Depressive symptoms and time perspective in older adults: associations beyond personality and negative life events

Abstract Objectives: To examine the extent to which time perspective, an individual’s habitual way of relating to the past, the present, and the future time frames, accounts for variations in self-reported depressive symptoms among older adults. Method: Four hundred two participants (60–90 years) completed the Center for Epidemiological Studies Depression scale (CES-D) and the Swedish Zimbardo Time perspective Inventory (S-ZTPI). The influence of personality as reflected by the Temperament and Character Inventory (TCI) and self-reported negative life events (NLEs) were controlled for in hierarchic regression analyses. Results: The six S-ZTPI dimensions accounted for 24.5% of the variance in CES-D scores beyond age and gender. Half of the variance remained when the TCI factors and NLEs were controlled for. Past Negative, Future Negative, and Past Positive (inverse association) were the significant unique predictors. Significant age interactions were observed for two S-ZTPI dimensions, with a diminished association to depressive symptoms for Future Negative and a magnified association for Present Fatalistic with higher age. Conclusions: The results demonstrate a substantial relation between facets of time perspective and depressive symptoms in old age. They also indicate an age-related shift in the relative importance from concerns about of the future (Future Negative) to the present (Present Fatalistic) with increased age. In young old-age, when the future is more ‘open’, future worries (Future Negative) may be a more frequent source of distress. In late senescence, perceived threats to autonomy (e.g. physical health problems and cognitive deficits), as reflected by higher scores on Present Fatalistic, may instead have more bearing on mood state.


Introduction
Population aging posits several important challenges to society, including maintenance of good health in the older population (Christensen, Doblhammer, Roland, & Vaupel, 2009). One of the most widespread health problems is depression (Byers, Yaffe, Covinsky, Friedman, & Bruce, 2010). Late life depression is more strongly related to suicide than depression at earlier ages (Fiske, Wetherell, & Gatz, 2009) and has been associated with other adverse outcomes, for example: accelerated cognitive decline (Lockwood, Alexopoulos, & van Gorp, 2002), increased dementia risk (Diniz, Butters, Albert, Dew, & Reynolds, 2013), more complicated recovery from medical illnesses such as cardiovascular disease (Hare, Toukhsati, Johansson, & Jaarsma, 2014), and increased risk of nonsuicide mortality due to, for example, impaired physical and social functioning, and greater self-neglect (for a review, see Blazer, 2003). Thus, there is a great need to identify factors that advance knowledge of depressive symptoms in older adults.
One concurring factor has been suggested to be an individual's time perspective (Desmyter & De Raedt, 2012). Time perspective refers to how individuals subjectively evaluate and focus on the past, the present and the future (e.g. , and there are indications that aspects of time perspective are related to well-being (R€ onnlund, Åstr€ om, & Carelli, 2017) as well as mental health problems in older adults (Desmyter & De Raedt, 2012). The main focus of the present study was to examine how time perspective is related to current depressive symptomatology in older adults.

Aging and time perspective
Scholars within different theoretical perspectives have integrated facets of time perspective, arguing that it is closely linked to the aging process and may have important consequences for mental health at old age (Carstensen, 2006;Shmotkin & Eyal, 2003;Westerhof, Bohlmeijer, & Webster, 2010). Social selectivity theory (SST) posits that older adults become less focused on the future and more focused on the immediate present, as their future time horizons shrink (Lang & Carstensen, 2002). This in turn influences social and emotional behavior, in the sense that older adults tend to trim their social networks so as to increase closer and emotionally meaningful relationships (Lang & Carstensen, 2002). Studies within the framework of reminiscence, have further shown that thinking about the past in terms of life review and reminiscence in old age, may be both positively and negatively related to mental health (for a review, Bohlmeijer, Roemer, Cuijpers, & Smit, 2007). As concerns depression, a specific form of reminiscence referred to as 'bitterness revival' (Westerhof et al., 2010), which involves rumination about past injustices, has specifically been linked to depression among older adults (Cully, LaVoie, & Gfeller, 2001).
The majority of past studies have though focused on the influence of either the past or the future, without considering the totality of older adults' views of the past, the present, and the future time frames. Hence, previous studies failed to assess the relative strengths of the other dimensions within individual temporal profiles. In the present study, time perspective was conceptualized according to Zimbardo and Boyd's (1999) time perspective theory, by which time perspective is defined as 'the often non-conscious process whereby the continual flows of personal and social experiences are assigned to temporal categories, or time frames, that help give order, and meaning to those events' , p.1271. Within this theoretical framework, not only the impact of the past, the present, and the future is simultaneously taken into account, but central is the valence and relative focus put on each time frame.

