Download as pdf or txt
Download as pdf or txt
You are on page 1of 6

Brain Research Bulletin, Vol. 47, No. 5, pp.

537–542, 1998
Copyright © 1999 Elsevier Science Inc.
Printed in the USA. All rights reserved
0361-9230/99/$–see front matter

PII S0361-9230(98)00088-4

Modelling motion sickness and subjective vertical


mismatch detailed for vertical motions
J. E. Bos* and W. Bles

TNO Human Factors Research Institute, Soesterberg, The Netherlands

[Received 24 January 1998; Accepted 24 January 1998]

ABSTRACT: In an attempt to predict the amount of motion issue of this article. To verify this model, quantitative motion
sickness given any kind of motion stimulus, we describe a sickness data from the literature are used. From all manifestations
model using explicit knowledge of the vestibular system. First, of motion sickness, crosscoupled Coriolis effects and seasickness
the generally accepted conflict theory is restated in terms of a
conflict between a vertical as perceived by the sense organs
are best documented to date. From these, however, sickness num-
like the vestibular system and the subjective vertical as deter- bers in relation to the stimulus parameters are only available for
mined on the basis of previous experience. Second, this con- seasickness. These data are reviewed first.
cept is integrated with optimal estimation theory by the use of Prediction of sickness incidence is classically achieved by
an internal model. If detailed for vertical motions only, the mathematical fits to experimentally obtained data. O’Hanlon and
model does predict typical observed motion sickness charac- McCauley [12] and McCauley et al. [11], for example, subjected
teristics, irrespective the parameter setting. By adjusting the
over 500 subjects to vertical sinusoidal motions in a ship-motion
nonvestibular parameters, the model can also quantitatively be
adapted to seasickness data from the literature. With this con- simulator. Amplitudes ranged from 0.0278 to 0.55 g (root mean
cept, sickness severity hypothetically can also be predicted for square [RMS]) and frequencies from 0.083 to 0.7 Hz. Exposure
other motions, irrespective of their origin and complexity. times were limited to 2 h. In these experiments the percentage of
© 1999 Elsevier Science Inc. vomiting subjects, called the motion sickness incidence (MSI),
was rated as a function of the duration, amplitude and frequency of
KEY WORDS: Motion sickness, Vestibular system, Model. the imposed motion. These experimental data were fit with a
mathematical function, which is the basis of the ISO 2631/3
standard.1 The percentage of sick people after 2 h of endured
INTRODUCTION motion at different frequencies and amplitudes is plotted in Fig. 1.
Building on the conflict theory by Reason and Brand [14], we How the MSI cumulates in the preceding period is shown in Fig.
postulated that motion sickness correlates with a conflict between 2 for three different accelerations and one fixed frequency.
a vertical as perceived by the sense organs and the subjective Griffin describes [7] the well-being of humans under all kinds
vertical as determined on the basis of previous experience [2]. of motion, including vertical motions in the frequency range from
How to relate this conflict with motion sickness severity and 0.1 to 0.5 Hz, in which range well-being is the opposite of motion
implement the concept as a whole in a mathematical model is the sickness.2 Although Griffin uses the term “vomiting incidence” to

* Address for correspondence: Dr. J. E. Bos, TNO Human Factors Research Institute, P.O. Box 23, 3769 ZG Soesterberg; The Netherlands. Fax: 131
346 353977; E-mail: bos@tm.tno.nl
1
The mathematical description fitted though the experimental data is given by
MSI 5 100F~ z a!F~ z t9!.
with

F~ z! 5
1
Î2 p E z

2`
e 2 ⁄2x dx
1 2

where
logt 2 1.46 zt 2 rza loga 2 m a~ f !
zt 5 , z t9 5 ,z 5
0.76 Î1 2 r 2 a 0.47 Î1 2 r 2
r 5 20.75, and m a 5 0.87 1 4.36 log f 1 2.73 log2f. McCauley uses g-units (RMS) for acceleration, minutes for time and Hz for frequency.
2
The description that fits the observed data best according to Griffin is

SE D
1/ 2
T
MSI 5 K a w~t! 2dt
0

537
538 BOS AND BLES

FIG. 1. Motion sickness incidence (%) after 2 h of endured motion versus FIG. 2. Motion sickness incidence versus exposure time at one frequency
frequency and acceleration. Each dot represents an observed average over of 0.25 Hz. (Adapted from Ref. 11.)
20 subjects. (Adapted from Ref. 11.)

