Studies that have compared SMT across age groups generally agree with each other in that they find a significantly slower tempo in older compared to younger people (Vanneste et al., 2001; Baudouin et al., 2004; McAuley et al., 2006; Turgeon and Wing, 2012; see Table 1). It is merely the magnitude of slowing with age, as well as the exact developmental course, that were somewhat different across studies. For example, while some studies found quite substantial differences in the ITIs of participants of different ages (1,072 ms for old participants; 654 ms for young participants; Baudouin et al., 2004), this difference is smaller in other studies (747 ms in old participants; 536 ms in young participants; Vanneste et al., 2001) and even subtler in others (McAuley et al., 2006, found the SMT of people aged 75+ to be 648 ms, 632 ms for participants between 60 and 74, and 522 ms for people aged 39–59).

Regarding the developmental course, some authors demonstrated that it is especially in very old age that a slowing of SMT is visible. McAuley et al. (2006) found a cubic relationship between age and SMT, suggesting that this variable slows particularly late in life (i.e., after the age of 75). Similarly, in study of Turgeon and Wing (2012), a slowing of spontaneous motor rate was visible particularly in participants aged 75 and above.

Consistency of spontaneous tapping, which is usually expressed in terms of CV, does not appear to be as affected by age, as we predicted. McAuley et al. (2006) and Vanneste et al. (2001) found that old and very old people at their preferred tempo tapped as consistently as young ones. Along similar lines, using linear regression, Turgeon and Wing (2012) did not find age to predict a significant proportion of variance of CV. Thus, at a tempo that participants choose themselves, differences in terms of consistency have not been observed between young and old people.

As SMT reflects one’s natural rate of rhythmic motor activity, it is often thought of as being related to preferred perceptual tempo. McAuley et al. (2006) investigated the relationship between SMT and preferred perceptual tempo, in a study in which they presented rhythmic sequences of different speeds to their participants and asked them to judge whether each sequence was too fast, too slow, just right (relative to their favorite speed), or anything in between. They found that preferred perceptual tempo slowed with age along with SMT and that these two variables were highly correlated. One may then conclude that SMT, which appears to significantly slow down with age, might reflect the slowing of one’s internal clock in old age.

Other types of unpaced tapping are those that require participants to tap in a repeating, continuous fashion as fast or as slow as possible, to see at which upper and lower limit participants are able to maintain a regular tap. As demonstrated by a large number of studies, when given the instruction to tap regularly as fast as possible, older people tapped at a slower rate than younger people (Nagasaki et al., 1988; Cousins et al., 1998; McAuley et al., 2006; Turgeon et al., 2011). This fastest tempo may slow down most evidently from the age of about 70 years onwards (McAuley et al., 2006). Conversely, McAuley et al. (2006) also made their participants tap as slowly as possible at a constant rate, in which case older people sped up more than younger people. It appears, then, that aged people have a narrower range of tempi at which they can consistently produce taps than young people.

The variability of taps at a fastest regular speed has not always been investigated. Where it has, CV was shown to be affected by aging, although this effect was weaker than the effect of aging on interval size. More specifically, Turgeon et al. (2011) found age to account for 7% of the variance in CV scores and for 32% of the variance in ITI, which might reflect decreased attention in old age.

In tasks requiring participants to synchronize their tapping with an external regular signal, differences between young and old have been less clear than in the unpaced tapping tasks described above (see Table 2). First of all, several studies did not find differences between these two groups in terms of variabilities or accuracies (Krampe et al., 2005; Drewing et al., 2006; Turgeon et al., 2011). Contrary to that, Nagasaki et al. (1988) did find CV to correlate with age at all the intervals examined (between 200 and 1,000 ms), but did not find a significant effect of age on asynchronies.

One study (Thompson et al., 2015) did report differences between old and young participants in both variabilities and accuracies. In this study, older adults (age range 51–80) tapped to a regular beat with a higher variability and with a larger asynchrony than younger and middle-aged adults (age range 18–43). These results may seem surprising, especially in light of the relatively young age of their old adults group (M_Age = 63.67; for comparison, Drewing et al., 2006 did not find such age-related effects in the group of people aged 78–88). However, the difference found by Thompson et al. (2015) may, at least in part, be explained by musical experience. Since the authors were interested in how musical experience influenced beat synchronization, they recruited many musically experienced participants (in the young and middle-aged adult groups, there were 32 people with and 11 people without musical background). However, musical background was not assessed in the group of older adults, and therefore old and young adults may have not been matched on this variable. In fact, the respective asynchronies found in this study (around 40 ms before the beat for the older group and around 15 ms before the beat for the younger groups) are remarkably close to values that have previously been found in research comparing musicians and non-musicians (Aschersleben, 2002). Assessing musical training in all groups, including older adults, would have made it possible to verify whether young and middle-aged adults’ superior performance was due to musical experience rather than age, and also to investigate whether musical training could be a neuroprotective factor in aged people.

