Introduction

This article investigates how interval timing might be affected by motor activity. Interval timing involves making temporal duration judgments in the range of 500ms to several minutes [1][2][3][4]. In contrast to precision timing, which is involved in many of our automatic motor acts, interval timing requires cognitive resources, and is typically assessed through the use of explicit judgments about interval magnitudes.

Several studies have shown that the accuracy of interval timing judgments can be altered by factors such as cognitive load and stimulus characteristics (see [5] or [3] for reviews). For example, higher stimulus intensity dilates the sense of time [6]. The same is true of stimulus flicker frequency [7]. Stimulus movement and the velocity of this movement also influence interval timing judgments [8][9][10]. Brown [8] demonstrated that a moving stimulus perceived for a fixed amount of time is perceived as being of longer duration than a stationary stimulus perceived for the same amount of time. The author also demonstrated that faster speeds increased perceived time to a greater degree than slower speeds. Tomassini et al. [10] extended these results to show that they held not only for visual but tactile stimuli as well. The position of the stimulus in a repetitive sequence of the stimulus also plays a role in our perception of its duration. The first stimulus is judged to last longer than the other stimuli in the sequence [11][12].

Another factor found to influence estimates of interval duration, has been attention and cognitive load [12][13][14][15][16]. Block, Hancock, and Zakay [16] after having analyzed the results from a total of 117 studies found a striking interaction between the type of time judgment requested and cognitive load. High cognitive load increases your estimates in the case of retrospective timing, whereas high cognitive load decreases your estimates in the case of prospective timing.

In contrast to the above studies, however, very few investigations have explicitly attempted to explore how motor activity can affect interval timing. A notable exception is the work of Haggard et al. [17], who found that subjects systematically reported that the interval between a button press and the resulting stimulus onset was shorter than its actual duration. In other words, the motor act (i.e., pressing the button) led to an underestimation of the time between the button press and the stimulus onset. This compression of the time interval between the motor act and the appearance of the stimulus can even lead to a reversal in the judgment of which event occurred first [18]. It should be noted, however, that the button press in these studies plays a ‘causal’ role i.e. stimulus onset is caused by the button press. This is in contrast to our study where the participants press a button in response to a stimulus appearance. Other more indirect signs of the influence of the motor domain come from studies that show that head position can influence timing judgments [19] or that patients with neuromuscular disorders such as dystonia [20] or Parkinson’s disease [21] also display inaccurate temporal judgments.

In the current investigation, we reasoned that if interval timing judgments can be influenced by visual, auditory or tactile sensory input, the same might be true of motor activity. Strong a priori support for this hypothesis comes from the fact that several of the neural structures important in interval timing are also very important in motor activity; especially, the basal ganglia [22][23][24][25] and the supplementary motor cortex [3][4] [26][27][28]. The former structure has been found to be more important in explicit timing tasks while the cortical premotor areas are involved in both explicit and implicit perceptual timing tasks or ‘temporal expectation’ [29]. Given the shared neural substrates for motor activity and interval timing judgments, it is reasonable to ask if motor activity could influence our interval timing judgments. Indeed, some researchers have even suggested that infants develop their sense of timing through motor activity [30][31][32]. Furthermore, if the hypothesis on adult temporal judgments is true, would high-amplitude motor activity have a greater influence on interval timing than low-amplitude motor activity? Stronger muscular contractions require higher motoneuron activity [33]. Penfield & Rasmussen (1952) [34] showed that there is a somatotopic organization in the precentral gyrus of the human brain. Movements involving more than one limb would therefore involve a greater area of the motor cortical map than those involving just one limb. In addition, certain cognitive activities recruit motor areas of the brain without eliciting visible motor activity. Motor imagery is one such activity [35][36]. Thus, if real actions impact on temporal judgments, then this may also be true of imagined actions.

