Multiple factors may change the perception of brief time intervals. The nature of the stimulus marking time is fundamental. Whether an interval is filled or empty, or marked by auditory or visual stimuli, for instance, will affect the experience of time (Grondin, 2001). Even within the family of filled intervals marked by visual stimuli, another critical factor affects the experience of time: the emotional value of the stimuli (Droit-Volet & Meck, 2007; Lambrechts, Mella, Pouthas, & Noulhiane, 2011; Tipples, 2008, 2011).

Recently, there has been a significant amount of scientific work showing that the emotional value of stimuli—mainly the facial expressions of emotion—exert influence on time perception (Droit-Volet, Fayolle, Lamotte, & Gil, 2013; Droit-Volet & Gil, 2009). Indeed, experience of time is often reported to depend on the output of an internal clock mechanism. In the contemporary time-perception literature, this clock is often reported to be a pacemaker–counter process (Grondin, 2010). This pacemaker emits pulses whose accumulation by the counter determines the experience of time. Moreover, this accumulation is under control of attentional processes, with more attention dedicated to the passage of time, resulting in larger accumulations and, therefore, in longer perceived durations.

When time has to be estimated, it has to be marked by some sensory signals. If these signals contain some information linked to emotion, they will exert influence on the estimation of time. Indeed, this influence of emotional information on perceived duration may arrive from two routes: via the pacemaker itself or by the attention mechanisms involved in the processing of time (Gil & Droit-Volet, 2012). For instance, compared with seeing neutral faces, seeing angry faces results in a subjective lengthening of time (Gil & Droit-Volet, 2011b; Tipples, 2008). In this case, it could be assumed that observing threatening faces generates arousal, which is argued to speed up the pacemaker component of the internal clock. With more pulses emitted, more pulses are accumulated, and time is therefore perceived as longer. On the other hand, some emotions may rather exert influence on perceived duration through their action on the attention to time. For instance, using faces displaying shame for marking time is reported to reduce perceived duration (Gil & Droit-Volet, 2011c). These authors explain this process by the fact that shame, which is a social–moral emotion, captures attention by a reflexive activity of self-awareness and directs attention away from the interval to be timed. Consequently, fewer pulses are accumulated and time is perceived as being shorter than it would have been with marking signals emotionally neutral.

Some other data in the literature indicate that the link between emotion and time perception requires additional nuances. According to Angrilli, Cherubini, Pavese, and Manfredini (1997), an emotion eliciting both a negative valence and low arousal should lead to shorter perceived duration. Simply observing a face expressing disgust should not generate too much arousal, but the valence of what is observed remains negative. Although some underestimation of time should be observed in this condition, Gil and Droit-Volet (2011a) reported that there is no tendency to underestimate time when disgusted faces are observed. As far as we know, no other data are available about the effect on time perception of observing disgusted faces, and more investigation is certainly needed before accepting the null hypothesis (no effect of disgusted faces on time perception). On the other hand, when nonattractive pictures of food are presented, time is underestimated (Gil, Rousset, & Droit-Volet, 2009). There is no reason to believe that presenting disgusting food pictures generates less arousal than presenting disgusted faces.

When disgusting images of body mutilations are presented, the effect on time perception is quite different. Images of body mutilations not only elicit disgust (negative valence), but also provoke high arousal (Angrilli et al., 1997). Gil and Droit-Volet (2012), with verbal estimates (typing a value on a keyboard), and Angrilli et al. (1997), with a line marking and a reproduction method, observed that time is lengthening when mutilation pictures are shown. In these studies, the effect is mainly attributed to arousal.

Angrilli et al. (1997) worked with 2-s, 4-s, or 6-s intervals and reported an arousal by duration interaction effect. High arousal leads to the overestimation of time at 2 s, but to an underestimation at 6 s, an effect attributed to attention. In Gil and Droit-Volet (2012), different participants were assigned to one of three emotion conditions—namely, disgust, sadness, and disgust/fear—and in each case, the arousal condition (high vs. low, in addition to neutral) was manipulated. The presentation of arousal conditions and durations (seven values from .2 to .8 s, or from .4 to 1.6 s) were randomized. Their results revealed a lengthening effect of emotion—compared with a neutral condition—an effect that was increased when arousal levels were higher.

