Experimental studies of task demands and processing performance in driving and driving simulation

(1)

VHS,-,,om

155 1990

Experimental studies of task demands and

processing performance in driving and

dri-vin9 simulation

Lisbeth Harms

Paper presented on the 22. INTERNA T/ONAL CONGRESS

OF APPL/ED PSYCHOLOGY, Kyoto 21 26 July 1.990.

Symposium: " Vision in Vehicles "

% Väg''06/1 Efi/(' Statens väg- och trafikinstitut ( VTI) 581 01 Linköping

(2)

(3)

VTIsrtryck

155 1990

Experimental studies of task demands and

processing performance in driving and

dri-vin9 simulation

Lisbeth Harms

Paper presented on the 22. INTERNA T/ONAL CONGRESS

OF APPLIED PSYCHOLOGY, Kyoto 27 26 July 7990.

Symposium: " Vision in Vehicles "

w,VägUCI) af/k Statensvväg- ochtrafiii'knsttut/VTI) 581 01 L1nköping

(4)

(5)

1. OBSERVATIONS OF CAPACITY LIMITATIONS IN DRIVING

The capacity limitations of the human information processing system are often illustrated with a reference to the everyday experience that the demands of a driving task may determine the extent to which a driver can perform other mental activities simultaneously:

"Finally, the transient variations in the effort that a subject invests in a task determine his ability to do something else at the same time. For example imagine that you are conducting a conversation while driving an automobile through city traffic, you normally interrupt conversation. Physiological measures would certainly indicate a surge of arousal at the same time, corresponding to the increased demands of the driving task."

(D. Kahneman, 1973)

"Imagine yourself entering the freeway during the rush hour sim ultaneously driving and talking to a friend or listening to the radio. As you get on the entrance ramp, you start shifting your eyes back and forth, checking for gaps in the main flow of traf

fic. However, somewhere in the process, your mind became so oc

cupied in making the transition, that you completely stopped listening to your friend or the radio only to realize it once you were again in the main flow of traffic."

(D. Shinar, 1978)

"Tasks are assumed to demand resources for their performance, and these resources are limited in their availability. There-fore, when the joint demand of two tasks exceeds the available supply, time sharing efficiency drops and will be more likely to do so as the difficulty of either component task increases. For example, conversation with one s passenger in a car will normal ly be disrupted if the demands of the concurrent driving task are increased by poor visibility or heavy traffic. Alternative-ly, driving performance may degrade as the conversation becomes extremely interesting."

(6)

These examples, and several others could be listed, are based on evidence that heavy task demands may prevent a driver from at-tending to an irrelevant activity, such as a conversation, while driving. It is likely to suggest that a heavy load on the dri-ver s processing system may also impede the processing of infor mation relevant to the driving task.

Another common assumption is that driving skill obtained through driving practice will lead to automatic information processing, which proceeds without attentional control and does not stress the limited capacity processing systems of the driver (Rasmus-sen, 1986 and Summala, 1988). Automatic processing is therefore supposed to override the capacity limitations of the human processing system.

2. THE CONCEPT OF AUTOMATIC AND CONTROLLED PROCESSING

Extensive amounts of consistent practice have been demonstrated to change the processing strategy of experimental subjects from a capacity demanding serial one to an automatic strategy probab ly based on parallel information processing (Shiffrin & Schnei-der, 1977 and Schneider & Shiffrin 1977). While for instance ve hicle control can obviously be subject to automation, it may be questioned whether real traffic provides sufficiently consistent practice conditions for the automation of the drivers' situa-tional control (see for instance Fisk, Ackerman & Schneider, 1987).

Assuming that variation in the traffic environments may induce measurable variation in the drivers' cognitive load, we decided to investigate the influence of ordinary variations in the driv ing environment on skilled drivers' cognitive load (Harms 1986, 1989 and 1990).

(7)

3. MEASUREMENTS OF DRIVERS' SECONDARY-TASK PERFORMANCE IN DIFFERENT DRIVING ENVIRONMENTS.

Three driving routes were selected for the field study. All the routes included both sections of highway driving and sections of driving through a village area. The drivers cognitive load was measured by a secondary task method (see Harms, 1986).

The drivers were presented with simple calculation tasks while driving, their reaction times to the tasks were measured and lo-calized at the actual position of the driving route. The longer the observed calculation time the higher was the driver's cog-nitive load supposed to be at that position. The driving speed was measured and localized at the successive positions of the driving routes.

The driving experiment was performed both before and after a major reconstruction of the village sections of the driving

routes and the study included a total of 6 data collections. It was a general finding that the drivers' cognitive load was higher on the village sections of the driving routes than on the highways (i.e. their mean calculation time was longer in village areas than on highways) whereas their driving speed was higher on the highways than in the village areas. As can be seen in Figure l the pattern of results was similar for all the driving routes and for measurements conducted both before and after the reconstruction of the village sections.

