1. Introduction

The Trail-Making Test is a well-established, expedient and widely used cognitive assessment tool. In TMT part A, subjects are asked to connect randomly distributed numbers in an ascending order; in TMT part B, they are required to connect randomly distributed letters and numbers in an alternating, ascending order (1, A, 2, B, 3, …). In either case, the outcome measure is completion time. Converging evidence from many studies suggests that individual differences on TMT-A are mainly attributable to the speed of perceptual processing, the speed of motor programming and the proficiency of visual search; differences on TMT-B are additionally related to working memory and task switching abilities [e.g.; 1].

A number of studies reported that performance on the Trail-Making Test decays in old age. Deficits are more pronounced on TMT-B, but even on TMT-A, 40-59 year olds are about 17% to 25,5% slower than healthy 20-29 years olds [2-5]. The present study evaluates possible reasons for the age-related slowing on TMT-A. In fact, all three determinants described by Sanchez-Cubillo et al. [1] are degraded in old age and thus may contribute to age-related slowing on TMT-A: Compared to young adults, older persons exhibit a lower processing speed [6, 7], both at the perceptual [8-10] and at the motor level [11-13] and a poorer visual-search ability, particularly when complex visual scenes are presented [14-16]. The latter deficit was attributed to difficulties in the spatial localization of task-relevant information [17], an increased sensitivity to distracting stimuli [16, 18, 19; see however 20] and problems of attentional control [21, 22; see however 15]. It has also been shown that in visual search, older individuals rely more heavily than young ones on contextual information [23, 24]. It therefore is conceivable that age-related deficits are ameliorated when visual search is embedded in a familiar context.

To distangle the contribution of visual search from that of basic perceptuo-motor processing in TMT-A, we implemented a version of the Trail-Making Test in which visual search is minimized. In this TMT-C, numbers are laid out in an easily predictable sequence, from left to right line by line, as in reading. Impaired perceptuo-motor processing should manifest both on TMT-C and on TMT-A, while impaired visual search should manifest on TMT-A alone.

A second purpose of our study was to analyse older person’s deficits on TMT-A at the level of eye movements. We are not aware of earlier work on age-related changes of the gaze pattern during the Trail-Making Test, but changes have been observed with other paradigms. Compared to young adults, older participants execute more fixations [25, 26] with longer fixation durations [27, 28], look at critical upcoming locations earlier and longer [29-32], have problems to suppress undesired saccades [33-35] and – although they have no difficulties to disengeage their gaze from inspected locations [36, 37] – they return their gaze to such locations more often [26, 38]. We therefore expected that during TMT-A as well, older participants will have no problems to detach their gaze from the last-inspected target but will refixate that target more often, will execute more fixations with longer durations before attaching their gaze to the next target, and look at that target for a longer time.

Trail-Making Tests have been administered in a number of variants besides the standard written version, e.g., in an oral [2], an eyes-only [39], a keyboard [40] and a walking version [41]. Here we introduce a new version in which trail-making is embedded in a barcode scanning task, a typical activity for workers in a goods receiving department. This allowed us to compare older persons with daily experience on this task to older persons without such experience. We hypothesized that by embedding the test in a familiar context, we will facilitate performance of the experienced relative to the unexperienced older group.

2. Methods

Thirty participants (19 women, 11 men) were recruited for the study. Ten were young adults with no particular experience in barcode scanning (YA, 22,5 ± 2,42 years of age), ten were older adults with no particular experience in barcode scanning (OA, 52,5 ± 6,29 years of age) and the remaining ten were older employees from the goods receiving department of a wholesale company who had 24,55 ± 7,93 years of professional experience that included barcode scanning for about 3 hours per day (OE, 50,2 ± 7,77 years of age). The daily activities of all three groups were composed of a combination of physical and cognitive tasks. At the time of testing, no subject reported any orthopedic or mental disease, and had normal vision. All participants gave their written informed consent. The study was approved by the local Ethics Committee of the German Sport University.

Participants stood in front of a poster (31,1 inch horizontal x 46,8 inch vertical) on which 20 rectangles of four different sizes were drawn (9,25 x 5,12 inch, 11,61 x 6,69 inch, 4,33 x 5,12 inch, 5,12 x 9,25 inch). Each rectangle contained a barcode (3,94 x 1,57 inch) and a number (1 to 20) to the left of it. This poster layout is illustrated in Fig. 1. Participants were told that the rectangles represent “parcels” whose barcodes they have to scan with a handheld scanner (Albasca MK-1000ZB, as used at work by OE). They held the scanner in their dominant right hand, scanned first the parcel numbered “1”, and proceed in ascending order until they scanned parcel number “20”. The use of the scanner was explained to unexperienced subjects beforehand. All participants were able to reach all barcodes solely through arm and trunk movements.

