Word diagrams in teaching classical conditioning.

An essential component of concept teaching is to present students with a clear specification of the structure of a concept, which consists of a statement of the critical features of a concept and how those features are related. Clear specifications of concept structure have been empirically validated in research (a) showing the value of concept definitions (Anderson & Kulhavy, 1972; Johnson & Stratton, 1966; Miller & Weaver, 1976) and (b) showing the value of analysis statements, explanations of how the concept structure is fully represented in examples and incompletely represented in nonexamples (Grant, McAvoy, & Keenan, 1982; Tennyson, Steve, & Boutwell, 1975). Whereas concept definitions describe the concept in abstract terms, analysis statements relate the words in a concept definition to concrete examples of a concept. For example, a definition of a square is a four-sided figure in which the equal sides intersect one another at right angles. In this context, the analysis statements would emphasize to stu dents how the verbal descriptors in this concept definition (e.g., "four-sided," "equal sides," "right angles") are present in specific examples of squares and are absent in nonexamples of squares (e.g., rectangles, nonsquare parallelograms). Applications of concept learning principles to instruction in behavior-analysis concepts have typically relied on standard text to present concept definitions and analysis statements to students (Grant, 1996; Miller, 1980; Peterson, 1978; Reese & Woolfenden, 1973). In the present experiment, the concept definition and the analysis statements were supplemented by diagrams. The purpose of the study was to determine if diagrams that represent concept definitions and examples improve student learning.

Much of the richness of the discipline of behavior analysis is in the study of complex behavior-environment relations or contingencies. This richness, however, poses one of the major difficulties in teaching behavior-analysis because complex relations between behavior and the environment are difficult to describe using prose definitions. Recognizing this problem, several observers have proposed diagramming systems for clarifying contingency relations (Goldwater & Acker, 1995; Hummel, Kaeck, Bowes, & Rittenhouse, 1994; Malott, 1992; Mechner, 1959; Snapper, Kadden, & Inglis, 1982). Relatedly, educators have also advocated using diagrams to represent complex behavior-environment relations in order to help students learn these concepts (Goldwater & Acker, 1995; Mattaini, 1995; Michael & Schafer, 1995).

The advocates of diagrams to teach behavior-analysis concepts have advanced persuasive arguments in favor of using diagrams that have considerable intuitive appeal. However, research into the effectiveness of diagrams in communication and teaching has been lacking. The present experiment examined the effectiveness of a simple word diagram in teaching classical conditioning. Basic word diagrams have long been used to illustrate classical conditioning in introductory psychology texts and texts in the psychology of learning (e.g., Keller & Schoenfeld, 1950, p. 19, p. 31), even though the use of these diagrams has never been empirically validated. In an informal survey of behavior-analysis textbooks, Goldwater and Acker (1995) expressed concern that recent textbooks have omitted diagrams.

In the present study, students used diagrams as part of a self-instructional concept-teaching program. The diagrams supplemented ordinary text as a means of showing how the definition applied to specific examples of classical conditioning.

Method

Participants

Sixty university students from introductory psychology participated to fulfill a course requirement. Their grades were independent of performance in the study. All participants complied with the experimental procedures.

General Procedure

When students arrived at the research site they obtained a materials package that included written directions. Initially, students read a 500-word introductory lesson on classical conditioning concepts. This lesson defined and exemplified unconditioned reflexes, conditioned reflexes, unconditioned stimuli (US), unconditioned responses (UR), neutral stimuli (NS), conditioned stimuli (CS), and conditioned responses (CR). The lesson also compared and contrasted classical conditioning with both pseudoconditioning and operant conditioning.

After reading the introductory lesson, the students read a conceptual exercise consisting of examples and nonexamples of classical conditioning. Each of the examples was based on a published account of human classical conditioning. One of the nonexamples illustrated operant conditioning and one illustrated pseudoconditioning. Beneath each item was an "analysis" that identified the item as an example or a nonexample. For examples, the analysis identified the US, UR, NS, CS, and CR. For nonexamples, the analysis explained why the item lacked the critical features of classical conditioning and when applicable identified the item as an instance of operant conditioning or pseudoconditioning.