(8)

M E A NDR IV IN G S P E E D Im /s eC I

-14oo

24 22_

-12oo

20

18 -1ooo

16-14

800

12

10 /,C]

/////

GM)

tr

1 PI

\/

Figure 1.: The mean speed and mean calculation time of drivers

on the 3 different driving routes on Highways (H) and in Village areas (V). Solid lines indicate observations before the recon-struction of the village sections of the driving routes, dashed lines indicate observations after the reconstruction.

M E A N R E A C T I O N T I M E (m se c)

(9)

Moreover, driving speed and the drivers cognitive load were found to be inversely related. For lOO-metre sections of the driving routes a high negative correlation was found between driving speed and the drivers cognitive load on all driving

routes (see Table 1).

Experiment 1 Experiment 2

Route 1 .61 .56

Route 2 -.76 .6O

Route 3 .79 -.55

Table l.: Product-moment correlation between. driving speed and the drivers cognitive load on the 3 driving routes before

(Experiment 1) and after (Experiment 2) the reconstruction of

the village sections driving routes.

The finding that the drivers cognitive load varied between highways and village sections. of the driving routes supported the assumption that the driving task may demand a varying amount of cognitive resources for its performance. Moreover, the result suggested that even common variations of the driving environment influenced the drivers cognitive load.

4. DISCUSSION of THE FIELD STUDY

The implication of this finding is ambiguous. It is likely that the observed increase in the drivers mean calculation time re-flected a general decrement of their processing efficiency. Thus it may be assumed that this should also have impeded the processing of information relevant to the driving task. On the other hand, as the drivers performance in responding to the primary task was not evaluated, it can be argued that both the speed reduction and the allocation of additional resources to the driving task may have prevented heavy task demands from impeding the processing of information relevant to driving.

(10)

For the development of a reasonable concept of safe driving it is important both to clarify the relationship between the var iation in the drivers cognitive load and their performance in responding to the primary task of driving. Furthermore it is important to decide which variations in the driving environment influence the driving performance.

Do delayed responses to calculation tasks while driving actually imply a decrement of drivers general information processing performance, such that the processing of information relevant to driving may also be impaired ?

The answer to this question requires an experimental set-up in which the demands of the driving task can be varied systemati cally, and it requires an appropriate evaluation of the influ ence of these task variations both on the drivers performance in responding to the primary task of driving and to the secondary task of calculating.

For further investigations of the influence of task demands on subjects performance, the method used for the field study was transferred to the laboratory conditions of a driving simulator. The first simulator studies aimed at testing which task variati-ons would lead to variation in the subjects cognitive load and also how the task variations would influence the subjects pri-mary task performance.

5. MEASUREMENT OF SUBJECTS DUAL-TASK PERFORMANCE IN DRIVING SIMULATION.

Two experiments were conducted in the VTI-driving simulator

(Harms, 1990, for a technical description of the driving

si-mulator see Nordmark et.al. 1985). The use of a driving simu-lator implied that driving on a real road was substituted for driving on a simulated one. Environmental variation was simu-lated by presenting visual items near the simusimu-lated road while a subject operated the driving simulator (see Figure 2).

(11)

V

Figure 2.: The possible positions of name-items (upper right) and color-spots (near the road line) in Experiment 1. The figure also shows that while the position of the road varied the visual items appeared on fixed positions in the visual display.

The visual items were subdivided in targets and distractors and subjects were required to be careful in responding when targets were presented and not to respond to distractors. They were told to drive as they would do on a real road, and to perform the calculation task at their own pace.

Both experiments used a factorial design with 2 different

sti-mulus types (color items or name items) and 2 task variants

(easy or difficult). The combinations of the types and variants of tasks resulted in 4 different driving conditions. Moreover, the practice conditions varied for the two tasks (see figure 3 for an illustration). In Experiment 1 the color task was prac ticed consistently whereas the name-task used varied practice, in Experiment 2 the practice conditions were reversed.

?

D D

(12)

____ ___,

'

:?

___ ZZ? '

D ;]

EASY NAME/ EASY COLOR

___

__

;

L

':

J

//

___E i

,

DIFE, NAME/ EASY COLOR

[_

H D

EASY NAME/ IIFF, COLOR

_,

?