fore, creates an almost impossible task. Alternatively, motion


indicate the percentage of vomiting people, we only use the term sickness characteristics may be uncovered by taking into account
MSI. The big difference between the latter description (model) and vestibular characteristics explicitly, thus yielding a more simple
that of McCauley is that complex vertical motions (multifre- approach to the problem in effect. The basic underlying mecha-
quency) are permitted, whereas McCauley only predicts the MSI nism can be modelled, and motion sickness severity then not only
for pure sinusoidal motions. In addition, Griffin covers motion follows for one type of motion but naturally comes out for all sorts
durations up to 6 h, and he also uses data obtained on passenger and combinations of motion, irrespective of their complexity.
ferries. This Griffin model is the basis of the British Standard For that purpose, Bles et al. [2] stated that all situations that
6841. For sinusoidal motions up to 2 h, the McCauley and Griffin provoke motion sickness are characterised by a condition in which
descriptions agree fairly well, although in rare cases they may the sensed vertical, as determined on the basis of integrated infor-
differ by 25% [3]. Within the present framework it is worth mation from the eyes, the vestibular system and the nonvestibular
emphasizing that the MSI shows an optimum at a frequency of proprioceptors, is at variance with the subjective vertical as ex-
about 0.16 Hz. pected on the basis of previous experience. Here motion sickness
Both models, however, lack any vestibular basis, where the should not be confused with vertigo and disorientation, and the
vestibular system plays a crucial role in the incidence of motion statement does not concern secondary or indirect effects. The
sickness [14]. Moreover, the seasickness predictions are restricted sensed and the subjective vertical represent vectors indicating the
to vertical motions, where vessel motions other than the heave can magnitude and direction of our estimates of gravity. To make this
also be provocative [15]. Even for vertical motions only, accurate statement quantitative, we have to resolve the gravitational com-
prediction of MSI in complex broadband motion has shown to be ponent from the gravitoinertial acceleration (GIA), for which we
difficult [9]. On the other hand, we assume that all manifestations rely on Mayne [10]. We also need a way to model the subjective
of motion sickness, including seasickness, and, for example, cross- vertical based on previous experience, for which we use a model
coupled Coriolis effects, have one and the same underlying mech- proposed by Oman [13]. To model the control of active body
anism. To give a predictor function that merely describes all such motions, Oman proposed a concept where the difference between
data then, each of the three translational and three rotational end-organ afferents and afferents generated by an internal model
degrees of freedom of motion should be evaluated separately and that predicts the body state on the basis of motor commands is
in combination with respect to their provocativeness for motion minimised. This difference, or conflict, is conceptually a multisen-
sickness, with an abundancy of possible interactions. Moreover, sory conflict and thus links up with Reason and Brand’s multisen-
the interindividual variability is large, necessitating large numbers sory conflict theory. Our approach introduces a model that mini-
of subjects for reliable statistical analysis. This approach, there- mises a single vector difference between a vertical as constructed

where T is the motion duration, and a w (t) 5 w( f )a(t) is the frequency weighted acceleration in m/s2, with weighing values w:
f
w5 for 0.1 # f , 0.125
0.125
w 5 1.0 for 0.125 # f , 0.25
w5 S D
0.25
f
2
for 0.25 # f # 0.5
w50 otherwise
According to Griffin, K is 1/3 on the average for a mixed (male and female) population. Griffin uses SI-units, that is m/s2 for the acceleration, s for time
and Hz for frequency.
MODELLING MOTION SICKNESS 539