To explore these issues, we carried out an interval timing task in which participants were asked to respond to the presentation of regularly appearing visual stimuli. Reductions in the reaction time to the visual presentations were taken as an indicator of an improved ability to predict the appearance of the next stimulus. Several studies have now shown that response times are reduced when subjects expect the appearance of a stimulus at a particular time (temporal expectancy) [24][29][37][38][39]. To study the effect of motor activity on temporal expectancy, we compared the interval timing performance of groups that trained with or without motor activity. The different training groups were as follows: (1) a group training with a simple motor task (SMT), (2) a group training with a complex motor task (CMT), (3) a group training with motor imagery (MI), (4) a group training with visual presentation only (VI), (5) a group training with a complex motor action but irregular stimulus timings (IS) and finally (6) a group with no training (CRTL). As the areas of the brain involved in motor activity are also important for temporal expectancy, we expected that the groups training on interval timing using motor responses (SMT, CMT, MI) would have a different sense of timing compared to those without any motor activity (VI, CRTL). We expected in addition, that the amplitude of motor activity would have an effect on this modification. In other words, we expected to see significant differences in the gains obtained from the SMT, CMT and MI training. If the changes observed in the groups that had trained with motor activity were actually due to an improved capacity to perform a perceptual task and not just due to an increased ease with performing a motor act, we would also observe differences in the IS group compared to the SMT, CMT or MI group.

Methods

Procedure

The full design and procedure are illustrated in Fig. 1. The experiment consisted of a pre-training session (i.e., a "familiarization phase") lasting 10 minutes, during which button-press responses to the regular appearance of the visual target were recorded. This was followed by a training session (approximately 25 minutes) during which the participants were once again presented with regular appearances of the stimulus. However, during training, responses could involve more complex motor tasks such as reaching or leaning over. We describe the different training conditions in more detail below. Finally in the post-training session (10 minutes), exactly the same button-press that had been used in the pre-training task was repeated. This enabled a comparison of the response times before and after the training sessions.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Experimental setup and design.

https://doi.org/10.1371/journal.pone.0119187.g001

Results

The effect of motor training on interval timing was assessed by comparing the post-training phase response times with those obtained in the pre-training phase (Fig. 2). The response times were analyzed with a mixed ANOVA. Session (pre-training, post-training) and Interval Duration (3, 5, 7 s.) were within-participant measures while Training Group (CMT, SMT, MI, IS, VI, CTRL) was a between-participant factor. Tukey HSD post hoc tests were carried out where appropriate.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Pre- and post-training response times (mean ± std) for all the participants at every stimulus frequency.

The experimental groups tested were from top to bottom, the CMT, SMT, MI, VI, IS, and CTRL.*p<0.05, **p<0.01, ***p<0.001.

https://doi.org/10.1371/journal.pone.0119187.g002

The ANOVA revealed a main effect of Training Group (F(5, 594) = 8.3, p<0.0001, η² = 0.36) and Session (F(1, 594) = 138.6, p<0.0001, η² = 0.42), as well as a significant interaction of Training Group x Session (F(5, 594) = 45.8, p<0.0001, η² = 0.1). There was no main effect of Interval Duration (F(2, 1188) = 0.59, p = 0.55, η² = 0.01). No significant differences in response times were observed between the different groups in the pre-training phase (p>0.05, Tukey HSD posthoc). The Training Group x Session interaction could be explained by the fact that pre-training and post-training phase response times were only significantly different when motor training or motor imagery was involved (Fig. 2), i.e., for the CMT, the SMT and MI groups (p<0.05, Tukey HSD posthoc). Response times (RT) following training were reduced in these three groups, reflecting an improved temporal expectation for the appearance of the upcoming stimulus. The mean gains were 40 (SD 19) ms, 33 (SD 20) ms and 11 (SD 9) ms for the CMT, SMT and MI groups respectively (Fig. 3). Reductions in response times for the remaining groups were not significant (The mean values had in fact increased by 1ms for the VI and IS groups and decreased by 1ms for the CTRL group).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Percentage decrease in the response time (mean ± std) in the post-training test phase when compared to that in the pre-training phase.

Gain ratios are reported for all the experimental groups (x-axis) and every test frequency (3s with rounds, 5s with squares and 7s with diamonds). *p<0.05, ***p<0.001

https://doi.org/10.1371/journal.pone.0119187.g003

There were no significant differences between the gains in the CMT and SMT groups (p>0.05, Tukey HSD). However, post-training phase response times for both the SMT and CMT groups were significantly shorter than those in the MI group (p<0.001, Tukey HSD posthoc). Although, the RT improvement was not as great as for the motor training groups, the MI group was also found to have significantly shorter test phase response times than all the non-motor-training groups (p<0.05, Tukey HSD posthoc) (Fig. 3).

