According to Kantowitz and Simsek (2001), a key assumption of secondary task methodology is that the primary task is uninfluenced by addition of the secondary task. This argument tends to hold for airplane pilots because they are highly trained to set the priority on flying the plane (e.g., Bartolussi et al., 1986), but must be evalu- ated anew when vehicle drivers become the test population. To this end, secondary task demands were varied by implementing seven different list lengths, two different control locations, and two different display locations.
6.3.1 List length
The number of searchable entries in the list did not change during one single condi- tion in all experiments. Thus, participants did not have to adapt to a changing number of entries as would be the case in a common hierarchical menu. Nevertheless, there
were significant performance differences between the number of list entries in short lists. The most obvious explanation for this is that shorter lists (with two and four en- tries) impose less task demands on the operator. Another explanation might be the number of displayed entries on the screen, which was four in all experiments. As the first position in the list was not yet a searchable list entry, only the list with two entries appeared on the screen in its entirety. In order to access the last entry of a list of four entries participants only had to scroll one item down. The necessity of scrolling fur- ther than this obviously entailed more visual demand, thus degrading driving per- formance. Motor demands imposed by navigating through the list via a control might have added to an increase in task demands as well, especially when participants passed the target item in the list and had to change direction of cursor movement. In keeping with this proposition, Rauch, Totzke, and Krüger (2004) suggest that infor- mation systems for use during driving should access deep menu structures, implying that with a deeper menu structure, the number of list entries on each level is reduced. However, this advantage probably holds only for an overall quantity of data small enough to not require many levels in menu hierarchy, as the navigation problem (i.e., getting lost or using an inefficient pathway to the goal) becomes more and more treacherous as the depth of the hierarchy increases (Paap & Roske-Hofstrand, 1986). Furthermore, while MacGregor, Lee, and Lam (1986) suggest that a minimum of eight alternatives per page is optimal, with some indication that the optimal could be considerably greater than eight, their results were not gained in a driving environ- ment. An explanation why differences between list lengths of more than eight entries did not become significant might be that participants reacted to increased list lengths by extending the duration of the secondary task: as overall task demands were ap- parently higher for longer lists, the strategy of extending the duration of the secon- dary task had to be used excessively in experiments 1 and 2, indicated by high per- centages of missed trials. But, as rather little changes in driving performance be- tween the different list lengths or list positions reveal, quite successfully: While driv- ing performance remained relatively constant throughout the four different list lengths in experiment 2 and the three different list positions in experiment 1, misses in- creased significantly. Jordan and Johnston (1993) and Summala (1996) report similar findings.
6.3.2 Location of secondary task controls
The effect of increasing the proximity of secondary task controls to the driver by lo- cating them on the steering-wheel was smaller than expected. The enhancing effect occurred only during less demanding situations and did not become significant. The moment task demands increased (mainly due to increasing curve radii), any advan- tages vanished: task completion times were not shorter and lane deviations were not less frequent or smaller in amplitude. Only the effect of car velocity on error rate, self- estimated effort and subjective ratings of secondary task interference and attention allocation was significantly effected by control location, indicating that at least subjec- tively experienced workload levels decreased during fast conditions. Apparently, reaching distances are not a major factor effecting driving performance.
6.3.3 Display proximity
The proximity compatibility principle (Barnett & Wickens, 1988; Wickens & Andre, 1990; Wickens & Carswell, 1995) suggests that a task requiring high processing proximity should be designed with high display proximity. In other words, if two infor- mation sources are used to a high extent within the same task, the according display components should be located close together (defined in spatial terms, i.e. cm). Fur- thermore, Wierwille (1993a; 1993b) suggests that single ‘check glances’ with a dura- tion of about 300ms do not have a degrading affect on driving performance and al- though it is true for visual tasks that only one single object can be focused at the time, other objects in the surrounding area of this focused object may be covered by indirect vision (Alm et al., 1997; Wickens & Carswell, 1995). If these assumptions were correct, then driving performance in experiment 5 should have been superior to all other experimental settings, as both the driving task and the list task were dis- played on only one monitor, thus increasing the proximity between the two sources of information to a maximum. This was indeed the case for the factor searching period, as driving performance improved significantly (especially during the last fifth of the search). It was also true for the effects of list length on task duration times, self- estimated effort and subjective ratings of secondary task interference. Furthermore ratings of attention allocation revealed that there was a significantly higher priority on the driving task during fast trials in experiment 5. Thus, increasing the display prox-
imity showed some improving effects on driving performance but could not account for increasing task demands imposed by frequently changing road conditions.
6.3.4 Searching period
The closer the participant got to the target item when navigating in the list, the higher the task demands (revealed by significant driving performance losses). Apparently, executing a secondary task while driving, namely searching in a list, does not impair driving homogenously at any point of the search. Evidently, the increase in visual demand of the secondary task in order to locate a specific item in the list (contrary to just roughly scanning the visible entries for the current position in the alphabet) is responsible for the significant rise in secondary task demands and hence for de- creases in driving performance. In these situations, participants were obviously too focused on the secondary task to detect changing task demands of the driving task, thus being incapable of executing a proper reaction fast enough. Furthermore, late discovers of changes in primary task demands might have caused quick, but less accurate reactions, thus causing higher lane deviation measures in the experiments. In real driving environments similar reactions might even resulted in a loss of vehicle control.
6.4 Subjective experience
Drivers’ self reports revealed their ability to recognise decreases in driving perform- ance when executing various secondary tasks while driving. They also demonstrated a clear increase in effort and secondary task interference when task demands in- creased. The values varied along the different conditions as participants could only react to increasing task demands by prolonging the secondary task duration which was quite effective to protect driving performance but did obviously not reduce work- load levels. The strategy to decrease task demands in a more rapid manner would possibly be the reduction of speed which was not possible in these experiments. Fur- thermore, participants seemed to be well aware of driving performance degrading shifts in task priority towards the secondary task.