After completing the conceptual exercise, students took a postest consisting of six novel (i.e., previously unseen) items, three examples of classical conditioning (Ellson, 1941; Razran, 1949; Vaitl, Gruppe, & Kimmel, 1985) and three nonexamples. One nonexample was an instance of operant conditioning, one was an instance of pseudoconditioning, and one was an instance of neither operant conditioning nor pseudoconditioning.

For each item, each student was required to classify the item as an example or a nonexample and to analyze the presence or absence of the critical features (i.e., US, UR, NS, CS, and CR). For the nonexamples each student was also asked to identify them as examples of operant conditioning or of pseudoconditioning. The posttest instructions did not specify whether the student should draw diagrams in answering the items.

Experimental Treatments and Design

There were two experimental treatments: Matched versus unmatched examples/nonexamples and the use of diagrams. In the matched condition, the conceptual exercise consisted of 14 example/nonexample pairs presented side by side. The examples (Bierley, McSweeney, & Vannieuwkerk, 1985; Cannon & Baker, 1981; Cannon, Best, Batson, & Feldman, 1983; Clarke & Hayes, 1984; Dekker, Pelser, & Groen, 1964; Doerr, 1981; Efron, 1964; Hayduk, 1980; Kasatkin & Levikova, 1932; McConaghy, 1970; Quarti & Renaud, 1964; Switzer, 1933; Watson & Rayner, 1920; Wolpe, 1958) illustrated classical conditioning. The matched nonexamples were modified versions of the examples changed so that they did not illustrate classical conditioning. Of the 14 nonexamples in the matched condition, 6 were examples of pseudoconditioning, 3 were examples of operant conditioning, and 5 were examples of neither pseudoconditioning nor operant conditioning. With minor modifications, the examples and nonexamples of classical conditioning used in the present st udy are included in Grant and Evans' (1994, pp. 412-417, pp. 506-512) self-instructional exercise over classical conditioning.

In the unmatched condition, the conceptual exercise consisted of 14 items, 10 examples and 4 nonexamples. The unmatched exercise was constructed by omitting the matched example or nonexample, as appropriate. One of the nonexamples illustrated operant conditioning and one illustrated pseudoconditioning. The design of the unmatched condition followed Miller and Weaver's (1976) concept-teaching recommendations, which designate that 30% of the items should be nonexamples. The comparison of the matched and unmatched treatments was included because the matched treatment permitted students to see 40% more diagrams than the unmatched treatment, enhancing any possible effects of diagrams.

In the diagram condition, the introductory lesson and each example in the conceptual exercise contained a diagram representing the classical conditioning. In the nondiagram condition, the diagram was omitted. Figure 1 represents the diagram used in the introductory lesson. The remaining diagrams were similar to Figure 1, differing only in the specific stimuli and responses contained in the example. As illustrated in Figure 1, the lesson emphasized the predictiveness of the CS in signaling the US as being the key factor in conditioning rather than temporal contiguity between the CS and the US.

Students were randomly assigned to one of the four conditions (15 participants per condition), formed by crossing lesson format (matched examples and nonexamples versus unmatched examples and nonexamples) with diagrams (present versus absent).

Interrater Reliability

Two independent raters scored all the posttests. For the classification task, the raters scored each item as correct or incorrect on the basis of whether the student had properly identified examples and nonexamples. For analysis-task examples, the raters individually scored the correctness of the student's identification of the US, UR, NS, CS, and CR. For analysis-task nonexamples, the raters scored the correctness of the student's identification of nonexamples as instances of operant conditioning, of pseudoconditioning, or of neither operant conditioning nor pseudoconditioning. If the first two raters disagreed, a third rater cast the deciding vote.

Interrater reliability was calculated by dividing the number of agreements by the number of agreements plus the number of disagreements. On the classification task, reliability of posttest grading was .99. On the analysis task, the reliability coefficient was 1.00 for nonexamples. On the analysis task, the mean reliability coefficient for examples ranged from .95 to .97 with an overall mean of .96

Results and Discussion

A 2 x 2 x 2 x 2 (diagrams/nondiagrams x matched/unmatched format x examples/nonexamples x classification/analysis) repeated-measures analysis of variance was conducted on the posttest data. The proportion of responses correct per item served as the dependent measure for purposes of data analysis. On the posttest, the students answered nonexample items (M = .76 correct responses per item) correctly more often than example items (M = .65 correct responses per item), F(1, 56) = 6.90, p < .05. In addition, the students correctly responded more often to the classification task (M = .825) than to the analysis task (M = .59), F(1, 56) = 112.56, p <.01.