Up until now, all studies we discussed here described paradigms in which people synchronized their movements to an isochronous sequence of repeating single beats. And in fact, while a large part of those studies discussed their results in terms of their relevance for music perception and cognition, one could argue that their stimuli were not strictly musical. People move spontaneously to a musical beat (Leman et al., 2017), even though this beat is a perceptual construct that does not have a clear physical correlate. Beats sometimes co-occur with musical notes, but not necessarily: A beat can occur on a silent event (McAuley, 2010). Extracting a beat from a complex auditory signal, such as music, while automatic, might be a quite different process than merely synchronizing to an explicit beat. And therefore, this process might be differently affected by aging and NCDs. Indeed, rhythm difficulties might be more easily identified with music than with simple, repeating tones (Sowiński and Dalla Bella, 2013; Falk et al., 2015). Another reason to look at music in addition to metronomic sequences is that generally, people tend to tap with a higher asynchrony (that is, their taps precede the corresponding stimulus onset) when synchronizing with a metronome compared to music (Thaut et al., 1997; Aschersleben, 2002), a result that has recently been confirmed in elderly people (Ghilain et al., 2020a,b), but still lacks a definitive explanation. It has been suggested that subdivision of intervals between beats leads to a reduction of negative asynchronies and their variability, and that synchronization may be facilitated by the recurrence of different pitches, event duration, or intensities (Repp, 2003). In this way, music could be seen as an extreme case of subdivision, and the observed pattern of reduced asynchronies in tapping to music might be no different from what has been observed in tapping to a subdivided metronomic sequence. In either case, however, no studies exist to the best of our knowledge that compare old and young participants on an SMS task requiring participants to tap along with an auditory complex, that is, musical, stimulus.

In synchronization–continuation tasks, participants first synchronize with a metronome and then continue tapping at the same rate when the sound has stopped. Generally, only the performance in the continuation phase is analyzed (Wing and Kristofferson, 1973). Regarding the question of how older adults perform in this paradigm, the results are somewhat mixed (see Table 2). Turgeon et al. (2011) used a synchronization–continuation paradigm in which the target tempi per individual were determined based on one’s SMT in the preceding spontaneous tapping task (see Section “Spontaneous Motor Tempo”). At the intervals IOI = 600 ms and IOI = 900 ms, age did not significantly predict variability or accuracy.

Other studies also did not find any differences in ITI or variability on synchronization–continuation tasks using tempi between 300 and 700 ms (Duchek et al., 1994; Vanneste et al., 2001). One study found intact performance in old subjects even at intervals as short as 150 and 225 ms (McAuley et al., 2006), but a higher asynchrony (speeding up) at the slowest target interval (1,709 ms) in their oldest group (75 years and above), but not in the group of 60–74-year-olds. Conversely, Carment et al. (2018) reported an increase in variability among older subjects at the IOI of 333 ms, but not at 500 and 1,000 ms. Another study found differences in both variability and accuracy among older participants at a target interval of 1,500 ms, but not at 500 ms or 1,000 ms (Bangert and Balota, 2012). These results lend support to the purported existence of two systems involved in temporal perception, an automatic and a cognitively controlled one, the latter of which is particularly implicated in the processing of slow intervals and presumably used less efficiently by older people.

Krampe et al. (2010) conducted a study with a dual-task design in which participants were required to tap to a faster (IOI = 550 ms) or slower (IOI = 2,100) tempo, while performing variants of the NBack WM task (Dobbs and Rule, 1989). In this task, participants were exposed to a sequence of visual stimuli and were asked to indicate when the current stimulus matched the one presented two steps earlier in the sequence. While this paradigm falls a bit outside the order of simple synchronization–continuation paradigms presented here, it is relevant for two reasons. First, the authors also report performance under single-task conditions, that is, while performing synchronization–continuation without a concurrent second task. In this case, there was no difference in variability or accuracy among age groups, even at the slow tempo of IOI = 2,100 ms, which is in contrast to the speeding up among oldest subjects in study of McAuley et al. (2006) study and the lower consistency and accuracy observed in oldest participants of Bangert and Balota (2012). This discrepancy might in part be explained by age differences: While older participants of Krampe et al. (2010) had a M_Age of 67, healthy old participants of Bangert and Balota (2012) were on average 75 years old and McAuley et al. (2006) included only people aged 75 or older in their oldest group. Therefore, it indeed appears to be difficult at least for very old people to synchronize successfully with a fading stimulus presented at a slow tempo (of at least an IOI = 1,200 or more).