output can be optimized, such that it agrees best with all informa-
tion manifest to the system. Depending on the amount of noise,
this weight factor is low (much noise, low accuracy, slow tracking)
or high (little noise, high accuracy, fast tracking). There are several
possibilities for a mathematical representation of the internal
model. One possibility would be the use of some sort of look-up
table, or “neural store” [13]. Filling in this table, however, is a
FIG. 3. Determination of the sensed vertical.
tedious task, because different kinds of motion should be included.
Probably the most simple way is just to copy the transfer func-
tion(s) of the mechanisms of which the internal model has to
predict the output. That is, the part of the internal model that
generates the subjective vertical can be represented by the transfer
on the basis of multisensory information and another subjective function of the vestibular system, plus eventually some central
vertical as constructed by an internal model instead of the intersen- nervous system processing, like the low-pass filtering to obtain the
sory conflict itself. By adding an appropriate nonlinear conflict vertical.
transfer function followed by a cumulation function, the model is Because it is this subjective vertical that we are actually aware
apt to predict realistic results under a broad range of stimulus of, we can make an estimation of the low-pass time constant
conditions. How this works out for passively endured vertical responsible for its behaviour. If all but the vestibular inputs are
motion is elaborated in detail. excluded,4 this behaviour can be measured by recording verbal
responses, joy-stick manipulations or adjustments of a visual ver-
SUBJECTIVE VERTICAL CONFLICT AND MSI tical during accelerations due to linear motion or centrifugation. If
Sensed Vertical we suppose a large weight value in this case, therefore providing
a fast tracking of the sensed vertical, the behaviour of the subjec-
In stationary conditions the GIA is the gravitational accelera- tive vertical is mainly determined by the low-pass filter of the
tion itself and verticality is thus determined easily by the otoliths. sensed vertical, thus allowing determination of the time constant t.
In determining verticality when the head is moving, we have to Guedry [8] summarises some of the experiments where t appears
deal with the physical impossibility to distinguish between the to be typically larger than 10 s. Mayne [10] even takes this
gravitational acceleration and accelerations due to body move- constant to be 20 s. However, in experiments performed by de
ments. Because gravity is constant and body accelerations are Graaf et al. [6] where they used moderate accelerations of less than
normally of short duration, Mayne [10] suggested that the constant 1 g added to the gravitational component on a sled and in a
gravitational acceleration can after all be estimated by low-pass centrifuge, time constants typically less than 10 s were found,
filtering of the GIA, that is, for all three orthogonal components sometimes even as small as 2 s. Taking an average of about 5 s for
(i 5 x, y or z) of this vertical (v) we can write in Laplace form the time constant of the low-pass filter under normal conditions
seems therefore reasonable.
1
ni 5 a [1]
ts 1 1 i
Conflict
where a represents the vertical acceleration as sensed by the
otoliths. To get a proper three-dimensional result, these compo- We have so far postulated a model that uses motion as input
nents should be earth fixed. If then, in addition, the head is tilted and that yields, as its output, a discrepancy, or conflict, between
(a rotation), a transformation U and its inverse U21 can take care the sensed and the subjective vertical. To mathematically model
of the corresponding adjustment of the coordinate system in which this additional processing, the conflict should be quantified. Be-
the low-pass filtering is executed. This principle is shown in Fig. cause both verticals can be represented as vectors with a magni-
3 and is described more extensively by Bles and Bos [1] and Bos tude and a direction, the magnitude of their vector difference (c) is
and Bles [4].3 The result of this processing is called the “sensed a scalar and suits the purpose well. Then, conditions where both
vertical.” verticals remain parallel but differ in magnitude, like on large
vessels at sea, will result in a conflict. Large conflicts can also
result in conditions where both verticals are of about the same
Subjective Vertical
magnitude but differ in direction, like in crosscoupled Coriolis
Analogous to Oman’s model for controlling body posture [13], effects with ultimately oppositely directed verticals.
a better estimate of the true vertical can be obtained by applying On the one hand, this conflict helps the system to generate a
optimal estimation theory. It is quite conceivable that the brain better prediction of the gravitational component of the GIA, but on
uses a similar strategy. Then an internal model based on previous the other hand the existence of a conflict itself is an unwanted
experience makes a prediction or estimation of the vertical using condition that should be eliminated. With respect to this side
the sensed vertical, visual information, motor commands, nonves- effect, we furthermore assume a quantitative relationship between
tibular proprioceptive information and eventually cognitive inputs. the conflict (c) and the severity of motion sickness without spec-
If a weighted amount of the discrepancy between this and the ulating on the nature of its underlying mechanism. Given this
sensed vertical is fed back to the input of the internal model, the assumption we can then test the model by predicting motion

3
Model results calculated so far appeared to be equal to those calculated by the three-dimensional equivalence of Mayne’s principle [see also 5], which
is given in vector notation:
dv a 2 v
5 2v3v
dt t
4
In general, the input to the vestibular system equals that to many nonvestibular proprio- or graviceptors, and the contributions to the subjective vertical
by these subsystems are therefore indistinguishable.
540 BOS AND BLES

FIG. 5. Subjective vertical conflict model for passive vertical motion.