The Zimbardo time perspective inventory
Along with their theoretical framework, Zimbardo and Boyd developed the Zimbardo Time Perspective Inventory (ZTPI; , a self-report questionnaire, which is currently the most frequently used multidimensional measure of time perspective. The ZTPI has five subscales: (1) Past Negative, reflecting negative and aversive views of the past; (2) Past Positive, which measures sentimental and nostalgic views of the past; (3) Present Hedonistic, capturing immediate pleasure seeking with little consideration of future consequences; (4) Present Fatalistic, which reflects a fatalistic and hopeless attitude towards the present; and (5) Future, which measures a general future orientation, including optimism and a strive for future goals and rewards. In the present study, we used the Swedish version of the ZTPI (S-ZTPI; Carelli, Wiberg, & Wiberg, 2011). S-ZTPI is identical to the original inventory with regard to the past and present dimensions, whereas the future dimension is extended so as to differentiate between two future factors: Future Positive and Future Negative. Future Positive is virtually identical to the Future scale of ZTPI, whereas Future Negative includes thoughts and attitudes toward the future that comprise negative expectations. To capture a negative future time perspective may be particularly important to get a more complete understanding of how time perspective is related to depression. Theoretical (Abramson, Metalsky, & Alloy, 1989;Beck, 1967) as well as empirical work (Korn, Sharot, Walter, Heekeren, & Dolan, 2017) have highlighted that negative views of the future (i.e. pessimism, hopelessness) are key symptoms or even causes behind depression. For example, according to Beck's cognitive theory of depression, the information processing of a depressed individual is centered within a cognitive triad, involving negative thoughts about oneself ('I am worthless'), the environment ('the world is unfair'), and the future ('nothing will ever get better'), which, in turn, both cause and sustain depressive mood (Beck, 1967). Some scholars have further underlined that negative prospection (i.e. negative mental representations of the future) might be a core causal mechanism behind depression (Roepke & Seligman, 2016). In fact, according to Roepke and Seligman (2016) 'the entire cognitive triad may actually boil down to negative future thinking ' (p. 25). From Roepke and Seligman's point of view, negative views of the future might be what leads to depression because negative views about oneself and the environment would be less discouraging if the individual could envision a more positive future.

