Age Differences in Negative, but Not Positive, Rumination

Abstract

Objectives

The main objective of this study was to determine whether there are age differences in positive and negative repetitive thought (ie, rumination).

Method

Young adults (ages 19–39; n = 114) and older adults (ages 60–85; n = 88) completed measures of negative and positive rumination. Bayesian analyses were used to determine whether age differences were present for both negative (young > old) and positive (old > young) rumination.

Results

There was extremely strong evidence for age differences in negative rumination, with lower scores in older adults. In contrast, the evidence was in favor of the null hypothesis for positive rumination.

Discussion

Age-related positivity is better characterized as decreased dwelling on the meaning of negative moods, rather than increased attention to positive ones.

It is now well established that older adults experience fewer or less intense negative emotions in their daily lives than do younger adults. High arousal negative emotions, such as anger, worry, and stress, show a marked pattern of decline with age (Stone, Schwartz, Broderick, & Deaton, 2010), though sadness may be more stable (Stone et al., 2010). In addition, older adults show low rates of mental illnesses (Kessler et al., 2005), and lower levels of neuroticism (Roberts, Walton, & Viechtbauer, 2006) than younger and middle-aged adults. Overall, although the young may be advantaged in physical and neurocognitive function, older adults are advantaged in emotional stability.

One explanation for age differences in emotional function is that older adults may be particularly motivated to regulate emotional experience. This hypothesis has its roots in Socioemotional Selectivity Theory (SST), a theory linking motivation and future time perspective across adulthood (Carstensen, 1987). According to SST, young adults focus on goals that encourage information gathering for possible future gain, whereas older adults focus on emotion regulation to facilitate a more meaningful and positive current experience.

Research on SST has operationalized emotion regulation in a number of ways. Most relevant for this study is the positivity effect, a bias in information processing in which older adults place relatively greater emphasis on positive than negative information in their memory and attention. Much of the positivity effect research calculates this bias by examining age × valence interactions, or subtracting memory for negative material from memory for positive (Reed, Chan, & Mikels, 2014). These methods make a mechanistic accounting of the effect difficult (Reed et al., 2014): it could occur because older adults show decreased attention to negative stimuli, increased attention to positive stimuli, or both. In fact, an earlier meta-analysis that compared information processing of positive and negative information separately (Murphy & Isaacowitz, 2008), suggested that reduced attention to negative information, rather than increased attention to positive, was likely driving the effect. In addition, attention and memory biases in laboratory studies are indirect measures of emotion regulation at best, and it is yet unclear whether these biases are the result of emotion regulatory differences, or byproducts of other aging-related differences (Gavazzeni, Andersson, Backman, Wiens, & Fischer, 2012; Isaacowitz & Blanchard-Fields, 2012).

Other research on older adults’ emotion regulation has more directly examined age differences in strategy use. This research suggests that older adults may use “maladaptive” strategies (eg, strategies whose use is higher in people with mental health problems; Aldao, Nolen-Hoeksema, & Schweizer, 2010) less frequently than younger adults (Nolen-Hoeksema & Aldao, 2011; Schirda, Valentine, Aldao, & Prakash, 2016). Our research focuses on one putatively maladaptive strategy: rumination, or the repetitive focus on the content, causes, and consequences underlying one’s affective state (Nolen-Hoeksema, Wisco, & Lyubomirsky, 2008).

Rumination in response to negative affect (negative rumination) has a long history of study and is implicated in the onset and maintenance of major depressive disorder and other related psychological concerns (Aldao, et al, 2010). Though rumination has not been a major focus in aging research, at least one study suggests that older adults report lower levels of negative rumination than do younger adults (Nolen-Hoeksema & Aldao, 2011). In addition, although most research on emotion regulation and aging tends to focus on the downregulation of negative emotions (Isaacowitz, Gershon, Allard, and Johnson, 2013), rumination increases or maintains an affective state.

Although most of the literature on rumination is focused on response to negative mood, recent research has begun to focus on rumination in response to positive states (Feldman, Joormann, & Johnson, 2008). Positive rumination is defined as “the tendency to respond to positive affective states with recurrent thoughts about positive self-qualities, positive affective experience, and one’s favorable life circumstances” (Feldman et al., 2008, p. 509). Positive rumination increases positive mood, but its status as adaptive or maladaptive may depend on whether a person has an existing vulnerability for bipolar disorder. That is, although some aspects of positive rumination are increased in people with bipolar disorder, positive rumination is also associated with reduced depression symptoms (Feldman et al., 2008). To the best of our knowledge, no prior studies have investigated age differences in positive rumination.

