The explicit judgment of long durations of several minutes in everyday life: Conscious retrospective memory judgment and the role of affects?

Sylvie Droit-Volet; Sophie Monceau; Mickaël Berthon; Panos Trahanias; Michail Maniadakis

doi:10.1371/journal.pone.0195397

Abstract

In this study, individuals estimated interval times of several minutes (from 2 to 32 minutes) during their everyday lives using a cell phone they kept with them. Their emotional state, the difficulty of the activity performed during this interval, and the attention that it required were also assessed, together with their subjective experience of the passage of time. The results showed that the mean time estimates and their variability increased linearly with increasing interval duration, indicating that the fundamental scalar property of time found for short durations also applies to very long durations of several minutes. In addition, the emotional state and difficulty of the activity were significant predictors of the judgment of long durations. However, the awareness of the passage of time appeared to play a crucial role in the judgment of very long duration in humans. A theory on the emergence of the awareness of the passage of time and how it affects the judgment of interval durations lasting several minutes is therefore discussed.

Citation: Droit-Volet S, Monceau S, Berthon M, Trahanias P, Maniadakis M (2018) The explicit judgment of long durations of several minutes in everyday life: Conscious retrospective memory judgment and the role of affects? PLoS ONE 13(4): e0195397. https://doi.org/10.1371/journal.pone.0195397

Editor: Warren H. Meck, Duke University, UNITED STATES

Received: November 10, 2017; Accepted: March 21, 2018; Published: April 3, 2018

Copyright: © 2018 Droit-Volet et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: All relevant data are within the paper and its Supporting Information file.

Funding: The authors received funding for this work from H2020 European Research Council, Horizon 2020 research and innovation action, TIMESTORM, H2020-FETPROACT-2014.

Competing interests: The authors have declared that no competing interests exist.

Introduction

The human brain is a clock allowing us to accurately measure time [1,2,3]. However, the question is: what is time? Is an internal clock capable of measuring all interval durations independently of the amount of time to be estimated? Most researchers who adhere to the internal clock models accept the idea that the same clock mechanism makes it possible to measure a wide range of interval durations going from a few hundred milliseconds to several minutes [4]. However, they have almost all focused their investigation of time judgment in humans on durations shorter than one minute. It is therefore too early to generalize the timing mechanisms involved in the processing of short durations to that of long durations. One main aim of our experiment was to characterize the judgment of time in the minute range in everyday life.

Very few studies have systematically investigated the timing of interval durations of several minutes in humans using a prospective paradigm in which the participants are explicitly instructed to judge the duration of the upcoming stimulus or event. This is primarily due to methodological obstacles related to the difficulty of finding subjects who are prepared to sit in front of a computer during a long, boring laboratory task, and this several times per day. As a result, the few studies that have examined durations of several minutes have tested interval durations of less than 3–5 minutes [5,6,7,8]. And in cases where longer durations have been used, they have often tested only one or two durations, each of which was presented once (one trial) [9,10,11]. For example, Tobin et al. [11] assigned participants to one of three duration conditions (12, 35 or 58 minutes) and explained to them that they would be interrupted during a video game and asked to judge the length of time they had been playing for. In such a task, it was impossible to measure the variability of time estimates because doing so would have required several trials [12]. This temporal measure is nonetheless essential if we are to test the conformity of minute-duration judgments to the scalar property of time, which is a fundamental property of interval timing. According to this property, the variability of temporal estimates increases (SD) proportionally to the length of the durations, with the result that the coefficient of variation (SD/mean) remains constant when the duration range changes [13]. In the present study, we therefore decided to test a series of interval durations in the minutes range over several trials, but to do so in an ecological condition, i.e., in everyday life, by using a cell phone that the participants kept with them for the day. They therefore went about their daily activities during the intervals.

