Embedding Contingency Judgement Experiments into Everyday Life: An Ecologically Valid Test of the Outcome Density Effect March 2016 Rachel M. Msetfi, Donal O’Sullivan & Pepijn Van de Ven University of Limerick
Action-Outcome Contingency ΔP = 5/(5+15) - 5/(5+15) = 0 Low OD ΔP = 15/(15+5) - 15/(15+5) = 0 High OD Contingency sensitivity: e.g., Allan & Jenkins, 1980; Wasserman et al., 1993 Outcome density effects: e.g., Dickinson, Shanks & Evenden, 1984
Key associative explanations: 1)Non-depressed respond to variation in background context which is extinguished allowing action-outcome association to become strong; Depressed people do not (e.g., Msetfi et al., 2005) 2)Decreased activity levels related to depression result in decreased action – outcome pairings (e.g., Matute, 1996; Blanco et al., 2013) 3)Increasing depressed activity levels via experimental manipulations (Byrom et al., 2015) or pharmacological intervention (Msetfi et al., 2016) enhances perceived control via effect on context Traditional explanations Healthy optimism Illusion of Control Expectation of having control Motivation to maintain self esteem Depressive realism Accurate view Expectation of no control No motivation to maintain self esteem Traditional explanations Healthy optimism Illusion of Control Expectation of having control Motivation to maintain self esteem Depressive realism Accurate view Expectation of no control No motivation to maintain self esteem Adapted from: Alloy & Abramson (1979) Low Ecological Validity
Studying context & depression effects Serotonin depletion model of depression & learning Low 5-ht impairs place context learning (Wilkinson et al. 1995); Low 5-ht enhances discrete cue context learning (Cassaday et al., 2001) Type of context: "Information which must be held in mind... that it can be used to mediate task appropriate behaviour" (Cohen et al., 1999) Type of context: “...many different types of stimuli, such as physical background cues, interoceptive state, and time.” (Bouton, 2006)
Do these effects generalise to a noisy, dynamic, extended context? Theoretical implications RWM: – Standard: Cxt kTrials – Extended: (Cxt kTrials) + (Cxt M- * 4kTrials) – *Dynamic: (Cxt M, N, O, P, Q, R, S, kTrials) – *Dynamic/Extended: (Cxt M, N, O, P, Q, R, S, kTrials) + (Cxt M-, N-, O-, P-, Q-, R-, S-, -T * 4kTrials) * assumptions: sequential movement through contexts; action / no action not equally sampled across contexts
Predictions: OD effects α Action =.8; α Cxt =.2; RWM Simulator: Mondragon, Alonso, Fernandez, & Gray (2012, ver 4)
RW Predictions: Context – action discrimination α Action =.8; α Cxt =.2; Pos value = Cxt>Act; 0 value = Cxt=Act; Neg value = Cxt < Act RWM Simulator: Mondragon, Alonso, Fernandez, & Gray (2012, ver 4)
Current study Challenge: Embed contingency judgement task in everyday activities against real world dynamic, noisy, extended, context Solution: Mobile phone administration spread over full day whilst participant engages in normal activities Ecological validity: Task similar to ongoing user- phone interactions – Alert > Access > Action > Outcome – Notification > Click on the link > Friend liked your tweet ♯ ALS2016
Judgement app 40 4s-trials spread over 8-hour day with variable ITIs, M=12mins Each trial mimics real phone interaction: alert > action > outcome
Contingency Judgements 5 blocks of 8-trials; Ratings made after each block -100 to +100 scale Action: Rate your control over the sound Context: Rate the control of external causes over the sound
Pilot Study (2): Method Design: – OD: Low OD vs. high OD, zero contingency – Depression: Low vs high BDI (≤5/≥6) Participants N = 31 ( data collection incomplete ) – Low BDIs: n=17, M BDI =1.82, SE BDI =.82 – High BDI: n=14, M BDI =9.54, SE BDI =.89 – Matched age, est IQ, digit;
Results: Validity ΔP experienced close to 0 and no difference between groups OD experienced was as programmed Action rate against non-missed trials was as instructed & consistent with previous research
Results: Action Control ratings & OD effects Findings High OD > Low OD Low BDIs > High BDIs Main effects: OD F(1, 27) = 5.80, p =.02 BDI F(1, 27) = 2.904, p =.09 OD * BDI F(1, 27) <1 Predictions
Pos value = Cxt>Act; 0 value = Cxt=Act; Neg value = Cxt < Act Results: Context-action Discrimination Findings -Low BDIs discrimination in line with dynamic context predictions -High BDIs inconsistent Main effects: OD F(1, 27) = 8.12, p =.008 BDI F(1, 27) = 4.08, p =.05 OD * BDI F(1, 27) = 2.51, p =.12 Predictions
Discussion Low BDIs: Significant OD effect; Action and context ratings consistent with dynamic context RW predictions High BDIs: Significant OD effect but lower, more negative judgements in all conditions; Action-context discrimination with low OD does not follow predicted pattern. ?Preliminary analysis of trial-by-trial behaviour & experience, suggestive of increased cell b frequency (action-) and less distinction between OD in high BDI group ?High BDIs tendency to use negative / preventative end of the judgement scale. Is this an artefact of specific judgement tool?
Preliminary conclusions Outcome density / illusion of control and depressive realism effects do generalise to realworld, everyday contingencies Effects maintained in noisy, dynamic, extended contexts in line with an associative model Future work: larger samples; manipulate contingencies and everyday stressors over time; monitor contingency related mood & affect changes over time
Acknowledgements: University of Limerick students Alison Garvey & Amy Walsh Questions welcome, feedback much appreciated!