The second reason for which study of Krampe et al. (2010) study is relevant to this review is that the dual-task nature of the paradigm might provide relevant information regarding the cognitive mechanisms involved in tapping at slow frequencies. Dual-tasking caused people regardless of age to speed up at the slow tempo, but additionally caused old participants to speed up at the fast tempo. Similarly, variability was significantly higher in older adults (at fast and slow tempi), but only in the dual-task condition. The authors’ interpretation is that maintaining temporal precision and stability, even at a tempo of IOI = 550 ms, might be a quite automatic process in younger people, but might cost older people more attention and WM resources, in line with the compensation hypotheses mentioned above. Without any cognitive load, they can deploy that attention and WM resources to perform the task as well as their young counterparts. However, with fewer of these resources available, their performance will drop. Indeed, as we discussed in Section “Brain and Cognitive Correlates of Healthy Aging and Neurocognitive Disorders,” it may be the cumulative effect of slow tempo, cognitive load, and age-related competition for prefrontal resources due to compensatory rewiring of the motor system that is expressed in the lower performance of older people.

Perhaps the clearest finding regarding aging and SMS is a lower SMT in older people, which has been demonstrated in people in their 60 s and above but may be most apparent from the age of around 75 years onward. Similarly, preferred perceptual tempo appears to slow with age and to correlate with SMT. In contrast, variability appears to be intact in old participants tapping at a self-chosen speed. Moreover, the range of rates at which aged participants can tap regularly is narrower than for young participants, with a slower fastest tempo and a faster slowest tempo.

Paced tapping tasks have not tended to reveal differences in variability or accuracy between old and young subjects. Exceptions are synchronization–continuation tasks where participants had to maintain a relatively slow rate beyond 1 s, in which older people tended to speed up and tap less consistently, especially very old people from around 75 years old. Higher variability has also been reported at fast intervals of 333 ms and below and some research points to an interaction between age and cognitive load even at intervals that are close to people’s natural pace.

Having discussed the relative slowing of SMT with age, we now discuss how this variable is affected by NCD. A few sources have suggested that unpaced tapping may become more variable and SMT may slow in NCDs, especially in advanced stages (see Table 3). Roalf et al. (2018) found people with AD and MCI to tap with a higher variability compared to age-matched healthy controls, and variability was negatively associated with their score on the Mini-Mental State Examination (MMSE; Folstein et al., 1983), a widely used test for screening cognitive function among the elderly. People with AD produced significantly fewer taps during 1 min of tapping at a comfortable rate compared to healthy participants, whereas the speed of participants with MCI did not significantly differ from either of the other groups. The difference in produced ITI between the AD group and healthy older adults was significant, albeit not very large (469 and 441 ms, respectively). Similarly, Rabinowitz and Lavner (2014) found that in a group composed of patients with MCI and patients with a diagnosis of dementia, variability was higher in that group than in healthy controls. Moreover, patients tapped at a slower speed (747 ms) than their healthy counterparts (581 ms). Additionally, MMSE score was found to be correlated with mean ITI, suggesting a slowing of SMT with disease progression. In contrast, some other studies compared people with and without NCD on a spontaneous motor tapping task and did not find any differences in ITI or variability (Martin et al., 2017; Ghilain et al., 2020a). In these studies, subjects tapped at a comfortable speed for 30 s or for 30 ITIs, respectively. The apparent disagreement in the literature may be explained by different instructions that were given. The durations in which people performed the SMT task were shorter in Roalf et al. (2018; six blocks of 10 s) and Rabinowitz and Lavner (2014; one block of 15 s). In just 10 or 15 s, even when told to tap at a comfortable speed, people might feel pressure to produce as many taps as possible. Indeed, the produced ITIs were relatively low, even compared to the ITIs found in the studies on SMT in healthy aging (see Table 1). Although speculative, it may be that people with NCD react to this pressure differently than healthy older adults. Additionally, cognitively impaired people might potentially have had more problems understanding the task instructions. If, due to lack of clarity of the task requirements, people with NCD tapped more hesitantly (slower and more variably) in the beginning of a task, this would be reflected in their overall scores more heavily if the whole trial was just 10 or 15 s long.