Intuitively, one might think that it is impossible to track the length of very long durations of several minutes without the help of our watches. The few results for very long durations are inconsistent with this idea because they have shown that it continues to be possible to discriminate durations in the minutes range. Indeed, although studies have only compared 2–3 interval durations, they have found that longer durations are systematically judged longer than shorter ones, for example 24 min is judged longer than 8 min [10], or 35 min longer than 12 min [11]. In addition, in the series of long durations tested, the shortest ones have tended to be overestimated and the longest ones to be underestimated [9,10,14], or overestimated to a lesser extent [11]. Despite the inconsistency in the direction of time estimates, these findings are consistent with Vierordt’s law, with short durations being reported as longer and long durations as shorter than they actually are. This finding cannot therefore result from intrinsic properties of durations in the minutes range since it is also observed for shorter durations of just a few seconds. Instead it reflects the central tendency of judgment described by Hollingworth [15], according to which the judgment of durations is not absolute, but relative to the median of the other durations presented in the series. In the Bayesian framework, this is explained by the influence of temporal context (preceding duration values) on judgment of a given duration [16].

As regards the variability of temporal judgment, only Lewis and Miall [14] have examined the scalar property of time in a study involving a small number of subjects (N = 8) and durations going from 68 s to 16 minutes. In their study, the coefficient of variation appeared to be quite large (CV = 0.35). However, and more critically, they observed a violation of the scalar property of time, with a non-constant CV across durations. However, as the authors explained, “it seems prudent to treat this initial finding with caution and even some suspicion” (p. 1900). This finding probably related to the task used (temporal reproduction), as well as to the secondary task used for the presentation and reproduction phases. In addition, the discontinuity in the CV, which might have indicated some changes in the timing mechanisms at approximately 1.5 s [17], was not found in their results in the critical intervals at about 1.5 s or 3–4 s.

Whatever the results, it is obvious that different mechanisms exist for different ranges of interval durations. We can presume that the estimation of very long durations cannot result from a number of counts (pulses) accumulated in a “timer” in an uninterrupted and regular way over a period of several minutes. The number of pulses accumulated is also dependent on the way attention is paid to time [18], and it is unlikely that the subjects are able to maintain the same level of attention over a very long period of time [11]. In addition, at the level of seconds, the mechanisms involved in the processing of durations are already different from those involved in the processing of shorter durations of the order of milliseconds, with more attention and working memory processes being devoted to the former [19,20]. As discussed later, although the mechanisms involved in the processing of long durations in everyday personal experiences have not as yet been identified, we can assume that additional processes related to long-term memory (episodic memory) are involved in the measurement of these durations.

Recently, Droit-Volet and her collaborators examined the relation between the judgment of durations and the subjective experience of the passage of time [21,22]. The passage of time judgment (PoTJ) is the awareness of the flow of time and its variation in comparison to the conventional representation (knowledge) of the constant rhythm of an external clock [23,24]. It results from a reflexive, moment-by-moment introspection on the self in time (self-time), i.e. the consciousness of the relationship between the pace of self-time and time in the external world. They observed that the judgment of durations is not related to PoTJ in the short duration ranges: 350 ms vs. 1650 ms [21], and 2.8 s vs. 33 s [22]. However, a link between these two forms of time judgment appears for durations in the range of minutes (2, 4, 6, 8 min) [22]. As we discuss later, this suggests that the judgment of long durations and the PoTJ would share some of the high-order cognitive mechanisms involved in the recovery of long durations in long-term memory, i.e. the retrospective memory retrieval of the non-temporal content of the temporal interval. Therefore, the PoTJ during the long interval to be estimated, as well as the contextual content of this interval, were also assessed in our study after each estimate of interval duration. Some studies have shown that the PoTJ depends on the level of emotion and the activity performed at the moment of the PoTJ. Participants do indeed perceive time as passing faster when they feel more excited and aroused. Although it has been observed less consistently across studies, the difficulty of the activity and the attention that it captures are also significant predictors of PoTJ. Consequently, to further examine the determinants of judgments of very long durations, we assessed the PoTJ, the affective states felt and the activity (difficulty, attention) performed during the interval duration. In our study, which was designed to examine the prospective time judgment of very long durations, the participants were therefore given a verbal estimation task in which they had to judge several interval durations, from 2 to 32 min, during their everyday life and also had to assess the feeling the passage of time, the emotion felt and the activity performed during these interval durations.