FIG. 4. Hill function with b 5 0.3 (chosen so only for demonstration


purpose). The parameter n is varied from n 5 1 (. . .), n 5 2 (——), n 5 PREDICTIONS FOR PASSIVE VERTICAL MOTION
5 (– – –) to n 5 20 (–z–z–).
Simplified Model
If for reasons of simplicity we ignore all sources of information
generating the subjective vertical but the vestibular part, the model
sickness severity (e.g., in terms of the MSI, from the model’s reduces to well-quantifiable quantities. This simplification implies
motion input parameters, and see if they (dis)agree with observed that motor commands do not result in a change in body posture and
data). the only way the body can move is then by external enforcement.
Within such passive subjects, or passengers, both the sensed and
Nonlinear Transfer Function the subjective verticals are still determined, despite the lack of
self-control of body motion. If we further restrict to vertical
There is no linear relationship between the MSI and the mag- motions, the sensed and subjective verticals only vary in their
nitude of the vector difference c. People cannot get sicker than magnitude and can simply be resolved by a one-dimensional
sick, a maximum that is reached asymptotically. For this reason, low-pass filter according to Eq. 1. Because the canals are inactive
the McCauley data (Fig. 1) suggest that larger conflicts are trans- and the coordinate system remains stationary, U and U21 from Fig.
formed by some logarithmic function. Small conflicts at the other 3 may be taken equal to the identity matrix. Under these conditions
end can eventually be transformed exponentially. A function that the model looks as shown in Fig. 5, and, if it is right, it will predict
satisfies both needs is the Hill function: seasickness for passengers.
In this simplified model we take the transfer function of the
~c/b! n otoliths (OTO in Fig. 5) to be a unity gain. Then it is not essential
h5 [2]
1 1 ~c/b! n whether a represents the imposed vertical acceleration (in m/s2) or
the otolith organ afferent response (in spikes/s). Both the sensed
For c $ 0 and some values of n, this function is shown in Fig. 4.
vertical (v) and the subjective vertical (v̂) as resolved by an
The benefits of this function are fourfold: it allows for a normal-
internal model (dotted box) are determined by low-pass filters that,
ization (0 # h # 1); the steepness of the function can be chosen
if someone is fully adapted to a certain situation, apply the same
with the parameter n, varying from logarithmic-like (n 5 1) to a
time constant t. Because the verticals are now represented by
step or threshold function (n 5 `); there is an indifferent point b:
scalars, the conflict itself is quantified by the difference of these
h(b) 5 0.5; and for even n the function is nonnegative. For the
scalars. Within the internal model, the integrator takes into account
present study n 5 2 (solid line) serves the purpose that for small
the fact that if the conflict is zero, the subjective vertical should
conflicts h is increasing exponentially, whereas for larger conflicts
remain unaltered instead of becoming zero too. For this reason, the
the function is logarithmic. The value of b can be chosen such that
feedback weight K has the dimension s21. Depending on the
the model fits the experimental data best in a quantitative sense.
amount of inevitable noise added to the different components of
the model, there is an optimum value for this weight for each noise
Cumulation condition. The nonlinear conflict transfer function and the second-
The last step in the model is a cumulation of the quantity h over order cumulation function are included as described above.
time to get the motion sickness severity. However, it does not
cumulate ad infinitum for the same reason as above: Vomiting is Output
considered to be the ultimate manifestation of sickness and the
MSI can therefore not exceed 100%. The sickness severity reaches To compute time responses for any input function, the model
a maximum asymptotically, and it returns to zero again after was implemented in a computer simulation tool. For sinusoidal
cessation of the conflict. Both facts can be taken into account by a motions only, that is when a is replaced by a sinv t, the transfer
leaking integrator. As is evident from Figure 2, at least a second- function offers a simple analyzing tool, and model predictions can
order filter should be chosen, where we ignore the slightly different then be compared more easily with the McCauley and Griffin
cumulations at different acceleration levels. So, when using the results described in the introduction. This output function can be
MSI as the sickness index, at first approximation we take determined based on

P 1 1 1
MSI 5 h [3] c 5 v 2 v̂, with v̂ 5 a and v̂ 5 Kc [4]
~ m s 1 1! 2 ts 1 1 ts 1 1 s
where P explicitly takes into account the maximum percentage of If it is furthermore assumed that for t , 0, the subject is motion-
motion sick people under the given circumstances. less, and if onset effects are ignored and if we take the average of
MODELLING MOTION SICKNESS 541

