Introduction

Emotional facial expressions convey important social information, as they allow us to predict information about the emotions and intentions of others in our environment. Interestingly, although some expressions convey clear information that predicts positive (happy) or negative (angry) outcomes, other expressions (surprise) are associated with both positive (e.g., birthday party) and negative (e.g., car accident) outcomes. Without contextual information that clarifies this valence ambiguity, these expressions can be used to delineate valence bias, or one’s tendency to interpret ambiguity as positive or negative (Kim et al. 2003, 2004; Neta et al. 2013, 2009; Neta and Tong 2016). This valence bias seems to represent a trait-like difference that is stable across time (Neta et al. 2009) and across stimulus type (Neta et al. 2013; Neta and Tong 2016). However, little attention has been paid to whether or not this bias is malleable; that is, what is the extent to which this stable bias can be shifted?

Across longer timescales, there is evidence of such a shift. Previous research has demonstrated a developmental trend of valence bias across the lifespan, with children showing a more negative bias (Tottenham et al. 2013) and older adults showing a more positive bias (Neta and Tong 2016; Shuster et al. 2016). The age-related (positive) bias in older adulthood is an extension of the age-related positivity effect, which has been observed in cognitive domains such as attention (e.g., Mather and Carstensen 2003, 2005; Mather et al. 2005) and memory (e.g., Charles et al. 2003; Kennedy et al. 2004; Mather and Carstensen 2003), such that as people get older, positive information is processed more favorably than negative information. Even in response to facial expressions, older adults perceive more positive affect in faces whose expressions are morphed across valence boundaries (happy/sad, happy/angry, and happy/fearful; Kellough and Knight 2012). The framework of this bias toward positivity is grounded in the socioemotional selectivity theory (Carstensen 1993). According to this theory, time perspectives influence goals such that an extended time perspective (associated with early life stages, e.g., young adulthood) prioritizes future-focused, preparatory goals such as gaining knowledge, exploring, and experiencing novelty; whereas a limited time perspective (associated with later life stages, e.g., older adulthood) prioritizes present-focused goals aiming to achieve well-being and emotional gratification (Carstensen 1993, 1995; Carstensen et al. 1999). This theory has been applied within a population of college students by comparing seniors, who naturally had a more limited time perspective with regards to their college experience, to freshmen (Pruzan and Isaacowitz 2006). Specifically, seniors spent less time viewing sad images and reported higher levels of positive affect than freshmen. Taken together, this theory postulates that time perspectives, not age per se, are what causes the motivational shift contributing to increased positivity in older adults.

Consistent with this hypothesis, numerous studies have found that when younger adults adopted a limited time perspective, or when older adults adopted an extended time perspective, the age-related positivity effect diminished (e.g., Cypryańska et al. 2014; Fung et al. 1999; Kellough and Knight 2012; Pruzan and Isaacowitz 2006). For instance, when choosing social partners under normal circumstances, younger adults prefer novel social partners (favoring exploratory goals) while older adults prefer familiar partners (favoring emotionally gratifying goals). However, instilling an extended time perspective (living 20 years longer due to medical advancement) in older adults (Kellough and Knight 2012), or a limited time perspective (immigration to another country on one’s own) in younger adults shifted preferences for partners in the opposite direction (Fung et al. 1999).

Here, we tested the effect of time perspectives on valence bias when resolving dual-valence ambiguity (surprised faces). We hypothesized that a limited time perspective (i.e., ending a life stage as an undergraduate) would shift participants’ motivation to focus on emotionally satisfying goals, and thus shift valence bias in the positive direction. Contrastingly, an extended time perspective (starting college life) would have the opposite effect: shifting motivation to exploratory goals, and thus shifting valence bias in the negative direction.