Time perspective and mental health in older adults
Studies utilizing the ZTPI have demonstrated that individual differences in time perspective in young adults are associated with a variety of important outcomes, such as risk taking (Jochemczyk, Pietrzak, Buczkowski, Stolarski, & Markiewicz, 2015), substance use (Keough, Zimbardo, & Boyd, 1999), and health-promoting behaviors (Griva, Tseferidi, & Anagnostopoulos, 2014); but also to aspects of psychological distress, such as anxiety and depression (van Beek, Berghuis, Kerkhof, & Beekman, 2011;. More specifically, Past Negative has been associated with anxiety, depression and neuroticism (van Beek et al., 2011; and Present Fatalistic has been related to depression . Research on time perspective (using the ZTPI or the S-ZTPI) and mental health in older adults is however sparse and it is not fully clear whether the foregoing results generalize to the older population. To date, only one study explored the relationship between time perspective and depressive symptoms in older adults, using Zimbardo and collaborators framework (Desmyter & De Raedt, 2012). Interestingly, Desmyter and De Raedt (2012) found that Present Fatalistic was related to depressive symptoms, whereas higher Past Negative was related to general negative affect, lower life satisfaction, and depressive symptoms. It should be noted that the aforementioned study used ZTPI and not S-ZTPI, and hence, negative views of the future (i.e. Future Negative) were not taken into account.
Another study (R€ onnlund et al., 2017) that examined time perspective and aspects of mental health (specifically subjective well-being) in older adults, reported a possible age-related pattern of relationship between time perspective and subjective well-being. In the study of R€ onnlund et al. (N = 447, 60-90 years), Future Negative was strongly (and inversely) related to subjective well-being in the young-old group (60-75 years), whereas among the oldest group (80-90 years old), Future Negative showed a small and non-significant association with subjective well-being. This finding points to a diminished relevance of Future Negative for subjective well-being in late senescence. A moderating effect of age was also reported by Simons and colleagues (Simons, Peeters, Janssens, Lataster, & Jacobs, 2016), who examined relationships between perceived happiness and time perspective in younger, middle-aged and older adults (N = 525, 20-87 years). They found that Past Negative and Present Fatalistic were inversely related to happiness, but that these relationships weakened with increasing age.

The present study
The aim of the present study was to examine the relationship between time perspective and depressive symptoms in a sample of older adults (60-90 years). The study extended on previous work (i.e. Desmyter & De Raedt, 2012) in several respects. First, the subdivision of the unitary Future dimension in ZTPI into positive and negative (i.e. by use of the S-ZTPI) allowed examination of the extent to which Future Negative adds to the prediction of depressive symptoms. Second, given prior indications that the relationship between facets of time perspective and mental health (e.g. subjective well-being and happiness) may be moderated by age, we examined interaction terms (i.e. age Â time perspective dimension) as part of regression models. Third, to further clarify the relationship between time perspective and depressive symptoms, we controlled for influence of variables (personality traits and exposure to negative life events) that should be important from viewpoint of validity, but not considered in prior studies of depression/depressive symptoms and time perspective. The inclusion of negative life events and personality traits as control variables was motivated by extensive prior research establishing negative life events and certain personality traits (e.g. neuroticism, harm avoidance) as major risk factors for depression (for reviews, see Fiske, Wetherell, & Gatz, 2009;Klein, Kotov, & Bufferd, 2011).
Based on previous research we hypothesized that Past Negative and Present Fatalistic would be associated with depressive symptoms and that Past Positive would be inversely related to depressive symptoms. Finally, based on reported associations between negative views of the future and depression (although not in the framework of time perspective), we hypothesized that Future Negative would be a predictor of depressive symptoms over and beyond associations with the other S-ZTPI dimensions.

