Cross‑domain analogies as relating derived relations among two separate relational networks
Authors
Ruiz, F. J., Luciano, C.
Journal
Journal of the Experimental Analysis of Behavior
Abstract
Two experiments show that participants can establish analogies between two separate relational networks; performance on the experimental task correlated with standard measures of analogical reasoning. Results support that analogy can be produced by relating derived relations between different networks (RFT).
Detailed Summary
Context and objectives
Analogy is broadly described as the relating of two situations based on sharing a common pattern of relationships among their constituent elements. From cognitive approaches, empirical studies have described many features of this type of reasoning, but these accounts do not specify the history needed for developing an analogical reasoning repertoire and are not focused on isolating analogical reasoning from other types of responding. From Relational Frame Theory (RFT), analogy involves establishing a relation of coordination among common types of trained or derived relations. Previous studies within this framework had analyzed within-domain analogies (i.e., relating relations within the same relational network). However, no published study had analyzed cross-domain analogies (i.e., relating derived relations among separate relational networks). Cross-domain analogies are particularly relevant because they appear to be more difficult to produce and recognize, have greater generative potential, and are common in everyday life and applied settings such as psychotherapy. This study had two main objectives: (1) to design a procedure allowing the recognition of analogical relations between two separate relational networks on the first attempt, and (2) to correlate participants' performance with a standard measure of analogical reasoning to evaluate the ecological validity of the model.
Method
Participants
Experiment 1 included 12 undergraduate students from the University of Almería (age range = 18–24 years; 8 females, 4 males). Experiment 2 included 10 undergraduate students from the same university (age range = 20–26 years; 6 females, 4 males). No participants had previous experience with stimulus equivalence research or knowledge of RFT literature. They received a canteen voucher exchangeable for breakfast or a snack.
Design
Both experiments consisted of two parts. Part I comprised two standardized intelligence tests: the D-70 (Dominoes Test; a nonverbal general intelligence measure) and the DAT-5 VR (Verbal Reasoning Scale of the Differential Aptitude Test; a 40-item analogical reasoning test). Part II was the computerized experimental task, programmed in Visual Basic 6.0, comprising five phases: (1) Series 0 Conditional Discrimination Training (nonsense syllables: 3-member equivalence classes A-B-C); (2) Relating Trained Relations Test (similar-similar and different-different trials with compound stimuli); (3) Series 1 and 2 Conditional Discrimination Training (abstract shapes and Greek letters: two separate relational networks of 3 equivalence classes with 2 members each); (4) Series 1 & 2 Combinatorial Relations Test (Experiment 1 only); and (5) the critical Analogical Test, consisting of two parts: relating combinatorial relations of sameness and relating combinatorial relations of distinction between the two relational networks. Phase 4 was eliminated in Experiment 2.
Materials and stimuli
Visual stimuli were black shapes framed in white 6 × 6 cm squares on a grey background. Three stimulus series were used: Series 0 (9 nonsense syllables), Series 1 (9 abstract shapes), and Series 2 (9 Greek letters). Series 1 and 2 stimuli were physically dissimilar to prevent relation establishment based on nonarbitrary properties. The procedure used simultaneous matching-to-sample with 3 comparisons.
Procedure
Participants first completed the intelligence tests (approximately 1 hour) and then the computerized task (approximately 50 minutes). In Phase 1 (Series 0), 6 conditional discriminations were trained using matching-to-sample with feedback (CORRECT/WRONG). Participants 1–9 received sequential training and participants 10–12 received errorless training. In Phase 2, the discrimination of relevant features of the analogical test format was assessed through 8 trials without feedback (4 similar-similar and 4 different-different) with compound stimuli. Movement feedback was implemented to facilitate discrimination. In Phase 3, 4 conditional discriminations were trained with Series 1 stimuli (F1G1, F2G2, F1H1, F2H2) and 4 with Series 2 stimuli (M1N1, M2N2, M1O1, M2O2). In Phase 4 (Experiment 1 only), combinatorial (equivalence) relations of both series were tested without feedback. In Phase 5 (Analogical Test), the first part presented 4 similar-similar trials (the sample was a combinatorial relation of sameness from Series 1 and comparisons were sameness and distinction relations from Series 2) and the second part presented 4 different-different trials (the sample was a distinction relation from Series 1 and comparisons were sameness and distinction relations from Series 2). The passing criterion was responding correctly to all trials.