Therefore, the SMT tasks used in Ghilain et al. (2020a) and Martin et al. (2017), affording their participants more time to establish a regular tapping pattern reflective of their internal speed, might be a better representation of their real SMT. Indeed, the rates observed in their studies, which lay roughly between 700 and 950 ms, are close to SMT values that have been observed in physiological aging (see Table 1). It therefore does not appear that during spontaneous tapping, ITI and consistency deteriorate in NCD, at least not in the mild to moderate stages of NCD that participants in all the studies quoted above tended to be in. More research, particularly including people with more severe NCD, is needed to establish an effect or absence of effect of NCD on SMT.

As with healthy aging, some studies have looked into the variability and speed at which people with major and mild NCD tap when asked to tap regularly as fast as possible.

Under these conditions, Kluger et al. (1997) identified no difference in terms of produced tempo between patients with MCI and age-matched healthy participants, but found people with mild AD to tap at a lower speed. In contrast, Goldman et al. (1999) did not find an effect of AD on produced fastest tempo. Variability was not taken into account in these studies. Colella et al. (2021) did find tapping variability to increase in people with MCI, but they did not find a difference in tempo between the groups. Taken together, these studies suggest that while a decrease in fastest tempo at which people can tap is only seen in advanced stages of NCD, regularity of fast tapping already appears to decrease in people with mild NCD. This is an interesting parallel to the results found by Roalf et al. (2018), who also observed decreased variability in people with MCI but decreased absolute tempo only with a diagnosis of AD in their SMT task.

To the best of our knowledge, there do not exist any studies comparing people with and without NCD on the ability to tap in a regular fashion as slowly as possible. Since paced tapping at a slow tempo might rely more heavily on WM and attention, we might infer that people with NCD, who tend to be impaired in these domains, might speed up and/or tap with a higher variability when instructed to keep a regular pace as slowly as possible, a prediction that remains to be tested.

The performance of tapping along with an auditory metronome has been investigated in people with NCDs in few studies. There might be several reasons for this, but we presume that many of the tasks described in the last few sections could not have been conducted in the same fashion with people with NCDs, especially in groups of patients with major NCD. In this group, particular attention must be paid to avoid stressful, unpleasant, artificial, and invasive laboratory situations. The listening and movement production tasks described in the previous sections might not be suitable for this patient group, and instead research with multimodal stimuli that creates a social or quasi-social situation might be conducive here (Desmet et al., 2017; Lesaffre et al., 2017). An example of this is the paradigm we described in Ghilain et al. (2020a,b), and Hobeika et al. (2021) in which participants were instructed to tap along with music or a regular metronome (IOI = 800 ms) while a musician, either seated across from the participant or projected onto a life-sized screen, vocalized and tapped along with the same stimulus. Under these conditions, benefitting from the presence of a musician, no differences in asynchrony or variability were found between people with and without NCD (see Table 4). It must be mentioned that the patients in this study were recruited from a day hospital rather than an inpatient care facility, so they might reflect a relatively independent and mildly impaired NCD group. Their average MMSE score of 20 was just on the fringe between mild and moderate cognitive impairment (Folstein et al., 1983), so we cannot exclude the possibility that a sample of more heavily cognitively impaired patients would show deficits in tapping performance compared to healthy older adults. The lack of effect of NCD might also be explained by other methodological variables, such as the use of only one tempo (close to elderly people’s SMT, see Table 1) or the impact of social entrainment related to the presence of a partner during the task. Since this paradigm is adapted to people with major NCD, it would be interesting to have a group of more cognitively impaired people perform this task.

Besides this, there do not exist many studies evaluating SMS with complex auditory (musical) stimuli, although, as we argued in Section “Synchronization Without Continuation,” it is not only interesting to directly compare SMS to metronomic sequences and SMS to music, due to the presumably different mechanisms involved in beat extraction, but providing musical stimuli can also be of particular value in creating an experimental situation that will make people with NCDs feel comfortable and that has some ecological validity.

One study that did uncover NCD-related differences in SMS was that by Henley et al. (2014). In this study, participants with different variants of frontotemporal dementia and AD synchronized with a metronome at an IOI of 1,500 ms. This study did not find differences in accuracy but higher variability in participants with behavioral variant frontotemporal dementia, but not in participants with other variants of frontotemporal dementia or AD, compared to healthy age-matched adults.