Figure 1. Illustration of a poster with the barcodes

Each participant scanned a total of six posters. Two of them represented the control version of our Trail-Making Test, TMT-C: rectangles were arranged in ascending order, such that participants simply had to move the scanner from left to right row by row, as in reading. They were instructed to complete the task at their favored speed. Two other posters represented the unspeeded version of TMT-A, which we call TMT-A_U: rectangles were arranged in a mixed order (different for each poster), such that after scanning one barcode, participants had to search for the barcode with the next-higher number. They were again instructed to complete the task at their favored speed. The remaining two posters represented the speeded version of TMT-A, called TMT-A_S: it differed from TMT- A_U in that participants were asked to proceed as fast as possible. All participants completed the six TMT posters in balanced order, with the constraint that the very first poster represented TMT-C.

Throughout testing, participants’ eye movements were registered at 30 Hz with a head-fixed registration system (SMI Eye Tracking Glasses Device 1.5 and SMI Recording Unit Device 2.0). The data were subsequently analysed with commercial software (BeGaze 3.4) to yield the following dependent variables:

• Completion time: interval from scanning the first to scanning the last parcel

• Gaze path: length of a path connecting all fixations during completion time

• Prescan time: interval from the first fixation of a to-be-scanned parcel to its scan

• Postscan time: interval from the scan to the last fixation of a scanned parcel

• Refixations: Number of gaze returns to a previously scanned parcel

• Transit time: interval from the last fixation of a scanned parcel to the first fixation of the next to-be-scanned parcel

• Transit fixation count: number of fixations during the transit time

• Transit fixation time: mean duration of fixations during transit time

The first three variables were averaged across both posters, and the other variables were averaged across all parcels of both posters. The outcome was submitted to a 2 x 3 analysis of variance (ANOVA) with the between-factor Group (YA, OA, OE) and repeated measures on the factor Condition (C, U, S). Factor level C was not used for transit fixation time and count, since the majority scores were zero in TMT-C. Significant effects were explored with Fisher’s LSD post-hoc tests.

3. Results

Table 1 summarizes the ANOVA outcome for all dependent variables, and Table 2 the post-hoc decompositions of the pertinent significant effects of Condition (last two columns), Group (last row for each variable) and Condition*Group (remaining cells). One variable, transit fixation time, yielded no significance at all. Three variables yielded significance for Condition only; they are illustrated in Fig. 2. The remaining four variables yielded significance for Group, and two of them also for Condition*Group; they are depicted in Fig. 3.

Table 1. Analyses of variance for all dependent variables. Cell entries are F-VALUES, with n.s *,**and *** indicating p>0.05, p<0.05, p<0.01 and p<0.001, respectively. Degrees of freedom are F(2,27) for Group, F(2,54) for Condition and F(4,54) for Interaction, except for transit fixation count and transit fixation time, where they are F(2,27) for Group, F(1,27) for Condition and F(2,27) for Interaction

Table 2. Outcome of post-hoc analyses for Group (first three columns, row “all conditions”), Condition (last two columns) and the interaction term (first three columns, rows “C”, “C” and “S”). Cells corresponding to non-significant ANOVA effects are marked by “N/A”

Figure 2. Means of three dependent variables for the three TMT Conditions and the three Groups (OE, OA YA). Error bars are standard deviations

Figure 3. Means scores of four dependent variables, separately for each group and condition. Error bars are standard deviations

According to these figures and tables, there were no group differences for any variable in condition TMT-C. All scores increased from TMT-C to TMT-A_U and TMT-A_S; an additional increase from TMT-A_U to TMT-A_S emerged for only two variables, postscan time and number of refixations, but notably not for the variable total time. Most importantly, the increase from TMT-C to the other two TMT conditions was more pronounced in older than in young participants for the variables completion time, gaze path and transit time, but the increase didn’t differ between OA and OE.

4. Discussion

Our study evaluates whether age-related decrements of TMT-A are modulated by task experience, and to what extent they reflect specific deficits of visual search rather than generalized sensorimotor slowing. To this end, we analyze not only the standard outcome measure, total completion time, but also include measures of the subjects’ gaze pattern.