The lengthening effect reported by Gil and Droit-Volet (2012) is quite clear, but a different look at the data reveals something intriguing. In all cases, the estimated duration of intervals was lower than their actual duration. Because the longest intervals were obtained with mutilation images (disgust/high arousal condition), it is in this condition that timing could be considered to be the most accurate. In the experiments proposed below, different emotional conditions will be compared. In addition to analyzing the differences in perceived duration (underestimation vs. overestimation) caused by emotions, it will be possible to determine a performance level, since an interval discrimination task will be used. Adopting this discrimination method for investigating the effect of emotions on time perception is not only important for providing an occasion to determine their relative performance levels. Indeed, when one wants to determine the effect of emotions on perceived duration, the choice of the temporal tasks might be a critical factor. Gil and Droit-Volet (2011b) investigated with five temporal methods the potential lengthening that could be expected from the presentation of angry faces. They reported an impact of angry faces on time perception with bisection, verbal estimation, and production tasks, but not with generalization and reproduction tasks.

In brief, there is something unclear about the effect of disgust on time perception. Although time is underestimated with the presentation of unattractive pictures of food, time is overestimated when pictures of mutilated bodies are shown. What is more, although the presentation of pictures showing emotional facial expressions leads to temporal distortions, apparently viewing disgusted faces does not (Gil & Droit-Volet, 2011a). Moreover, there is only one report of the effect of disgusted faces on time perception, and it is known that the methods used to study the effect of attention on time perception may influence the verdict about this effect.

The purpose of the present study is to provide an additional test—a different method—for estimating any temporal distortion due to the observation of disgusted faces or disgusting mutilation pictures. The method used in the present study consists of relative judgments rather than absolute judgments (such as the verbal estimate method used by Gil & Droit-Volet, 2012). Moreover, the present study is also an occasion for testing the effect of the emotional content of images on discrimination levels. This last issue is quite important considering that one interpretation of the data of Gil and Droit-Volet (2012) leads to the idea that timing is more accurate when pictures of mutilated images are used.

Experiment 1

In Experiment 1, neutral faces (N), disgusted faces (F), and disgusting mutilation (M) pictures serve to mark successive intervals (400 ms vs. 482 ms) to be discriminated. The intervals are presented back to back, in a random order, with a brief interstimulus interval (ISI) in between.

In the study, there will be conditions in which the same type of image (both N, both F, and both M) or different types will be used to mark the duration of the first and second stimuli. It will be possible to determine the overall tendency, for each type of image, to respond that the second is shorter or longer than the first when both stimuli are N, F, or M (baselines), and to contrast these results with conditions in which the image types differ.

More specifically, whereas the ISI will last 1 s, the period following the presentation of the second image will be brief, because the message asking the participant to respond will appear on the screen after 200 ms. This manipulation should have an impact on the involvement of attention in the discrimination process. On the one hand, it is known that the processing of an interval might not end exactly at the end of the physical signal marking time (Grondin, 1993; Nakajima, 1987). On the other hand, it cannot be excluded that an attentional component is at play when these signals carry some emotional content. For instance, Öhman, Flykt, and Esteves (2001) have shown that evolutionarily relevant threatening stimuli, such as snakes, can capture attention. What is more, it is also known that emotionally relevant stimuli, especially of an extreme negative valence, can be more difficult to disengage from than neutral images (Biggs, Kreager, Gibson, Villano, & Crowell, 2012).

In the processing of temporal information, attention plays a critical role (Brown, 2008; Grondin & Macar, 1992; Macar, Grondin, & Casini, 1994). Indeed, according to the attentional gate model (Block & Zakay, 2008), attention has two roles. In the process of pulses’ accumulation, described earlier, it determines, via a switch component, the moment when the timekeeping activity starts and stops. Moreover, the fact of paying more or less attention to time, as opposed to paying attention to nontemporal information (as in a dual-task condition) during the interval to be timed, determines the perceived duration of an interval because more pulses are accumulated when more attention is devoted to time.

Since participants are not subjected to a dual-task situation (there is no task to complete other than estimating time), we may assume from a resource viewpoint that, even when using images of body mutilation, no pulses will be lost. However, a mutilation image could be much more difficult to disengage from than a neutral image (Biggs et al., 2012). Assuming that the pacemaker is free running and its pulses are accumulated as long as the stop signal is not internally clear (the participants are still dwelling on what they just saw), interrupting the processing of the second of two successive images after 200 msec should be more damaging if this image is a picture of mutilation than if it is a neutral image, the latter being easier to disengage from.

Consequently, if two successive mutilation images are presented, the second one should be perceived as shorter if the delay before the next event, 200 msec, is briefer than the ISI (1 s). Such a line of reasoning also predicts that if two neutral images are presented successively, there should be no such tendency to perceive the second one as briefer. Also, if neutral and mutilation images are compared, the overall difference in perceived duration should be larger when mutilation images are presented first than when neutral images are presented first.

The predictions that can be derived from using disgusted faces are more difficult to establish. As for mutilation images, there is some content to analyze. However, it should be easier to disengage attention from these faces than from mutilation images. Therefore, any effect on perceived duration should be minimal, or at least weaker, than that obtained when viewing mutilation images.