Bangert and Balota (2012), whose study was mentioned in Section “Synchronization–Continuation” about healthy aging, had their participants tap at a rate of 500, 1,000, and 1,500 ms, after the metronome stopped. Besides comparing the performance of young and healthy old participants, they also compared the latter group’s performance to that of people with very mild (mean MMSE = 27) and mild dementia (mean MMSE = 24). At the interval of 500 ms, they identified no group differences in terms of variability or accuracy (see Table 4). At 1,000 ms, people with dementia tapped with a greater variability but at the same accuracy as their healthy counterparts, whereas at 1,500 ms both consistency and accuracy were reduced in cognitively impaired people. These effects were stronger in people with mild compared to very mild dementia. Complementing these findings, Carment et al. (2018) found unimpaired performance with MCI and AD (mean MMSE = 22) at tapping rates of 500 ms and 1,000 ms, but increased variability at the most rapid tempo of 333 ms. Duchek et al. (1994) employed an IOI of 550 ms and found merely a non-significant trend toward higher variability. Nichelli et al. (1993) did not test their participants at shorter intervals, but found AD patients’ variability to be larger than healthy participants’ at an interval of 1,000 ms, corroborating findings of Bangert and Balota (2012). Along similar lines, Henley et al. (2014), in a synchronization–continuation task at 1,500 ms, found lower accuracy (speeding up) and greater variability in people with behavioral variant frontotemporal dementia, and greater variability in people with AD compared to healthy participants. Based on these studies, it appears that ITI and/or variability in people with NCD are only impacted at relatively long or relatively short intervals, presumably due to an impacted cognitively controlled timing system. However, it is important to point out that the participants in all these studies had on average very mild or mild NCDs.

In contrast, the patients in the study by Martin et al. (2017) tapped with a higher variability than healthy controls, even at a comfortable rate (in this study, the synchronization–continuation task was performed at a rate that depended on their SMT in the preceding spontaneous tapping task; see Section “Spontaneous Motor Tempo”). The apparent discrepancy between this study and the ones mentioned before might be explained by degree of impairment: The patients in study of Martin et al. (2017) had an average MMSE of 19, which indicates moderate cognitive impairment and which was lower than the MMSE scores reported in any of the studies mentioned above. Taken together, these results suggest that SMS performance in people with mild NCD might decline only at relatively long or short tempi, whereas in more severe cognitive impairment, people might tap less precisely even at tempi that are close to their natural speed. Nonetheless, more research about timing abilities in people with moderate or even severe cognitive impairment is needed.

Overall, there is not a lot of evidence suggesting that tapping at a comfortable rate reveals differences between people with and without NCD. On the other hand, tasks requiring regular tapping as fast as possible have demonstrated increased variability even for people with mild NCD and increased variability as well as slower tapping in people with major NCD. The fact that changes in variability during unpaced tapping are visible even in people with mild NCD are relevant in terms of using motor tasks as a diagnostic tool: While motor speed might only reliably identify people with major NCD, tapping variability could be used to identify people who are at a risk for developing a more serious impairment. Indeed, in MCI, motor impairments are predictive of developing AD (Aggarwal et al., 2006; Ghilain et al., 2020a).

Presently, there is little research studying the effect of cognitive impairment on performance in simple tapping (synchronization without continuation) tasks, but the extant literature suggests that people with frontotemporal NCD might tap less precisely than AD patients and healthy controls at a tempo of 1,500 ms. More research is needed with more different tempi and different levels of cognitive impairment. More commonly, synchronization–continuation studies have been used, and suggested that people with NCD might speed up and show increased variability at slow IOIs from around 1,000 ms, as well as show increased variability at short IOIs below 500 ms. There is some suggestive evidence that even at a comfortable rate, people with a moderate or severe NCD might tap less precisely. Etiology of the NCD may play a role, with a possible interaction between tempo and type of NCD, but more research is needed with various groups of patients with NCD due to different causes, and different levels of impairment.

In short, there is a need for more research investigating SMS abilities in people with moderate and severe cases of NCD. At the same time, however, designing tasks adapted to people with NCD is not trivial: a great deal of effort has to be done to ensure people understand the instructions and to create experimental paradigms that are not stressful, invasive, or unpleasant to the participants. With the existing research, differences between people with and without NCD must be evaluated critically: It is important to verify that they tap into the mechanisms a study claims to test, or whether observed differences may reflect differences in motivation, comfort, or having well understood the instructions. Finally, a fruitful avenue of research would be more studies using musical stimuli of different levels of rhythmic complexity, not only to test the ability to infer and maintain an implicit beat, but also to create a situation that is engaging and motivating, ideally involving a social or quasi-social element resembling real-life musical interaction (Desmet et al., 2017; Lesaffre et al., 2017; Ghilain et al., 2020a,b; Hobeika et al., 2021).