European Journal of
Applied
Eur J Appl Physiol (1987) 56:163--168
Physiology
and Occupational Physiology 9 Springer-Verlag 1987
Occipital midline EEG and subjective rating of task difficulty as indices of mental task strain Toshio Kakizaki Department of Industrial Physiology, National Institute of Industrial Health, 21-1, Nagao 6-chome, Tama-ku, Kawasaki 214, Japan
Summary. Three measures of mental task strain, namely performance level, occipital midline beta2 (Oz[]2) amplitude and subjective rating of task difficulty, were taken in 34 healthy male students aged 19--24 years as they performed either a paced or a self-paced calculating task for 4 h. The performance levels were so constant throughout the work periods that they were of no use in evaluating mental task strain. The Oz[]2 amplitudes increased significantly with the execution of both tasks, and a significant increase took place during the self-paced task work periods. From this, a critical value of Oz[]2 amplitude for a heavy task strain was estimated to be about 3.5 ~tV (the difference between the values at work and at rest). The subjective ratings of task difficulty rose linearly with work time in both tasks. The subjective task strain attributable to work time was found to exceed the critical level long before the objective measure. In addition, other results suggest that the Oz[]2 activity attributable to task execution is related to the capacity for achieving the task. Key words: Task performance -- Beta-2 amplitude -- Subjective difficulty rating -- Mental task strain
Introduction As noted by Fraser (1983) and Levi (1984), occupational strain, or by their definition "stress", has become an important problem in workers' health care. Recently, interest has been directed towards establishing a specific method for measuring mental work strain (Hopkin et al. 1979; ISO 1981) and towards alleviating heavy strain on workers.
Although some subjective, behavioural and physiological measures for detecting signs of mental strain have been suggested, none of these has yet been developed into a useful means for evaluating mental strain (Moray 1982; Fried et al. 1984). Since the task strain (Kakizaki 1984) in workers during the performance of an intrinsic task must be characteristic of the work strain, it is necessary first to develop some measure for task strain, and then to establish a specific evaluation method for it. Previous studies (Kakizaki 1984; 1985), in which the usefulness of three measures for evaluating mental task strain was examined, have shown that the occipital midline beta-2 (Oz[]2) amplitude and the subjective rating of task difficulty are potentially useful indicators of mental task strain. Moreover, Kakizaki (1984) found that the critical levels for heavy strain were about 8.5 ~tV for Oz[]2 amplitude and about 4.5 for subjective difficulty rating. Since these critical levels were obtained from an experiment in which a paced calculating task was imposed upon the subjects for a short period under different workloads, it may be necessary to corroborate these values with another estimation procedure. The present paper assumes the usefulness of the above measures as indices of mental task strain, and applies them in experiments in which a calculating task is imposed upon the subjects for 4 h under two different conditions.
Methods Preliminary experiment. To ascertain if neck EMG potentials could infiltrate the occipital midline (Oz) EEG during the performance of a calculating task such as that adopted in the present study, use was made of neck EMG recordings obtained (but not published) during previous experiments (Kaki-
164 zaki 1984). In these experiments, 24 healthy male students (19--23 years of age) performed a paced calculating task at eight grades of workload (30, 40, 50, 60, 70, 80, 90, 100%), "workload" being defined as the external task load expressed as a percentage of work capacity (Welford 1978; Kakizaki 1984). The unipolar neck E M G signals were derived from a circular electrode on the skin of the nape of the neck in the midline. The neck E M G and OzEEG signals were recorded simultaneously during rest and task periods, and analysed by the same process as that described below for EEG recordings.