The major significant result was the diagrams/nondiagrams x examples/nonexamples x classification/analysis interaction, F(1, 56) = 8.86, p < .01. Newman-Keuls tests indicated that students using diagrams correctly classified nonexamples (M = .93) more frequently than did students not using diagrams (M = .83), p < .01. In addition, students using diagrams correctly analyzed examples (M = .61) more frequently than did students not using diagrams (M = .49), p < .01.

Of the 30 students in the diagram group, 15 (drawers) happened to draw their own diagrams in answering the posttest example items and the other 15 (nondrawers) did not. In order to examine differences between these two groups, a 2 x 2 x 2 (drawers/nondrawers x examples/nonexamples x classification/analysis) repeated-measures analysis of variance was conducted. Students correctly answered classification-task items (M = .84) more often than they correctly answered the analysis-task items (M = .62), F(1,28) = 4.25, p < .05.

Drawers answered posttest items (M = .80) correctly more often than did nondrawers (M = .66), F(1, 28) = 6.30, p < .05. Also significant in this analysis was the drawers/nondrawers x examples/nonexamples x classification/analysis interaction, F(1, 28) = 8.93, p < .01. Newman-Keuls tests indicated that drawers correctly classified examples (M = .82) more often than did nondrawers (M = .69), p < .05. Drawers classified nonexamples correctly (M = 1.00) more often than did nondrawers (M = .87), p < .01. Finally, drawers correctly analyzed examples (M = .75) more often than did nondrawers (M = .47), p < .01.

To summarize, the major finding of this study was that diagrams of classical conditioning improved students' learning of that concept. On the posttest, students given diagrams (a) more often correctly classified novel nonexamples of classical conditioning and (b) more often correctly. analyzed novel examples of classical conditioning by being able to correctly identify the US, UR, NS, CS, and CR.

The beneficial effects of diagrams provide general support for advocates of the use of diagrams to teach behavior-analysis concepts (Goldwater & Acker, 1995; Malott, 1992; Mattaini, 1995; Michael & Schafer, 1995). Although there are differences in the proposed systems of diagrammatic representation, all the systems share the representation of temporal sequences of events in spatial dimensions on the printed page. Students in the present study benefited from relatively simple diagrams that illustrated, in spatial dimensions, the temporal sequences among the US, UR, CS and CR. Behavior-analysis procedures generally involve temporal sequences of events (e.g., the effects of antecedents and consequences on responses) and it may be that diagrams make these sequences easier to understand and to learn by representing them in spatial dimensions. In comparison to the systems for diagramming behavior analysis concepts (Goldwater & Acker, 1995; Hummel et al., 1994; Malott, 1992; Mechner, 1959; Snapper et al., 1982), the diagrams the students used in the present study were simple ones that required no specific knowledge of diagramming symbols. Students may stand to benefit even more from learning more sophisticated diagramming systems.

The diagrams used in the present study provide support for the general proposition that diagrams illustrating concept structure are a useful adjunct to concept teaching methods. Waddill, McDaniel, and Einstein (1988) suggest that research concerning the instructional effectiveness of visual aids should identify the specific contexts and conditions that are appropriate for presenting visual aids. The present study suggests that diagrams are appropriate at the definitional and example analysis stages of concept teaching. This finding is especially important because research and advice in concept teaching (Grant, 1986; Merrill, Tennyson, & Posey, 1992; Tennyson & Cocchiarella, 1986) has generally emphasized only standard prose concept definitions. It may be that word diagrams should often augment the use of standard prose in teaching concepts.

The beneficial effect of diagrams on the example-analysis task indicated that diagrammatic representation was particularly effective in teaching students to identify the relationships among the stimuli and responses in classical conditioning. The finding that diagrams improved nonexample classification indicates that diagrams were also particularly effective in reducing errors of overgeneralization or overextension, in which nonexamples of a concept are classified as examples.