Experiment 1

Methods

Procedure

The first study examined the effects of a limited time perspective on valence bias. Participants completed two sessions approximately 1 week apart. In the first session, we measured valence ratings of angry, happy and surprised faces (baseline). In the second session, participants repeated the task, but only after they were given the limited time perspective manipulation adapted from Kellough and Knight (2012): “Additionally, we also want you to imagine that you are a graduating senior, and today is the last day that you will be a student at [the University of Nebraska-Lincoln]. You made it to graduation day! In a few days, you will be leaving Lincoln and your life as a college student is coming to a close. Think about what you will do with the time you have left. Keep this new perspective in mind while you judge the following facial expressions.”

Both sessions contained four blocks, each with 24 trials (12 surprise, 6 happy, 6 angry). Participants saw each image for 500 ms, followed by an ISI (varying 1500–3500 ms), and rated faces as positive or negative using a two-alternative forced choice button response, consistent with previous work (Neta et al. 2009, 2011, 2013; Neta and Whalen 2010). Participants were able to make a response at any point throughout the trial (total available response time varied from 2000 to 4000 ms), and were asked to respond as quickly and accurately as possible. There was a 10 s break between each block.

Stimuli

The stimuli were taken from previous work (Neta et al. 2013), including a set of 48 faces, 24 with an ambiguous valence (surprised), and 24 with a clear valence (12 angry and 12 happy). Of the 48 faces, 34 discrete identities were used (17 male, 17 female) posing angry, happy, and surprised expressions: 14 identities (7 male, age 21–30 years old) from the NimStim standardized set (Tottenham et al. 2009), and 20 identities (10 male, age 20–30 years old) from the Karolinska Directed Emotional Faces database (KDEF; Lundqvist et al. 1998). The face stimuli were split into two subsets (24 images each) that were not significantly different from each other in terms of valence ratings (p > .2), based on data from a previous study (Neta and Tong 2016). One subset was used in Session 1 (baseline; before the manipulation), and the other subset was used in Session 2 (after the manipulation). The order in which each subset was used in each session was counterbalanced across participants.

Participants

The intended sample size was informed by the sample size of previous research in the same area of study (Neta and Whalen 2010). Thirty-nine participants volunteered; three were excluded due to high error (> 40%) for angry and happy ratings. The final sample included 36 participants (Mean age = 20.17, SD = 3.13, 23 females) with a mix of class standing (15 Freshmen, 11 Sophomores, 5 Juniors, 4 Seniors, 1 unknown). All participants identified as Caucasian/non-Hispanic, with the exception of one participant who identified as Asian/Pacific Islander. None were aware of the purpose of the experiment, and all were compensated for their participation through monetary payment or course credit. Written informed consent was obtained from each participant before the session, and all procedures were approved by the University of Nebraska Committee for the Protection of Human Subjects.

Results

Valence ratings

Participants rated clear items accurately (Session 1 average accuracy: angry = 95.57%, 95% CI [93.53–97.61%]; happy = 95.88%, 95% CI [93.77–97.99%]; Session 2: angry = 95.80%, 95% CI [93.49–98.11%]; happy = 96.92%, 95% CI [95.34–98.50%]). Similar to previous work (e.g., Neta et al. 2009, 2013; Neta and Whalen 2010), valence ratings for surprised faces were coded using percent negative ratings, which is calculated as the percent of trials that surprise was rated as negative, out of the total number of surprise trials, excluding omissions (see Neta et al. 2009). Consistent with previous work (Neta et al. 2009, 2013), there were individual differences in ratings of surprised faces (Session 1 mean = 64.84%, SD = 25.16%, 95% CI [56.33–73.36%]; Session 2 mean = 55.32%, SD = 27.62%, 95% CI [45.97–64.66%]).

A composite rating combing responses from angry and happy (clearly valenced) trials was created in order to directly compare clear trials to surprise trials. The composite rating is an average of angry and happy accuracy scores (percent of angry trials rated as negative and percent of happy trials rated as positive, excluding omissions). However, because there is no correct answer for surprised faces, this composite rating excludes data from surprise trials, which were coded using percent negative ratings. An Order (Session 1, Session 2) × Valence (Clear, Ambiguous) repeated measures ANOVA revealed a significant Order × Valence interaction [F(1,35) = 6.32, p = .02, partial η2 = 0.15], such that ratings of surprised faces were more positive in Session 2 (after the manipulation) compared to Session 1 (p = .02, Cohen’s d = 0.40), but there was no change in ratings for clearly valenced faces (p = .41; Fig. 1a).

