More Questions than Answers: Continued Critical Reanalysis of Fredrickson et al.'s Studies of Genomics and Well-Being

Abstract

We critically re-examine Fredrickson et al.'s renewed claims concerning the differential relationship between hedonic and eudaimonic forms of well-being and gene expression, namely that people who experience a preponderance of eudaimonic well-being have gene expression profiles that are associated with more favorable health outcomes. By means of an extensive reanalysis of their data, we identify several discrepancies between what these authors claimed and what their data support; we further show that their different analysis models produce mutually contradictory results. We then show how Fredrickson et al.'s most recent article on this topic not only fails to adequately address our previously published concerns about their earlier related work, but also introduces significant further problems, including inconsistency in their hypotheses. Additionally, we demonstrate that regardless of which statistical model is used to analyze their data, Fredrickson et al.'s method can be highly sensitive to the inclusion (or exclusion) of data from a single subject. We reiterate our previous conclusions, namely that there is no evidence that Fredrickson et al. have established a reliable empirical distinction between their two delineated forms of well-being, nor that eudaimonic well-being provides any overall health benefits over hedonic well-being.

Fig 1. Extensions of Fredrickson et al.’s [2] Figure 1C.

In this figure, we have augmented Fredrickson et al.’s [2] Fig 1C—from which the first two pairs of bars in our figure are directly taken—with two new pairs of bars. The third pair from the left, labeled “Discovery Study (corrected),” indicates the fold difference values that are produced by the application of Fredrickson et al.’s [1] repeated-regression method to the discovery sample, after correction of the variable “White” for participant SOBC1-1299 from “4” to “0”. The rightmost pair, labeled “Discovery Study (SOBC1-1293 omitted),” indicates the fold difference values produced by the omission of participant SOBC1-1293 (cf. discussion in the text of the present article).