Experiment 1 (Exp. 1). Ten healthy male students, 19--23 years of age participated in Exp. 1, which was carried out between 1 and 6 p.m. Each subject sat at a desk with his eyes open in an air-conditioned room under constant ambient conditions within the ranges of 23.5--26.0~ and 68--79% relative humidity. After practicing the Calculating Work Tester (Shibata Kagaku Ltd, Tokyo, Japan) operation for 15 rain, each subject's work capacity for the calculating task was determined as described previously (Kakizaki 1984). The subject rested for 10 rain, and then performed a paced calculating task at a workload of 70% for 4 h, and again rested for 10 rain. During the rest periods, the subject was asked to relax and do as little thinking as possible. The work time was divided into eight 30 rain task periods separated by 5 rain rests. During the rests, the subject was asked to indicate the subjective task difficulty near the end of each task period on the category scale given below. The average percentage of correct responses to the tasks assigned during each 30 min task period was adopted as a performance standard. A 10 s EEG recording was made from the Oz near the end of each rest and each task period. The experiments on all the students took 10 days.
T. Kakizaki: Indices of mental task strain
EEG analysis. A circular electrode was pasted on the occipital midline (Oz) of the subject's scalp with reference to an electrode on the earlobe (At or A2). A ground electrode was placed in the centre of the forehead. Using the EEG-4217 (Nihon Kohden Ltd, Tokyo, Japan), a 10 s EEG recording was made by setting the frequency characteristics of the amplifier to a time constant of 0.3 s for the low range and to 30 Hz for the high range. The EEG signals were then fed to an MCE-1100 analyser (Nihon Kohden Ltd, Tokyo, Japan) which analyses the EEG signals in five bands (~: 2 4, 0: 4--8, c~: 8--13, [3~: 13--20, [32:20--30 Hz) at intervals of 0.5 Hz by means of a fast Fourier transform, and prints out the average amplitude during each 10 s recording for each of the five bands. Head movement and teeth clenching were kept to a minimum while the E E G was being recorded. In Exp. 1 and Exp. 2, no record was made of the neck EMG~ because the preliminary experiment had shown that neck EMG is unlikely to interfere with the OzEEG during such a calculating task.
Category scale for task difficulty ratin 9. As in previous studies (Kakizaki 1984; 1985), the following category scale, similar to the ratings of perceived exertion (RPE) proposed by Borg (1970), was used to measure mental task strain. 1. Not difficult at all 2. Minimally difficult 3. Somewhat difficult 4. 5. Difficult 6. 7. Considerably difficult 8. Very difficult 9. Extremely difficult
Experiment 2 (Exp. 2). Twenty-four male students 19--24 years of age participated in Exp. 2 at the same time and in the same room as Exp. 1, the ambient conditions being within the ranges of 23.5--25.5~ and 60--75% relative humidity. First, the subject practised operating the Calculating Work Tester for 15 min. After 10 rain rest, the subject performed a selfpaced calculating task for 4 h, and again took a 10 rain rest. The subject was instructed to perform the task continuously at a pace not requiring excessive effort. The work was discontinued briefly every 30 rain so that the subject could grade the task difficulty near the end of each task period, but no effective rest time was taken. The number of attempts and the average percentage of correct responses during each 30 rain task period were used to measure performance. A 10 s E E G recording was obtained three times from the Oz near the end of each rest and each task period and the three EEG amplitude values were averaged. Extreme E E G amplitudes due to head movement or clenching the teeth were discarded, where this occurred only two of the three values were averaged. The experiments on all the subjects occupied a period of twelve days.
Calculatin 9 task. The subject was required mentally to perform the addition or subtraction of two 2 digit numbers displayed on the operating unit of the Calculating Work Tester, and to record the last 2 digits of the answer by pressing two n u m b e r keys. In the case of the self-paced calculating task, the display was changed by pressing the E (enter) key following the number keys. For the paced calculating task, the display time was automatically controlled at regular intervals. The subject was forbidden to repeat a calculation, but was able to correct a mistake by pressing the C (clear) key. A quiet 1 kHz sound signal of 0.2 s duration was sounded when an error was made.