Finally, in the visual training (VI), the irregular target (IS), and in the control group (CTRL) there were no significant RT improvements between the pre and post-training phases. The lack of significant RT improvement (p>0.05, Tukey HSD posthoc) observed in the irregular stimulus (IS) group indicates that the gains observed in the CMT, SMT and MI groups were not simply due to a general sensorimotor facilitation associated with making (or imagining) a movement.

All participants had 100% accuracy in the secondary task in all conditions. This indicates that there were no major differences in attention levels between the training groups. No responses prior to the appearance of the stimulus were observed in any of the conditions. This suggests that the RT improvements observed in the test phase were not simply due to a systematic (and intentional) underestimation of when the stimulus was going to appear.

Discussion

These experiments indicate that repetitive, time locked motor activity is able to improve temporal expectancy. Groups that had incorporated motor activity into their training protocol with regularly spaced stimuli showed shorter post-training phase response times than the group with only visual training. The shorter response times suggest better preparation and more accurate anticipation of the appearance of the dot on the screen (i.e., improved implicit interval timing). Since we did not find any significant differences between the pre-training and post-training phase response times for those who received no training at all (CTRL), the performance improvement in the motor and motor-imagery training groups can be attributed specifically to the training phase. This raises the questions of how and why?

Could the improved performance of the motor training and motor-imagery groups be due to differences in attention or vigilance in these groups? To rule this out, we added a secondary task to the basic interval timing task. This task involved asking participants to count the number of green colored targets that appeared during the training sequences. Performance on this secondary task was constant and high across all training conditions, suggesting that there were no significant differences in the level of alertness across the different groups.

Another possible explanation for the improvement in reaction times observed in the motor training groups is some kind of general “sensorimotor facilitation” for performing a motor act. We use the word ‘general’ here because the motor tasks used in the training were quite different from those used in the test phase. Specifically, testing involved pushing a button on a joy stick with the thumb, whereas motor training involved whole arm or body pointing towards a target. However, this possibility was also ruled out by a lack of significant improvements in response times in the IS group. This group performed the same movement as those in the CMT group, but with an irregular unpredictable stimulus. Moreover, the total number of movements in this group was the same as in the motor-training groups with a regular stimulus. If the improvements obtained with the other motor training groups had been just due to an improvement of the speed of the motor act itself, without any perceptual anticipation, we should also have seen an improvement in the response times of the IS group, which was not the case.

One interesting observation is that the response-time improvements obtained in the simple motor activity (SMT) group were not significantly different from those in the complex motor training (CMT) group. This may lead one to make the conclusion that movement amplitude does not play a role in temporal expectancy. This is probably not accurate however, since as we explain in the paragraph below, the reduction in temporal expectancy with the MI group was significantly less than for the SMT or CMT groups (Fig. 3). Motor imagery involves the activation of motor area, but at a lower amplitude than what is observed during real movement [42]. Since temporal expectancy is not significantly different between the SMT and CMT groups, but is significantly lower in the MI group (Fig. 3), this indicates a nonlinear dependence on the amplitude of activation in the motor areas. A previous study by Gavazzi et al [43] has also shown how temporal judgments can change based on the specific motor act that is used to reproduce the stimulus duration.

The small, but significant performance improvement that was observed in the motor-imagery group suggests that simply activating the mental and neural structures necessary to support movement, but without an actual movement is sufficient to produce improved accuracy in interval timing. These results also help to rule out the possibility that simple repetitive haptic stimulation is responsible for the observed improvements.

One might be a little surprised by the lack of any significant reduction in response times for the group that only underwent visual training (VI group). Meegan et al [44] reported that participants who had taken part in perceptual training also showed improvements in a motor task involving interval timing. The reason for the difference in our results may lie primarily in the type of stimulus that was used. Meegan et al [44] had used an auditory stimulus while the sensory stimulus in our study was visual. A study by Cicchini et al [45] would also seem to corroborate this explanation. These authors found that more accurate interval judgments were made when auditory tones rather than visual stimuli were used.