Method

Participants

The sample consisted of 60 participants (Mean Age = 28.44, SD = 6.46; 40 women and 20 men). The results for one participant were excluded due to a technical problem with the cell phone. A consent form describing the constraints necessarily involved in the experiment was signed by each individual before his or her participation in this study. This study was conducted in accordance with the 1964 Helsinki declaration and was approved by the Sud-Est VI Statutory Ethics Committee of France. The participants received 20 euros for their participation.

Material

The participants were given a Motorola G Androit Jelly Bean smartphone, with a program written by CATech (http://lapsco.univ-bpclermont.fr/catech) at Clermont Auvergne University. This program managed all the events: alerts, temporal tasks, questions, recording of responses. The participants gave their responses by pressing on the touch screen of the cell phone. The stimulus at the onset and the offset of the interval duration was a sound (LA, 440 Hz).

Procedure

The participants were randomly assigned to one of the two duration conditions: 2/8-min (30 participants) and 8/32-min (29 participants). In each duration condition, there were 7 target interval durations to be estimated. The interval durations were 2, 3, 4, 5, 6, 7, and 8 min in the 2/8-min condition, and 8, 12, 16, 20, 24, 28 and 32 min in the 8/32-min condition. Each target duration was presented 3 times, together with 5 other durations randomly chosen between 1 and 16 for the 2/8-min condition, and between 4 and 64 minutes for the 8/32-min condition. This represented a total of 26 trials. The 26 trials were presented randomly during the day, the inter-trial interval being between 10 and 17 min (M = 13.5, SD = 2.3). The procedure was therefore similar in the two duration conditions, except for the length of the interval durations to be estimated and the consequence of this on the duration of the session, with the trials ranging from 7.00 a.m. to 7.30 p.m. for the 2/8-min condition and from 7.00 a.m. to 11.55 p.m. for the 8/32-min condition.

For each trial, an alert given by the cell phone indicated that a new trial was to be performed. The participant touched the phone screen to stop this alert. The word "ready" then appeared and the participant again touched the phone screen to trigger the trial. The participants therefore heard two sounds separated by the interval duration to be judged. The length of the first sound was 2 s and the second stopped when the participants touched the cell phone in order to respond. The participants then gave their estimate of the interval duration by pressing on the touch screen of the cell phone and received no feedback.

After each verbal estimate, the participants answered 7 questions on their cell phones. The first (1) was the question on the speed of the passage of time compared to the notion of a constant clock rhythm (PoTJ): “how is time passing for you between the 2 sounds, compared to the time of the clock, from "much slower" to "much faster?”. The participant responded to this question on a 7-point scale going from "much slower" to "much faster". The next 4 questions related to the affect they had experienced during the interval between the two sounds: (2) “excited/stimulated” (Arousal), (3) “relaxed/calm” (Relaxation), (4) “happy” (Happiness), and (5) “sad” (Sadness). The last two questions then related to the activity performed between the two sounds: (6) if it was difficult and (7) if it had captured all their attention. For the affective and the activity questions, the participants responded on a 7-point scale from “not at all” to “very much”.

The day before the experimental session, the participants were instructed that they would have to judge the duration between two sounds and two demonstration trials were administered with an interval of a few seconds. They were told only that the durations would be between 1 and 16 minutes for the 2/8-min condition and between 4 and 64 min for the 8/32-min condition. The participants activated their cell phones on the morning of the day of the test session and kept their cell phone with them during the whole day. They were required to stay at home during this specific day. They were also instructed to remove all clocks (watches, cell phone clock, clock) because looking at a clock would have biased the scientific results. They were only allowed to consult a watch or their cell phone under exceptional circumstances between two trials if they judged this to be absolutely necessary. The participants, who were paid, agreed to comply with these experimental constraints before participating in this study. They were also instructed not to count, even though this would have been impossible for the very long durations. As indicated by the results, the variability of temporal estimates did not remain constant but increased with the duration length, as expected due to the lack of a counting strategy [25].