Results
Preliminary experiment The mean amplitudes of the neck E M G for each workload of the paced calculating task are shown in Figs. la and lb: the amplitudes of all the spectral bands during the calculating work are smaller than those at rest (R). Since neck muscle activity must be less under rest conditions, the neck E M G would not then be expected to interfere with the OzEEG, and one can conclude that neck E M G probably does not infiltrate into the OzEEG during the performance of such a calculating task. The figure also shows that the sample regression coefficients of the amplitudes of the [32, [31, oc, and 0 bands are significant if a statistical test is applied only to those obtained during the task periods. Consequently, the correlations between the mean values of these four bands of the neck E M G and the O z E E G obtained during the task periods result in significances as follows; r=0.974"* for [32, r=0.961"* for [31, r=0.970"* for c~, r=0.838" for 0.
T. Kakizaki: Indices of mental task strain
,9[
0>~18
Experiment 1
o
o
W- o ~=16'I
~ z ,
~o_151 ~
165
o
z,
~
z,
detta
[
131~
30 40 50 60 70 80 90 100 Workload (%)
23 21 o
atpha
<~
~20
~j9 o17 ~16 -~15 ~~ 13
o
R
30 40 50 60 70 80 90 100 Workload (%)
Fig. la, b. Mean neck EMG amplitudes for each workload of the paced calculating task. Significance of the sample regression coefficient: t=3.70" for theta, t=9.70"** for alpha, t = 8.82"** for beta-l, t = 8.87"** for beta-2, d.f.=6. * p<0.05, 9* p<0.01, *** p<0.001
"6
T
9~ r-h I I - %tl ' I N87t1 I o89tl
0
0.5
1.0
15 2-0 25 Work period(h)
3.0
35
4.0
The mean percentages of correct responses for the paced calculation for each 30 min task period are shown in Fig. 2. Although the performance levels seemed to decrease in the middle of the work periods, none of the differences between levels are statistically significant. Therefore, no change in the phase of task strain is inferred from this performance measurement. Figure 3 illustrates the mean levels of the Oz[32 amplitudes that were measured once near the end of each rest (R) and each task period. As seen in the figure, the paced task brought about a significant rise in the [32 level. This [32 level stayed constant throughout the work periods and returned to the pre-task level during the 10 min rest period. These constant [32 amplitude levels suggest that the level of task strain is constant throughout the work periods. The subjective ratings of task difficulty rise linearly with the elapsed work time (Fig. 4). The sample regression coefficient is highly significant as shown in the figure. Taking account of the consistent workload, the performance levels, and the [32 levels throughout the work periods, this result suggests that the subject correlated work time with task difficulty. Therefore, the subjective task strain becomes greater in proportion to the total work time. Referring to the critical level (about 4.5) of subjective rating for a heavy strain described in the previous paper, the task strain on the subjects may be judged to increase when the work time exceeded 2.0 h.