Participants
The study comprised of participants from the Betula prospective cohort study, a population-based, prospective study on aging, cognition, and health (Nilsson, et al., 1997;2004). The Betula study was launched in Umeå, Sweden, in 1988, and participants were initially selected from the population registry of Umeå municipality based on stratified (age, gender) random sampling (for a detailed description of the Betula study in terms of selection criteria, sampling, procedure, and measurements, see Nilsson et al., 1997;2004). The Betula study includes six samples (S1-S6) and data have been collected at six test waves (T1-T6), with five years in between every test wave. The present analyses involved participants in S1 (included in the Betula study at T1) and S3 (included in the Betula study at T2). The reason why S1 and S3 were included in the present study, was that these samples were included in all the test waves from which data in the current study were drawn (T2-T6), and were consequently the only two samples who had data on the main study variables (i.e. time perspective, depressive symptoms, personality, and negative life events). Data on time perspective and depressive symptoms were collected at the sixth test wave (T6, 2013-2014), at which the participants were 60, 65, 70, 75, 80, 85 or 90 years old and S-ZTPI was first included in the test battery. Data on negative life events were collected from T2 to T5 and personality at T3. The included questionnaires were administered at one of the two assessments in each test wave.
Comparison of baseline characteristics of S1, which was comparable to S3 (R€ onnlund, Nyberg, B€ ackman, & Nilsson, 2005), with that of the target population (Umeå municipality), and a comparison between participants and nonparticipants, suggested that the sample had adequate population validity in terms of major demographic variables, including education, income, and marital status (Nilsson et al., 1997). Participants scoring <24 on the Mini-Mental State Examination (MMSE; Folstein, Folstein, & McHugh, 1975) were excluded from the study (n = 9). Participants who had failed to complete one or more of the included measurements were additionally excluded (n = 45). The final sample consisted of 402 participants (M = 70.12 years, SD = 7.36; 55% females).
A summary of select background variables of the participants can be found in Table 1. The gender distribution was fairly equal across the age groups, and there were fewer participants in the older groups, particularly in the two oldest groups (85 and 90 years). Years of education at inclusion in the Betula study, MMSE score and scores on the SRB synonym test (Dureman & S€ alde, 1959) were related to age; such that higher age was related to fewer years of education (r = À0.411, p < 0.001), lower MMSE score (r = À0.178, p. < 0.001), and lower SRB score (r = À0.187, p < 0.001).

Sample attrition
As noted, the included samples were found to be representative of the target population at time of inclusion. However, provided that the samples were subject of attrition over a period of 20 (S3) to 25 years (S1) and thereby reduced from 1966 to 402 participants, we deemed it important to examine the population validity of the final sample of returnees. For this purpose, we performed a binary logistic regression analysis with participant (returnee; coded as 1) and nonparticipant (dropout; coded as 0) as the outcome. Select demographic factors (age, gender, years of schooling at entry), sample (sample 1 = 0, sample 3 = 1), and score on the SRB synonym test (Dureman &  S€ alde, 1959) at entry of the study, were entered as the predictors. As expected given the age range of the samples, higher baseline age was associated with lower of odds of participation, OR = 0.91 (95% CI: 0.90-0.93), p < 0.001, but neither gender, OR = 0.92 (95% CI: 0.72-1.18), p = 0.52, nor years of formal education, OR =0.98 (95% CI: 0.94-1.02), p = 0.25, were significant predictors. In line with prior observations (R€ onnlund et al., 2005), dropout was significantly associated with baseline cognitive level as indicated by higher odds of participation with higher synonym test score, though, OR = 1.12 (95% CI: 1.08-1.16), p < 0.001. Omission of synonym score or years of schooling did not alter levels of significance for neither of the variables (a check being motivated by a moderate association between the variables, r = 0.49, p < 0.001). Thus, except for a relationship with age and slightly better baseline level of performance on the synonym test in participants who were included in the present study, the final sample exhibited little selectivity in regard to sex distribution and education.

Swedish Zimbardo Time perspective inventory
S-ZTPI (Carelli et al., 2011) is a Swedish validation and extension of the original ZTPI . S-ZTPI is a 64-item self-report measure that comprises six subscales: Past Negative, Past Positive, Present Fatalistic, Present Hedonistic, Future Negative, and Future Positive. Items are answered on a 5-point Likert scale ranging from very uncharacteristic (1) to very characteristic (5). S-ZTPI has demonstrated adequate reliability with internal consistency ranging from 0.70 for the Future Positive scale, to 0.84 for the Past Negative scale, and test-retest reliability at 0.60 to 0.85 (Carelli, et al., 2011). Analyses of convergent validity included comparisons with the General Decision Making Styles scale (Scott & Bruce, 1995) and the Barratt Impulsiveness Scale (Patton, Stanford & Barratt, 1995), demonstrated correlations in expected directions (Carelli et al., 2011). In the present sample, Cronbach's a ranged from 0.64 for Future Positive to 0.82 for Past Negative.