Method

Participants

Eighteen students and employees of Université Laval (four men and 14 women) participated in the experiment (ages 19–40 years; mean = 24.05; median = 23). They received a monetary compensation of Can$10 for the 60-min session.

Apparatus and stimuli

Participants were seated in a dimly lit room, directly facing a 16-in. monitor. Three types of visual stimuli were used: images of neutral emotion expression, images of faces expressing disgust, and images of disgusting mutilated human bodies. Five samples of each stimulus type were used. The images of mutilated bodies were sampled from the International Affective Picture System (Center for the Study of Emotion and Attention, 1995: images 3010, 3062, 3063, 3064, 3130). The pictures of faces were selected from the Montreal Set of Facial Displays of Emotion (MSFDE: Beaupré, Cheung, & Hess, 2000). The experiment was under control of an IBM computer.

Procedure

This one-session experiment contained nine blocks of 50 trials. During each trial, two intervals, one of 400 ms and the other of 482 ms, were presented in a random order. These intervals were filled, that is, marked by the onset and the offset of the visual stimulus. There was a 1-sec empty interval between the presentations of images. Moreover, there was a 200-ms delay between the end of the presentation of the second image and the appearance of the message on the screen asking the participant to respond.

Participants were asked whether the second interval was shorter or longer than the first interval. They responded on the computer keyboard by pressing either “1” for shorter or “3” for longer. For each of the blocks, one of nine specific emotion pairs was presented. The pairs were N–N, M–M, F–F, N–M, M–N, N–F, F–N, M–F, or F–M. Each participant was first presented with the three conditions in which the stimuli were identical. The order of presentation for each participant is reported in Table 1.

Table 1 Stimulus counterbalancing (in Experiments 1 and 2)
Full size table

Dependent variables and effect sizes

There are two dependent variables of interest. One is the probability of responding “long,” which indicates if the second interval is perceived as particularly short or long in one of the experimental conditions. The other one is the probability of responding correctly, which indicates the discrimination levels (or sensitivity). For instance, in the N–N condition, a participant may respond “long” in 70 % of the instances in which the second interval is actually long, and 20 % “long” when the second interval is short. This participant would have a total probability of responding “long” of 45 % ((70 + 20)/2) and a total probability of responding correctly of 75 % (i.e., the mean of 70 % of “long” responses when the second interval is long, and of 80 % of “short” responses when the second interval is short).

In order to quantify the size of the effects and of the significant differences, the following indices were used. For the effect calculated from the analyses of variance (ANOVAs), both the traditional partial eta-squared (η p 2) and the generalized eta-squared (η 2 g) were reported. The latter is a measure of the size of the effects as they would have appeared in a between-subjects design (see Bakeman, 2005). A variant of Hedge’s g was used to quantify the size of the differences between the experimental conditions. The relevant Cohen’s d statistic was simply equal to the difference of the means divided by the arithmetic average of their standard deviations.

Results

Perceived duration

In the first step of the analyses, the conditions in which both stimuli marking time were identical (i.e., N–N, M–M, or F–F) were compared. Figure 1 shows the mean results (probability of responding “long” and of responding correctly) in each condition. The left part of Fig. 1 reveals a strong tendency to respond “short” more often than “long” in the M–M condition. A one-way repeated-measures ANOVA revealed that the differences among conditions are marginally significant [F(2,34) = 2.84, p = .076, η p 2 = .143, η 2 g = 0.059]. Moreover, M–M was the only condition under which the probability of responding “short” was significantly greater than 50 %, p < .001.

Fig. 1
figure 1

Mean probability of responding “long” (left panel) or correctly (right panel) in conditions in which both images are from the same category in Experiment 1. Bars are standard errors. N = neutral faces; F = disgusted faces; M = mutilation images

Full size image

Figure 2 (upper panel) shows the probability of responding “long” in each of the conditions in which stimuli marking time were dissimilar. For each pair (e.g., N–M vs. M–N), the order was compared. A series of t-tests with dependent samples revealed that participants responded much more often that the second stimulus was short in the M–N than in the N–M condition [t(17) = 2.136, p < .05, g =.870]; the difference between the N–F and F–N conditions was not significant (p = .314); and participants responded much more often that the second stimulus was short in the M–F than in the F–M condition [t(17) = 3.042, p < .01, g = 1.150].