Results

Temporal estimates of long durations

Mean duration.

The mean estimated duration was calculated for the target interval durations in the 2/8-min and the 8/32-min condition (Fig 1). An initial statistical analysis was conducted on this temporal measure that did not find any effect of the sex factor. This factor was thus excluded from the subsequent analyses. An ANOVA was performed on the mean estimated duration using IBM SPSS, with the duration condition as between-subjects factor and the interval durations as within-subjects factor (7 interval durations). This ANOVA showed significant main effects of interval duration, F(6, 342) = 99.62, p = .0001, η²_p = .64, and duration condition, F(1, 57) = 787.27, p = .0001, η²_p = .93, as well an interval duration x duration condition interaction, F(6, 342) = 10.50, p = .0001, η²_p = .16. For each duration condition taken separately, there was a linear effect of interval duration (2/8-min, F(1, 29) = 372.12, p = .0001, η²_p = .93; 8/32 min, F(1, 28) = 243.80, p = .0001, η²_p = .90), indicating an increase in the mean temporal estimates as the duration value increased. In other words, time discrimination was maintained for very long durations of several minutes. However, the post-hoc comparisons using the Bonferroni test in the 8/32-min condition revealed that the mean of the estimated durations did not differ between the interval durations of more than 24 min (24 vs. 28, 24 vs. 32, 28 vs. 32, p > .05), although these durations differed from the other shorter interval durations. Similarly, in the 2/8-min condition, there was no difference between the temporal estimates for the two longest interval durations (7 vs. 8 s, p > .05).

Download:

Fig 1. Mean duration.

Mean estimated duration plotted against interval duration in the 2/8 and the 8/32-min condition.

https://doi.org/10.1371/journal.pone.0195397.g001

Standardized error.

To examine the size of the deviation of the time estimates from the target durations, we calculated an index of standardized error that was the difference between the time estimates and the target duration, divided by the target duration. The ANOVA conducted on this new measure showed a significant main effect of the duration condition, F(1, 57) = 15.44, p = .0001, η²_p = .23, indicating that the overestimation of long durations was greater in the 2/8-min condition than in the 8/32-min condition (0.36 vs. 0.15). However, there was also a significant main effect of the interval duration, F(6, 342) = 12.58, p = .0001, η²_p = .18, and a significant interaction between the interval duration and the duration condition, F(6, 342) = 12.58, p = .0001, η²_p = .18. The ANOVA conducted on each duration condition taken separately showed no significance of the interval duration, F(6, 174) = 1.97, p = .072, in the 2/8-min condition, while this factor was significant in the 8/32-min condition, F(6, 168) = 29.65, p = .0001, η²_p = .51. As illustrated in Fig 2, in the 2/8-min condition, the interval durations were all overestimated, with the time estimates being significantly different from zero (p < .05 for all t tests) and the same magnitude of overestimation occurring across the different interval durations. By contrast, in the 8/32-min duration condition (8/32 min), the time estimates for the shortest interval durations were also different from zero and overestimated (8 min, t(28) = 8.08; 12 min, t(28) = 5.62; 16 min, t(28) = 5.18, all p < .05), while those for the longest durations were underestimated (28 min, t(28) = -2.52; 32 min, t(28) = -4.80, p < .05), with a central time judgment at 20 and 24 min (t(28) = 1.55, t(28) = 1.17, respectively, both p > .05).

Download:

Fig 2. Standardized error.

Standardized error plotted against interval duration in the 2/8 and the 8/32-min condition.

https://doi.org/10.1371/journal.pone.0195397.g002

Standard deviation.