TABLE 1
MODEL PARAMETERS

21
t 5 5 s, K55s , b 5 0.7 m/s2, m 5 12 min, P 5 85%

the sinusoidally modulated MSI only, a partial fraction expansion


leads to

MSI 5
2 SD
b S S D D
P
2
11
12
t
m
t
1 1 e 2m [5]
E
with

S
E 2 5 a 2 t 2v 2 2 2K t 1
K2
v2
11 D 21
[6]

Table 1 summarizes the parameter choice that gives quantitative FIG. 7. Predicted MSI (%) after 2 h of vertical sinusoidal motion versus
results adequate for the present purpose. frequency and acceleration.

Results
It was verified that MSIs given by expressions 5 and 6 agreed DISCUSSION AND CONCLUSIONS
with results obtained from the simulation program when a simple
sinusoidal input was used. To get some insight in how the model A sense of “one g,” both in direction and amplitude, is essential
works, Fig. 6 shows intermediate results where only the frequency for spatial orientation and control purposes. A major contribution
of a sinusoidal input increases from 0.1 to 0.5 Hz. Here it can be to the determination of the vertical is given by the vestibular
seen that with increasing frequency the sensed vertical decreases in system. Because the otoliths are only capable of sensing the
amplitude while the amplitude of the subjective vertical shows an resultant of all accelerations exerted on them, the gravitational
optimum somewhere about 0.2 Hz. Note also that the phase component can only be inferred from the resultant. Experience can
difference between the sensed and subjective vertical shows an help to improve the accuracy of the estimated vertical. For mod-
optimum, a lag of 180°, in the neighbourhood of this frequency. elling purposes, two verticals are postulated, one as perceived by
Because of these two factors, the conflict and its Hill transform the senses and one as predicted by an internal model. Both verti-
show an optimum about 0.16 Hz. cals are thus generated using multisensory information. We fur-
For pure sinusoidal motion with the frequency and amplitude thermore assume a nonlinear relationship between a single dis-
within the ranges as described by McCauley et al. [11] and a fixed crepancy between these two verticals and the severity of motion
duration of 2 h, MSIs were calculated by Eqs. 5 and 6 and plotted sickness. As opposed to the multisensory conflict theory [14], we
in Fig. 7. Here the optimum in the MSI about 0.16 Hz is even more therefore assume that intersensory conflicts and motion sickness
obvious than it was in Fig. 6. Also the asymptotical MSI increase severity do not correlate. Although detailed for vestibular inputs
versus stimulus amplitude is clear. At frequencies off-optimum, only, our model is applicable also when visual and other proprio-
the exponential increase in MSI versus stimulus amplitude is also ceptive inputs are added. If so, all inputs determine one single
discernible. All these characteristics show a striking resemblance sensed vertical, and the internal model does so for the subjective
with those of the McCauley data in Fig. 1. vertical. Also under these conditions, motion sickness is related to
the magnitude of the vector difference as postulated. If, for in-
stance, visual input is added, both verticals can be determined far
more accurate in normal conditions with concordant information
and Coriolis effects are much less provocative, which indeed is an
actual fact.
For vertical passive motions only, the model slims down into
one dimension. This one-dimensional model is described explic-
itly, and it predicts seasickness as one of the manifestations of
motion sickness. We restricted to relatively short exposure times,
avoiding adaptation, habituation or plasticity of the system. There
are only a few well-interpretable parameters within the model, of
which some are fixed because they control the functioning of the
vestibular system itself, whereas others are fit to adapt the model
quantitatively to the observed MSI data. When we deprived the
system from all but vestibular information, the internal model, in
fact, can only rely on the sensed vertical. From Fig. 5 it can be
inferred that the sensed vertical then lags the GIA, and the sub-
jective vertical in turn lags the sensed vertical. How this works out
FIG. 6. Intermediate model results. Top: input acceleration (. . .), sensed can be seen in Fig. 6. At low frequencies the amplitudes of both
(– – –) and subjective vertical (——). Bottom: conflict (——) and Hill- verticals are about equal, and their phase difference is small,
transformed conflict (. . .). resulting in a small conflict. At intermediate frequencies (about 0.2
542 BOS AND BLES

