Abstract
The present article reports on the relationship between various, physiological and sociological factors on a person’s individual perception of time in life. Specifically, 200 participants (100 males, 100 females) were solicited from diverse, university-centered communities. These volunteers completed a series of questionnaire-based evaluations and also had physiological recordings of heart rate and core temperature taken. For the measure of time-in-life we used an amended version of the Lines test which compared the individual’s perception of their current time-in-life against the actuarial expectation of their lifespan. Results confirmed a strong inverse relationship between and individual’s age and this measure of perceived life duration. The gender of the individual had an important impact on their perceived lifespan. There were also indications of differences contingent upon the time of day at which the test was administered. However, there was little evidence of any linkage between lifespan perception and the physiological indicators recorded. The results are discussed in terms of possible cognitive and sociological determinants of individual’s perception of their present life location.
Introduction
It is one of the truisms of aging that time seems to pass evermore quickly the later in life one supposedly is. While one hears some discussion and sees reports of this phenomenon among those in their teens and early 20 s (Joubert, 1984; Lemlich, 1975; Walker, 1977), it is a much more frequent topic of conversation of retired individuals. Nor is this acceleration of time with age a ‘new’ or modern phenomenon (see e.g., Guyau, 1890; James, 1890). In the 18th century, Thomas Campbell (1777–1844) composed a poem entitled ‘The River of Life’ about this effect (see Hancock, 2002) and Cannon Henry Twells (1823–1900) offered his iconic ‘Time’s Paces,’ on this topic in which he observed: When as a child I laughed and wept, time crept, When as a youth I waxed more bold, time strolled. When I became a full grown man, time ran. When older still I daily grew, time flew. Soon I shall find, in passing on, time gone. . .
In cases where the above influences have been investigated, it is the case that by-and-large, the dependent time perception measures have most frequently focused on the perception of short intervals up to one minute in duration. Obviously, such assessments prove to be convenient experimental measures to administrate but problematically, an individual’s perception of very short intervals of time may or may not be consistent with that same individual’s perception of extended periods of time up to years, decades, and indeed a lifetime’s duration (and see Grondin, 2008). To evaluate this latter form of perception, one has to use instruments which address long-term memory rather than assessments of short-term accuracy with respect to discrete clock intervals of seconds or minutes (cf., Cottle, 1976).
Thus the purpose of the present investigation was to evaluate the effect of personal, physiological, and social characteristics on the perception of time in life. The latter provides a measurement of one’s own personal assessment of where one is in the prospective lifetime. The hypotheses for the present work postulated that there would be a strong and consistent effect for an individual’s chronological age; this based upon the prior observations of subjective acceleration of time with age. However, it was also postulated that there would be a consistent effect for the sex of the individual based on some previous empirical findings (Hancock, 2010) as well as an overall appraisal of the literature on temporal sex differences (Hancock, 2011; Johnson, 1964). Similarly, it was postulated that body temperature and time of day would have some effect while heart rate was expected to exert little influence. Finally, we tested a proxy of geographic representation of social class through the use of two differing data collection locations. Confirmation of any or all of these effects would suggest some avenues through which to understand the mystery of extensive individual differences in time perception (Doob, 1971), especially the apperception of time in life.
Experimental method
A convenience sample of individuals (n = 200; 100 males, 100 females) was taken from around the environs of two major southeast University communities. No specific attempt was made to distinguish or solicit participants on the basis of sex, age, race, or socio-economic class. Experimental data, as described below, was collected by a number of experimenters, both male and female since previous experimentation (Hancock, 2010) had shown that the sex of the experimenter did not affect results of the present assessment procedure. The data were recorded on a single data sheet and the collection period for any individual did not exceed more than five minutes in duration. The whole procedure was approved by the appropriate Institutional Review Board and the participants were treated according to the principles of the American Psychological Association’s (APA) ‘Ethical Treatment of Research Participants.’ The following sequence of information was elicited.
Demographic profile
A series of demographic questions were first asked in order to identify: i) the self-declared age of the individual; ii) their sex; iii) the time-of-day at which the test occurred; as well as iv) the specific date of the test. In addition to these queries, also recorded was; v) the specific data collection site, i.e., the northern or the southern University environs.
