Pilot Study Method
The Pilot Study was designed to establish agreement on the coding scheme and determine if coded measures of self-‐discrimination predicted wellbeing. This
statements then correlating the frequency of these statements with the subjective wellbeing measures. The assumption was that if someone was able to take
perspective on their experience [SX], they would be less fused with the content of their experience, which would equate to a decrease in rigid self-‐conceptualisation [SS]. That is a measure of FLEX would equal SX-‐SS and measures of FLEX would correlate positively with measures of wellbeing. This is based on the idea that flexibility is a combination of mindfulness or perspective taking and defusion from cognitive verbal content (Hayes et al. 2012b).
Pilot Study Results
Reliability
No inter-‐rater reliability measures were taken while coding the initial set of interviews, n = 15. Inter-‐rater reliability measures were taken during subsequent rounds of the study. These reliability measures are discussed at Round One and Three below.
Analyses
In my initial set of coded interviews, n = 15, notable associations became apparent between code frequencies, calculations of FLEX and measures of Satisfaction with Life and Psychological Well Being.
There was a strong correlation between FLEX and Satisfaction with Life, r = .63, p = .01, which indicated 40% of the variance in Satisfaction with Life could be explained by FLEX. While it is impossible to infer causation from a correlation, I assume if people are more able to objectify experience and identify with it as transitory rather than literal truth, they would also more likely be satisfied with life. Thus, it seemed more likely that flexible self-‐discrimination would cause satisfaction with life than the other way around. It was hard to imagine how satisfaction with life would cause more flexible self-‐discrimination. This was supported by a negative, strong, and statistically significant correlation between Satisfaction with Life and SS, r = -‐.61, p = .02, indicating that 37% of the variance in Satisfaction with Life may be explained by
SS. This suggested that the more rigidly one identifies with symbolic content, the less satisfied with life they will be.
There was a moderate positive, but non-‐significant, correlation between FLEX and Psychological Wellbeing, r = .36, p = .19. Although this was nonsignificant, with this small sample the effect size was such that it warranted further investigation. People may be more likely to experience psychological wellbeing if they are able to defuse from psychological content and identify with it as passing experience. This tentative finding was supported by a moderate non-‐significant negative correlation between Psychological Wellbeing and SS, r =-‐.35 p = .20.
Multiple regression analyses were conducted to examine the predictors of Satisfaction with Life and Psychological Wellbeing each with seven predictors entered together into the model: Age, FLEX, Openness, Neuroticism, Agreeableness, Extroversion and Conscientiousness. The model did not significantly predict
Psychological Well-‐being (F = 1.5, p = .31) but the predictors accounted for 57% of the variance in well-‐being. Similarly, the model did not significantly predict
Satisfaction with Life (F = 3.5, p = .07) but accounted for 23% of the variance in Satisfaction with Life. FLEX emerged as the only predictor of Satisfaction with Life (β = .96, p = .05) and the only predictor of Psychological Wellbeing (β = .96, p = .05).
Discussion (Pilot Study Coding)
Because the sample size was so small for this exploratory phase, many of the correlations between self-‐discrimination codes and other measures were non-‐
significant. Across all measures taken FLEX was moderately and positively correlated with Psychological Well Being, Satisfaction with Life and Positive & Negative Affect. FLEX also showed a moderate negative correlation with Stress and Anxiety. This trend was mirrored by a moderate negative correlation between SS and the measures of Psychological Well Being, Satisfaction with Life and Positive & Negative Affect. SS showed a moderate positive correlation with Stress and Anxiety. While these
these results supported the emerging validity of the coding scheme and suggested that further coding would be worthwhile.
The International Personality Item Pool measures: Openness, Neuroticism,
Agreeableness, Extroversion and Conscientiousness were all negatively correlated with FLEX. Except for Neuroticism, which was not expected. The assumption was that these measures would positively correlate with FLEX. More investigation would explain this.
SX1 was positively correlated with Psychological Wellbeing and Positive & Negative Affect. SX1 was negatively correlated with Neuroticism and Stress. This trend again supported the idea that being able to objectify inner behaviour (thoughts and feelings) is important for wellbeing.
The results for SP-‐then were contrary to expectations given that this code tended to be negatively correlated with measures of wellbeing and positively correlated with measures of Stress, Anxiety and Depression. The assumption was that if someone were holding inner experience flexibly, these relations would be reversed.
Round One
Round One Method
Following the Pilot Study considerable thought was given to further refining the definition of the codes and their application to text. The aim was to ensure as far as possible a sound theoretical underpinning based on RFT.