Another ANOVA was performed on the variability of time estimates (SD) (Fig 3). This analysis showed a significant main effect of the interval duration, F(6, 342) = 5.44, p = .0001, η²_p = .09, and of the duration condition, F(1, 57) = 99.93, p = .0001, η²_p = .64, while the interval duration x duration condition interaction did not reach significance, F(6, 342) = 0.84, p = .54. Consequently, the variability of time estimates was larger in the long than in the shorter duration condition (2.15 vs. 4.21), and it linearly increased with the length of the interval durations whatever the duration condition, F(1, 55) = 17.37, p = .0001, η²_p = 0.23. In accordance with the scalar property of time, when the SD was divided by the target duration (coefficient of variation) (Table 1), these effects were no longer significant. This was true of the interval duration, F(6, 342) = 1.61, p = .14, the duration condition, F(1, 57) = 1.08, p = .30, and the interaction between them, F(6, 342) = 0.34, p = .91. In each duration condition considered separately, we did indeed find a linear effect of interval duration on SD for both the 2/8-min condition, F(1, 29) = 71.79, p = .0001, η²_p = .71, and the 8/32-min condition, F(1, 28) = 6.72, p = .015, η²_p = .19, with no effect of interval duration being observed on the coefficient of variation, F(6, 174) = 1.26, p = .28, F(6, 168) = 0.83, p = .55, respectively.

Download:

Fig 3. Variability of time estimates.

Standard deviation of estimates plotted against interval duration in the 2/8 and the 8/32-min condition.

https://doi.org/10.1371/journal.pone.0195397.g003

Download:

Table 1. Temporal measures of judgment of interval durations in the 2/8-min and the 8/32-min duration condition.

https://doi.org/10.1371/journal.pone.0195397.t001

In addition, as the same interval duration of 8 min was used in the two duration conditions, we conducted a one-way ANOVA for each temporal measure for this specific interval duration. This showed no difference in the temporal variability between the duration conditions for the SD, F(1, 57) = 0.59, p = .44, or the coefficient of variation, F(1, 57) = 0.34, p = .56. The 8-min duration was nevertheless judged longer in the 8/32-min (M = 12.20, SE = 0.52) than in the 2/8-mn condition (M = 10.40, SE = 0.51), F(1, 57) = 6.15, p = .02, η²_p = .097, and a higher standardized error was observed, F(1, 57) = 5.93, p = .02, η²_p = .094, thus confirming the influence of temporal context [16,26].

The predictors of time estimates of long durations

On each trial, the participants reported their judgment of the passage of time during the interval duration (inter-sound interval), their affective state and the level of difficulty of the activity performed during this interval, and the extent to which it captured their attention. We analyzed which factor might potentially be a significant predictor of their time estimates (mean estimated durations) by using multilevel modeling (MLM) in the SPSS software (for the same procedure with similar measures and a description of this model, see [21,22,23,24]). Since the temporal measure (mean estimated durations) changed as a function of the duration condition, we decided to conduct these analyses for each duration condition taken separately.

For both the 2/8-min condition (Table 2) and the 8/32-min condition (Table 3), the best predictor of the verbal estimation of durations lasting several minutes was the passage of time judgment (PoTJ) (p = .0001). The interval duration was therefore judged shorter when time was experienced as passing faster, and increased when it seemed to pass slower. The other factor that significantly predicted the estimation of interval durations of several minutes was the affective state, i.e., the level of arousal (excitement/stimulation) felt during the interval (2/8-min, p = .04; 8/32-min, p = .004). The higher the arousal level was, the shorter the interval duration was judged to be. In other words, the interval duration was judged longer when the self-reported level of arousal decreased. Our results suggest that the estimation of long durations of several minutes was not linked to other factors, such as those related to the affective valence (happiness, sadness) or the level of attention devoted to the current activity (difficulty, attention). However, when all duration conditions were included in the same statistical analysis (Table 4), these factors also reached significance (p < .05), except for the attention paid to the current activity, which just failed to reach significance (p = .09). Consequently, when a larger number of trials was considered, the estimated durations became shorter as the level of task-difficulty increased. As far as affective valence is concerned, the time estimates lengthened with the level of happiness, and shortened with that of sadness. The analyses for the standardized error are not reported, but found that the PoTJ was the main significant predictor of inter-individual differences in the magnitude of temporal judgment errors (Estimate = -0.055071, SE = 0.0114, t = -4.84, p = .0001).