Socio-economic and self-perceptual assessment
As well as the demographic, physiological, and temporal data, subsequent inquiries allowed for the assessment of certain facets of self-perception and socio-economic status. Participants were asked whether they were currently married, their highest achieved level of education, their current employment status, as well as whether they had members of their immediately family (i.e., mother/father or immediate siblings) still living. They were also asked to mark off on a continuum to what degree they perceived their social network as supportive, as well as their own self-perceived present state of health. Finally, in order to approach the question of the geographic distribution of psychological propensities (Yoon, 1991), individuals were asked to identify the street and city in which they lived. If the individual identified themselves as a student they were asked for their permanent home address and not their campus residence address (since the data were recorded on college campuses this was a frequent rectification). Specific care was taken not identify actual addresses by number but only by cross-street locale. It was the matrix of these collective data that were subsequently the subject of the analysis which follows. After the completion of these demographic questions, the next data to be recorded concerned reflections of the individual’s physiological status which are described below.
Physiological measurements
The first physiological parameter to be recorded was the participant’s heart rate. This was accomplished through the use of a wrist watch recording device, namely; a Reebok™ model ‘Heart Rate Beat Calculator’ manufactured by Reebok International Limited (Canton, MA, USA). Contemporary with this recording, the participants also provided a measurement of their inner ear temperature. This value, which is a proxy for the core temperature measurement at the tympanic membrane (Hancock, 1980, 1981) was derived using a handheld digital display model ‘Thermoscan Instant Thermometer’ manufactured by Thermoscan Inc. (San Diego, CA, USA). The model was designed to be used with reusable probe covers. Body temperature measurements taken in this way rely on infrared heat generated by the eardrum allowing the measurement to be taken quickly in about one second. Such an assessment procedure involved minimal discomfort. Measurement through the ear is also beneficial as the method eliminates the possibility of participant cross-contamination due to the lack of mucous membrane contact.
Measurement of the perception of age
In order to evaluate the perception of the individual’s present age, a revised version of Cottle’s (1976) line test derived by Hancock (2010) was employed. The participant is presented with a 10-inch horizontal line across a standard sheet of paper which represents a time continuum. In the center of the line appears a single vertical line that represents the present moment, here termed the ‘now’ mark (or specious present, see James, 1890). Participants were then asked to mark a perpendicular line to the left of the ‘now’ mark at a position where they perceive their birth to be. Following this same procedure, they were asked to strike another perpendicular line to the right of the ‘now’ mark where they anticipated that their prospective death would occur. This procedure allows the experimenter to subsequently calculate the length of the perceived past lifetime, the length of the perceived prospective lifetime and also to compare the participant’s perception of where they are in their own life versus the actuarial expectation of their lifespan derived from data provided by the U.S. Department of Health and Human Services (2007). There are two potential methodological issues here which have previously been raised and addressed. First, there is no inherent ceiling effect in the marking of the line such that ‘running out of space’ could be responsible for a curtailed range of opportunity as the age of the individual increases. This methodological issue has been previously addressed and resolved (see Hancock, 2010). A second issue concerns participant selection. That is, if participants are selected who are already older than their actuarial cohort life expectancy, then they are unable to mark a location on the line greater than the 100%+ that would be calculated for their actuarial comparison. Thus individuals older than their lifespan expectancy were not included in the present sample.
It is from the measurements taken from this modified Lines Test that outcome dependent measures of life perception were derived. Specifically, these dependent measures were i) the total line length that the individual struck off; ii) the length of the segment represented by the past, that is, life up to this point in time; iii) the length of the segment represented by the future, that is, the perceived prospective lifespan remaining [note specifically that the combination of ii) and iii) necessarily connotes the length represented in i)]. The fourth dependent measure, iv), was the percentage of lifetime perceived at the present time. To derive this, the experimenter took the length of the marked line representing the past segment (i.e., dependent variable ii) and divided it by the length of the marked line for the whole lifetime (i.e., dependent variable i). This provides a value denoting where the individual perceives themselves to be in their own life. The final dependent variable, v), represents a difference score between this latter value and an actuarial appraisal of where that individual is according to current cohort figures. This latter value is taken by comparing the individual’s perception of time in life (i.e., dependent variable v) with their actuarial life expectancy as expressed in the tables provided by the U.S Department of Health and Human Services (2007). This latter measure is here referred to as the difference measure and connotes the primary dependent variable for the study.