Fig. 2
figure 2

Mean probability of responding “long” (upper panel) and mean probability of responding correctly (lower panel) in conditions in which both images are from different categories in Experiment 1. Bars are standard errors. N = neutral faces; F = disgusted faces; M = mutilation images

Full size image

Sensitivity

The same comparisons as the ones described in the previous section were used to approach the data. Figure 1 (section on the right) indicates that the discrimination level is lower in the M–M condition than in the N–N or F–F conditions. However, a one-way ANOVA revealed that the differences are not significant (p = .286).

Figure 2 (lower panel) illustrates the discrimination level in each condition in which images differ within each pair. A one-way repeated-measures ANOVA was conducted, which compared the average probability of responding correctly for the N–F and F–N pairs versus the M–N and N–M pairs versus the F–M and M–F pairs. It revealed that there were no significant differences between these conditions (p = .462). Moreover, there was no difference between the F–M and M–F conditions (p = .533) and between the N–M and M–N conditions (p = .243). Note, however, that participants tend to be better in the F–N than in the N–F condition [t(17) = 1.826, p = .085, g = .231].

Finally, the mean discrimination level in the N–N condition (M = .740) is significantly higher than the mean discrimination level (M = .682) involving the N condition with F or M stimuli (N–F, F–N, N–M, and M–N) [t(17) = 3.094, p = .007, g = .472].

Discussion

Experiment 1 reveals that when the same type of stimulus is presented back to back in the discrimination task, the probability of responding that the second interval is shorter or longer than the first one is not 50 %. It is in the case of mutilation that this tendency is the most important. The results also reveal that, when compared with neutral or disgusted faces, duration is perceived as much longer when disgusting mutilation images are presented.

For now, let us focus on the fact that when both stimuli are from the same category, there is an overall tendency to respond that the second image is presented for a briefer period than the first. In the time-perception literature, when the probability of response about the relative duration of two intervals presented successively does not equal .5 (for each of “shorter” and “longer” responses), there is a time order error (TOE, also known as a “presentation order error”). Responding more often that the second of two intervals is shorter (overestimation of the first) corresponds to a positive TOE (Hellström, 1979, 1985), a finding reported, for time perception, when the intervals to be discriminated are brief (<1 s; see Eisler, Eisler, & Hellström, 2008).

What is interesting here is the fact that the probability of responding short is .4 in the M condition, but .45 in the N and F conditions. It seems that there are two effects at play here. One could be strictly psychophysical (TOE), and the other related to emotion. It cannot be excluded that the overall effect was caused by the brief 200-ms delay before the message requesting a response appeared on the screen. This brief-delay effect would be more damaging when the second image marking time shows body mutilation, since such an image probably requires a longer time for disengaging attention (Biggs et al., 2012).

Experiment 2

The present experiment is designed to test the potential effect of using a very brief delay between the second image and the message on the screen asking for a response. This experiment is a replication of the first, but with a longer delay after the presentation of the second image. When the second interval is marked with a mutilation image, having a longer delay should reduce the damage caused by the time required for disengaging attention. If attention is critical in the process, the tendency to perceive mutilation images as longer, if any, should be reduced.

Method

Participants

Eighteen students and employees of Université Laval (six men and 12 women) participated in the experiment (ages 20–37 years; mean = 24.94; median = 22.5). They received a monetary compensation of Can$10 for the 60-min session. One participant produced chance performance for every experimental condition, so his/her data were excluded from further analyses.

Apparatus, stimuli, and procedure

The method used was exactly as in Experiment 1, except that the delay between the second image and the message on the screen asking for a response was 1 s. In other words, the ISI and this delay were the same.

Results

Perceived duration

Figure 3 shows the mean probability of responding “long” and of responding correctly in each condition. The left part of Fig. 3 reveals a tendency to respond “short” more often than “long” in the M–M condition, but no such effect in the two other conditions. More than 55 % of the 900 responses (from all participants) in the M–M were “short” (but note that for the probability of responding “short” more often than 50 %, p = .06). A one-way repeated-measures ANOVA revealed that the differences among conditions were not significant [F(2, 32) = 1.894, p = .167, η 2 p = 0.106, η 2 g = 0.029].

Fig. 3
figure 3

Mean probability of responding “long” (left panel) or correctly (right panel) in conditions in which both images are from the same category in Experiment 2. Bars are standard errors. N = neutral faces; F = disgusted faces; M = mutilation images

Full size image

Figure 4 (upper panel) shows the probability of responding “long” in each of the conditions in which stimuli marking time were dissimilar. For each pair (e.g., N–M vs. M–N), the order was compared. A series of t-tests with dependent samples revealed that participants did not respond significantly more often that the second stimulus was short in the M–N than in the N–M condition [t(16) = 1.781, p = .094, g = .550]; the difference between the N–F and F–N conditions was not significant (p = .857); and participants responded more often that the second stimulus was short in the M–F than in the F–M condition [t(16) = 2.266, p = .038, g = .821].