Fig. 1
figure 1

Mean valence ratings in Experiments 1–3. a In Experiment 1, surprised faces were rated more positively in Session 2 (limited time perspective) compared to Session 1 (baseline), but there was no change in ratings of clearly valenced faces. b In Experiment 2, no significant differences in valence ratings were found. c In Experiment 3, surprised faces were rated more negatively in Session 2, 3, and 4 (extended time perspective) compared to Session 1 (baseline), but there was no change in ratings of clearly valenced faces. Error bars represent standard error

Full size image

Reaction time (RT)

As with valence ratings, we combined RTs for angry and happy faces into a composite for clearly valenced faces. An Order (Session 1, Session 2) × Valence (Clear, Ambiguous) repeated measures ANOVA on RTs revealed only a main effect of Valence [F(1,35) = 113.47, p < .001, partial η2 = 0.76], such that RTs were longer for surprised than clearly valenced expressions, consistent with previous work (e.g., Neta et al. 2009, 2011).

Experiment 1 discussion

As predicted, after the limited time perspective manipulation, ratings of clearly valenced faces (angry, happy) were unchanged but surprise was rated more positively. This suggests that, consistent with socioemotional selectivity theory (Carstensen 1993), a limited time perspective prioritizes present-focused goals aimed to achieve well-being and increases positive affect (Carstensen 1993, 1995; Carstensen et al. 1999). This time perspective manipulation was robust to our task of rating ambiguously valenced facial expressions, even within subjects that were simply asked to imagine limited time remaining in their college careers, as in previous work (Pruzan and Isaacowitz 2006). However, one alternative explanation for our findings is that surprise ratings might become more positive with time and/or practice with the task. Indeed, more time (operationalized as longer reaction times within a trial) is associated with more positive ratings of ambiguously valenced images (Neta and Tong 2016). Therefore, in Experiment 2, we set out to test this effect of repeated exposure on valence ratings of surprise.

Experiment 2

Methods

Procedure and stimuli

Similar to Experiment 1, participants completed 2 sessions approximately 1 week apart. The sessions were identical to Experiment 1, except that in Session 2 (7.36 days on average after Session 1), there was no additional manipulation (i.e., participants were given the same instruction as in Session 1 without any time perspective manipulation). This allowed us to examine the effect of repeated performance on ratings, and to determine whether the change observed in Experiment 1 was attributed to the time perspective manipulation or other factors such as practice.

Participants

Thirty-six participants volunteered; one participant was excluded due to high error (> 40%) in rating angry faces. The final sample included 35 participants (Mean age = 20.06, SD = 1.35, 21 females) with a mix of class standing (11 Freshmen, 13 Sophomores, 8 Juniors, 3 Seniors), all identifying as Caucasian/non-Hispanic. None were aware of the purpose of the experiment, and they were all compensated for their participation through monetary payment or course credit. Written informed consent was obtained from each participant before the session, and all procedures were approved by the University of Nebraska Committee for the Protection of Human Subjects.

Results

Valence ratings

Similar to Experiment 1, participants rated clear items accurately (Session 1 average accuracy: angry = 95.91%, 95% CI [93.86–97.96%]; happy = 92.64%, 95% CI [89.05–96.24%]; Session 2: angry = 94.95%, 95% CI [92.15–97.75%]; happy = 94.55%, 95% CI [91.78–97.32%]). Valence ratings for surprised faces were coded using percent negative ratings, as described previously. Consistent with previous work (Neta et al. 2009, 2013) and with Experiment 1, there were individual differences in ratings of surprised faces (Session 1 mean = 64.98%, SD = 17.66%, 95% CI [58.92–71.05%]; Session 2 mean = 61.82%, SD = 23.98%, 95% CI [53.58–70.06%]).