Center for epidemiologic studies-depression scale
A Swedish translation (Gatz, Johansson, Pedersen, Berg, & Reynolds, 1993) of the Center for Epidemiologic Studies-Depression scale (CES-D; Radloff, 1977) was used to measure depressive symptoms. CES-D measures current level of depressive symptomatology, with specific focus on the affective component of depression, i.e. depressed mood. The scale comprises 20 items that are answered according to frequency of occurrence: rarely or none of the time; some or little of the time; occasionally or moderate amount of time; most or all of the time. CES-D has demonstrated excellent psychometric properties, with internal consistency at 0.85 (Radloff, 1977) and test-retest reliability of 0.87 (Miller, Anton, & Townson, 2008), as well as adequate construct validity (Weismann, Sholomskas, Pottenger, Prusoff & Locke, 1977). Psychometric evaluation of the Swedish version of CES-D has shown equivalent results (e.g. Haynie, Berg, Johansson, Gatz, & Zarit, 2001). Internal consistency in the current sample was 0.83.

Life event inventory
The Life Event Inventory (Perris, 1984) includes 56 life events about private life, working life, relatives and friends, health, and the deaths of relatives or close friends. For each event, the respondent is asked to indicate whether the event has occurred in the past five years; if the event was expected or unexpected; if the event was experienced as negative or positive (assessed on a five-point scale, coded as 1 = very positive, 2 = positive, 3 = neutral, 4 = negative, 5 = very negative); if the respondent could control whether the event occurred or not; and finally, whether it was easy or difficult to adjust to the event. In the current study, we only included events that were rated as negative or very negative, and these ratings were summed for each participant (Sundstr€ om, R€ onnlund, Adolfsson, & Nilsson, 2014).

Temperament and character inventory
Personality was measured by a Swedish translation (Br€ andstr€ om et al., 1998) of the Temperament and Character Inventory (TCI; Cloninger, Svrakic & Przybeck, 2006). TCI is based on Cloninger's psychobiological theory of personality, which conceptualizes personality by four temperament dimensions and three character dimensions (Cloninger et al., 2006). Traits adhering to the temperament dimension of TCI are considered as more genetically based and relatively stable, whereas the character traits are more shaped by experience and social learning. The temperament dimension consists of Harm Avoidance (anxiety proneness, pessimism, shyness), Novelty Seeking (novelty seeking, impulsive decision making, low frustration tolerance), Reward Dependence (dependence on approval of others and social dependence), and Persistence (ambitiousness, diligence). The character traits include Self-Directedness (self-determination, ability to maintain goaldirected behavior and to adapt one's behavior according to context), Cooperativeness (social integration, empathy, helpfulness), and Self-Transcendence (tolerance of uncertainty, transcendence). The TCI is a self-report questionnaire that contains 238 items that, in the version of the inventory currently used, are answered by true/false statements.

Statistical analyses
SPSS version 23 was used for the statistical analyses. Two hierarchic regression analyses (with forced entry) were conducted to investigate the main research questions of the study. In the first analysis, time perspective as a predictor of depressive symptoms and the hypothesized moderating influence of age, were examined. Before performing the analysis, all variables were transformed to z-scores and interaction terms were computed by multiplying z-score of age with each time perspective subscale (e.g. Dawson, 2014;Simons, et al., 2016). In the subsequent hierarchic regression, age and gender were entered in a first block, in the second block the time perspective subscales were entered, and in the third block the interaction terms (age Â time perspective subscale) were entered.
In the second regression analysis, we examined the association between time perspective and depressive symptoms beyond influence of personality and negative life events. In the first block, age and gender were entered, in the second block, personality traits in the temperament dimension of TCI were entered. The character traits were entered in the third block, negative life events were entered in the fourth block and in the final, fifth block, all time perspective scales were entered. Multicollinearity was assessed by examination of tolerance values for all predictor variables (tolerance >0.20 indicating no issues with multicollinearity; Menard, 1995).

