The perception of time
DOI:10.1093/acprof:oso/9780199228768.003.0021
Abstract and Keywords
This chapter discusses changes and developments in the study of time perception in psychology. It discusses some of the highlights from studies of time perception, concentrating principally on developments since the 1960s. It provides some background to contemporary timing research by discussing a small amount of historical work including the internal clock models in the 1960s and 1970s, the resurgence of research on human time perception in the 1990s, and work on the neural basis of timing during the 1980s.
In this chapter, I will discuss what seem to me some highlights from the study of time perception, concentrating principally on developments since the 1960s, and mostly on ones that are much more recent. The reader unfamiliar with the field is warned that the material discussed here represents only a small sample of the potential research available and reflects my personal choices, and sometimes personal involvements. For example, an important distinction in contemporary timing research is that between prospective and retrospective timing. The former involves situations where people are alerted in advance that timing is important (e.g. ‘Hold down this button for one second’). The latter involves situations where unexpected questions about time are asked (e.g. ‘How much time has passed since you started reading this chapter?’). Most modern opinion regards these two types of timing as being distinct, with prospective timing involving some sort of internal timer, and retrospective time judgements being based on the quantity of something non-temporal (e.g. the amount of information processing, or ‘contextual change’) that has occurred during the time period judged. The vast majority of both classical and recent research has concentrated on prospective timing, and this area will be my sole concern here. For a discussion of some issues in retrospective timing, see Wearden (2005).
In successive sections I first provide some background to contemporary timing research by discussing a small amount of historical work, then proceed, successively, to sections on the rise of internal clock models in the 1960s and 1970s, the resurgence of research on human time perception in the 1990s, and work on the neural basis of timing, dating from the 1980s onwards, before a brief conclusion.
In the beginning: 1864–1960
The study of time perception, at least in its commonest form, that of the judgement of the duration of stimuli or events, has a long history. Lejeune and Wearden (submitted for publication) identified Höring (1864) as the first (p.266) publication on the subject, and this was closely followed by the extraordinary book Der Zeitsinn nach Versuchen (‘The Experimental Study of the Time-sense’) by Höring's teacher, Vierordt, in 1868. In the later nineteenth century, interest in time perception was keen, with many studies focusing on the potential existence of an ‘indifference point’, a duration that was neither over- nor under-estimated, considered by some to be a basic psychological unit of time. Nichols (1891) and Woodrow (1930) reviewed much of this early research.
In spite of its venerability, the study of time perception since the nineteenth century has often teetered on the brink of virtual extinction, sometimes occupying only a tiny number of active workers. Before the 1960s, where the story of this chapter properly begins, notable highlights were the work of François and Hoagland on body temperature and time judgements (for a modern review see Wearden and Penton-Voak 1995), and the studies by Goldstone and colleagues on judgements of the duration of auditory and visual stimuli (see Wearden et al.1998). A figure who deserves particular mention is Paul Fraisse, not only a specialist in time perception (see his magisterial book from 1964), but also probably France's best-known experimental psychologist in any domain for many years. His prestige helped to keep the fragile flame of time perception alive in French-speaking countries during periods when Anglophone psychologists had little or no interest in the subject.
An exception to the general neglect of time perception in the USA and UK was work on timing in animals, deriving mainly from research by Skinner (1938), which by the end of the 1950s had resulted in an enormous body of experimental results (for just some of them see Ferster and Skinner 1957). However, for reasons connected with the ideological basis of Skinnerian psychology, this work was divorced from the study of time perception itself (Lejeune et al. (2006) discuss the history of Skinnerian research in the timing field), and was not connected with it until much later, as will be seen below.
Internal clocks: 1960–1990
The idea that people and animals perform some sorts of time judgement by using an internal clock-like device, a sort of ‘organ’ for the perception of duration, can be traced back at least to the 1920s (Wearden and Penton-Voak 1995), but the years between 1960 and 1990 saw its principal flowering. In 1963, Michel Treisman produced what everyone would agree is a landmark in the development of internal clock theory. Treisman (p. 18) modestly writes that his model ‘derives from suggestions which have been made before…and attempts to put them together in not too arbitrary a fashion…”, but in fact his model was so sophisticated and advanced for its time that it was only equalled, (p.267) but perhaps not surpassed, by a rather similar proposal of Gibbon, Church, and Meck (1984), which still dominates many studies of time perception.
In Treisman's model, the raw material for time judgements comes from an arousal-sensitive pacemaker, which sends pulses to a counter. The pulses are assumed to be periodic. As well as the pacemaker and counter, the model also involves a store of ‘reference’ durations (e.g. temporal standards or other values needed for the timing task in hand), and a comparator mechanism. Comparison of values in the counter and the store determines behavioural output, or time judgements.
