Abstract
Previous research has found inaccuracy of temporal perception in increased working-memory loads, but the direction of this inaccuracy has not been measured. A time reproduction task was used to investigate the direction of inaccuracy in a temporal N-back task with three levels of increasing difficulty. Findings suggest that increasing working memory loads cause under-estimation of temporal perception. Results are also consistent with previous research in finding a greater inaccuracy with increasing memory loads.
1. Introduction
The ability to accurately estimate time is essential in our daily lives and fundamental in holding an accurate temporal representation of the world. However, unlike our other senses, time has no designated sensor (Gibson, 1975), resulting in temporal perception being subjective and easily modulated by environmental stimuli, giving rise to inaccuracy. Moreover, the temporal perception of the same duration can vary depending on its non-temporal features such as size, brightness and, motion (Eagleman, 2008). Due to these many influencers on our perception of time, there has been a large surge in popularity of research in recent years, leading to vast numbers of review articles, special issues and bibliographies requiring the publication of reviews of reviews (Block & Grondin, 2014; Grondin, 2001). This exemplifies the relevance for ongoing research in this field.
1.1. The Scalar Timing Model of Time Perception
Scalar Timing Model (STM; see Figure 1.) is the most cited model used in understanding the cognitive process underlying temporal perception of durations in the millisecond-second range (Gibbon, Church, & Meck, 1984). It assumes an internal clock mechanism is responsible for time perception, and is defined by the operation of three stages: (1) clock (timing), (2) memory process mechanism (remembering), and (3) comparator (deciding; Gibbon & Church, 1984; Matell & Meck, 2004).
Figure 1. Schematic diagram of the scalar timing model depicting the clock, memory and decision stages.
Note. Adapted from Allan (1998).
The clock stage is comprised of a pacemaker-accumulator device. The pacemaker emits pulses at a constant rate, these are accumulated in a counter which is controlled by a switch mechanism (see clock stage in Figure 1.; Gibbon, 1991). The total count of the pulses determines the perceived duration of the interval. The functioning of the switch mechanism is controlled by an attentional gate, which is modulated by the amount of attention given to processing time. The count value of the pulses from the clock is then transferred to the working memory where it is retained. The reference memory stores pulse count values accumulated from previous trials (see memory stage in Figure 1.). The final stage involves the comparator, which computes a ration comparison between the current count value stored in the working memory, and the remembered count value from the reference memory (see decision stage in Figure 1.). The model attributes errors in time perception to memory and decision processes (Block & Zakay, 1996; Allan, 1998).
The Memory stage has a limited capacity in its ability to retain temporal information. In a sequence of durations held in the working memory, those added first, decay quicker than those added later, which are prioritised (Allen, Baddeley, & Graham, 2014). This recency effect is sensitive to cognitive loads which can cause the latter durations to decay quicker. Working memory and executive functions have been found to be engaged during temporal perception tasks (Radua, Del Pozo, Gómez, Guillen-Grima and Ortuño, 2014), and increased cognitive demand was associated with greater engagement of brain regions associated with temporal perception. This suggests that manipulation of cognitive loads placed on the working memory can affect the accuracy of time perception.
1.1.1 Types of memory engaged in temporal performance. Prospective memory is engaged when participants are presented with a duration and are aware that a temporal judgment will be required of them. Alternately retrospective memory is where awareness occurs after experiencing the target duration (Kvavilashvili et al., 1996). Research investigating prospective memory for temporal perception is able to yield Time-based prospective memory requires the duration to be retrieved from the working memory after a specific time period has elapsed. In contrast, event-based prospective timing requires the duration to be retrieved when cued to do so (Sellen et al., 1997). Studies have shown that event-based prospective timing tasks have better temporal performance rates because time does not have to be monitored in addition to the cognitive loads inflicted by the task (Kliegel et al., 2008). Therefore, such a task will be able to place exclusively task-related cognitive loads on the working memory and provide control over the independent variables (i.e. task conditions) affecting temporal performance.
1.2. Investigating Temporal Perception
When investigating temporal perception, researchers make two methodological decisions concerning the paradigm used to measure performance, and the temporal range examined (Block, 1990). A widespread issue recurrent in the literature is the lack of justification given for the paradigms chosen to measure temporal perception (Mioni, Stablum, Prunetti, & Grondin, 2016). The results of time perception studies depend on the paradigm employed (Matthews & Meck, 2014), careful selection of the methodology is therefore crucial. This makes a comparison between findings difficult because different paradigms measure different indices of temporal performance (Block, 1990; Mioni, 2013).
1.2.1. Prospective methods of assessing temporal performance. Grondin (2010) identified the four tasks commonly used in measuring prospective timing: time discrimination, verbal estimation, time production, and time reproduction. Time discrimination tasks require participants to view and compare two sequentially presented intervals and decide whether they match. Verbal estimation tasks require participants to experience an interval and then translate it into chronometric units. Alternately, time production tasks require participants to view a chronometrically labelled visual interval and convert it into a subjective duration. Lastly, time reproduction tasks require participants to reproduce the duration of a previously presented interval.