Experimental results
The present data set was subjected to a number of analyses from which numerous significant trends were extracted. These were divided into the respective dependent variables for the study. The first of these to be evaluated was the primary measure, which is the difference score. Described as dependent variable v) above, this is the difference between someone’s perceived life location versus their actuarial expectation.
Difference score
The most evident effect for this respective difference score was that for age. As can be seen from Figure 1, as the age of the individual increased, the degree to which they underestimated their actuarial life expectancy increased accordingly. This effect was consistent for both male and female subjects; although, as Figure 1 illustrates, the linear trend for males and females was not coincident. The regression equations for these lines permit us to derive a measure which we (Hancock, 2010) have previously termed the ‘life indifference interval.’ This value is the point at which the individual estimates their prospective longevity as directly matching that of their actuarial expectancy for their lifespan. More simply, this is the point where each of the linear equations crosses the zero value on the difference axis. As can be seen from Figure 1, this lifetime indifference interval was 26.0 years for males and 30.4 years for females. This compares with a previous investigation (Hancock, 2010) in which the male indifference interval was 24.3 years, while the prior investigation showed a female indifference interval of 27.9 years. These findings are thus in substantial agreement and especially with the consistent male versus female difference which was +4.4 years for females in the present study, as compared to +3.6 years in the prior study. The stability of this difference is evident and important since the data derive from a comparison of 320 individuals in the first study and 200 different individuals in the present investigation. The overall indifference interval for the whole sample in the present study, both male and female, was 28.2 years, while the overall indifference interval from the previous study was 26.0 years (and see Hancock, 2010). The degree to which this close correspondence represents a meaningful observation, as opposed to simply a close correspondence between the respective age samples in the two studies is discussed below.
The difference score between the empirically observed self-perception of time in life versus the actuarial projection of that same individual’s lifespan plotted as function of their sex and age in years. The respective regression lines are: Females = −0.915age + 27.898 (R2 = 0.503); and Males = −0.595age + 15.49 (R2 = .157).
Given the differences in the linear equations for males and females, and our previous observation of sex differences in this difference score (see Hancock, 2010, Figure 3), we conducted a simple t-test evaluation of the proposition of reliably significant sex differences in the present sample. Despite the fact that there was again a similar mean outcome as previously observed (i.e., that males underestimated their life location [−3.437%] more than their females counterparts [−0.571%]) on this occasion, the difference did not reach an accepted level for the overall difference score (p > .05). There were, however, consistent sex differences in other of the outcome dependent variables as discussed below.
There is an alternative perspective from which to conceive the present data which is suggested by the work of Rubin and Berntsen (2006). Their specific procedure asked individuals to express whether they felt ‘younger’ ‘older’ or ‘the same’ as their chronological age. From these responses, they could then plot the percentage of individuals in each category according to their age, as parsed by sequential decades of aging. Fortunately, our present data permits a direct comparison with their findings. To accomplish this, we had to define a bandwidth of difference scores which represent the ‘same as’ category and for this we chose a ±10% window. This then generated data for differing age groups. Rubin and Berntsen (2006) questioned 1,470 individuals and so had a large sample in each decade of aging. For our present sample of 200 individuals, we split them into groups of below 30 years of age, 30–40 years of age, 40–50 years of age, and finally 50+ years of age. The resultant graph is presented in Figure 2.
The proportion of individuals in each age group who considered themselves younger (>−10%,) the same as (−10% to +10%) or older (<10%) than their chronological age. Percentages calculated as difference scores expressed in Figure 1. The number of participants in each respective category was: (<30 = 122; 30–40 = 27; 40–50 = 29; >50 = 22). A significant time of day effect in the observed difference score for perceived versus actuarial lifespan expectancy. The fitted, third order polynomial curve serves to suggest a circadian effect in this perception. Further information as to responses in the later hours of the evening may well be needed to fully establish such an effect.