Fig. 4
figure 4

Mean probability of responding “long” (upper panel) and mean probability of responding correctly (lower panel) in conditions in which both images are from different categories in Experiment 2. Bars are standard errors. N = neutral faces; F = disgusted faces; M = mutilation images

Full size image

Sensitivity

The same comparisons as the ones described in the previous section and in Experiment 1 were used to approach the data (Fig. 3, right part). A one-way ANOVA revealed that the differences between the M–M, N–N, and F–F conditions were not significant [F(1.489, 23.826) = .789, p = .431, η 2 p = .047, η 2 g = .0119].

Figure 4 (lower panel) illustrates the discrimination level in each condition in which images differ within each pair. A series of paired-samples t-tests were conducted in order to compare the average probability of responding correctly for the N–F and F–N pairs, for the M–N and N–M pairs, and for the F–M and M–F pairs. There was no difference between the F–M and M–F conditions (p = .146, g = .239), between the N–M and M–N conditions (p = .865, g = .032), nor between the F–N and N–F conditions (p = .160, g =.317).

Finally, the mean discrimination level in the N–N condition (M = .704) is not significantly higher than the mean discrimination level (M = .655) involving the N condition with F or M stimuli (N–F, F–N, N–M, and M–N) [t(16) = 1.939, p = .070, g = .379].

Experiments 1 versus 2

Another way of approaching the results is to make a direct comparison between experiments, taking into account the fact that stimuli were presented in the long–short (L–S) versus short–long (S–L) order.

A 2 (Order) × 9 (Image condition) × 2 (Experiment) ANOVA, with repeated measures of the first and second factor, was conducted on the proportion of long responses. None of the effects (main or interaction) involving the experiment factor were significant. Indeed, only the Stimulus × Order interaction was statistically significant [F(8, 264) = 3.695, p < .001, η 2 p = .101, η 2 G = .0127]. Because the probability of responding that the second interval is longer will of course be higher in the S–L order, the rest of the analysis (order and image condition) is focused on the probability of responding correctly (sensitivity).

The same 2 (Order) × 9 (Image condition) × 2 (Experiment) ANOVA, with repeated measures of the first and second factors, was conducted on the proportion of correct responses. The analysis showed no experiment effect [F(1, 33) = .613, p = .439, η 2 p = .018, η 2 G = .006]; furthermore, this factor did not interact significantly with the order factor [F(1, 33) = 2.481, p = .125, η 2 p = .070, η 2 G = .014] or with the image factor [F(8, 264) = .979, p = .453, η 2 p = .029, η 2 G = .0034]. Moreover, there was no significant triple interaction effect [F(3.54, 116.91) = .297, p = .858, η 2 p = .009, η 2 G = .004].

There was a main effect of order [F(1, 33) = 7.713, p =.009, η 2 p = 0.189, η 2 G = 0.042] and of the image condition [F(8, 264) = 3.679, p < .001, η 2 p = 0.100, η 2 G = 0.013] and, most importantly, there was a significant interaction effect between these factors [F(3.54, 116.91) = 5.218, p < .001, η 2 p = 0.137, η 2 G = 0.0589]. Table 2 reports the mean probability of responding correctly for each image condition in the short–long and long–short conditions.

Table 2 Mean probability of responding correctly (and standard errors of the mean) for each image condition with the short–long and long–short sequences. Data from Experiments 1 and 2 are pooled together. In these experiments, participants had to report whether the second of two intervals was shorter or longer than the first.
Full size table

For decomposing this interaction, we first looked at each stimulus pair. For the M–M, M–F, and M–N conditions, the differences between the short–long and the long–short conditions are statistically significant, t(34) = 4.404, p < .001, g = 1.176, t(34) = 4.006, p < .001, g = .980, t(34) = 4.010, p < .001, g = .987, respectively. In none of the other stimulus pairs was the difference statistically significant (p > .1).

We also approached the interaction by looking specifically at what happened in the S–L and in the L–S conditions. In the short–long condition, there was a statistically significant difference between the F–F and M–M conditions, t(34) = 3.082, p = .004, g = .430; between the F–M and M–F conditions, t(34) = 3.257, p = .003, g = .684; and between the M–M and N–N conditions, t(34) = 3.600, p < .001, g = .543. After Bonferroni’s correction (significance level set at .0042), there was no significant difference between the M–N and N–M conditions, t(34) = 2.596, p = .014, g = .603; between the F–F and N–N conditions, t(34) =.348, p = .730, g = .063; and between the F–N and N–F conditions, t(34) = .812, p = .412, g = .196.