Results
Descriptive statistics (means and standard deviations) and zero-order correlations of the study variables are presented in Table 2.
As can be seen in the table, depressive symptoms were significantly and moderately correlated with Past Negative and Future Negative. Significant, but smaller, correlations were found between depressive symptoms and Present Fatalistic, and between depressive symptoms and Past Positive (inversely related). As concerns measurement of personality and negative life events, a small, positive correlation was found between Harm Avoidance and depressive symptoms, whereas Self-Directedness was moderately and negatively related to depressive symptoms. Finally, depressive symptoms were moderately correlated with negative life events.
In the next step of the analysis, we examined the extent to which the S-ZTPI scales accounted for variance in CES-D scores (see Table 3). Potential influence of basic demographic factors (age, gender) was controlled for by entering them in a first block. The six S-ZTPI dimensions were entered next. In the third step, the interaction terms (age Â S-ZTPI subscale) were entered. Inspection of tolerance values indicated no issues with multicollinearity (all values >0.46; cf. Menard, 1995).
The p-values indicated that the block of demographic variables was significant, and the significant b-estimates for age and gender demonstrated an age-related increase of CES-D scores, and higher scores in women, respectively. Of major concern at present, the S-ZTPI dimensions accounted for a total of 24.5% of the variance in CES-D scores, beyond the demographic variables, with significant unique associations for Past Negative, Past Positive, Future Negative, and Future Positive.
In the third step, two of the interaction terms were significant, age Â Present Fatalistic and age Â Future Negative. To further examine the nature of these interactions, we divided the sample into three age groups, young old (60-65), middle old (70-75) and oldest old (80-90) and ran separate correlation analyses between depressive symptoms, Present Fatalistic, and Future Negative in each age group. These analyses showed that for the oldest age group (80-90), there was only a small and nonsignificant correlation between Future Negative and depression (r = 0.15, p = 0.17), with the highest value for the youngest group (r = 0.49, p < 0.001), and slightly lower value for the 70-75 year-olds (r = 0.42, p < 0.001). The opposite pattern was found for Present Fatalistic; weak and nonsignificant associations with depression in the 60-65 year-olds (r = 0.12), stronger correlations in 70-75 year-olds' age group (r Table 2. Zero-order correlations and descriptive statistics (means and standard deviations) of the study variables. Variable  = 0.20, p = 0.01) and the oldest age group (r = 0.27, p = 0.02).
In the final set of analyses, we examined the extent to which time perspective accounted for variance in depressive symptoms after controlling for personality and negative life events. The variables controlled for were entered in a stepwise fashion (in blocks) before entry of the S-ZTPI dimensions in the final step. Values of tolerance indicated no issues with multicollinearity (all values >0.75). The outcome of the analyses are summarized in Table 4.
In Model 2, scores on the four temperament scales of TCI were entered. These variables accounted for 4.9% of the variance over and beyond age and gender (see Table  4, first step/block for values pertaining to the model including demographic variables only). Higher Harm Avoidance and Novelty Seeking were significant predictors of depressive symptoms. The TCI character scales were entered in the third step, accounting for an additional 9% of the variance. Specifically, higher Self-Directedness predicted lower CES-D scores. Adding negative life events in the fourth step accounted for another 5.3% of the variance, with a greater number of accumulated events being predictive of more symptoms.
Critically, time perspective added in the fifth, and final step, accounted for an additional 11.7% of the variance in depressive symptom score. The full model including all variables accounted for a total of 34.6% of the variance in depressive symptoms. Gender (age marginally significant at p = 0.052), Self-Directedness, negative life events, Past Negative, Past Positive, and Future Negative were the significant predictors of depressive symptoms in the final model.