An Order (Session 1, Session 2) × Valence (Clear, Ambiguous) repeated measures ANOVA on valence ratings revealed no significant Order × Valence interaction [F(1,34) = 0.85, p = .37, partial η2 = 0.02; Fig. 1b].

Reaction time (RT)

An Order (Session 1, Session 2) × Valence (Clear, Ambiguous) repeated measures ANOVA on RTs revealed a main effect of Valence [F(1,34) = 55.24, p < .001, partial η2 = 0.62], such that RTs were longer for surprised than clearly valenced expressions, as in Experiment 1. There was also an Order × Valence interaction [F(1,34) = 5.70, p = .02, partial η2 = 0.14], such that RTs for clearly valenced expressions were faster in Session 2 than Session 1 (p = .004, Cohen’s d = 0.53), but there was no such effect for surprised faces (p = .60).

Experiment 2 discussion

We found that merely asking people to return a week later and repeat the valence rating task, without adopting a time perspective, did not change ratings of surprised faces. This is consistent with work showing that valence bias is stable over 1 year (Neta et al. 2009). Further, although there appears to be some effect of practice on rating clearly valenced faces (reaction times were faster in Session 2 than Session 1), there was no evidence of practice effects for surprised faces. These results lend further support for the hypothesis that a change in time perspective shifts valence bias since simply repeating the task did not result in such a shift.

However, another alternative explanation for the Experiment 1 findings is that the prompt (imagining graduation) induced a state of positive affect, which, rather than the limited time perspective per se, resulted in more positive ratings of surprise. Moreover, it is unclear how long the effect lasts (i.e., if ratings of surprise return to baseline right away or if there are more long-lasting effects on ratings). To test both of these effects, we ran another study to examine the reverse effects of time perspective (extended time perspective) that included four sessions across a longer time frame. In this experiment, participants were also asked to rate the valence of the prompt to determine if this influences ratings of surprise.

Experiment 3

Methods

Procedure

In contrast to the methods of Experiments 1 and 2, participants completed 4 sessions in a 2-week period. Similar to previous experiments, Session 2 occurred approximately 1 week after Session 1 (average = 7.2 days). Then, Session 3 occurred an average of 3.1 days after Session 2, and Session 4 occurred an average of 4 days after Session 3 (overall, approximately one additional week elapsed between Sessions 2 and 4). Consistent with previous experiments, participants rated angry, happy, and surprised faces as positive or negative (baseline), based on a gut reaction. In Session 2, participants repeated the task, but only after they were given the extended time perspective manipulation: “Additionally, we also want you to imagine that you are an entering freshman, and today is the first day that you will be a student at [the University of Nebraska-Lincoln]. You made it to college! In a few days, you will be starting classes in Lincoln and your life as a college student is coming to a beginning. Think about what you will do with the time you have here. Keep this new perspective in mind while you judge the following facial expressions.” Finally, in Sessions 3 and 4, participants were asked to complete the valence rating task on a new set of stimuli with no mention of the extended time perspective manipulation.

Stimuli

Because we added two sessions, we doubled the stimuli used, adding stimuli from the Umea University Database of Facial Expressions (Samuelsson et al. 2012). These stimuli consisted of 29 discrete identities (ages 17–39) with the highest hit rate (i.e., most accurately labeled expressions according to normative ratings). As such, some of the 29 identities were represented in only one or two of the expression conditions (angry, happy, surprise), and others were represented in all three conditions, resulting in 48 discrete stimuli. Of these 48 faces (24 males, 24 females), there were 24 faces with an ambiguous valence (surprised expression), and 24 faces with a clear valence (12 angry and 12 happy). This set was split into two subsets (24 images each) that were not significantly different from each other in age (p = .11) or hit rates (p = .38). Together with the stimuli used in Experiments 1 and 2, we had four subsets of images (each included 6 angry, 6 happy, and 12 surprised faces), which were used for each subset of the four sessions. The order in which each subset was used was pseudorandomized such that participants would see a new set in each session, and images in each subset were presented in a pseudorandomized order for all participants.