Treisman's model contains the basic mechanism of a pacemaker–accumulator clock, which was later used by Gibbon et al. (1984), but also shows that in order to generate any kind of timing judgements more than the basic clock mechanism is needed, with both some kind of store of reference times and, most importantly, some comparison mechanism also intervening to produce behaviour. Treisman's work thus situates a simple clock mechanism within the framework of a more complex cognitive system involving both memory (store) and decision (comparator) mechanisms.
An enduring mystery is why publication of Treisman's article failed to produce an immediate upsurge of interest in time perception. The work was published as a monograph supplement to Journal of Experimental Psychology, a premier outlet for experimental research, so was presumably highly visible, and, although the work had some mathematical aspects, it was not particularly obscure or difficult to understand, particularly by the standards of the psychophysics of the period. However, the neglect of Treisman's achievement began almost from the date of its first appearance. It was not referred to in Cohen's (1964) popularization of research on Time Psychology published in Scientific American just after the appearance of his article, nor was it referenced for further reading. On the other hand, Treisman's own account of his model in a ‘popular science’ publication (Treisman 1965) cited Cohen (1964), but this article did not appear to lead to widespread interest in Treisman's own work.
A model very similar to that of Treisman was developed by Gibbon and colleagues in 1984—their scalar expectancy theory (SET). This model also had a pacemaker–accumulator clock as its basis, as well as reference and working memory stores and decision processes. The main difference between the models lay in their application: Treisman was concerned to account for data from studies of time perception in humans (of which there was little that he could use), whereas Gibbon et al. were interested in explaining data from studies of animal timing, vast amounts of which had accumulated since the 1930s—see Lejeune et al. (2006) for a review of some early work by Skinner and others; (p.268) this article also discusses some competitors to SET in the domain of animal timing.
Animal psychology furnished innumerable experimental studies, usually with very orderly data, that could be used to evaluate Gibbon et al.'s model. In addition, a characteristic of SET-based research in the 1980s was a marked physiological flavour. A particular highlight is psychopharmacological research by Meck (1983, 1996), which used drug effects to dissociate the different parts of the SET system. For example, data suggested that the rate of the pacemaker of the clock was affected by dopamine levels (‘ticking’ faster when dopamine levels were higher), whereas the reference memory store (where animals stored ‘important’ times, such as those associated with reinforcement) was manipulable by changing acetylcholine levels. I shall return later to attempts to understand the neural basis of timing. Another feature of SET, and one that it shared with Treisman's original work, was a concern with exact quantitative modelling of data, using mathematical analyses or computer simulation.
Much research with animals established that their timing behaviour frequently conformed to the ‘scalar properties’ of time (for a review see Lejeune and Wearden 2006). The first scalar property is mean accuracy, the requirement that the internal ‘estimate’ of some real time, t, is on average exactly equal to t. The second is the scalar property of variance, the requirement that the standard deviation of time ‘estimates’ varies linearly with their mean, a form of Weber's Law. Lejeune and Wearden (2006) illustrate how these properties are measured in data from animals, and they also identify some situations in which the scalar properties are violated. Given the success of SET as an account of animal timing, it seemed only natural to attempt to apply this model to human time perception. Although the possibility of scalar (i.e. Weberian) timing in humans had been discussed much earlier (even in the nineteenth century), the application of ideas related to SET started only towards the end of the 1980s.
Human time perception: 1990–present
In 1988, Wearden and McShane published an article reporting data from humans who produced time intervals ranging from 0.5 to 1.3 s without counting. The behaviour obtained was in almost perfect accord with the two scalar properties of time: the mean time produced tracked the time requirement almost perfectly, and the standard deviation was an almost constant fraction (about 0.13) of the mean. This was followed in 1991 and 1992 by four articles (Allan and Gibbon 1991; Wearden 1991a,b, 1992) more explicitly employing the (p.269)theoretical apparatus of SET (i.e. the clock, memory, and decision processes) to account for human performance on tasks of bisection and temporal generalization, both methods that are analogues of procedures previously used with animals (e.g. Church and DeLuty 1977; Church and Gibbon 1982). An advantage of these analogue methods is that they necessarily ask participants what seem to be very simple questions about time judgements. In bisection, people receive examples of ‘short’ and ‘long’ standards (e.g. tones 200 and 800 ms long, respectively) and then have to decide to which standard subsequently presented comparison durations are more similar. In temporal generalization, a single standard (e.g. a tone of 400 ms) is presented, and the participant must then judge whether subsequently presented comparisons are, or are not, the standard. These procedures can be used with slight variations to study timing in children, and data from both provide measures that can be used to evaluate conformity to the scalar properties of time (Wearden and Lejeune 2008).