The diagrams were of no help in the nonexample-analysis task, in which the students were required to specify whether nonexamples were instances of operant conditioning or pseudoconditioning. Because the diagrams did not represent these concepts, the diagrams could not be expected to improve student performance in analyzing nonexamples. However, the diagrams were also of no help in classifying examples, and they should have been of assistance if the diagrams had acted to improve the students' abilities to identify the components of classical conditioning. Students in the diagram group who drew diagrams in answering the posttest items were better at example classification than were diagram group students who did not draw diagrams. This result suggests that diagrams improved both example and nonexample classification for those students who actually made use of the diagrams. Although suggestive of additional benefits of diagrams, the correlational nature of the comparisons between the drawers and nondrawers makes it difficult to come to any firm conclusions. For example the drawers could have simply been more motivated, which led them to draw diagrams and do better on the posttest, without any necessary functional relationship between drawing and improved test performance.

The effectiveness of diagrams leads naturally to considerations concerning how teachers can and should implement diagrams in their written instructional materials, lectures, and web-based materials. Although this issue is beyond the scope of the present data, some rough guidelines have emerged from the author's work. First, because diagrams tend to focus the student's attention, using diagrams may be more effective to teach relatively difficult concepts like classical conditioning that students need to spend more time on than simpler concepts. Second, many diagrams illustrate temporal sequences of events and these types of diagrams lend themselves well to use in overheads in lectures because the instructor is able to point out subsections of the diagram and explain the component processes the diagram illustrates. This kind of step-by-step highlighting of parts of diagrams is also increasingly possible through computer-based and web-based instruction. Diagrams are particularly well suited to web-based instruct ion because they are relatively easy to read on a computer screen, unlike long passages of text. Third, as suggested by the comparisons of the drawers and nondrawers in the current study, there may well be important benefits in teaching and encouraging students to draw diagrams. Diagramming methods may enable students to organize and recall concepts and principles more effectively than traditional methods such as reading and rereading and reciting text.

References

ANDERSON, R. C., & KULHAVY, R. W. (1972). Learning concepts from definitions. American Educational Research Journal, 9, 385-390.

BIERLEY, C., MCSWEENEY, F. K., & VANNIEUWKERK, R. (1985). Classical conditioning of preferences for stimuli. Journal of Consumer Research, 12, 316-323.

CANNON, D. S., & BAKER, T. B. (1981). Emetic and electric shock alcohol aversion therapy: Assessment of conditioning. Journal of Consulting and Clinical Psychology, 49, 20-33.

CANNON, D. S., BEST, M. R., BATSON, J. D., & FELDMAN, M. (1983). Taste familiarity and apomorphine-induced taste aversions in humans. Behavior Research and Therapy, 21, 669-673.

CLARKE, J. C., & HAYES, K. (1984). Covert sensitization, stimulus relevance and the equipotentiality premise. Behavior Research and Therapy, 22, 451-454.

DEKKER, E., PELSER, H. E., & GROEN, J. (1964). Conditioning as a cause of asthmatic attacks; a laboratory study. In C. M. Franks (Ed.), Conditioning techniques in clinical practice and research (pp. 116-131). New York: Springer.

DOERR, H. O. (1981). Cognitive derivation of generalization stimuli: Separation of components. Bulletin of the Psychonomic Society, 17, 73-75.

EFRON, R. (1964). The conditioned inhibition of uncinate fits. In C. M. Franks (Ed.), Conditioning techniques in clinical practice and research (pp. 132143). New York: Springer.

ELLSON, D. G. (1941). Hallucinations produced by sensory conditioning. Journal of Experimental Psychology, 28, 1-20.

GOLDWATER, B. C., & ACKER, L. E. (1995). A descriptive notation system for contingency diagramming in behavior analysis. The Behavior Analyst, 18, 113-121.

GRANT, L. (1986). Categorizing and concept learning. In H. W. Reese & L. J. Parrott (Eds.), Behavior science: Philosophical, methodological, and empirical advances (pp. 139-1 62). Hillsdale, NJ: Lawrence Erlbaum.

GRANT, L. (1996). Positive reinforcement A self-instructional exercise. [WWW Document]. URL http://server.bmod.athabascau.ca/html/prtut/reinpair.htm

GRANT, L., & EVANS, A. E. (1994). Principles of behavior analysis. New York: Harper-Collins.

GRANT, L., MCAVOY, R., & KEENAN, J. B. (1982). Prompting and feedback variables in concept programming. Teaching of Psychology, 9, 173-177.