1 0'20 sea 1 O-ZOOOm 12000 ' ZZSOmg 2 25012 000 ml

\ /\A j? ) xf T xf

TASK NO'TASK TASK CONDlTlONS: THE [+ POSSIBLE COMBINATIONS

PRESENTATION EONOlTION

OF COLOR TASK AND NAME TASK DIFFIEULTY

Figure 3.: Before each of the 4 task conditions the subjects were presented with the targets of the next task condition and 2 kilometres of free driving. Within each road interval of 250 metres 1 task was presented. The minimum stimulus separation was

(13)

Experiment 1

Subjects were presented with a color discrimination task with one and the same target color used for all experimental sessions (consistent practice). The subjects were required to move their foot from the gas pedal to the brake pedal whenever the target spot was presented. Task difficulty was manipulated by the simi larity between the color of the target and the color of the distractor. The difficulty of the name-task was manipulated by the number of different names involved in the task. Target names were replaced in the set of distractor names after each experi mental session and new target names were selected (varied prac tice). The subjects were required to respond to the name-task by turning on the headlamp control whenever a target name was pre

sented.

Results of Experiment 1

It was found that subjects' performance was thoroughly unaffec ted by the variation in color difference between target and dis tractor color. The subjects' speed, their mean reaction time to both the visual task and to the calculation tasks, and their error rate were the same for both variations of discriminability (Figure 4, p. 11). On the contrary, the number of names involved in the name-task influenced subjects' performance considerably. As can be seen in Figure 4 the difficult name condition, including more names than the easy one, resulted in lower mean speed, longer mean reaction time and a higher error rate in responding to the target names. Moreover, the subjects' mean calculation time was greater for the difficult name condition but this effect was just marginally significant (p <

(14)

12 A 11 ~ 7 if) 9 ö O S% V 10 __ LU 2 F % ; 09 %% UJ "__ 0,4 (I 2 < 0.8 If 0.3L Lu < E CI: d: O 2 ~ O a: o7I gcme A)? [__ 1 x _1 E D E D E D E D

NAME COLOR NAME COLOR

I I I i | "* 1.10 I ,. | 0 E I o 100 ; moo g₈ A I m £ I g E I I

E

_w

,

Q

₆ LU I (

%

I

&

CD 80 " I " 0.8 Z Z | L1<J > E | 2 G | % 70 *" I " O.7 LU I 2 I I so) , £$

%

1

E D E D E D E D

Figure 4.: Upper panel: Mean reaction time and error rate for the

easy (E) and difficult (D) conditions of the visual tasks (color or name)of Experiment 1. Lower panel: Mean driving speed and mean

calculation time for the easy (E) and difficult (D) conditions of

(15)

Experiment_g

Experiment 2 used the same metod and the same stimuli

as Ex

periment 1 but the practice conditions were reversed. The

color-task then used varied practice and the name-task used consistent

practice. Another important variation was that the difficu

lty of

both the tasks was varied by the number of items used for

the

task (see Figure 5)

F?

_._______ |

EASY NAME/EASY COLOR

M I

_{D J}

IQ

EASY NAME/ DIFF. COLOR

& ÖJ

DlEENAME/EASY COLOR

DTF F. NAME / DIEF. COLOR

FEET

1 020 sec. i 0-2000m 12000 ' 2250m 2250 , TZO

OOm

TASK NO-TASK TASK CONDlTlONS: THE I+ POSSlBLE COMBTNATTONS

PRESENTATION CONDlTlON OF COLOR TASK AND NAME TASK DIFFlCULTY

Figure 5.: The procedure and task conditions of Experiment 2.

As

can be seen from the figure the difficulty of both tasks varied

(16)

Results of Experiment 2

Both the tasks of Experiment 2 were found to influence the

subjects performance similarly (Figure 6, p. 14). The subjects mean speed varied between conditions of difficulty and their mean reaction time and error rate in responding to the visual tasks increased as a function of task difficulty, but only the color-task for which varied practice was used caused a

significant increase in the drivers mean calculation time (p <

(17)

1.4 13% ' 7 {5 1.2 5 O (1) (I) LU g 10» F. Z 9 l 1 209 UJ CC z O4~ (( llJ 5 oar E 03 (i [I En 02 o 07 %: 01 »

$

L

L ,__J |

E D E D E D E D

1 I | 1,40 | | 1 | 1 | V 20 _| * 1. I | EE

i

%

| " 1.10 8 l 33 | UJ | 2_|. 100 _ 1 - 1.00 z | Q A l å * 2 E | Lu x 90 _ l "* 0.9 (I vO | 5 LU 5 l 2 % I o 80 " l 0-8 Z 1 2 (I 1 O 1 & 70 +- | 0.7 L 1 E _| | 60 l -1 0.6

4(

'

$

E D E D E D E D

Figure 6.: Upper panel: Mean reaction time and error rate for

the easy (E) and difficult (D) conditions of the visual tasks

(color or name) of Experiment 2. Lower panel: Mean driving speed and mean calculation time for the easy (E) and difficult (D) conditions of the visual task (color or name) in Experiment 2.