Participants

Forty-two participants volunteered; six were excluded due to attrition. The final sample included 36 participants (Mean age = 20.28, SD = 3.34, 34 females) with a mix of class standing (7 freshmen, 16 sophomores, 5 juniors, 8 seniors), all identifying as Caucasian/non-Hispanic. None were aware of the purpose of the experiment, and they were all compensated for their participation through course credit. Written informed consent was obtained from each participant before the session, and all procedures were approved by the University of Nebraska Committee for the Protection of Human Subjects.

Results

Valence ratings

Similar to Experiments 1 and 2, participants rated clear items accurately (Session 1 average accuracy: angry = 95.11%, 95% CI [92.77–97.86%]; happy = 92.61%, 95% CI [95.27–89.38%]; Session 2: angry = 96.08%, 95% CI [93.19–98.31%]; happy = 91.18%, 95% CI [94.98–87.87%]; Session 3: angry = 97.84%, 95% CI [96.58–98.85%]; happy = 93.37%, 95% CI [95.79–90.75%]; Session 4: angry = 98.16%, 95% CI [97.01–99.46%]; happy = 92.38%, 95% CI [97.04–87.31%]), and valence ratings for surprised faces were coded using percent negative ratings. Consistent with previous work (Neta et al. 2009, 2013) and with Experiments 1 and 2, there were individual differences in ratings for surprised faces (Session 1 mean = 67.45%, SD = 22.06%, 95% CI [59.98–74.91%]; Session 2 mean = 81.98%, SD = 16.99%, 95% CI [76.22–87.72%]; Session 3 mean = 84.93%, SD = 17.95%, 95% CI [78.85–91.01%]; Session 4 mean = 76.63%, SD = 16.66%, 95% CI [70.99–82.27%]).

An Order (Session 1, Session 2, Session 3, Session 4) × Valence (Clear, Ambiguous) repeated measures ANOVA on valence ratings revealed a significant Order × Valence interaction [F(3,33) = 28.49, p < .001, partial η2 = 0.72], such that ratings of surprised faces were more negative in Sessions 2, 3 and 4 (after the manipulation) compared to Session 1 (Session 2: p < .001, Cohen’s d = 0.74; Session 3: p < .001, Cohen’s d = 0.87; Session 4: p = .006, Cohen’s d = 0.47), and then became more positive in Session 4 compared to Sessions 2 and 3 (Session 2: p = .03, Cohen’s d = 0.17; Session 3: p < .001, Cohen’s d = 0.48), but there was no change in ratings for clearly valenced faces (p’s > 0.3; Fig. 1c).

To compare these findings to Experiments 1 and 2, we ran an additional analysis from just the first two sessions (since Experiments 1 and 2 contained only 2 sessions). An Order (Session 1, Session 2) × Valence (Clear, Ambiguous) repeated measures ANOVA revealed a significant Order × Valence interaction effect [F(1,35) = 21.60, p < .001, partial η2 = 0.32], showing that ratings of surprised faces were more negative in Session 2 (after the manipulation) than Session 1 (p < .001, Cohen’s d = 0.74), but there was no change in ratings for clearly valenced faces (p = .67).

Reaction time (RT)

An Order (Session 1, Session 2, Session 3, Session 4) × Valence (Clear, Ambiguous) repeated measures ANOVA on RTs revealed only a significant main effect of Valence [F(1,35) = 83.58, p < .001, partial η2 = 0.71], such that RTs were longer for surprised than clearly valenced expressions, as in previous experiments.

Again, to compare these findings to Experiments 1 and 2, we ran an additional analysis on RTs from just the first two sessions. An Order (Session 1, Session 2) × Valence (Clear, Ambiguous) repeated measures ANOVA revealed a main effect of Valence [F(1,35) = 70.33, p < .001, partial η2 = 0.67], mimicking previous experiments, such that RTs were longer for surprise than clear expressions.

