Facets of Optimism: Comment on

Abstract

In light of the analyses of Scheier et al. (2020) concerning differential associations of optimism and pessimism measures with physical health, we argue here that whether optimism and pessimism are bi-polar, lying on separate ends of a spectrum, or whether they represent two separate dimensions is a conceptual, rather than an empirical, question. Differential associations of various indicators, or indicator sets, with health may indeed be of interest, but there are also different ways of grouping indicators other than whether items are worded to correspond to optimism or pessimism. We do nevertheless believe that the analyses of Scheier et al. (2020) can be useful in helping guide intervention development in precisely the ways they suggest.

We thank Scheier et al. (2020) for insightful analyses concerning optimism, pessimism and physical health. In this comment, we raise three important issues for additional consideration: (1) the question of whether optimism and pessimism are “bi-polar” is conceptual, not empirical; (2) items may be grouped for reasons other than because they map on to underlying optimism versus pessimism concepts; (3) analyses of associations of individual indicators or indicator-groups can provide important practical insight, as indeed they do in Scheier et al. (2020).

Whether Optimism and Pessimism are "Bi-polar" is a Conceptual Question

Whether optimism and pessimism are bi-polar, lying on separate ends of a spectrum, or whether they represent two separate dimensions has been an ongoing debate. Factor analyses with the Life Orientation Test-revised (LOT-R) suggesting two factors might arguably be explained by differential responses to negatively- versus positively-worded items (Monzani et al., 2014; Chiesi et al., 2013). Regardless of this possibility, however, questions of meaning simply cannot be settled by factor analysis (Maraun, 1998). The question of “bi-polarity” here is conceptual, not empirical. If optimism is defined as “a disposition towards having expectations that the future will be good” and pessimism as “a disposition towards having expectations that the future will be bad” then these are opposed “bi-polar” concepts. The same person cannot be both optimistic and pessimistic about the same thing at the same time. Someone may be optimistic about certain domains of life (e.g. family) and pessimistic about others (e.g work), but optimism and pessimism cannot coexist in the same person with the same object. One cannot say both that “he is optimistic about all of life” and also that “he is pessimistic about all of life.” Both statements cannot logically be true. In principle, someone could both expect many good things and many bad things, but then we generally would not refer to such as state as optimism or pessimism. One cannot both “expect more good things to happen to me than bad” (LOT-R; Scheier et al., 1994) and yet simultaneously “expect more bad things... than good.” One may be neither optimistic nor pessimistic, but one cannot be both optimistic and pessimistic about the same thing in the same way. This is a conceptual point, and no empirical analysis could overturn this. Scheier et al. (2020) are careful not to draw such conclusions. However as previously noted (Kubzansky et al., 2004; Herzberg et al, 2006), some investigators have in fact interpreted their own similar results in precisely this manner.

It is Possible to Group Different Items from a Scale in Different Ways

The analyses of Scheier et al. (2020) suggest that the three negatively worded LOT-R “pessimism” indicators are more strongly associated with physical health than the three positively worded “optimism” indicators. As discussed by Scheier et al. (2020) and also below, this may have important practical implications. However, this does not de facto imply that optimism and pessimism exist as separate independent latent entities. Regardless of the conceptual definitions, study participants can respond to specific items on optimism and pessimism in different ways that could result in these indicators being differentially associated with health. A complex underlying psychological reality gives rise to how people respond to items related to the meaning of the items and how they are phrased. It is possible that only the more negatively worded items adequately assess pessimism-pole of the construct.

However, there are multiple different reasons for grouping items. Consider associations between the individual LOT-R indicators and stroke, presented in Table 1 below, using Health and Retirement Study data, with pessimism items reverse-coded. The analyses present odds ratios (ORs) relating 2006 LOT-R items to stroke risk through 2014, adjusting for age, race, and sex. One can see, as per Scheier et al. (2020), that the first three positively worded optimism items are, on average, more weakly associated (mean-OR=0.91) with lower odds of developing stroke than are the three negatively worded items (mean-OR=0.88). However, other ways of grouping items are possible. One might consider comparative statements of more good than bad (item iii) versus absolute statements that make no such comparison (all other items). In this case the comparative indicator has stronger protective association with stroke (OR=0.87) than the absolute indicators (mean-OR=0.90). Causal effects do not necessarily correspond to the statistical factor structure: a single statistical factor is consistent with differential causal effects of indicators (VanderWeele, 2020); conversely, two distinct factors may in fact have the same causal effects. However, evaluating item groupings and considering different ways of grouping them can provide insight into concept of optimism and how it relates to outcomes.

Table.

Associations between optimism (LOT-R) indicators and incident stroke (2006–2014) in the Health and Retirement Study.

Item Indicator	Odds Ratio (95% CI)
(i) I’m always optimistic about my future	0.91 (0.86–0.97)
(ii) In uncertain times, I usually expect the best	0.96 (0.90–1.02)
(iii) Overall, I expect more good things to happen to me than bad	0.87 (0.82–0.93)
(iv) If something can go wrong for me, it will	0.86 (0.82–0.91)
(v) I hardly expect things to go my way	0.87 (0.82–0.92)
(vi) I rarely count on good things happening to me	0.91 (0.86–0.96)

Analyses adjusted for age, race, and sex

Analyses of Indicator-Groups or Individual Indicators Can Provide Important Practical Insights

Arguably the most insight will be gained with theoretically informed groupings or item-by-item analyses, especially when simultaneously controlling for all other indicators (VanderWeele, 2020). Analyses concerning individual indicators or indicator-groups effectively give clues as to the most relevant aspects of the underlying reality (VanderWeele, 2020). Scheier et al. (2020) point out that their analyses may have important practical implications. They note, “Perhaps existing interventions that focus more on lessening pessimism such as those involving cognitive restructuring will be more successful in promoting better health than will those that place a greater weight on promoting optimism.” Likewise, it is possible that interventions to promote “comparative” optimism (expecting more good than bad) by changing attentional focus or providing greater resources and agency may be more effective for health than promoting “absolute” optimism (always expecting good) which may sometimes result in unrealistic optimism (Shepperd et al., 2017). Analyses with observational data cannot guarantee differential effects of interventions. Observational analyses may still be subject to confounding; and effects of interventions will often alter a host of other things. However, one must start somewhere with intervention design; and analyses, such as those provided by Scheier et al., may indeed give clues as to the best place to begin.