Any mechanisms proposed to explain how motor activity could improve temporal expectancy would have to be of a non-specific nature. This is because both the temporal spacing of the stimuli and the responses to them were different in the training and the test phases. Specifically, the intervals to be timed in the test phase (3, 5, and 7s) were not the same as those used in the training phase (4, 6, and 9s). And, more importantly, the motor activity used in the testing phase was very different from those used in the training phase.

One possible explanation for our findings may lie in what Eagleman [5] proposed when attempting to explain why the first interval in a sequence appears to last longer than subsequent ones. This shortening of perceived durations of successive stimuli in comparison to the first, have now been observed by several researchers [5][11][12][46][47]. Eagleman [5] suggests that this phenomenon is due to the well-known observation of decreased neural firing in the brain following repeated stimulation due to neural adaptation. Motor training may increase the general level of activity in areas that are also implicated in interval timing. This may then reduce the effects of repeated stimulations. Such an explanation would hold both for models that ascribe timekeeping to a dedicated neural pacemaker structure [48][49][50][51] as well as for models that attribute timing to a more distributed spatiotemporal neuronal map [1][2][32][52]. Of course, if this were true, then one might ask why no response-time improvements were observed in the irregular stimulus (IS) condition. One answer is that the habituation produced in the IS condition with an irregular stimulus may have been significantly less than in the conditions with a temporally predictable stimulus.

Another possible explanation for the facilitatory effect of motor activity on interval timing is that the repeated pairing of the two leads to heterosynaptic Hebbian association [53] and, as a result, some motor neurons are recruited for interval timing. This would then allow for the combination of cues from two different modalities, the visual and the motor, hence improving the final interval estimation that has to be made. It has been suggested that the brain uses a Bayesian framework to integrate multiple sensory cues [54][55][56][57]. In this study we suggest that not only sensory input, but also input from the motor system may contribute to improving perceptual and cognitive estimates. It should also be noted that motor activity not only provides the signals for the execution of a movement but also generates ‘sensory’ information on the movement in the form of the efference copy [58][59][60]. As Hebbian association are based on the temporal proximity of two signals [61][62][63], the spatial proximity of the motor and interval timing systems would facilitate an association of the two.

The current study opens up several questions. If the improved interval estimation was obtained through the combination of stimuli from different modalities, it is possible that associating an auditory stimulus with the presentation of the white dots in our experiment may also have reduced reaction times. Indeed several studies have now shown that temporal estimation relies more heavily on auditory than visual information [64][65][66]. Future comparative studies will have to be carried out in order to determine if associating an auditory tone with the presentation of the white dots would improve interval timing estimates more than the association with motor activity.

Finally, it should be noted that the motor task used in the training did not achieve its effects through corrective feedback. An example of this would be catching a ball. In this case, a failed catch would indicate among other things, a failed interval time judgment. The error feedback can then be used to re-calibrate the timing system. Several researchers have shown that the re-calibration of interval timing judgments based on prior errors are in agreement with the predictions of Bayesian integration [67][68]. While motor activity clearly has this potential, it did not play such a role in our study. All motor activity during training took place after stimulus presentation, and, therefore, could not have provided error feedback to adjust timing judgments. This particular aspect sets our study apart from the Haggard investigation [17].

Future studies will be needed to verify if the effects that we have observed with motor activity also hold for other protocols of interval timing judgments. For example, would temporal expectancy also be improved if we used filled intervals for the stimulus? Other than the implicit tasks that were used in this investigation, future studies should also involve more explicit timing tasks such as the reproductions of temporal intervals, comparisons of time intervals or the detection of rhythmic or irregular sequences. Finally the timing performance improvements observed when using motor activity must be compared with those that might be obtained by associating other sensory stimuli—for example, auditory stimulation—with interval timing judgments. This would help us to determine whether the activation of the motor areas provides specific, additional benefits compared to the activation of other sensory areas during interval timing.

In conclusion, our study demonstrates that repetitive, time locked motor activity is able to improve temporal expectancy. This is the case even when the motor activity is non-causal with respect to the stimulus appearance and does not provide error feedback that could be used to improve the interval timing performance.

Supporting Information