For the long–short conditions, the following differences were found to be statistically significant: N–M vs. M–N, t(34) = 2.565, p =.015; and F–M vs. M–F, t(34) = 3.749, p < .001, g = .785. After Bonferroni correction, none of the following differences were significant: N–M vs. M–N (p =.015), F–F vs. M–M (p = .454), F–F vs. N–N (p = .887), F–N vs. N–F (p = .466), and M–M vs. N–N (p = .383).

Discussion

The main observation that can be extracted from Experiment 2 is the fact that extending from .2 to 1 s the delay before the message asking for a response reduced the TOE, but it did not erase the tendency to report that viewing disgusting mutilation images leads to an impression that duration is longer.

Of particular importance is the additional analysis integrating both experiments, in which the order (S–L vs. L–S) was taken into account. Firstly, this analysis reveals that participants responded correctly less often in the M–M condition than in the N–N and F–F conditions, but only when the shorter duration (400 ms) was shown first. Secondly, they also responded less accurately in the short–long than in the long–short conditions, only when a mutilation picture was shown first (M–M, M–F, and M–N conditions). Thirdly, since there was no main or interaction effect of the experiment factor (and the associated effect sizes were very weak), it follows that the delay between the offset of the second stimulus and the response prompt has no incidence on the participants’ correct response rate. Finally, in addition to these three points, the pooled data analysis reveals that the proportion of correct responses was lower in the M–F than in the F–M conditions when the shorter duration was shown first, whereas it was the opposite when the longer duration was shown first; and was lower in the M–N than in the N–M conditions when the shorter duration was shown first; but the difference between those two conditions was not significant when the longer duration was shown first.

As was mentioned earlier, one of the possible explanations for the fact that the proportion of “long” responses was below 50 % in the M–M condition of the first experiment pertains to attentional factors. When one is shown a mutilation picture, his or her attention could be captured by that stimulus. As such, the participant may not notice the stimulus’ offset, which would result in an overestimation of its duration. If the extent of this overestimation depends on the length of the interval following the offset of the stimulus, it could thus be argued that the perceived duration of a first mutilation picture was more overestimated than that of a second mutilation picture because the ISI of the first experiment was longer than the delay after the offset of the second stimulus. However, the third finding reported in the preceding paragraph does not support this explanation. Indeed, since the length of the delay after the offset of the second stimulus was set as equal to that of the ISI, there should have been a significant effect of the experiment factor in the analysis of the pooled data.

Note that this explanation does not necessarily mean that attentional factors did not contribute, at least partly, to the temporal distortions observed in Experiments 1 and 2. If showing a first mutilation picture induces a tendency to underestimate the duration of the second stimulus, it cannot be excluded that this first image captured the participants’ attention in such a way that they simply missed the onset of the second stimulus.

Experiment 3

In this experiment, we use a method in which the interstimulus interval and potential TOE effect should not be a concern. Only one image is presented on each trial, and participants are asked to respond, after each presentation, whether the image presented was short or long. This experiment will provide a new direct comparison of the perceived length of N, F, and M images. The present method is again based on relative judgments about the different conditions, but the role of the memory component should be different here.

Method

Participants

A total of 20 students and employees (12 female, eight male) of Université Laval volunteered for this experiment (ages 18–39 years, mean = 26.5, median = 25). They received a monetary compensation of Can$5 for their participation in the 20-min. session.

Apparatus and stimuli

These aspects of the method are exactly as in Experiments 1 and 2.

Procedure

In the present experiment, instead of using a forced-choice procedure, as in Experiments 1 and 2, a single-stimulus method is employed (Grondin, 2008). The experiment started with a memorization phase in which a picture was presented on the computer monitor for 400 ms (short) or 482 ms (long); each duration was presented three times, in a random order. After each presentation, a text display informed the participant whether this duration was the short or the long interval. Then, for five experimental blocks of 60 trials, pictures were presented on the screen for a time corresponding either to the short or the long duration. The participants had to indicate whether it corresponded to the short or long duration by pressing different keys on the keyboard. There was no message on the screen asking for a response; participants responded when they were ready. Within each block, there were 20 presentations (10 short and 10 long), in a random order, of each of three picture types: N, F, and M. Therefore, individual performances in each picture condition were based on 100 trials, 50 with short and 50 with long intervals.

Results

As indicated in Fig. 5 (upper panel), participants responded “long” much more often when a mutilation image was presented than when other images were presented [F(2, 38) = 9.39, p < .01, η p 2 = .331, η 2 G = .140]. Also, participants responded “long” significantly more often when the mutilation images, instead of disgusted or neutral faces, were presented when the stimulus was long (Fig. 5, lower panel, dashed columns) [F(2, 38) = 6.70, p < .01, η p 2 = .261, η 2 G = .0784] and when it was short (Fig. 5, lower panel, white columns) [F(2, 38) = 8.55, p < .01, η p 2 = .310 η 2 G = .183].