What is very evident from this illustration is that it matches that expressed by Rubin and Berntsen (2006, Figure 1) very closely. Indeed, our data feature the most volatile region in that the percentage values in each category tend to stabilize after 50 years of age with the vast majority of individuals believing themselves to be ‘younger’ than their chronological age with only a diminishingly smaller percentage expressing themselves as feeling older than their chronological age. The data here suggest that the functions that we have shown as linear for both males and females in Figure 1 may in fact, be more veridically represented by a curvilinear equation; a proposition we have thus tested. The linear model explained 31% of the variance (i.e., R2 = .312) [F(1, 199) = 89.88, p < .0001]. Introduction of the additional degree of freedom of the curvilinear model now explained 33% of the variance (i.e., R2 = .331), however, that latter step was also significant [F(2, 199) = 48.742, p < .0001]. The increment in variance account for by the curvilinear term was also statistically significant [ΔF(1, 197) = 5.54, p = .02; ΔR2 = .019]. Thus, despite the relatively modest gain in variance explained, the latter model should perhaps by adopted, especially as this outcome accords with certain previous findings with larger samples, albeit the pattern is derived from different methodological approaches (see Rubin & Berntsen, 2006, but see also Friedman & Janssen, 2010; Wittman & Lehnhoff, 2005). There were, however, additional patterns expressed in the difference scores as related to the independent variables examined.
For the difference score, there was an effect for time-of-day. This influence, identified with a fitted polynomial equation is illustrated in Figure 3.
Here we can see what, at first blush, appears to be a circadian function. However, we have to be very careful about any such conclusion at the present time as: firstly, there are relatively few observations taken at the later times of day, thus any trends are liable to be highly influenced by small numbers of individual data points. Secondly, one can fit various equations to this data set where the present polynomial tends to appeal because it follows a tentative circadian pattern. Moreover, there was no reliable observation of an associated variation in body temperature rhythm which might be anticipated to accompany any such circadian change. However, it is possible that the large individual differences in the present sample acted to mask such a circadian effect in body temperature. It is important to note that this pattern was not a result of a covariate with age. That is, the current trend does not arise because of a particular selection pattern in testing for example older individuals later in the day. In the current data, age of participant was spread fairly evenly across all times of testing. At present, it is best to conclude that this is a suggestive trend, worthy of more extended investigation. For the difference score, there were no significant influences of heart rate or body temperature. Thus, we move on to the basic dependent variables that eventually coalesced into the difference score and the first of these was overall line length. There were no effects of any of the demographic questionnaire dimensions on the perceived difference score.
Overall line length, past segment, and future segment
Unlike the difference score, the influence of age was not significant on overall line length. Rather, the only factor here which exerted such an effect was the sex of the participant. To evaluate this, we conducted a t test which found significant differences [i.e., t(198) = 2.055, p < .05]. In this case, females generated line lengths that were longer than their male counterparts (i.e., Females, Mean = 17.4, SD = 4.94; Males, Mean = 15.6, SD = 7.0, all measures in centimeters). Given that the present life expectancy in the United States is 80.8 years for females and 75.6 years for males (CIA World Fact Book, 2009), the percentage difference in current life expectancy of 6.44% is tolerably similar to the difference (10.34%) in line length observed. Also, as is clear from the data, females were less variable than males; although to some degree there is an inherent ceiling effect in the length of line, it is possible to use in the present procedure. For the value of past life segment, there was a dominant effect for age where, not unnaturally, the length of past segment marked off increased with the chronological age of the participant. For the length of the future segment, there were effects for both age, in which the length of the segment decreased with age, while there was also an effect for sex [t(198) = 2.413, p < .02]. Here again, the females participants marked a longer but less variable line compared to male participants (i.e., Females, Mean = 10.1, SD = 2.50; Males, Mean = 9.0, SD = 3.86, all measures in centimeters). Of particular interest was a significant effect for perceived state of health. Here, the healthier the individual perceived themselves to be then the longer the overall line length that they marked [F(1, 198) = 11.619, p < .001; B = 0.646 (SE = 0.19), Beta = 0.235] and this significant effect extended to both the segment representing the past [F(1, 198) = 7.3, p < .01; b = 0.33 (SE = 0.122), Beta = 0.19] as well as the segment representing the future [F(1, 198) = 9.618, p < .002; B = 0.316 (SE = 0.102), Beta = 0.215].