Discussion
This study investigated the relationship between time perspective and depressive symptoms in a sample of older adults, including the possibility that age moderates the relationship between dimensions of time perspective and depressive symptoms. We furthermore examined the relationship of time perspective and depressive symptoms following control of personality and negative life events.
The results provide strong support of the notion that depressive symptoms are linked to time perspective (Anagnostopoulos & Griva, 2012;van Beek et al., 2011) and that this pattern generalizes to older adults (Desmyter & De Raedt, 2012). Regarding the time perspective dimensions, our results confirm earlier findings of a strong positive relationship between Past Negative and depressive symptoms (e.g. Desmyter & De Raedt, 2012, van Beek et al., 2011, and an inverse relationship between Past Positive and depressive symptoms (van Beek et al., 2011). Thus, the results highlight the important link between a past negative focus and depressive symptomatology, whereas access to and appreciation of positive aspects of the personal past may buffer against a negative mood state. In line with the hypotheses, Present Fatalistic showed a positive association with symptoms overall, but not in the regression analyses (i.e. of the whole sample) including the other S-ZTPI dimensions.
Additionally, our results demonstrated that the Future Negative scale, not considered in the study by Desmyter and De Raedt (2012), was a prominent predictor of depressive symptoms, highlighting the importance of this subscale in research on mental health correlates (Blomgren, Svahn, Åstr€ om, & R€ onnlund, 2016;R€ onnlund et al., 2017;Åstr€ om, Wiberg, Sircova, Wiberg, & Carelli, 2014). The link between Future Negative and depressive symptoms is in keeping with Beck's cognitive triad of depression, in which negative views of the future is considered an important aspect that may both underlie and sustain depressive mood. Interestingly though, the relationship between Future Negative and depressive symptoms appeared to be moderated by age; whereas a significant association between Future Negative and depressive symptoms was observed in the young-old (60-75 years), a weak (and nonsignificant) association was seen for the oldest participants (80-90 years). Interestingly, the opposite was observed for Present Fatalistic, which was more strongly related to depression in the oldest portion of the sample. In other Table 3. Summary of hierarchic regression of time perspective as a predictor of CES-D scores and age as a moderator words, apart from an age-invariant relation to Past Negative, the results suggest that advanced age may be associated with a shift in regard to what particular time perspective biases are related to depressive symptoms. The shift in weight from negative aspects of the future to negative aspects of the present in the oldest group, might be considered in light of what has been labeled as the 'fourth age' (Baltes, 1998). This life period (usually 80 years and older) is characterized by more marked biological deterioration, decreased health, and increased dependency. At this stage, future negative aspects may become pertinent, consistent with the significant positive association between age and Future Negative observed in the present sample, but are possibly adapted to, as some negative expectations in regard to the personal future are inevitable. By contrast, agency or perceived control (i.e. internal locus of control) and handling the perceived threats to these aspects of life, may be more of struggle for the old-old adults. An age-related increment in Present Fatalistic, evident in the present sample, is in keeping with patterns of age-related losses of perceived control using other measures (Drewelies, Wagner, Tesch-R€ omer, Heckhausen, & Gerstorf, 2017), which may be more predictive of a depressive mood than concerns about the future in old-old age. Physical health and social factors may be important to perceived control (Drewelies et al., 2017). Cognitive abilities might be important as well, and relevant to consider given evidence that cognitive deficits can be a forerunner of depression in old age (e.g. Perrino, Mason, Brown, & Spokane, 2010). Interestingly, Present Fatalistic was inversely related to executive functioning and fluid intelligence in young adults (Stolarski & Matthews, 2016). Thus, a common link between cognitive deficits, a present fatalistic attitude toward the present, and depressive symptoms might be particularly apparent in old-old individuals. Future studies will have to examine these possibilities.
Overall, variations in time perspective as operationalized by the S-ZTPI dimensions accounted for a substantial portion of variance in CES-D scores (i.e. beyond the demographic factors). In further analyses, TCI temperament and character dimensions and negative life events were additionally considered as predictors. As concerns TCI, the results confirmed that aspects of temperament and character are predictive of depressive symptoms. In line with previous research, higher Harm Avoidance and lower Self-Directedness were related to depressive symptoms (Celikel et al., 2009). In fact, Self-Directedness turned out as a unique predictor in the final regression model including the entire set of predictors, which is particularly noteworthy given the time-lag between measurement of the TCI dimensions and depressive symptoms. This was also the case for negative life events, which was measured longitudinally prior to the assessment of depressive symptoms. In accord with prior work, negative life events remained a significant predictor of depressive symptoms in the final regression model (Table 4). Interestingly, negative life events were associated with both Past Negative and Future Negative in the correlational analysis (see Table 2). This finding possibly reflects that even though an individual's time perspective is considered as fairly stable and traitlike , it may at least to some extent, be modulated by environmental influences, such as negative (Holman, Silver, Mogle, & Scott, 2016) or positive (Leist, Ferring, & Filipp, 2010) life events.
As should be expected, the TCI, in common with other measures of more traditional measures of personality like the 'Big Five' (Jochemczyk et al., 2015;Zhang & Howell, 2011;, shared variance with time perspective and hence, control of these factors reduced the association between time perspective and CES-D scores. Notwithstanding this fact, at least half of the variance accounted for by the S-ZTPI dimensions (i.e. in the baseline model adjusting for demographic factors only) remained following adjustment of the seven TCI dimensions as well as life event burden. Thus, the additional analyses further substantiated the link between time perspective and depressive symptoms established in prior research.
As concerns the demographic factors considered, the finding that female gender was associated with more depressive symptoms is consistent with substantial prior evidence that depression is more common among women compared to men (e.g. Djernes, 2006;Kuehner, 2003). Additionally, higher age was associated with higher scores on CES-D. This finding may seem surprising in light of epidemiological studies that have shown that depression is less common in old age (e.g. Kessler et al., 2005). However, this is the case for a diagnosed depressive disorder, whereas the opposite has been found for endorsement of depressive symptoms on symptom checklists, such as CES-D (Newmann, 1989). Older adults tend to report higher scores on inventories of depressive symptoms compared to younger adults (Newmann, 1989), possibly because some of the symptoms may reflect aspects of physical illness or bereavement, which is more common among older, compared to younger, adults (Blazer, 2003).