The Current Study

To investigate age differences in positive rumination, we used the Bayesian approach advocated by Lakens, McLatchie, Isager, Scheel, and Dienes (2018). Bayesian analysis quantifies whether a given result is more likely under the null hypothesis (no age difference) or under a specified alternative hypothesis (older adults report more positive rumination than young), rather than merely stating that an age difference is “statistically significant”. Bayesian analyses are valuable in developmental research, when finding no age differences can have important practical and theoretical importance (Lakens et al., 2018). A Bayesian approach is particularly appropriate for the mechanistic question outlined above: Are age differences in emotion regulation due to decreased attention to negative information, increased attention to positive information, or both? To show that younger and older adults do not differ in attention to positive information, traditional methods rely on finding that the null hypothesis (young = old) cannot be rejected. Retaining the null hypothesis, however, does not mean that the alternative hypothesis (old > young) must be rejected: the obtained result may be as likely under the alternative as under the null. In other words, a finding that’s “not statistically significant” using traditional methods may in fact be “inconclusive” (eg, have a Bayes factor around 1) using a Bayesian approach.

Given prior strong evidence in favor of reduced negative rumination at older ages using traditional statistical methods, we expected to replicate this finding using Bayesian analyses. Therefore, the primary goal of this study was to determine whether or not older adults engage in more positive rumination than young adults.

Method

The data reported here were collected as supplementary measures from two experimental studies conducted in our laboratory. Preliminary analyses did not indicate systematic differences between the two studies; we provide these analyses, along with more methodological detail on the studies themselves, in Supplementary material. No a priori power analysis was conducted for the analyses reported in this article; the sample size for each experiment was based on separate power analyses to detect the relevant experimental effects (using traditional ANOVA and ANCOVAs) with a power of at least .80.

Participants

The current analysis includes 114 young adults (64 women, 49 men, 1 not reported) ages 19–39 (M = 25.05, SD = 5.69) and 88 older adults (61 women, 27 men) ages 60–85 (M = 69.75, SD = 5.14), who were recruited from the surrounding community and paid $30 for participation. More detail about the participants may be found in Supplementary materials.

Materials

Negative rumination

To measure negative rumination, participants completed the Ruminative Response Scale (RRS; Treynor et al., 2003). The RRS is a 22-item self-report questionnaire with three factor-derived subscales: Depression (12 items; α = .90 in current study), Brooding (5 items; α = .80), and Reflection (5 items; α = .82). Brooding and Reflection measure different components of negative rumination, whereas Depression items are typically redundant with measures of depression symptomatology (Treynor et al., 2003). The Brooding subscale involves “moody pondering” about one’s own personal shortcomings, whereas the reflection subscale reflects attempts at understanding why one feels unhappy. Participants were asked to read each item and indicate the frequency in which they generally engage in each behavior when feeling “down, sad, or depressed” using a four-point Likert scale (1 = Almost never, 4 = Almost always). Responses were averaged with higher scores reflecting a higher level of negative rumination.

Positive rumination

To measure positive rumination, participants completed the Response to Positive Affect Questionnaire (RPA; Feldman et al., 2008). The 17-item self-report measure includes three subscales: Emotion-focus (5 items; α =.81 in the current study), Self-focus (4 items; α =.85), and Dampening (8 items; α =.81). The Emotion-focused and Self-focus subscales ask about positive rumination on both emotions and progress toward goals, respectively. In contrast, the Dampening subscale reflects the tendency to diminish or reduce positive affect, rather than rumination per se. Instructions ask participants to rate what they generally do, and not what they think they should do, when feeling happy, excited, or enthused using a four-point Likert scale (1 = Almost never, 4 = Almost always). Responses were averaged with higher scores reflecting a higher level of positive rumination.

Analysis

We briefly note that RRS-Depression and RPA-Dampening do not measure rumination as defined in the introduction, but we include them here for completeness. To test for age differences in each component of the RRS and RPA, we adopted a strategy comparable to Example 1 in Lakens and colleagues (2018). For all analyses, we used JASP 0.9.1.0 (JASP Team, 2018). A full accounting of the process, we used for determining the prior distribution in the Bayesian analyses is included in Supplementary materials, along with additional relevant references. Here, we briefly note that we chose an informed prior, with a half-normal distribution centered on zero with an SD = .30. For four of the measures (RRS-Depression, RRS-Brooding, RRS-Dampening, and RPA-Dampening), the prior was set to allow only negative effect sizes (indicating that older adults were expected to have lower scores than younger adults). For the other two measures (RPA-Emotion Focus and RPA-Self Focus), the prior was set to allow only positive effect sizes (indicating that older adults were expected to have higher scores than younger adults).

Results

Figure 1 presents prior and posterior plots for RRS-Reflection and RPA-Self Focus. Figure 2 presents a sequential analysis of these two measures, to allow readers to assess the robustness of the findings. These plots may be found for the remaining scales in Supplementary material.

Figure 1.