Duration judgment ratio
The final dependent variable to consider was one derived frequently in time perception research termed the duration judgment ratio (DJR). Here, the value is derived by dividing the person’s individual estimate of their perceived time in life by their actual chronological age. A value of one in the DJR represents an accurate estimation, a value of less than 1 indicates underestimation and a value over 1 indicates overestimation. Here, we again see effects of both age and sex. In the formal comparison across sex, there were significant differences [t(198) = 2.064, p < .04] such that females overestimated their lifespan with a mean DJR of 1.08 (SD = 0.44), while the males underestimated their lifespan at a DJR of 0.96 (SD = 0.40). Figure 4 shows the DJR effect in both age and sex differences with the respective regression lines for males and females illustrated.
Significant effects for age and sex on the Duration Judgment Ratio (DJR) defined as the perceived time in life divided by the actuarial designation of time in life for that same individual. Males are represented by the solid squares and the dotted line; females are represented by the open circles and the solid line. The respective regression equations are: Females = −0.0194age + 1.6883 (R2 = 0.343); and Males = −0.017age + 1.5038 (R2 = 0.251).
It could be argued that observed sex differences in different dependent variables derived from an inherent age difference between the sample of males and females elicited. However, sex effects are unlikely to be a result of such sample differences in age since the mean age difference between men and women here was less than 8 months apart as the mean changes between males and females are an order of magnitude greater than the percentage difference in age itself; it is unlikely that sex effects are simply a byproduct of age sampling. Overall, the current findings confirm the anticipated effect for age as a primary influence on the perception of time in life as well as some associated sex differences (and see also Friedman & Janssen, 2010; Gallant et al., 1991). There were, however, disappointingly weak effects for any of the physiological variables evaluated. The time-of-day of testing did show some potential influence but clarification of this effect awaits greater sampling across a wider range of times of day. There were no significant effects of any of the demographic questionnaire factors on the duration judgment ratio.
Discussion
The present experiment sought to extend previous findings concerning individual difference factors in the perception of time in life. In so doing, the present procedure allowed us to re-evaluate the major factors featured in the previous investigation (Hancock, 2010). It is gratifying that the substantive effects of age and sex have been here confirmed. The overwhelming effect in this data set and in the previous investigation concerned the effect of age. As individuals grow older, they sequentially perceive themselves to be ‘younger’ than contemporary life-expectancy projections would suggest. While 60 may indeed be the new 40, the overall regression is not quite as sanguine as this pronouncement but the trend certainly goes in this direction. Perhaps the trend further reflects the Garrison Keillor observation of a town where 90% of people believe they are above average or perhaps the data reflect a fear of dying. It is important to emphasize that the linear trend we have shown in Figure 1 could probably not continue into extreme old age without some form of asymptote, as indeed the pattern reported by Rubin and Berntsen (2006) would appear to indicate. Such a trend does not imply ‘speeding’ of time perception with age per se, but a consistent underestimation of one’s chronological age. Of course, we must always bear in mind that chronological age, i.e., the use of an external physical reference change codified in social formalizations such as calendars, may be rather poor representations of an individual’s functional age in terms of both physiological and psychological dimensions (and see Lecomte Du Nouy, 1937; Surwillo, 1964). If these functions are further confirmed, by for example procedures such as meta-analysis, it implies that the search for the supposed phenomenon of temporal speeding with age is either chimeric or must be found in the estimates of intervals of time less than those over the lifespan as investigated in the present work and allied investigations. What is shown in our own data, but is not featured in that of others, is the large differences between individuals. The scatter plots, especially those presented in Figures 1 and 4, illustrate that the nomothetic trends, whether linear or curvilinear across the lifespan, only derive from the collapsing of large intrinsic variations between people. Some degree of that variation is contingent upon the sex of the individual; however, this is only one factor in what appears to be a highly complex amalgam of effects that still await a full exposition. However, having noted these positive effects, it is to the sex differences that we now turn.