At this point in the study considerable thought was given to the definition and application of SS. As outlined above in the section on how SS evolved, I decided that SS would be applied to self-‐discrimination statements describing both primary and secondary attributes of the speaker. In the pilot study I had applied SS to self-‐
discrimination statements describing only secondary attributes. I retained my initial
they would be less fused with the content of their experience, which would equate to a decrease in rigid self-‐conceptualisation [SS]. I continued to calculate a measure of FLEX as equal to SX-‐SS and assumed measures of FLEX would correlate positively with measures of wellbeing.
To further test these assumptions the original 15 interviews, plus an additional five interviews (total n = 20), were recoded/coded with the revised coding scheme that included the adjusted interpretation of SS.
Round One Results
Reliability
Two coders independently coded four interviews and the inter-‐rater reliability was calculated using QDA Miner. Inter-‐rater reliability (Krippendorf’s alpha) was
calculated using the following specifications: the double coded segments had to overlap by at least 70% and common absences were counted as agreements. As each interview was coded, disagreements were discussed. After the first interview was double coded, inter-‐rater reliability was barely adequate with a Mean Overall Kappa of 0.45. At the fourth double coded interview prior to any discussion on
disagreements, inter-‐rater reliability was adequate with a Mean Overall Kappa of 0.65 (Table 3.3). After discussion about disagreements in each of the double coded interviews, agreements were reached and differences were reconciled for each code. In each case inter-‐rater reliability was then excellent. The nature of the discussions and resulting agreements are outlined above in the sections explaining how each code evolved. Fleiss (1981) recommended kappa of 0.4 to 0.75 be considered fair to good, while kappa > 0.75 be considered excellent.
Krippendorf's Kappa Pre Discussion Post Discussion Self-‐as-‐Process Now [SP-‐now] .82 1
Self-‐as-‐Process [SP-‐then] .57 .92
Self-‐as-‐Story [SS] .62 .97
Self-‐as-‐Perspective 1 [SX1] .52 1 Self-‐as-‐Perspective 2 [SX2] 0 0
Mean Overall Kappa .65 .97
NB. The SX2 code did not occur in this interview.
Table 3.3: Inter-‐Rater Reliability Round One Results for the fourth double coded interview.
Analyses
The revised coding scheme resulted in FLEX becoming less predictive of wellbeing. In particular, Satisfaction with Life was nonsignificant (r = .23, p =.32) compared with the correlations in the pilot study (r = .63, p = .01).
To compare these results with those of the Pilot Study, multiple regression analyses were conducted to examine the predictors of Satisfaction with Life and Psychological Wellbeing. For each measure seven predictors were simultaneously entered into the model: Age, FLEX, Openness, Neuroticism, Agreeableness, Extroversion and
Conscientiousness. These analyses confirmed that the revised coding of FLEX was substantially less predictive of wellbeing, controlling for personality and age. In the Pilot Study these predictors accounted for 23% of the variance in Satisfaction with Life and 57% of the variance in Psychological Wellbeing. In Round One they
accounted for 3% of the variance in Satisfaction with Life and 58% of the variance in Psychological Wellbeing. Standardised coefficients for FLEX indicated that FLEX no longer significantly predicted either Satisfaction with Life (β = 1.63, p = .13) or Psychological Wellbeing (β = .38, p = .08).
Discussion (Round One Coding)
Coding SS for both primary and secondary attributes of the self appeared to reduce its capacity to predict wellbeing. I decided to assign SS to statements where the speaker ascribed only a secondary attribute in a frame of coordination to the ‘self’, as was done in the pilot study, and not to statements about both primary personal or behavioural attributes.
Round Two
Round Two Method
Round Two of coding involved recoding the set of interviews, n = 20, with a particular focus on applying the revised SS code capturing self-‐discrimination statements
ascribing secondary attributes to the self. I continued with my initial assumptions that if someone were able to take perspective on their experience [SX], they would be less fused with the content of their experience, which would equate to a decrease in rigid self-‐conceptualisation [SS]. That is, a measure of FLEX would equal SX-‐SS and that measures of FLEX would correlate positively with measures of wellbeing.
Round Two Results
Analyses
While remaining nonsignificant, the relationships between FLEX and Satisfaction with Life and Psychological Well Being strengthened. The initial correlation between FLEX and Satisfaction with Life in the pilot study was strong and significant, r = .63, p = .01. At Round One the strength was reduced and significance of this relationship was lost, r = .23, p = .32. At Round Two, after recoding, the relationship was
moderately strong and marginally significant, r = .43, p = .06. Further, the
relationship between Satisfaction with Life and SS was initially strong, negative and statistically significant, r = -‐.61, p = .02. At Round One the relationship was weak, negative and nonsignificant, r = -‐.20, p = .39. At Round Two the relationship was moderately strong, negative and significant p < .10 ( r = -‐.38, p = .09).