Fig. 2. Mean percentages of correct responses and their standard errors for each 30 min period of the paced calculating task
7
~6
t = 6.67"**, d.f.=6
I
l_
3.5
4.0
"13
._c ~4 >
paced calculating task
RO
i
i
i
05
1.0
1.5
i
k
2.0 2.5 Work period (h)
i
i
3.0
35
S.E.I
J
4.OR
Fig. 3. Mean Oz~32amplitudes and their standard errors during the rest period (R) and during the paced calculating task
CO
0
05
1.0
15 2.0 2.5 Work period (h)
3.0
Fig. 4. Relationship between the length of the work period of the paced calculating task and the subjective rating of task difficulty. *** p<0.001
166
T. Kakizaki: Indices of mental task strain
Experiment 2 The mean performance levels of the self-paced calculation for each 30 min task period are shown in Fig. 5. Neither of the performance parameters change throughout the work periods. From these measurements, therefore, no change in the phase of task strain can be determined throughout the work periods. Figure 6 illustrates the mean levels of the Oz[32 amplitudes that were measured three times near the end of each rest (R) and each task period. The [~2 level rises significantly with execution of the self-paced task and remains constant up to 3.0 h, and then rises sharply. This change is statistically significant (p < 0.05). Thus task strain clearly becomes greater when work time exceeds 3.0 h. From this figure, the mean critical level of the [32 amplitude for heavy task strain can be estimated to be about 7.5 gV. Although this level is lower than that shown in the previous paper (about 8.5 pV), the mean resting level (about 4.0 lxV) is also lower than that reported in the previous paper (about 5.0 lxV). Therefore, the mean difference 400[
z
340t [ a; 95r
Y-~
[
l'l
0
05
1.0
]Jl
;
1.5 2.0 2.5 Work period (h)
I
3.0
3.5
4.0
Fig. 5. Mean performance levels and their standard errors for each 30 rain period of the self-paced calculating task
in [32 amplitude from the resting value is almost equal (about 3.5 ~V) in both cases. The same value was also obtained from the results shown in Kakizaki (1985). Accordingly, this evidence shows that this value (about 3.5 ,ttV) is quite reliable as a critical value for a heavy task strain. Moreover, the [32 amplitudes have little relation to work time, unlike the subjective ratings of task difficulty. As in the results of Exp. 1, the subjective ratings of task difficulty regress linearly with work time (Fig. 7). The sample regression coefficient is somewhat higher than that estimated in Exp. 1. Therefore, this confirms that, over the course of the work period, the subjects have increasing difficulty in performing the task, and that the subjective task strain becomes greater. According to the critical rating level (about 4.5) represented in the previous paper, it is suggested that subjective task strain becomes greater when work time exceeds 1.0 h. Thus, it can be seen that the subjective strain of a prolonged mental task exceeds the critical level long before the objective strain becomes critical. These results lead to the following conclusions. Neither the percentage of correct responses nor the attempted number of calculations is of use in evaluating mental task strain. The difference between the Oz[32 amplitudes at work and at rest can be used as a critical value for the strain of a heavy mental task. The subjective rating of task difficulty is associated with mental task strain insofar as work time is concerned.
Discussion Because performance levels hardly changed, even through the long work periods, they would be of little use in measuring task strain. In only one previous case (Kakizaki 1984), in which workloads
s
7 t= 7.51"**, d.f.: 6 y=0.63x +3.63
e "O
T
"6
"0
T / ~ ' f
/
.~5 E ell ,~5 N O
4 3
{
self-paced calcutating task i
R0
i
0.5
i
1.0
i
~
i
1.5 2.0 25 Work period (h)
i
3.0
i
3.5
i
4.0R
Fig. 6. Mean Oz[32 amplitudes and their standard errors during the rest period (R) and during the self-paced calculating task
0
0.5
1.0
1.5 2D 2.5 Work period (h)
3.0
35
4.0
Fig. 7. Relationship between the length of the work period of the self-paced calculating task and the subjective rating of task difficulty. *** p < 0.001
T. Kakizaki: Indices of mental task strain