The sex difference here is also of central concern since the life expectancy tables almost ubiquitously show greater lifespan expectancy for females compared with males. This is especially true in developed countries in which death in childbirth has been significantly reduced (the first country in the world where male life expectancy exceeds females is Niger (Male = 57.8 vs. Female = 56), the 160th country in terms of overall life expectancy. The female responses in the present experiment reflect this difference which is evident in perception as well as physiology. While the absolute difference has been attributed to the protective value afforded by the years of female productivity, the comparative psychological or perceptual difference appears to be founded either on a learned expectation or an inherent perceptual mindset that accords with the underlying physiological differentiation. In support of the latter contention, it should be recalled that the present investigative methodology uses no numerical comparisons. Thus, the present outcome is not the result of an overt mathematical calculation, although this form of mental calculus may well be one strategy that individuals use in making their marks on the single horizontal line. If this latter explanation is correct, it means that on average, this calculation is very accurate indeed.
Previously, the effect of age on the perception of short intervals of time has been demonstrated through the use of meta-analytic procedures (see Block et al., 1998). Here, and in accord with a previous, allied investigation, we have confirmed the effect of age on the much extended perception of time in life (Hancock, 2010). Further, meta-analysis has also served to show the influence of sex on the perception of short-term intervals (Block et al., 2000), as has a more extended evaluation of sex effects in numerous forms of timing capacity (see Hancock, 2011). Here, we report that such sex differences persist in estimates of the interval of a lifetime. These are thus established factors which affect an individual’s short-term and longer-term perception of time. Unfortunately, the primary physiological variables investigated here (i.e., body temperature and heart rate) showed no such effects. While body temperature has been linked to the perception of short intervals of time (e.g., Hancock, 1993; Hoagland, 1933; Wearden & Penton-Voak, 1995), we found no such association for the present estimation of lifetime. This could well be due to the large individual differences that are inherent in time perception; and indeed were the phenomenon upon which we were focused. However, it is more than likely that the one single estimate provided; which is a necessary facet of the present investigational technique, was insufficient to establish a stable effect. This sort of problem has been encountered before in nomothetic studies of time estimation (Hancock, 2011). The null finding for heart rate is not unexpected. There has been very little evidence of any for an associated between heart rate and short-interval time estimation, thus the expectation for an effect on longer intervals is not a surprise (although see Meissner and Wittmann, 2011). The influence of time-of-day seems a promising one but the present data set is most probably insufficient to be anything more than suggestive. It was especially interesting that the subjective apperception of overall life length as well as both the past life segment and the prospective (future) life segment co-varied with perceived health. That one’s general state of health has an impact on one’s cognitive outlook is surely not an original observation. However, the indication that health affects past, future, and overall perception of time in life must be considered a new and original observation from the present data. Exactly, how and why this mechanism operates is as yet not fully elucidated, however, this must surely relate also to the perception of short intervals of time during health and illness (e.g., Cohen, 1967).
In general, the psychological dimensions of people’s perception of time in life which is, in effect, the longest rational interval that they can estimate from personal experience, are incompletely articulated (Ukraintseva, 2001). Indeed, there are only a limited number of techniques that can address such long-term estimates. Those typically used for estimation of short intervals of up to approximately one minute in duration are, in general, ill-equipped to address these longer-term intervals. Neither have major known dimensions of cognitive influence (e.g., prospective vs. retrospective estimations) been pursued in this realm of extended estimation. While it might be suggested that differing neurophysiological and neuropsychological mechanisms mediate the estimate of intervals ranging from milliseconds to decades, any eventual account of human time perception must deal with the whole lifetime as well as the small componential moments of it. This work represents one of the first such steps down this important path.
Footnotes
Funding
This research received no specific grant from any funding agency in the public, commercial or not-for-profit sectors.
Acknowledgments
We would like to thank David Keellings and Katie Hood for help with the data collection in this experiment. Our grateful thanks go to Professor James Szalma for all his help with the statistical analysis of the results of this experiment.