1. Assessing and Training Categorization Repertoires in Young Children
Teresa Mulhern
Dr Ian Stewart
National University of Ireland, Galway
This research was funded by the NUI Galway Doctoral Research Scholarship

2. Classification
Stimuli are said to be in a class when a common set of responses are emitted in their presence (Barnes-Holmes, Hayes, Dymond & O’Hora, 2001).
Perceptual classes (physical properties – e.g., Gelman & Meyer, 2011)
Associative classes (abstract – e.g., Galizio, Stewart & Pilgrim, 2001)
Natural language classes (both abstract and physical – e.g., Adams, Fields & Verhave, 1993)

3. Hierarchical Classification
In hierarchical classification, which is a relatively advanced form of classification, classes are themselves categorised into higher order classes
Hierarchical classification has garnered less research than that of basic classification
Classes are themselves categorised into higher order classes
Hierarchical classification has gained less research attention than that of basic classification

4. Hierarchical Classification
Modelled within Relational Frame Theory (RFT) as patterns of relational framing (containment and hierarchical framing)
Gil, Luciano, Ruiz & Valdivia-Salas (2012)
Slattery & Stewart (2014)

5. Learning Hierarchical Classification (Relational Frame Theory)
Containment (Non-Arbitrary)
E.g., “The water is in the glass. What does the glass contain?” (given container and contained materials)
Containment (Arbitrary)
E.g., “The coin is in the box. What does the box contain?” (no stimuli present)
Hierarchy (Arbitrary)
E.g., “A lion is a type of animal. Does the class ‘animals’ contain lions?”

6. Study 1
Measure patterns of relational framing linked with categorization in 50 young typically developing children (aged years)
Non-arbitrary containment
Arbitrary containment
Arbitrary hierarchy
Correlate framing performance with linguistic and cognitive performance

7. Study 1 Stanford Binet, 5th edition (SB-5)
Children’s Category Test (CCT)
Peabody Picture Vocabulary Test, 4th edition (PPVT-4)
Class Inclusion
Can children respond to a stimulus as simultaneously belonging both to a class, as well as to a subclass contained in it

8. Study 1: Relational Framing Test
Non-Arbitrary Containment
Differently coloured, and differently sized boxes
Arbitrary Containment
Differently coloured same-size circles
Arbitrary Hierarchy
Nonsense syllables presented on a laptop with certain hierarchical relations between stimuli shown and pt. had to derive others

9. Study 1: Relational Framing Test
All three relational frames (192 questions in all) were measured for:
Mutual Entailment (16 questions each)
e.g., A is inside B; Does B Contain A?
e.g., B is a type of A; Does the class A contain B?
Transformation of Function of mutually entailed relations (16 questions each)
e.g., A is inside B. Brian likes A; Does B contain something that Brian likes?
e.g., B is a type of A. B’s have big eyes. Does the class A contain members that have big eyes?)
Combinatorial Entailment (16 questions each)
e.g., A is inside B, B is inside C; Does C contain A?
e.g., C is a type of B, B is a type of A; Does the class A contain C?
Transformation of function of combinatorially entailed relations (16 questions each)
e.g., A is inside B, B is inside C. Chloe likes A. Does C contain something that Chloe likes?
e.g., C is a type of B, B is a type of A. C’s like sweet food. Does the class A contain members that like sweet food?)

10. All participants were exposed to NAC, AC and AH relational responding testing with novel stimuli between 10 & 14 days after initial testing. All measures showed high reliability.

11. Study 1: Relational Framing Scores by Age Group
These data provide a developmental trajectory of repertoires of containment and hierarchical relational responding

12. Study 1: Correlations
Strong correlations between the participants age in months and relational framing performance.
Strong correlations between relational framing repertoires and cognitive and linguistic performance.
Moderate correlations between relational framing repertoire and categorization and class inclusion responding

13. Conclusion
Containment & hierarchical framing strongly related to linguistic and cognitive performance.
Insight into development of relational frames underlying categorization in young children.

14. Study 2 – Arbitrary Containment
Participants:
Three typically developing 5-year-olds (Mean age = 5 years 3.7 months)  Experimental Group
Three typically developing 5-year-olds (Mean age = 5 years 4.7 months)  Control Group
Experimental Design:
Combined MBD (across pts and components)
All pts were first assessed for non-arbitrary containment repertoires and then arbitrary containment repertoires to determine their eligibility for inclusion.
Also assessed using PPVT, CCT and class inclusion (before and 6 months after training) to determine impact of training on these variables.
2F; 1 M
Because of the small sample size, the comparison between the experimental and control groups with regard to effect on PPVT and / or CCT could not be statistically significant even if there’s a large effect. However, we did what we could under the very time limited circumstances. If data are suggestive that there was a difference then we could use that to justify doing a larger n study.

15. Pre-Assessment (Arbitrary Containment)
Percentage of Correct Arbitrary Containment Relational Repertoires
Participant ME
ToF ME CE
ToF CE
Total
EP1
68.75%
62.5%
50%
43.75%
56.25%
EP2
54.69%
EP3
CP1
75%
59.38%
CP2
57.81%
CP3
37.5%
All participants were then assessed for arbitrary containment repertoires. Participants with <60% correct responding were included within the current research.