Allan (1998) and Wearden (2003) provide general reviews of the application of SET to human time perception. Although constraints of space preclude me from entering into the details here, I will select what seem to me a few highlights of this work.
Firstly, the application of SET breathed new life into the old idea that humans possess a pacemaker–accumulator internal clock, and some studies have demonstrated that the putative pacemaker can be made to run faster (Penton-Voak et al. 1996; Treisman et al. 1990) or more slowly (Wearden 2008). Differential pacemaker speed has also been used as a potential explanation of the venerable auditory/visual differences in duration judgements (‘tones are judged longer than lights’—Goldstone and Lhamon 1974; see Wearden et al.1998), as well as the ‘filled duration illusion’, the finding that empty intervals (e.g. starting and ending with clicks) are judged as shorter than ‘filled’ ones (e.g. continuous tones) (see Wearden et al. 2007).
As well as the study of internal clock processes, the success of SET has stimulated research on the roles of memory and decision processes in timing (e.g. Jones and Wearden 2003, 2004; Wearden and Grindrod 2003) and, in general, has encouraged the view that a full explanation of timing behaviour requires the internal clock mechanism to be situated in a more elaborate cognitive structure, involving attention, memory, and decisions (e.g. Wearden 2004); this is basically an elaboration of the position taken by Treisman in 1963.
The quantitative modelling often associated with SET's explanation of animal performance also translated into studies with humans, and has led to the development of models of performance on many different timing tasks (p.270) (e.g. Wearden 1992, 1995; for a review see Wearden 2004). These models enable us to decompose performance into its underlying psychological components, so permitting differences between groups and conditions to be attributed with some confidence to differences in underlying mechanisms, and this has been exploited in another important development, the study of timing in different participant populations.
Perhaps the clearest progress in this area has come from studies comparing children of different ages, and adults, also sometimes of different ages. Droit-Volet (2003a) reviewed much of this work and illustrated how the use of SET enables researchers to draw stronger conclusions than just that younger children are poorer at timing than older children and adults. For example, several studies have shown that, rather than children having fundamentally different decision processes from adults, the variability of time representations decreases with age. In addition, some work suggests that the youngest children may systematically misremember standard durations as being shorter than they are (see Droit-Volet et al. 2001; McCormack et al. 1999), but that this tendency disappears from about 8 years of age. In addition, other work has implicated difficulties in paying attention to the timing task in younger children as being an important determinant of their performance (e.g. Droit-Volet 2003b; Droit-Volet and Wearden 2001).
At the other end of the developmental scale, some studies have used ideas from SET, or similar ones, to account for data from older adults (e.g. McCormack et al. 1999; Wearden et al. 1997). In Wearden et al.'s (1997) study, for example, modelling based on the principles of SET revealed that, at least in some cases, the variability of time representations increased systematically with increasing age and decreasing IQ. In general, however, effects of ageing on prospective timing are rather small, with older people generally exhibiting the same pattern of behaviour as younger ones, albeit with more performance variability. The common complaint of older persons that ‘time seems to pass more quickly’ with increasing age is difficult to explain simply in terms of the processes used to account for prospective timing, and the understanding of this ‘distortion’ of subjective time seems to need some alternative explanation, as well as novel sorts of research (Wearden 2005).
Time and the brain: 1990–present
Concern for the neural basis of time perception is not new, and speculation about brain mechanisms that subserve timing processes can be found as early as Hoagland (1933). Research in the early 1980s with animals, conducted mostly by Meck and colleagues, gave renewed impetus to the (p.271) search for the physiological basis of timing mechanisms (e.g. Meck 1983). Compared with researchers seeking the physiological basis of auditory and visual perception, those interested in time perception are disadvantaged by the absence of an ‘organ’ that provides the obvious starting point of the perceptual process both psychologically and, perhaps more importantly, neurophysiologically, thus allowing connections to be traced from the external organ to the brain, and permitting good first guesses as to the brain areas involved in the process. One reaction to this difficulty was to take the clock-memory-decision structure of SET literally and to try to map this on to brain structures.
Meck (1996) discusses an attempt to do this, with the substantia nigra pars compacta providing dopaminergically sensitive pacemaker neurones, whose output was collected by regions of the basal ganglia, with other areas of the brain providing memory and decision mechanisms. Although this attempt to ‘physiologize’ SET accounted for much data, more recent research has taken different directions, largely because of the alleged physiological implausibility of the pacemaker-accumulator mechanism that is the basis of SET (e.g. Matell and Meck 2004).