(18)

6. DISCUSSION OF THE SIMULATOR STUDY

The results of these first experiments on the relationship between task demands and subjects dual task performance in driving simulation did not lead to decisive evidence, but the pattern of results showed some consistency which is important to future investigations to test and to clarify.

A simple variation of similarity between target and distractor color was not found to influence subjects performance. The variation in color difference may have failed to increase the confusability between targets and distractors in this experi ment, but also the use of consistent practice may have contri-buted to the result. It is up to future research to decide about the relative importance of the two factors, discriminability and practice, to the subjects performance.

The other experimental manipulations used the number of items for the variation of task difficulty, and the pattern of results was rather consistent for these variations. The number of items involved in a task influenced the subjects performance in a ge-neral way. With many items, subjects performance in responding to the targets degraded and their driving speed went down. More over, with varied practice the subjects mean calculation time increased for conditions with many items, whereas with consist-ent practice the number of items did not influence the subjects calculation time and apparently the subjects performance in responding to the targets was less influenced by the the number of names than it was when varied practice was used.

This tendency might suggest that the subjects processing effi ciency was influenced by practice conditions and it may be speculated that much more practice would have resulted in a greater difference between the conditions of practice.

The use of both different subjects and different tasks in the two experiments complicated a clear interpretation of the ob served tendency and obviously further studies are required to qualify the influence of practice conditions on subjects

(19)

7. A CONCLUDING REMARK

The assumption that driving practice will lead to automation of driving performance implies that real traffic Should provide sufficiently consistent practice conditions for automatic pro-cessing to develop.

The result of the present study of driving in real traffic show ed that even common environmental variations such as driving from a highway through a village area caused measurable varia tion in the drivers' cognitive load. This result supported the suggestion that driving in real traffic demands a varying amount

of cognitive resources for its performance. The finding that

driving speed and the drivers' cognitive load were inversely related might even suggest that the task demands influenced the drivers' performance more generally. However, except for the ob served variation in driving speed drivers' performance on the primary task of driving was not evaluated in the field study. The study of task demands in driving simulation showed that with varied practice the task demands influenced subjects' performan-ce generally and systematically. Many items caused a decrement in the subjects' processing efficiency, influencing both their performance in responding to the visual items and in responding to the calculation tasks, and subjects' speed was considerably

lower for such conditions.

With consistent practice the subjects' performance was apparent ly less influenced by the number of items. It is likely to assume that much more consistent practice would have increased the subjects' processing efficiency even further and would finally have led to automatic processing whereas no such effect should be expected for conditions of varied practice.

The experimental design was not sufficiently sensitive to de monstrate clearly the importance of practice condition to the

subjects' performance. The present results showed some tenden-cies but more empirical evidence is needed to qualify the condi-tions necessary for the development of automatic processing in driving and its influence on the drivers' situational control.

(20)

References

Fisk, A.D., P.L. Ackerman & W. Schneider: Automatic and Controlled Processing Theory and its Application to Human Factors. In P.A. Hancock (ed.) Human Factors Psychology. North

Holland, 1987.

Harms, L.: Diver's Attentional Responses to Environmental

Variation: A Dual-task Real Traffic Study. In Gale et.al. (eds.) Vision in Vehicles l., North Holland, 1986

Harms, L.: Subjects' Task Performance, Speed and Cognitive Load

as a Function of Task Demands in Driving Simulation. VTI-report, Swedish Road and Traffic Research Institute, 1990.

Harms, L: Variation in Driver's Cognitive Load. Effects of Driving through Village Areas and Rural Junctions. (in press)

1990.

Kahneman, D.: Attention and Effort, Prentice Hall, 1973.

Nordmark, S., H. Jansson, M. Lidström and G. Palmkvist: A Moving Base Driving Simulator with a Wide Angle Visual System. TRB, 64. Annual Meeting. Transportation Research Board. 1985.

Rasmussen, J.: Information Processing and Human Machine Interaction. An Approach to Cognitive Engineering. North Holland 1986.

Schneider W. & R.M. Shiffrin: Controlled and Automatic Human Information Processing I. Detection Search and Attention. Psychological Review , 1, pp. 1-66, 1977.

Shiffrin R.M. & W. Schneider: Controlled and Automatic Human Information Processing I. Perceptual Learning, Automatic Attending and General Theory. Psychological Review, 2, pp. 127-190, 1977.

Shinar, D.: Psychology on the Road. The Human Factor in Traffic Safety. Wiley and Sons, 1978.

Summala, H.: Risk Control is not Risk Adjustment: The Zero risk Theory of Driver Behaviour and its Implications. Ergonomics, 4, pp. 491-507, 1988.

Wickens, C.: Engineering Psychology and Human Performance. A.

(21)

(22)

(23)

(24)