Fig. 5
figure 5

Probability of responding “long” in Experiment 3 when the interval is short (lower panel), when the interval is long (lower panel), and total (upper panel). Bars are standard errors. N = neutral faces; F = disgusted faces; M = mutilation images

Full size image

As for sensitivity, overall, the probability of responding correctly was the same in the three conditions (between .67 and .68), but as the reader can figure out from the figure and the preceding paragraph, there were significantly more correct responses in the mutilation condition when the intervals were long, and more incorrect responses when the intervals were short with the mutilation conditions. Note that, overall, there were many more correct responses when the intervals were short than when they were long. Even in the mutilation condition, more correct responses were given when the intervals were short.

General discussion

This series of experiments provides quite a clear picture of the impact of presenting disgusting mutilation pictures on perceived duration. As for discrimination level, the overall portrait is a little more blurred.

Perceived duration

Experiments 1 through 3 provide new evidence about the impact disgusting mutilation images have on time perception. Time is perceived as longer when viewing such images, compared with viewing faces expressing disgust or viewing neutral faces. This finding is consistent with that reported by Gil and Droit-Volet (2012), who asked their participants to type the value of the interval just presented, an interval lasting between .2 and .8 s, or between .4 and 1.6 s. In the present study, the demonstration involved duration discrimination, with both a forced-choice procedure where two successive intervals marked by different types of images (F, N, and M) are directly compared, and a single-stimulus procedure in which, after each image presentation (F, N, or M), a response is made on the basis of a duration representation in memory, a representation that is based on the mixture of preceding trials, that is, of different image conditions.

Gil and Droit-Volet (2012) reported in their investigation that the interval estimates were longer with images inducing higher levels of arousal. Although they provide a strong case for the role of arousal, their effect does not seem to be limited to arousal, because the images they used for generating high arousal did not lead to the same time estimates. They used images of snakes, which clearly induce fear and increase arousal, and they used mutilation images, which are disgusting and also increase arousal. Nevertheless, there was a clear difference between these conditions, fear and disgust: Intervals were estimated as longer with disgusting mutilation images. Clearly, arousal alone cannot account for this difference.

One alternative hypothesis for explaining this type of difference pertains to the potential role of attention mechanisms: Time perception is known to be sensitive to manipulations of attention (Brown, 2008; Grondin & Rammsayer, 2003; Macar et al., 1994). Moreover, stimuli such as pictures of snakes or of spiders can capture attention (Öhman et al., 2001). It is also known that disengaging attention from a stimulus with an extreme negative valence is very difficult (Biggs et al., 2012). This could mean that it is more difficult to end the timekeeping period when a mutilation image is presented because attention is still attracted by the image. In other words, the period of accumulation of pulses would be lengthened with mutilation images, and, consequently, time would be perceived as longer. Longer perceived duration with mutilation images is essentially what is reported in the present study. However, such an attentional perspective should have led to different results in Experiments 1 and 2. In Experiment 1, the processing of the second image was interrupted more rapidly than the processing of the second image in Experiment 2, because the requirement to respond appeared on the screen more rapidly in Experiment 1. If disengaging from a stimulus with an extreme negative valence is a critical determinant of perceived duration, then the damage caused by the different delays before the responding message in Experiments 1 and 2 should have been greater in the mutilation condition.

Overall, the different delays led to different probabilities of responding “long,” but the difference between Experiments 1 and 2 did not depend on the image conditions. Mutilation images were perceived as lasting longer than images from the other conditions in both experiments, and the magnitude of this difference was essentially the same. Therefore, our data do not support an attentional explanation for the fact that mutilation images are perceived as lasting longer than other images used in the present study. Of course, the fact of not observing a difference in magnitude between Experiments 1 and 2 (acceptance of the null hypothesis) cannot be considered a demonstration that attention is not a critical factor for explaining the effects of mutilation images on perceived duration. Other manipulations of attention may eventually be successful. But for now, the arousal explanation seems very robust. Increased arousal, from the pacemaker–counter viewpoint described earlier, results in more pulses emitted and, consequently, in longer perceived duration. Therefore, the effect of mutilation images would be linked to the activity of the pacemaker.