were set proportional to the capacity of each subject, could the percentage of correct responses be used as a performance measure of mental task strain. It is certain that the CNS is greatly involved in the occurrence of strain when performing a mental task, and scalp EEG analysis is the only noninvasive way of revealing transient activity changes in the CNS. However, no studies, except that by Ullsperger and Gille (1982), have been devoted to applying E E G analysis to the measurement of mental task strain. Although they suggested a relationship of the scalp 0 range to the calculating process during a mental arkhmetic task, our previous studies (Kakizaki 1984; 1985) showed that the [32 amplitude at Oz, among other bands, could be associated with both the quality and quantity (workload) of the mental tasks. Accordingly, only the data on the Oz[~2 band were given in the text. Furthermore, the present study has confirmed that the difference (about 3.5 ~V) between the Oz[32 amplitudes at work and at rest is reliable as a critical determinant of a heavy mental task strain. As for the upturn in [32 level towards the end of the work periods seen in Fig. 6, the following explanation is offered: the cortical functions of the subjects might begin to decline when work time exceeds 3.0 h, but this decrease could be compensated for by an increase in cortical activity, performance levels remaining unchanged throughout the work periods, the increase in cortical activity being manifested by the rise in 132 seen at the 3.5 and 4.0 h points of the work periods. As Moray (1982) noted, it is certain that many factors are involved in subjective feelings of task difficulty. Borg and his colleagues (1978) obtained high correlations between subjective and objective measures of task difficulty. Kakizaki (1984; 1985) also found that the subjective rating of task difficulty was closely related to the quantity (workload) of the calculating task, but was only weakly related to the nature of the mental task. These reports also show that the subjective rating of task difficulty is associated with time pressure in performing the calculating task. On this subject, Phillip et al. (1971) have also suggested that the subjective feeling of difficulty is essentially dependent on time stress in task performance. Furthermore, the present study shows that the subjective rating of task difficulty corresponds to the work time of the calculating task. In this way, the factors concerning the changes in Oz[~2 amplitudes and the subjective ratings of task difficulty during performance of
167 r = 0.65*** y= 0.75x- 1.25
o o oO
O
O
O
~3 ~
"6~2
o
0
0
0
E .col
z-0
4 5 6 Work capacity (s/problem)
7
Fig. 8. Correlation between the difference between the Oz[32 amplitude and its resting value and the work capacity for achieving the calculating task. *** p < 0.001
mental tasks have been determined. Consequently, these two parameters will be useful as indices in evaluating mental task strain. However, the critical values of these measures obtained in this series of studies would be effective in determining heavy mental task strain in only some groups of healthy Japanese male students, because there appears to be some differences in EEG activity and in strain sensitivity between sexes, between age groups, and perhaps between races.
Supplement. In a further experiment, the 24 subjects who participated in Exp. 2 rested for 5 min and then performed the paced calculating task at a rate of one problem every 8.6 s for 10 min. The Oz~2 amplitude was measured once during the rest period and three times during the task period. A significant inverse correlation was observed between the difference between the working and resting Oz[32 amplitudes and work capacity (Fig. 8). This suggests that Oz[32 activity attributable to task execution is dependent on the capacity for achieving the task.
References
Borg G (1970) Perceived exertion as an indicator of somatic stress. Scand J Rehabil Med 2:92 98 Borg G (1978) Subjective effort in relation to physical performance and working capacity. In: Pick H (ed) Psychology from research to practice. Plenum press New York, pp 333--361 Fraser TM (1983) Human stress, work and job satisfaction. ILO Occup Safety Health Ser 50:1--67 Fried Y, Rowland KM, Ferris GR (1984) The physiological measurement of work stress: a critique. Personnel Psychol 37:583--615
168 Hopkin VD, Parks DL, Rohmert W, Rault A, Soede T, Schmidtke H (1979) Final report of application group. NATO Conf Ser 3:469--495 International Organization for Standardization (1981) Ergonomic principles in the design of work systems. ISO 63851981(E): 1--4 Kakizaki T (1984) Relationship between EEG amplitude and subjective rating of task strain during performance of a calculating task. Eur J Appl Physiol 53:206--212 Kakizaki T (1985) Evaluation of mental task strain based on occipital beta activity and subjective rating of task difficulty. Eur J Appl Physiol 54:466--470 Levi L (1984) Stress in industry. ILO Occup Safety Health Ser 51:1--61
T. Kakizaki: Indices of mental task strain Moray N (1982) Subjective mental workload. Human Fact 24:25--40 Phillip V, Reiche D, Kirchner J (1971) The use of subjective rating. Ergonomics 14:611--616 Ullsperger P, Gille HG (1982) Investigation of mental work by means of event-related brain activity. Z Ges Hyg Grenzgeb 28:441--444 Welford AT (1978) Mental work-load as a function of demand, capacity, strategy and skill. Ergonomics 21:151-167
Accepted October 22, 1986