16. Method
The relationship between nonsense syllables were presented as text on a laptop with up to 4 stimulus sets.
E.g., “A blorg is inside a grap.”
Underneath this description is a question which assesses ME, CE, or ToF.
E.g., “Is a blorg inside a grap?”
A total of 64 questions

17. Baseline Baseline Assessment:
Each pt was assessed for AC relational responding across ME relations (16 questions), CE (16 questions), and ToF (32 questions).
Tested across 4 stimulus sets.
No feedback or positive reinforcement provided.
Pts were introduced to training if:
(a) baseline responding was stable, and/or
(b) the previous pt had completed that phase of training

18. Training Sequence Phase 1: Mutually entailed relations (2 stimuli)
Transformation of stimulus functions of mutually entailed relations (2 stimuli)
Phase 3:
Combinatorial entailed relations (3 stimuli)
Phase 4:
Transformation of stimulus functions of combinatorially entailed relations (3 stimuli)

19. Consequences
Positive reinforcement (tokens & praise) for correct responses in addition to specific feedback
FR4 exchange of tokens for stickers
Pt was given feedback for incorrect responses and re- exposed to the trial
If the pt beat his/her score from the previous session, they could then choose something from the stationery box

20. Progression The pt was exposed to training using one stimulus set
Once mastery criterion was met (i.e., 100%), the pt was then assessed for generalization
If the pt demonstrated generalization, he/she progressed onto the next phase of training, if not, they were re-exposed to training using a novel stimulus set
Then the next pt was introduced to that phase of training (dependent upon a stable baseline)

21. Combined Multiple Baseline Design Across Components and Participants for Arbitrary Containment
Current slide represents the combined MBD for Participants 1, 2 and 3 and provides generalization data

22. Multiple Baseline Across Relational Components Design for Participant 1
Includes 5 week Maintenance Data

23. Multiple Baseline Across Relational Components Design for Participant 2
Includes 5 week Maintenance Data

24. Multiple Baseline Across Relational Components Design for Participant 3
Includes 5 week Maintenance Data

25. Results Training sessions were conducted over a five-week period.
All pts generalised arbitrary containment responding to novel stimuli across all phases
Maintenance was also demonstrated for all 3 pts 5 weeks post-training.
Both Experimental and control groups will be retested on CCT, PPVT and class inclusion in June 2017

26. Conclusion
First study to successfully train arbitrary containment repertoires in young children
Generalization and maintenance for all 3 pts
Future research
Train arbitrary containment in younger children, or individuals with developmental disabilities
Determine impact of training on additional outcome measures (e.g., verbal ability, cognitive ability)

27. Study 3 – Arbitrary Hierarchy
Participants:
Three typically developing 6-year-olds (Mean age = 6 years 4.3 months)  Experimental Group
Three typically developing 6-year-olds (Mean age = 6 years 3.3 months)  Control Group
Experimental Design:
Combined MBD (across pts and components)
All pts were first assessed for arbitrary hierarchical repertoires to determine their eligibility for inclusion
Also assessed using PPVT, CCT and class inclusion (before and 6 months after training) to determine impact of training on these variables
2F; 1 M
Because of the small sample size, the comparison between the experimental and control groups with regard to effect on PPVT and / or CCT could not be statistically significant even if there’s a large effect. However, we did what we could under the very time limited circumstances. If data are suggestive that there was a difference then we could use that to justify doing a larger n study.

28. Pre-Assessment (Arbitrary Hierarchy)
Percentage of Correct Arbitrary Hierarchical Relational Repertoires
Participant ME
ToF ME CE
ToF CE
Total
EP1
62.5%
56.25%
50%
43.75%
53.13%
EP2
46.88%
EP3
37.5%
CP1
68.75%
54.69%
CP2
48.44%
CP3
42.19%
N.B. As with previous study, participants were told that the experimenter would repeat the question if they wanted, or were unsure. This was throughout the assessment and training procedure.
All participants were assessed for arbitrary hierarchical repertoires. Participants with <60% correct responding were included within the research.

29. Method
Identical to the previous study, however, hierarchical relations were described
Baseline, training sequence, consequences, and progressions were identical to that of Study 2

30. Combined Multiple Baseline Design Across Components and Participants for Arbitrary Hierarchy

31. Multiple Baseline Across Relational Components Design for Participant 1 (Arbitrary Hierarchy)
Includes 5 week Maintenance Data

32. Multiple Baseline Across Relational Components Design for Participant 2 (Arbitrary Hierarchy)
Includes 5 week Maintenance Data

33. Multiple Baseline Across Relational Components Design for Participant 3 (Arbitrary Hierarchy)
Includes 5 week Maintenance Data

34. Results Training sessions were conducted over a six-week period
All pts generalised arbitrary hierarchical responding to novel stimuli across all phases
Maintenance was also demonstrated for all 3 pts 5 weeks post-training
Both Experimental and control groups will be retested on CCT, PPVT and class inclusion in June 2017

35. Conclusion
First study to successfully train arbitrary hierarchical repertoires in young children
Generalization and maintenance for all 3 pts
Future research
Train arbitrary hierarchical repertoires in younger children, or individuals with developmental disabilities
Determine impact of training on additional outcome measures (e.g., verbal ability, cognitive ability)