Experiment 3 provides a different angle on the data. The same short and long intervals had to be categorized as short or long, but there were no direct comparisons of two images presented successively, before a response had to be made. In other words, there was no reason to have any bias, as may happen (TOE) when a forced-choice procedure is adopted (see, e.g., Englund & Hellström, 2012). Nevertheless, there was an overall tendency to respond “short” more often than “long” in the neutral and disgusted face conditions. In addition to revealing once more that using mutilation images leads to longer perceived duration, these results may be interpreted as a consequence of a relaxing effect (reduced arousal) when it is not a mutilation image that is presented. This means that, because of its impact on emotion, the weight of trials involving mutilation images would be greater in the building of the interval representation in memory serving for the comparison with the interval presented. Therefore, in addition to attention and arousal, part of the differences in perceived duration with emotional images may depend on the link between emotion and memory. Memory processes are often reported to be a main source of variance in the processing of time intervals (Gibbon, Church, & Meck, 1984).

Finally, the data also reveal that there is not much difference in perceived duration between the disgusted face and neutral conditions. It is mostly arousing facial expressions such as fearful or angry faces that seem to generate temporal overestimations (Gil & Droit-Volet, 2011a; Grommet et al., 2011; Tipples, 2011). Whereas the impact on time perception of experiencing disgust is strong, that of seeing disgusted faces remains minimal.

Sensitivity

The present study provides an opportunity to assess the sensitivity to time when emotional stimuli are used to mark intervals. In the study by Gil and Droit-Volet (2012), durations were estimated as longer when mutilation images were used, in comparison with other images. Since all estimates in that study were underestimations (i.e., lower than the real value presented to participants), it is in the conditions in which intervals were the least underestimated (longer estimates for mutilation images) that participants were the closest to real time (more precise).

The present study clearly shows that no gain in precision can be expected from using mutilation images. Discrimination levels were not higher in this condition than in others (with F or N images), whether a forced-choice (Experiments 1 and 2) or a single-stimulus (Experiment 3) method was used. The effect involving mutilation images on perceived duration was not paralleled with an effect on the discrimination level. If anything, when compared with the N–N and F–F conditions, the discrimination level was slightly lower in the M–M condition in Experiments 1 and 2 (forced-choice procedure). Indeed, one may have expected lower discrimination with mutilation images, considering the potential attentional disturbance that viewing such images may cause. Moreover, if perceived duration increases with mutilation images as a result of an increased accumulation of pulses, one may expect more response variability (a lower discrimination level) on longer intervals. That expectation stems from the assumption that variability increases proportionally with the magnitude of time, as would be the case would Weber’s law apply; this is usually accepted in the animal timing literature (scalar property), but is admittedly a disputable issue (Grondin, 2001, 2012).

Mixing categories leads to lower discrimination levels than using two neutral images, at least in Experiment 1 (p = .07 in Experiment 2). Compared with the condition in which two neutral faces were presented, the condition in which one N image and either an F or M image were presented decreases the overall level of discrimination. Whereas an attentional hypothesis is not kept for explaining the differences relative to perceived duration, the different discrimination levels when categories of images are changed within a trial may be due to the attentional mechanisms involved in the task. One may posit that, with a forced-choice procedure, there is not much cost when attention is switched from one neutral image to another neutral image during a given trial, as is the case in an N–N condition. However, there is probably a loss of efficiency when there is a need to switch from a neutral image to another type of image, either F or M, or from an F or M image to a neutral one; or even when switching between M and F images. In Experiment 1, the probability of responding correctly is reduced from around 74 % (N–N) to 68.2 % (see Fig. 2, lower panel); and in Experiment 2, from 70.4 % to 65.5 % (see Fig. 4, lower panel).

From a psychophysical viewpoint, it is also noticeable that the discrimination levels in Experiments 1 and 2, versus Experiment 3, are essentially the same. Because the N, M, and F images were randomized in Experiment 3, this must be compared with instances in Experiments 1 and 2 with conditions in which two different categories of images were compared. In these conditions, the discrimination levels are about 67 %, whether a single-stimulus or a forced-choice method is adopted.

Conclusion

The present data offer direct and clear evidence that in comparison with conditions in which neutral or disgusted faces are presented, duration is perceived as being much longer when images of mutilations are presented. This effect was obtained with both a forced-choice and a single-stimulus method. Also, overall, the difference for perceived duration between neutral and disgusted face conditions seems negligible.

Although the attentional hypothesis cannot be totally discarded, the effect obtained with mutilation image presentations is probably caused by the increased arousal associated with this type of image. This arousal increases the rate of a pacemaker’s emission of pulses, which in turn results in longer perceived duration. Experiment 3 indicates that, in order to understand the impact of intense emotion on perceived duration, a closer look at the impact of intense emotions on memory for time is an interesting avenue.

Finally, the discrimination level remains about the same, whether N, F, or M images are presented, but presenting two successive images from different categories (F, N, or M) leads to a lower level of discrimination than does presenting two images containing no emotional expression.