|
| Int'l Symposium - Experimental, Computational, and Observational Analyses of Complex Language Behavior |
| Monday, May 30, 2005 |
| 1:30 PM–2:50 PM |
| Stevens 4 (Lower Level) |
| Area: VBC; Domain: Basic Research |
| Chair: Denis P. O'Hora (University of Ulster) |
| Abstract: Behavioral accounts of complex language responding have sharply increased in complexity in recent years with the analysis of derived relational responding. In a landscape previously dominated by Skinner’s (1957) Verbal Behavior, novel behavioral approaches such as Relational Frame Theory (Hayes, Barnes-Holmes & Roche, 2001) have emerged. These new approaches ask new questions about human language behavior and to answer these new questions, new techniques are emerging. The current symposium includes four studies that employ a variety of techniques in the analysis of such behavior. The first two papers employ a basic experimental approach. The first paper examines interference effects that may occur due to functional equivalences between words in the English language, the second examines syntactic and semantic bootstrapping. The third paper employs computer modeling to examine productive sequential responding and highlights the advantages of this technique. The fourth paper introduces a time-dependent statistical technique called recurrence quantification analysis that will allow behavior analysts to deal with much larger corpora of behavior than typically possible in laboratory or applied settings. |
| |
| Orthographic and Phonological Interference in Derived Relational Responding |
| IAN THOMAS TYNDALL (American College, Dublin), Denis P. O'Hora (University of Ulster) |
| Abstract: In human language, words have visual (orthographic) and auditory (phonological) features that vary in similarity of function. Previous research suggests that the provision of formally (Stewart, Barnes-Holmes, Roche & Smeets, 2002) and functionally (Tyndall, Roche & James, 2004) similar incorrect comparisons may impede the establishment of equivalence classes. In the current study, four three-member equivalence classes were established using real words and then four conditions were employed during testing. In the first condition, incorrect comparisons (e.g., bug) were visually similar and rhymed with the sample (e.g., rug). In the second, incorrect comparisons (e.g., tow) rhymed with the sample (e.g., hoe) but were not visually similar. In the third, incorrect comparisons (e.g., rough) were visually similar to the sample but did not rhyme (e.g., bough). In the final condition, incorrect comparisons (e.g., bun) neither rhymed nor were visually similar to the sample (e.g., tip). Findings demonstrated that both orthogonal and phonological similarities interfered with demonstration of equivalence relations. These findings have implications for early language training and for a behavioral understanding of dyslexia. |
| |
| Semantic and Syntactic Bootstrapping: Experimental analyses |
| DENIS P. O'HORA (University of Ulster), Richard Dale (Cornell University) |
| Abstract: O’Hora and Dale (2004) outlined a behavioral model of syntactic and semantic boot strapping in terms of derived relational responding. The current paper reports two experiments in which such responding was examined. In Stage 1 of Experiment 1, participants were trained to choose an arbitrary shape (circle, square or triangle) in the presence of three nonsense syllables, A1, A2 and A3. In a second stage, participants were trained to choose the sequence, A1 then A2 then A3. Participants were then trained to choose B1 in the presence of A1, B2 in the presence of A2 and B3 in the presence of A3 and were tested for the derived sequence, B1 then B2 then B3. Novel stimuli were then included in this new derived sequence and acquired sequential functions by exclusion; N1 then B2 then B3, B1 then N2 then B3, B1 then B2 then N3. In a final test, participants were presented with N1, N2 and N3 and were required to choose one of the arbitrary shapes from Stage 1. A number of participants demonstrated this performance. Experiment 2 examined these performances in further detail. These procedures suggest alternative sources of productivity in early language training. |
| |
| Modeling Sequential Response Classes Sequentially: A Connectionist Approach |
| RICHARD DALE (Cornell University), Michael Spivey (Cornell University) |
| Abstract: Jordan (1989) and Elman (1990) devised neural network architectures for exploring behavior extended in time and are now widely applied in the cognitive sciences (Botvinick & Plaut, 2004; Christiansen & Chater, 1999; Gaskell, Hare & Marslen-Wilson, 1995). There is also a budding literature within behavior analysis on applying connectionist techniques to complex human behavior (e.g., Barnes & Hampson, 1994; Cullinan, Barnes, Hampson & Lyddy, 1994). These network architectures are particularly well suited to capture the properties of generalized sequential response classes (Green, Stromer & Mackay, 1991; Lazar, 1977; Wulfert & Hayes, 1986). In this paper, I present a series of simulations that produce generalized sequential responses in a simple recurrent network (Elman, 1990). By building models that map onto current theory and empirical findings, the model may serve a predictive role. For example, one may ask about limitations on the length of sequential response classes. Because experimenters are subject to considerable limitations on how long and persistently human participants can be trained, these models offer access to experimental circumstances not readily available to researchers. |
| |
| Uncovering Temporal and Structural Patterns in Complex Behavior: Categorical Recurrence Analysis |
| RICHARD DALE (Cornell University), Michael Spivey (Cornell University) |
| Abstract: In this paper, we review and extend a time-dependent statistical technique, recurrence quantification analysis (Zbilut & Webber, 1992), for observing patterns of behavior as they are organized in time. We adapt this technique with the goal of drawing general quantitative characterizations of complex behavior, and demonstrate its utility in studying language learning and structure. Initial analyses indicate this technique provides rich opportunity for investigating the development of language structure and performance. Despite the apparent noisiness of the child-caregiver interaction databases used, the analysis generates results consistent with previous research, and offers insight into the temporal organization of language acquisition in context. In particular, the temporal properties of the analysis reveal constraints on language learning unexplored in more common, static analyses (i.e., frame-based structural analyses, Mintz, 2003). This statistical method may be used across many areas of the behavioral sciences as a categorical technique for finding structure in units of measure at a higher time scale than what has yet been offered. These techniques are of particular interest to behavior analysts because of the explicit focus on the historical and dynamic nature of behavior. |
|
| |