HAYDUK, A. W. (1980). Increasing hand efficiency at cold temperatures by training hand vasodilation with a classical conditioning-biofeedback overlap design. Biofeedback and Self-Regulation, 5, 307-326.

HUMMEL, J. H., KAECK, D. J., BOWES, R. L., & RITTENHOUSE, R. D. (1994). Diagramming operant processes. The ABA Newsletter, 17, 4-5.

JOHNSON, D. M., & STRATTON, R. P. (1966). Evaluation of five methods of teaching concepts. Journal of Educational Psychology, 57, 48-53.

KASATKIN, N. I., & LEVIKOVA, A. M. (1932). On the development of early conditioned reflexes and differentiations of auditory stimuli in infants. Journal of Experimental Psychology, 18, 1-19.

KELLER, F S., & SCHOENFELD, W. S. (1950). Principles of psychology. New York: Appleton-Century-Crofts.

MALOTT, R.W. (1992). Saving the world with contingency diagramming. The ABA Newsletter, 15, 45.

MATTAINI, M. A. (1995). Contingency diagrams as teaching tools. The Behavior Analyst, 18, 93-98.

MCCONAGHY, N. (1970). Penile response conditioning and its relationship to aversion therapy in homosexuals. Behavior Therapy, 1, 213-221.

MECHNER, F (1959). A notational system for description of behavioral processes. Journal of the Experimental Analysis of Behavior, 2, 133-150.

MERRILL, M. D., TENNYSON, R. D., & POSEY, L. O. (1992). Teaching concepts: An instructional design guide. Englewood Cliffs, NJ: Educational Technology Publications.

MICHAEL, J., & SHAFER, E. (1995). State notation for teaching about behavioral procedures. The Behavior Analyst, 18, 123-140.

MILLER, L. K. (1980). Principles of everyday behavior analysis (2nd ed.). Monterey, CA: Brooks/Cole.

MILLER, L. K., & WEAVER, F H. (1976). A behavioral technology for producing concept formation in university students. Journal of Applied Behavior Analysis, 9, 289-300.

PETERSON, N. (1978). An introduction to verbal behavior. Grand Rapids, MI: Behavior Associates.

QUARTI, C., & RENAUD, J. (1964). A new treatment of constipation by conditioning: A preliminary report. In C. M. Franks (Ed.), Conditioning techniques in clinical practice and research (pp. 219-227). New York: Springer.

RAZRAN, G. (1949). Sentential and propositional generalization of salivary conditioning to verbal stimuli. Science, 109, 447-448.

REESE, D. G., & WOOLFENDEN, R. M. (1973). Behavior analysis of everyday life: A program for the generalization of behavioral concepts. Kalamazoo, MI: Behaviordelia.

SNAPPER, A. G., KADDEN, R. M., & INGLIS, G. B. (1982). State notation of behavioral procedures. Behavior Research Methods and Instrumentation, 14, 329-342.

SWITZER, S. A. (1933). Disinhibition of the conditioned galvanic skin response. Journal of General Psychology, 9, 77-100.

TENNYSON, R. D., & COCCHIARELLA, M. J. (1986). An empirically based instructional design theory for teaching concepts. Review of Educational Research, 56, 40-71.

TENNYSON, R. D., STEVE, M. W., & BOUTWELL, R. E. (1975). Instance sequence and analysis of instance attribute representation in concept acquisition. Journal of Educational Psychology, 67, 821-827.

VAITL, D., GRUPPE, H., & KIMMEL, H. D. (1985). Contextual stimulus control of conditional vasomotor and electrodermal reactions to angry and friendly faces. The Pavlovian Journal of Biological Science, 20, 124-131.

WADDILL, P. J., MCDANIEL, M. A., & EINSTEIN, G. O. (1988). Illustrations as adjuncts to prose: A text-appropriate processing approach. Journal of Educational Psychology, 80, 457-464.

WATSON, J. B., & RAYNER, R. (1920). Conditioned emotional reactions. Journal of Experimental Psychology, 3, 1-20.

WOLPE, J. (1958). Psychotherapy by reciprocal inhibition. Stanford, CA: Stanford University Press.

Please address requests for reprints and other correspondence about this article to Lyle K. Grant, Psychology Centre, Athabasca University, Athabasca, Alberta, Canada T9S 3A3.