an early version of the manuscript for their helpful comments. Caleb D. Hudgins in now at Rutgers University.