We implemented a control condition which minimized visual search, TMT-C, and found no evidence for an effect of age or experience on completion time or gaze. This might seem surprising, given that reports on age-related sensorimotor slowing [11, 6, 7] and changes of the gaze pattern (e.g., 26, 27, 30-33] abound in literature. One possible explanation is that the mean age of our older group was only about 50, compared to 60 – 70 in most earlier studies. When data from other studies (e.g., 42-44] are interpolated to predict sensorimotor speed at the age of 50 – capitalizing on the fact that speed declines in a close-to-linear fashion from the age of 20 on [45, 46] – the expected decrement is about 15%. In contrast, we observed no decrement at all for completion time in TMT-C. It therefore is unlikely that our older subjects were too young to exhibit noticeable sensorimotor slowing, and we rather propose an alternative interpretation for our findings: TMT-C might be resistant to sensorimotor slowing since it is a more natural task than the abstract laboratory paradigms in the above studies. We submit that all persons, even if they don’t routinely scan barcodes at work, regularly observe this activity e.g. in supermarkets and department stores. Indeed, there is experimental evidence that age-related slowing is absent in familiar, ecologically valid scenarios: no age deficits were observed when a reaction-time task was embedded in a realistic car-driving scenario, where subjects had to brake in response to unexpected traffic events [47-49]. Summing up, TMT-C might be resistant to sensorimotor slowing since it is more natural than the abstract laboratory paradigms typically used in literature.

TMT-C served as a control for the unspeeded and speeded variants of TMT-A, which additionally required visual search. Not surprisingly, this additional requirement led to a poorer performance on all registered variables: the gaze travelled along a longer path, rested longer on each scanned parcel and returned to it more often, spent more time between scanned parcels and as a consequence, completion time increased. Interestingly, these changes were comparable in TMT-A_U and TMT-A_S, except for two minor differences: in the latter condition, subjects’ gaze remained longer in the scanned parcel and returned to it less often. These differences, however, had no noticeable effect on completion time. In other words, we were unable to speed up visual search by means of verbal instructions.

The increase of completion time from TMT-C to TMT-A was much more pronounced in our older than in our young adults, which confirms earlier accounts about age-related slowing on TMT-A [2-5]. More importantly, the emergence of slowing on TMT-A but not on TMT-C suggests that the underlying deficit is related specifically to visual search rather than to sensorimotor processing in general. This conclusion is supported by our gaze data: compared to young adults, older persons didn’t increase their pre-and postscan time but their transit time was 0,899 s longer, which adds up to 17,081 s for 19 transits between 20 parcels and thus fully accounts for the 17,371 s increase of completion time. Older participants therefore spent more time looking for the next to-be-scanned parcel, but not fixating the presently scanned parcel. This prolonged search for the next parcel is paralleled by an increased length of the gaze path and an increased number – but not duration – of fixations.

The observed changes of the gaze pattern in old age fit well with earlier work which reported that during visual search, older participants execute more fixations [25, 26] with no increase of fixation duration [26] unless complex visual processing is required [27, 28]. As in our study, the number of refixations was unchanged in old age [18], or it was found to increase [38, 26]. Studies in which subjects had to negotiate obstacles while walking found that elderly persons’s gaze is directed at critical locations earlier and longer [29-32]; this increase was not reflected by a longer pre- or postscan time in our study, possibly because scanning was not relevant for posture and therefore didn’t call for in-depth processing. Summing up, our findings on age-related changes of the gaze pattern during visual search are consistent with literature.

Our study implemented a TMT variant which models the working-life experience of employees in a goods receiving department: the search for numbers was embedded in a barcode scanning task, and rather than drawing lines, subjects operated a scanner type which group OE actually used in their daily work. We expected that subjects from group OE would benefit from their working-life experience and perform better than those from group OA; however, such benefits didn’t emerge for completion time or any gaze parameter. It therefore appears that decades of experience with parcel scanning didn’t improve sensorimotor processing (TMT-C) or visual search (TMT-A minus TMT-C) in our experiment. One possible explanation is that OE felt uncomfortable to call up performance similar to their daily work, such that the benefits of experience were cancelled by the costs of distress. As an alternative explanation, our variant of TMT was not sufficiently similar to the daily work of OE, which demands not only scanning, but also simultaneous activities such as sorting, reading or checking the content of a parcel against an invoice. In any case, our findings provide no evidence that age-related deficits of visual search can be ameliorated by extensive training on similar tasks.

Taken together, our study confirmed that completion time on TMT-A increases in old age, attributed this increase to specific deficits of visual search rather than to generalized sensorimotor slowing, and documented the correlates of those deficits at the level of eye movements. We found no evidence that age-related deficits on TMT-A are reducible by practice.

ACKNOWLEDGEMENTS

Thanks are due to all participants and student assistants Kristin Henkelmann and Denise Schulze.