Verbal estimation tasks provide the least accurate responses because participants tend to round off their duration estimations to the second (Zakay, 1990). Time discrimination tasks measure accuracy rates, consequently providing limited information on temporal performance (Grondin, 2010). Conversely, time production and reproduction tasks show the direction of accuracy in addition to the amount. Time production tasks investigate individual differences at the internal clock level because they measure variation in the speed of the individual’s pacemaker (Block et al., 1998). In time reproduction tasks, temporal performance depends on cognitive processes as opposed to clock speed, because individual’s clock speed will remain the same both when experiencing the target, and in reproducing the duration (Block et al., 1998; Zakay, 1990). For the target duration to be retrieved and reproduced, it needs to be maintained and manipulated in the working memory. Temporal accuracy in a reproduction task, therefore, relies primarily on working memory capacity, so would be most suited to measuring temporal perception as a function of cognitive load ().
1.2.2. Which temporal range to assess? The STM predicts that standard deviation of the target interval increased linearly with its mean, this is known as the scalar property (Rakitin et al., 1998; Allan, 1998). For example, when timing a 5-sec interval, an individual’s judgment accuracy will be within 1-sec, but when timing a 10-sec interval, accuracy falls to 2-sec. Thus, an exponential function is used to select target durations to ensure that difficulty of discrimination between adjacent durations increases successively. Processing duration intervals under 1000-ms are automatic, preventing the use of a chronometric counting strategy (Wearden & McShane, 1988) whereas longer intervals (over 1-sec) require support from additional cognitive resources (Fraisse, 1984; Lewis & Miall, 2003). Therefore, durations of a few hundred-thousand milliseconds can be used to effectively engage cognitive processes (Matthews & Meck, 2014). In order to investigate temporal perception reliant on automatic processing, and select enough durations using an exponential function, durations must be selected from around the 1000-ms mark.
In a recent study, Chen and Huang (2016) developed a temporal version of the N-back paradigm to study the effects of increasing WM loads on temporal performance. A time discrimination task was used where participants were presented with sequentially presented visual stimuli and decided whether they matched the previous stimuli shown (1-, or 2-back). Responses in the 2-back task (defined as the high memory-load condition) caused greater inaccuracy in responses than compared with 1-back (low memory-load condition). Thus demonstrating that increased loads on the working memory caused greater inaccuracy in temporal performance.
1.3. Methodological Issues Identified in Chen & Huang (2016)
Chen and Huang (2016) used a time discrimination task, which restricted their measurement of temporal performance to just the amount of inaccuracy. A reproduction task would provide a more detailed account of temporal performance, by measuring the amount of inaccuracy and also the direction of the error. The present study will, therefore, use a temporal reproduction task to investigate whether the errors associated with greater memory load were under- or -over-estimated.
Chen and Huang (2016) used four target durations (100-, 200-, 400-, & 800-ms), and found that accuracy of 200- and 800-ms to increase in both low and high working memory loads. It would be worth investigating whether accuracy will continue to increase for the next target duration selected based on increasing exponential function. Previous studies have shown longer durations to be under-estimated (Eisler et al., 2008). In order to investigate whether this underestimation continued with task difficulty, a 1600-ms target duration was added to see if the trend in decreasing accuracy extended to the next exponentially greater duration. The present study will, therefore, examine target durations of 400-, 800-, and 1600-ms for further analysis.
In their review of the time perception reviews, Matthews and Meck (2014) direct future research to study aspects of cognition on temporal perception as a matter of urgency. Few studies have examined the effect of increasing working memory loads on temporal perception. The present study will include a 3-back (defined as very-high memory load) to place further demands on the working memory and examine the effect of both amounts of inaccuracy and direction of inaccuracy of temporal performance. Using three conditions will also enable a trend in results to be identified and thus enable predictions to be made about temporal performance.
Chen and Huang’s (2016) findings had medium statistical power (.50) and effect size (d = 1.15) due to their selected sample size (n = 14). The present study increased sample size (n = 26) to ensure that results had a high statistical power (.80), and a medium effect size (d = 1.15; Cohen, 1988; Rosenthal, 1996). This will increase the probability of finding a medium effect size with reasonable confidence (80%).
1.4. The Purpose of the Present study
This is the first to use a temporal version of the N-back task to measure reproduction durations. To do so, adaptations were made to Chen and Huang’s (2016) N-back task to measure reproduction durations, thus adding a novel temporal task to the literature. The task will be able to identify a trend in the direction of inaccuracy of temporal performance under the influence of increasing working memory loads. Three indices of temporal performance were measured (as in Mioni et al., 2016) to provide a deeper understanding of the impact of the independent variables. Where Chen and Huang (2016) used the spacebar key to record responses, the present study used the left-click mouse button. This is due to its microchip having better millisecond discrimination ability, enabling more accurate data to be collected.
The literature suggests that increasing working memory loads will impair the accuracy of temporal perception to a greater extent (Rauda et al., 2014). The present study will, therefore, expect to find greater inaccuracy produced in reproduction durations under higher working memory loads. Increasing working memory loads are also expected to cause an under-estimation of reproduction durations. Under-estimations of reproduction durations are expected to be found with increasing memory loads.