All figures were produced in JASP 0.9.1.0 (JASP Team, 2018). Prior and Posterior plots for one aspect of negative (Reflection—top graph) and one aspect of positive (Self-Focus—bottom graph) rumination. The prior is the distribution of possible effect sizes before the data were taken into account, based on the directional hypotheses (Old < Young for Reflection, Old > Young for Self Focus) and assumptions outlined in the text. The posterior distribution is the updated distribution of likely effect sizes after the data are taken into account. BF-0 is the Bayes factor stated in terms of strength of the alternative hypothesis, and BF0- is the same factor stated in terms of the null hypothesis. The median and 95% CI refer to the effect sizes in the posterior distribution. The pie chart is a graphic representation of the relative likelihood of the 2 hypotheses. See Wagenmakers and colleagues (2018) for more detail about JASP output.

Figure 2.

Sequential plots for one aspect of negative (Reflection—top graph) and one aspect of positive (Self-Focus-bottom graph) rumination. Sequential analysis assesses the Bayes factor after each data point is added. For the current sequential analysis, we ordered the data so that older and younger adults were entered in an alternating fashion, with the remaining younger adults entered last. Within the age groups, we used a random number generator to choose the order of entry. The evidence labels are produced in JASP, which are based roughly on Jeffries (1961). See Wagenmakers and colleagues (2018) for more detail.

Replicating prior research, there was extremely strong evidence in favor of the alternative hypothesis (young > old) for all three of the RRS subscales. Depression showed the smallest difference between the young (M = 1.96 SD = .60) and older (M = 1.65, SD = .46) groups, d = .57, B_H(0,.30) = 313.70. The difference between young (M = 1.97, SD = .66) and old (M = 1.60, SD = .38) was slightly larger for Brooding, d = .67, B_H(0,.30) = 2,590.41. Reflection showed the largest difference between young (M = 2.14, SD = .74) and old (M = 1.66 SD = .58), d = .71, B_H(0,.30) = 6,508.62.

For the RPA emotion focus and self-focus scales, the Bayes factors indicated that our data were approximately three times as likely under the null hypothesis than under the alternative hypothesis (old > young). Young adults reported slightly more emotion focus (M = 2.75, SD = .61) than older adults (M = 2.70, SD = .54), B_H(0,.30) = 0.29. Similarly, young adults reported slightly more self-focus (M = 2.10, SD = .66) than older adults (M = 2.03, SD = .63), B_H(0,.30) = 0.29. Young adults also reported slightly more dampening (M = 1.79, SD = .56) than older adults (M = 1.72, SD = .42), but the Bayes factor indicated that our data were as likely under the null as under the alternative hypothesis for this scale, B_H(0,.30) = 1.06.

Discussion

This study used Bayesian analyses to investigate whether there are age differences in positive rumination. Although age differences in negative rumination were large and robust, we found no age difference in the use of positive rumination. This supports a mechanistic account of the positivity effect based solely on reduced attention to negative mood states in older adults.

This finding has several implications. First, we believe our results are broadly consistent with the age-related differences in goals proposed by SST. We suggest, however, that age differences in rumination may be more driven by a decline in goals related to information gathering across age. That is, reflection on past negative events may help young adults gather information to avoid future mistakes, even as it raises depression vulnerability.

Second, these results may be useful for designing future controlled laboratory research on age differences in emotion regulation. As noted by the extremely large Bayes factors and accompanying sequential plots, age differences in negative rumination are robust and can be found with relatively small sample sizes. In contrast, the evidence in favor of the null hypothesis for positive rumination did not accumulate beyond “anecdotal” status until the total sample size was over 100. In other words, larger sample sizes may be needed to study age differences (or lack thereof) in positive emotion regulation compared to age differences in negative emotion regulation.

Third, past research suggests that age differences in emotional experience are not particularly large for sadness. The RRS, however, is framed specifically in terms of sadness. This presents somewhat of a puzzle, although we do note that rumination is about the response to sadness, and not the experience of sadness itself. RRS scores also tend to be elevated across a broad range of psychopathologies (eg, anxiety disorders), suggesting that that it may reflect a general response to negative mood. Future research could investigate whether this result varies as a function of the specific negative emotion being experienced.

Although this study has important strengths, there are several limitations. First, these data were collected for exploratory purposes that were secondary to the experimental questions of the original studies. Second, we made a theoretical choice to base our Bayesian analysis on the assumption that older adults would engage in positive rumination more than younger adults. This does not rule out the possibility that young adults engage in more positive rumination than older adults. Finally, it is important to note that this study is cross-sectional. Mental illness is strongly linked to shorter life, with a recent meta-analysis suggesting that having a mental illness reduces life expectancy by an average of 10 years (Walker, McGee, & Druss, 2015). Thus, lower rates of negative affect at older ages may in part reflect the earlier deaths of those with mental illness.

Funding

This work was supported by the National Institute on Aging grant 1R15AG051017-01 to L.E.