Interest in the neurological basis of time perception has produced a considerable increase of activity in the time perception field but, unfortunately, a simple summary of definite progress remains difficult to provide at the time of writing. However, a single article that summarizes some of the main ideas is that of Buhusi and Meck (2005). Some studies have used scanning techniques to identify the brain areas that appear to be involved in timing, but an obvious problem, alas without any obvious solution, is what control conditions should be used. For example, if certain brain areas are activated during a timing task, do these relate to timing per se, or to memory and decision processes that might be common to judgements of other types of stimuli? Various control procedures, of different degrees of subtlety and ingenuity, have been proposed (e.g. Macar et al. 2004; Nenadic et al. 2003), and it is true that certain brain regions (e.g. parts of the basal ganglia, the supplementary motor area of the cortex) appear as ‘usual suspects’ in many, but not all, timing tasks. However, the function of these activated areas is unclear, and some work using scanning or electrophysiological measures appears little more than a reaffirmation of materialism, the idea that brain processes underlie psychological processes, in some way that is currently difficult to specify.
Other studies have looked at timing in patients with various sorts of brain damage or brain degeneration, such as people with Parkinson's disease. Here, performance differences between patients and controls are sometimes (Harrington et al. 1998), although not always (Spencer and Ivry 2005), found, but (p.272) even when between-group differences appear marked, as in Malapani et al. (1998), it is not always clear how they can be explained. Obviously a persistent problem when comparing patient groups with controls is that any difference in performance on timing tasks may be due to differences in factors other than those relating to time perception per se, such as memory and attentional changes, or differences in motor performance.
An important theoretical shift linked to neuroscience-based considerations has been the general abandonment of the pacemaker–accumulator clock as a basic mechanism of timing, in spite of the fact that psychological models embodying such a process make many novel and accurate predictions about behaviour (e.g. see Wearden and Jones 2007), and its replacement either by nothing definite (i.e. research without any quantitative theoretical modelling, or any theoretical basis at all) or by some sort of oscillator-based process. Once again, constraints of space preclude discussion of the details of this sort of model, but the general idea is that the representation of time intervals is accomplished not by the accumulation of ‘ticks’ from a pacemaker, but in terms of vectors of oscillator states (Church and Broadbent 1990), or ‘coincidences’ of oscillator firing patterns (e.g. Matell and Meck 2004). Such models enable the timing of durations that are very long compared with the rate of firing of neurones (with periods of milliseconds), and avoid the problem of ‘unbounded accumulation’, that is, the problem of how the timing of long intervals (in animal experiments sometimes many minutes long) can be accomplished by neurone-like units that fire thousands of times in the interval to be judged.
Oscillator-based models essentially represent different durations by different patterns of neural activity, not by more neural activity for longer durations (as in an accumulator process), and this qualitative representation can cause problems (e.g. see Wearden and Doherty, 1995, for a discussion of the Church-Broadbent model). For example, two time intervals that are very different in real time may have very similar pattern representations, and the model may have no way, without ‘add-ons’ that specifically perform the function, of producing ordinal judgements (i.e. deciding whether one duration is longer or shorter than another as opposed to just different from it), although such judgements appear very easy for people to make, perhaps even easier than judgements of the equality of two event durations.
In general, then, research on the neural basis of timing has not yet uncovered the mechanism by which duration judgements are generated, in spite of considerable, and expensive, experimental effort and theoretical speculation.
(p.273) Concluding remarks
This chapter is intended to give the reader a brief overview of what seem to its author the most interesting developments in the study of time perception since the 1960s. As mentioned above, the choice of topics is a personal one, but I hope that researchers in the field would agree that at least some of the most important trends have been discussed here. Work on the experimental study of time perception in humans has made significant progress since the 1980s, with the possibility of the development of comprehensive quantitative models, based on SET and developments thereof, which may account not only for performance on standard timing tasks, but also for differences between different participant populations, as well as providing a bridge between modern theory and classical findings known since the nineteenth century (Wearden and Lejeune 2008). What the neural mechanisms of time perception are remains, at the time of writing, a very open question, with data suggesting the involvement of many areas of the brain, whose exact function in the timing process is currently unknown. However, in spite of many outstanding problems, the last 20 or 30 years have probably seen more progress in our understanding of time perception than in the previous 100, and the properties of what the early German experimenters called Der Zeitsinn, have been explored more extensively than the pioneers of the subject, such as Vierordt, could probably ever have imagined.
No comments:
Post a Comment