Ratings of the prompt

Participants were asked to rate the valence of the extended time perspective prompt on a scale of 1–5 (1 = extremely negative, 3 = neither negative nor positive, 5 = extremely positive). A one sample t-test was performed to determine if a significant difference existed between participant ratings of the prompt (average ± standard error = 3.41 ± 0.14) and the neutral midpoint of the prompt scale (neutral = 3). Results of this test show that ratings were significantly more positive than neutral [t(35) = 22.42, p < .001]. To test whether ratings of the prompt had any observed effect on valence ratings, correlational analyses were performed between ratings of the prompt and the rate of change in surprise ratings between Session 1 and all subsequent sessions, but no effects were found (all p’s > 0.10).

Comparing the samples across three experiments

Levene’s test revealed a trend toward inhomogeneity in variance of Session 1 (baseline) surprise ratings across our three samples (p = .09). As such, a Kruskal–Wallis test was performed to determine if the Session 1 ratings across the three experiments were significantly different from each other (i.e., if participants showed differences in their baseline). Ratings of clearly valenced faces [χ2(2) = 1.64, p = .44] and surprised faces [χ2(2) = 1.13, p = .57] were not significantly different across the samples.

Levene’s test revealed no significant differences in variance of age across our three samples (p = .25). Therefore, a one-way ANOVA was performed to assess age differences across the three samples, revealing no significant age differences (p = .94).

Experiment 3 discussion

Consistent with our hypothesis, after the extended time perspective manipulation, ratings of clearly valenced faces (angry, happy) were unchanged, but ratings of surprise became more negative. This is consistent with previous work (Carstensen 1993, 1995; Carstensen et al. 1999) that demonstrated that an extended time perspective prioritizes future-focused, preparatory goals associated with increased negative affect.

The effect of the extended time perspective on surprise ratings was relatively long-lasting. Specifically, the increased negativity was evident immediately after the manipulation (Session 2), and maintained at Session 3 (2–3 days later). Indeed, ratings did not begin returning to baseline until Session 4 (a week after the manipulation), and even then, ratings were more negative than baseline. It is worth reiterating that the additional sessions (Sessions 3 and 4) did not include any manipulation, and Experiment 2 demonstrates that repeat sessions do not significantly change the surprise ratings over a similar time frame.

Finally, participants rated the prompt as mildly positive. Given that ratings of surprised faces were more negative after the manipulation, this suggests that the shift toward negativity occurred despite the positive affect associated with the prompt. As such, the time perspective, per se, was successful at shifting valence ratings of surprised faces. Interestingly, both time perspectives (limited perspective describes graduation day and extended perspective describes the first day of college) describe events that have been previously demonstrated to elicit ambivalent responses, in that there is both excitement and anxiety when thinking about the future. As such, we do not expect that the ambivalence in mood induced would likely have a strong effect on ratings in either direction, much less different effects in the different experiments (more positivity with a limited perspective and more negativity with an extended perspective). However, because prompt ratings were not a direct measure of participant mood, the absence of a mood measure limits interpretative evidence as to how mood might have influenced valence ratings.

General discussion

These findings reveal that time perspectives are sufficient for shifting responses to ambiguity. Specifically, a limited time perspective resulted in more positive ratings of surprise (Experiment 1), and an extended time perspective resulted in more negative ratings of surprise (Experiment 3). This change in ratings was specific to surprised faces, as ratings of clearly valenced faces (angry, happy) were unchanged. Moreover, this effect cannot be attributed to mere experience with the task, as simply repeating the task without a time perspective manipulation did not change ratings (Experiment 2). Importantly, there were no differences in baseline valence bias or age across the three samples.