Limitations
Although the strengths of this study, including a fairly large, population-based sample and control of personality dimensions should be highlighted, it also has limitations. Even though we incorporated some longitudinal aspects (i.e., life event data had been collected across many years preceding measurement of the outcome variable), the associations between the main study variables (i.e. time perspective and depressive symptoms) were cross-sectional. This fact precludes conclusions regarding directionality of their influences. In accord with the theoretical framework behind the ZTPI, we assumed that certain time perspective biases predispose of depressive symptoms, but a reversed direction of influence, from depressive mood state to temporal perspective, might exist (cf. Holman et al., 2016); a model involving a bidirectional influence might possibly be the most accurate. Disentangling the influences requires longitudinal data with repeated measurement of both constructs. In regard to assessment of personality, it is warranted to point to the fact that data had been collected at an earlier time point. This likely biased prediction of depressive symptoms in favor of time perspective. Nevertheless, we deemed these data to be of important value, and the basic finding that time perspective dimensions accounts for variance over and beyond more traditional personality factors is consistent with results of a recent study involving mental health indicators in a younger sample (Stolarski & Matthews, 2016) Conclusions Taken together, the results of the present study demonstrated a substantial link between depressive symptomatology and time perspective, and further indicate that in particular high Future Negative and Past Negative might be important aspects to identify in older adults, as they might be risk factors for developing depressive symptoms. Novel findings included the observation that Future Negative accounts for variance in depressive symptoms in older adults in addition to that associated with the five dimensions in the original ZTPI. Interestingly, this may hold true except for the oldest old (80-90 years) where a relation between depressive symptoms and present fatalism may more apparent. Finally, the results demonstrated that even after control of personality traits and negative life events, time perspective still explained a considerable portion of the variance in depressive symptoms.

Disclosure statement
The authors report no conflict of interest.