Also, there was initially a moderate correlation between FLEX and Psychological Well Being, r = .36, p = .19. At Round One this relation became weaker and nonsignificant, r = .23, p = .34. After recoding at Round Two strength and level of significance in the relationship returned, r = .32, p = .17.
Multiple regression analyses were conducted to examine how well the measures predicted Satisfaction with Life and Psychological Wellbeing. For each measure seven predictors were simultaneously entered into the model: Age, FLEX, Openness,
Neuroticism, Agreeableness, Extroversion and Conscientiousness. Initially these predictors accounted for 23% of the variance in Satisfaction with Life. At Round One they accounted for less that 3% of the variance. After recoding at Round Two these predictors accounted for 25% of the variance in Satisfaction with Life. While overall these variables did not significantly predict Satisfaction with Life, F = 1.85, p = .17, the Standardised Coefficient of FLEX, β = 2.78, p = .02, again indicated that FLEX
significantly predicted Satisfaction with Life. For Psychological Wellbeing, initially and at Round One, these predictors accounted for 58% of the variance. At Round Two they accounted for 62% of the variance. In Round Two, these variables now
significantly predicted Psychological Wellbeing, F = 5.17, p = .01. The Standard Coefficient for FLEX, β = .44, p = 0.04, indicated that FLEX was a significant predictor of Psychological Wellbeing on its own. As a predictor FLEX was positively associated with both Satisfaction with Life and Psychological Wellbeing.
Discussion (Round Two Coding)
At Round Two of coding the moderate to strong correlations between FLEX and the other measures of wellbeing that were lost at Round One returned. The revised definition of SS was supported. SS was to be applied to statements where the speaker ascribed secondary attributes in a frame of coordination with the ‘self’ and not to statements about primary personal or behavioural attributes.
During coding I wondered if positive self-‐discrimination statements might be more positively related to wellbeing than negative self-‐discrimination statements. This
hypothesis was tested in Round Three of coding with the introduction of additional codes that distinguished positive and negative self-‐identity statements.
Further, a question about personal beliefs and self-‐rules arose. I recognised that such statements did not really fall into the category of SP-‐then, which led to a number of distinctions about personal beliefs and self-‐rules. Many of these statements were future oriented and specified: what the speaker held as literal ‘truths’ about how the world is; the speaker’s strategies for controlling or avoiding unwanted experience; or, their strategies for behaving in a way that will take them in a valued direction. While it was clear these statements were not SS they were not SP-‐then either. At this point I introduced codes for Control Oriented Self-‐Rules [COR] and Value Oriented Self-‐Rules [VOR] to capture such statements. To capture Personal Belief statements I decided to introduce codes for Control Oriented Personal Beliefs [COB] and Value Oriented Personal Beliefs [VOB] along with a suite of ‘Other’ codes in Round Four of coding, the results of which are discussed in Chapter 4: Development and Application of the Functional Self-‐Discrimination Measure of ‘Other’.
Questions about the nature of SX versus OX
It was at this point I drew a distinction between statements coded SX versus OX as a number of statements coded SX were in fact statements about another’s perspective. I acknowledged that such statements were not really SX. These were statements in which the speaker drew a distinction between another’s perspective and their own. Some statements were a rigid classification of one view being correct, usually their own, and the other’s view being incorrect. Other statements were more flexible in that the speaker distinguished two points of view, their own and another’s, without any rigid classification about the apparent correctness of either. This observation led to the definition and future application in Round Four of the OX code, which is
defined and discussed in Chapter 4: Development and Application of the Functional Self-‐Discrimination Measure of ‘Other’. At this point I decided to double code any statement that attracted an SX with one of the other codes such as SS or SP to indicate whether the perspective being taken was rigid or flexible.
Questions about SP
At the conclusion of Round Two of coding and analysis some further distinctions about the SP code were drawn. In addition to being applied to statements about current experience I decided to apply this code to statements that contained personal pronouns such as ‘I think’ as well as linguistic conventions that appeared to be
habitual and served to create space for thinking. Based on this distinction the SP code was expanded to include SP-‐hedge with SP-‐now and SP-‐then as defined above.
While the majority of the correlations and regression analyses’ in Round Two were nonsignificant and the number of interviews was small (n = 20), overall the results supported the validity of the coding scheme and provided enough evidence to continue applying the codes with refinements. The work that followed is discussed below in the Round Three Results and in Chapter 4 where Round Four Results are discussed.