Previous studies demonstrated a developmental trend for valence bias with children showing a negative bias and older adults showing a positive bias. The positive bias in older adults is an extension of the age-related positivity effect, which according to the socioemotional selectivity theory (Carstensen 1993, 1995; Carstensen et al. 1999), is related to a motivational shift in older adults who naturally adopt a limited time perspective. Children, on the other hand, naturally adopt an extended time perspective, relying more on future-focused, preparatory goals that prioritize acquiring knowledge, exploring, and experiencing novelty, and are associated with increased negative affect. Our findings suggest that, despite the stability of the valence bias over time (Experiment 2 shows stability over 1 week; Neta et al. 2009 shows stability over 1 year), the valence bias is malleable, and is sensitive to manipulations of time perspective. This is consistent with work demonstrating that ratings of surprise are vulnerable to explicit manipulations of context (Neta et al. 2011; Davis et al. 2016) and deliberation (Neta and Tong 2016), but extends this work by showing that (a) time perspectives, even within the brief college experience, are sufficient for modulating the valence bias, and (b) this manipulation has relatively long-lasting effects (i.e., at least 1 week; Experiment 3).

Interestingly, it is possible that the manipulation had a lasting effect such that participants adopted a new approach or decision-making process when resolving emotional ambiguity. But it is also possible that participants drafted associations between certain factors (e.g., testing room, research assistant) and the time perspectives in the prompt from the previous session(s), and were merely primed with those associations in subsequent sessions. It is difficult from these data to adjudicate the alternative explanations for this long-lasting effect. Future work will be needed to examine the source of the long-lasting effects, and perhaps, whether this might generalize to other ambiguous stimuli or other testing environments.

Finally, the otherwise consistency among sessions (e.g., same face stimuli, testing room, experimenter, instructions, etc.) suggests that the changes in ratings are related to the time perspective manipulation per se. However, the lack of a measure confirming that time perspective was indeed manipulated in each participant is a limitation in the current study. Having said that, the effect of time perspective is further supported by Experiment 2, in which we found no change in ratings when no manipulation was provided.

Time perspectives and the initial negativity hypothesis

Despite individual differences in valence bias that have been reported (Neta et al. 2009, 2013; Neta and Tong 2016; Kim et al. 2003), the more automatic interpretation is negative (Neta and Whalen 2010). Our working model (i.e., the initial negativity hypothesis) suggests, then, that positive interpretations may require an additional regulatory process that only some individuals adopt naturally. This model is consistent with previous work demonstrating that older adults, who naturally adopt a more limited time perspective associated with greater emotion regulation, show a more positive valence bias (Neta and Tong 2016). Relatedly, in the present work, we found that instilling a more limited time perspective also resulted in a more positive bias. In contrast, previous work has shown that children, who naturally adopt a more extended time perspective associated with weaker emotion regulation, may be relying on the automatic (negative) response (Tottenham et al. 2013). In the present work, we found that instilling a more extended time perspective also resulted in a more negative bias. Interestingly, comparing effect sizes between age-related changes (Neta and Tong 2016; Tottenham et al. 2013) and the present findings, it seems that the effects may have been smaller in the aging data compared to the time perspective data. This is not surprising given that the socioemotional selectivity theory posits that age-related changes in emotion are driven by changes in goals and time perspectives. As such, it could be that testing the effects of time perspectives is a more sensitive and direct measure of age-related changes (and perhaps less noisy) than age itself.

Notably, the present findings support a link between the positive valence bias and socioemotional selectivity theory. In other words, older adults’ greater positivity is attributed to their goals, rather than other mechanisms such as difficulties in emotion perception. Future work will shed light on how the aging brain overrides an automatic negativity, and will identify this developmental trajectory toward positivity by comparing the mechanisms underlying the bias in young adults to a population of subjects with a naturally enhanced positive bias (older adults). Specifying the strength and reliability of these putative regulatory mechanisms will allow for major theoretical advances on the factors that drive responses to uncertainty.

Conclusions

The current study examined factors that influence interpretations of ambiguity. We compared the effects of limited and extended time perspectives on valence bias. Results showed that (1) socioemotional selectivity theory extends to situations of ambiguity, (2) valence bias is malleable and sensitive to time perspectives, and (3) these effects are relatively long lasting. Further, the effect of time perspective was consistent with the natural developmental trend towards positivity. These findings suggest that where we are in life (i.e., age, outlook) can have profound effects on how we process emotions.