Analyses
Individual results for each participant across phases were analyzed descriptively, including number of trials, percentages of correct responses, and number of attempts. Pearson correlations were calculated between standardized test scores (D-70 and DAT-5 VR) and performance on the Relating Trained Relations Test. The Wilcoxon test was used to compare errors between similar-similar and different-different trials.
Results
Experiment 1. All participants completed Series 0 training. D-70 scores ranged from 21 to 37 (M = 29.82, SD = 4.57) and DAT-5 VR scores from 19 to 33 (M = 24.36, SD = 4.03). Eleven of 12 participants passed the Relating Trained Relations Test; only 1 on the first attempt (M attempts = 4.9). The number of attempts strongly correlated with DAT-5 VR scores (r = −.78, p = .008) but not significantly with D-70 scores (r = −.28, p = .43). Participants made significantly more errors on different-different trials (M = 3, SD = 1.78) than on similar-similar trials (M = .91, SD = 1.37; Z = −2.42, p = .007). Ten of 11 participants passed the Combinatorial Relations Test (9 on the first attempt). On the Analogical Test, all 10 participants who reached this phase passed it: 8 of 10 passed the first part (sameness) on the first attempt, and 8 of 10 passed the second part (distinction) on the first attempt. Overall, 6 of 10 participants passed both parts on the first attempt.
Experiment 2. All participants completed the training. D-70 scores ranged from 27 to 38 (M = 32.2, SD = 3.67) and DAT-5 VR scores from 16 to 31 (M = 24.3, SD = 5.69). All participants passed the Relating Trained Relations Test (5 on the first attempt; M attempts = 4.1). The number of attempts strongly correlated with DAT-5 VR (r = −.74, p = .014) but not with D-70 (r = −.09). On the Analogical Test (without prior exposure to the Combinatorial Relations Test), all 10 participants passed both parts: 7 of 10 passed the first part (sameness) on the first attempt, and 9 of 10 passed the second part (distinction) on the first attempt. Overall, 7 of 10 participants passed both parts on the first attempt.
Across both experiments, 65% of participants passed the complete analogical test on the first attempt.
Discussion and conclusions
This study was the first to demonstrate the derivation of cross-domain analogies within the functional-analytic approach to analogical reasoning based on RFT. Participants in both experiments related derived relations between two separate relational networks. Notably, in Experiment 2, participants recognized these analogies without being previously tested on the combinatorial relations of each network, providing a cleaner and more ecologically valid demonstration of the phenomenon, as people in everyday life frequently establish analogies across domains without having previously derived the relations involved. The strong correlation between the number of attempts to pass the Relating Trained Relations Test and DAT-5 VR scores (a standard measure of analogical reasoning) strengthens the ecological validity of this behavioral model. Interestingly, this correlation was specific to analogical reasoning (DAT-5 VR) and not to general intelligence (D-70), suggesting that performance on the Relating Trained Relations Test is functionally related to analogical reasoning abilities. The authors note as a limitation that the analogical test was brief (only 4 trials per part) and that future research should compare the production and recognition of within-domain versus cross-domain analogies and analyze the role of other shared features between relational networks.
Significance and contribution
This study contributes significantly to the functional analysis of analogical reasoning by experimentally demonstrating, for the first time, the derivation of cross-domain analogies (relating derived relations between separate relational networks) within the RFT paradigm. The procedure showed that most participants recognized these analogies on the first attempt—a result notable for its ecological validity since everyday analogies are frequently established without prior evaluation of the relations involved. The specific correlation with measures of analogical reasoning (but not general intelligence) provides evidence for the validity of this behavioral model of analogy, opening research lines on contextual variables controlling the generativity of analogical reasoning with implications for basic research on human cognition.
This summary was generated using Artificial Intelligence and may contain errors. Please refer to the original article.