Research in Coherence: Pitfalls, Developments, and Suggestions Sam Ashcroft, Lee Hulbert-Williams, Kevin Hochard & Nick Hulbert-Williams University of Chester
Introduction
Cognitive Dissonance Theory “The holding of two or more inconsistent cognitions arouses the state of cognitive dissonance, which is experienced as uncomfortable tension. This tension has drive-like properties and must be reduced.“ - Cooper, 2007; p.7
Relational Frame Theory “coherence or sense-making appears to function as a powerful reinforcer for relational activity” - Barnes-Holmes, Hayes, Dymond & O’Hara, 2001; p. 70
Relational Frame Theory “…once established, coherence and sense-making will serve as a continuously available reinforcer for derived relational responding” - Hayes et al., 2001; p. 48
B A C Directly Trained ‘Equals’ Relationships Combinatorial Entailment Nodes/Stimuli
B A C Directly Trained ‘Unidirectional’ Relationships Combinatorial Entailment Nodes/Stimuli
B A C Directly Trained ‘Unidirectional’ Relationships Ambiguous Relationship Nodes/Stimuli
Quinones and Hayes (2014) On ambiguous A-C test trials, participants responded systematically as though A>C or A<C B A C Directly Trained ‘Unidirectional’ Relationships
Quinones and Hayes (2014) On ambiguous A-C test trials, participants responded systematically as though A>C or A<C B A C Directly Trained ‘Unidirectional’ Relationships
Study One
Why run a study? Quinones and Hayes (2014) • Small sample • No inferential statistics • No additional measures (such as affect)
#### - Discriminative Stimulus ZKR CDO Stimulus A Stimulus B
$$$$ - Discriminative Stimulus ZKR CDO Stimulus A Stimulus B
The networks trained Three Coherent: A>B>C Three ‘Ambiguous A-C’: A>B<C
Experiment- Block - Stages - Training Stage Test Stage Trials -
The participants for Study One N = 80 65 Females Aged 18 to 58 (M=21.71, SD=5.55)
48 is the max score Cutoff
Close to 24 means strong, consistent A>C or A<C responding
Variable Ambiguous Coherent Mean p Generalized Mean (Standard (Standard Eta Squared Deviation) Deviation) Reaction times 1.29 (0.25) 1.15 (0.25) 5.99e-07* 0.383 on test trials Affect -0.31 (1.02) 0.27 (1.05) .0007* 0.200 Arousal 4.01 (1.83) 4.03 (1.64) .886 0.0004 Sense-Making 4.82 (1.61) 6.41 (1.65) 3.023-e08* 0.448 Small, medium and large GES would be 0.02, 0.13 and 0.26 resp spective vely y (Bake keman, 2005, p , 2005, p383) 383).
Relationship Screens Discriminative Stimuli Correct trained Bigger Bigger A, B A A B A>B Smaller Smaller A, B B A B Bigger Bigger B, C B B C B>C Smaller Smaller B, C C B C Design seems sound
Times Correct Times on Reinforcement Stimulus Screen (OS) Ratio (OS / C) (C) A 2 1 50% B 4 2 50% C 2 1 50% And yet, there is an alternative hypothesis…
Alternative Hypothesis (and resolution)
Bigger A B
Smaller B C
A C
But remember This is only an issue for coherent A>B>C networks This means that on coherent A>B>C networks, the ‘combinatorial entailment’ effect may actually be simple ‘pairing’ of stimuli and discriminatives Then I realised that this issue isn’t only inherent in my design, but in that of Quinones and Hayes (2014) and any other study using a unidirectional A>B>C format
End stimuli are the issue X > A > B > C > Y A > B > C Let’s add the circles and see… Remember, this is theoretical (but still important!).
X > A Bigger X A
X > A Smaller X A
A > B Bigger A B
A > B Smaller A B
B > C Bigger B C
B > C Smaller B C
C > Y Bigger C Y
C > Y Smaller C Y
X > A > B > C > Y X A C Y B
A C
3 Stim A B A C 5 Stim
Times on Times Correct Reinforcement Paired with Stimulus Screen (OS) (C) Ratio (OS / C) Discriminative X 2 1 50% Bigger A 4 2 50% Smaller, Bigger B 4 2 50% Smaller, Bigger C 4 2 50% Smaller, Bigger Y 2 1 50% Smaller
Study Two
#### - Discriminative Stimulus ZKR CDO Stimulus A Stimulus B Second Experiment
Experiment- Block - Stages - Training Stage Test Stage Trials -
The participants for Study Two N = 75 59 Females Aged 18 to 45 (M=20.83, SD=4.39)
Five nodes, and ONE discriminative Having five nodes and two discriminatives is near impossible to learn So, I dropped to one discriminative (i.e.“Bigger” is always the relationship) All the tables and ratios and ‘pairing circles’ are completely balanced This did something important which I will talk about if there is time
Close to 4 means strong, consistent A>C or A<C responding
Why do participants now not ‘make their own coherence’? Experiment One Experiment Two Bigger Smaller Bigger A C A C Y Q
Max score is 8 Cutoff
Ambiguous Coherent Variable p GES Mean (SD) Mean (SD) Reaction Time 1.47 (0.54) 1.42 (0.47) .283 0.030 Affect -0.11 (1.10) -0.03 (1.19) .418 0.017 Arousal 3.61 (1.84) 3.77 (1.52) .310 0.027 Sense-Making 3.84 (1.88) 3.92 (1.61) .527 0.011
3 Stimulus 5 Stimulus p GES p GES Reaction Time 5.99e-07*, 0.383 .283 0.030 A-C Trials Affect .0007*, 0.200 .418 0.017 Arousal .886, 0.0004 .310 0.027 Sense-Making 3.023-e08*, 0.448 .527 0.011
Ambiguous A-C Test Trials 3 Stimulus 5 Stimulus Strong A>C or A<C Weak A>C or A<C responding responding Coherent A-C Test Trials Confidence is through 3 Stimulus 5 Stimulus the roof? High correctness (ceiling?) Reasonable correctness
Take home messages Be wary of results from unidirectional A>B>C experiments, they may be inflated Five-stimulus X>A>B>C>Y networks resolve the inherent issue with A>B>C network designs Using one discriminative, and/or distractor stimuli in test stages seems to prevent artificial creation of ‘coherence’
Other Research within my PhD Physiological measures for A-C trials may be sensitive enough to assess differences on block types (near complete) Add feedback to the test stage so that you can further manipulate coherence / ambiguity / incoherence (complete) Ask participants to choose block types to do again, assessing appetitiveness (complete) Future: real stimuli, vignettes
For further information Please come and speak with me about Sam Ashcroft , PhD Student, Graduate Teaching Assistant any thoughts and feedback! That’s why I’m here! w: www.cbslab.uk | http://www.chester.ac.uk/psychology/cruph e: s.ashcroft@chester.ac.uk LinkedIn: www.linkedin.com/in/samashcroft GitHub Code: https://github.com/S-Ashcroft
Recommend
More recommend