Hz in the middle of the graph), the amplitude of the sensed vertical ever motion, irrespective its complexity or origin. If visual func-
has decreased due to the low-pass filter, whereas the amplitude of tion and eventually active body motion would explicitly be in-
the subjective vertical shows an increase. More important in this cluded, the predictive power of the model would even be further
frequency range is that the subjective vertical lags the sensed enhanced. With vision included, for example, the model could
vertical up to 180°, resulting in a maximum amount of conflict. At predict simulator sickness as well and predict the possible sickness
higher frequencies, the amplitudes of both verticals have become attenuating effect of an artificial horizon. Here we worked out this
insignificant, resulting in small conflicts again. We thus did not blueprint for seasickness specifically. But any type of motion-
implement a band-pass filter, but mainly due to a lag between the induced sickness is in our opinion just motion sickness predictable
sensed and the subjective verticals caused by a simple low-pass by one and the same model.
filter in the model, the MSI finally shows an optimum in the
frequency range. The location of this optimum is determined by ACKNOWLEDGMENTS
the low pass time constant t and by the weight K. Here we took t
fixed (known from subjective vertical measurements on a sled and Supported by the Royal Netherlands Navy. We express our gratitude to
in a centrifuge) and varied K such that the optimum was equal to Dr. Gresty and the referee for their comments on the original manuscript.
that of the McCauley and Griffin data. The MSI itself at these
optima, however, does not vary.5 It is worth emphasizing that the
value of t did not have to be violated to accomplish this. The role REFERENCES
of the nonlinear Hill transfer function on the conflict c is also
1. Bles, W.; Bos, J. E. Een ruimtelijk orientatiemodel. TNO Human
evident (see bottom of Fig. 6): boosting small conflicts and limit- Factors Research Institute, Soesterberg, the Netherlands, Report A-58;
ing large ones. The need of an exponential increase was indicated 1994.
already by Oman [13], and for that purpose he suggested the 2. Bles, W. Coriolis effects and motion sickness modelling. Brain Res.
square of the amount of conflict. Then, however, the apex of the Bull. 47:543–549; 1999.
MSI versus frequency curve peaks sharply like a tractrix. More- 3. Bos, J. E. Scheepsbewegingen en “motion sickness incidence” (MSI):
over, by only taking the inner product (squaring) of the conflict, Voorspelling van het percentage bewegingszieken op basis van verti-
MSIs will far be overestimated at large conflicts, where it should cale bewegingen. TNO Human Factors Research Institute, Soester-
reach a maximum asymptotically. The Hill transfer function caters berg, the Netherlands, Report A-71; 1995.
4. Bos, J. E.; Bles, W. Simulaties met het ruimtelijk orientatiemodel:
for both shortcomings.
mathematische invulling en numerieke evaluatie. TNO Human Factors
When using the parameters of Table 1, MSIs of Figures 1 and Research Institute, Soesterberg, the Netherlands, Report A-30; 1995.
7 are alike. But because Fig. 1 refers to a group average and Fig. 5. Glasauer, S. Das Zusammenspiel von Otolithen und Bogengangen im
7 refers to the sensitivity of an individual, we have to be careful Wirkungsgefuge der subjektiven Vertikale. Thesis Technical Univer-
with a direct comparison. It may even be doubted whether the sity Munchen, Germany; 1992.
vomiting of an individual is a sensitive measure of his or her 6. de Graaf, B.; Bos, J. E.; Tielemans, W.; Rameckers, F.; Rupert, A. H.;
individual sickness severity. Obviously it is not, and in this case Guedry, F. E. Otolith contribution to ocular torsion and spatial orien-
the MSI should be redefined in terms of a likelihood for emesis. tation during acceleration. Naval Aerospace Medical Research Labo-
The parameter P then represents the maximum likelihood of the ratory, Pensacola Fl. Technical Memorandum 96-3; 1996.
7. Griffin, M. J. Handbook of human vibration. London: Academic Press;
individual under consideration to vomit under any condition. If 1990.
group averages are concerned, these will in general show a wider 8. Guedry, F. E. Psychophysics of vestibular sensation. In: Kornhuber,
curve, mainly due to a spread in frequencies of the individual H. H., ed. Handbook of sensory physiology. Vol VI/2. Vestibular
optima. Because we do not have any statistics on these interindi- system. Berlin: Springer-Verlag; 1974:3–154.
vidual differences, further optimization of the model at this stage 9. Guignard, J. C.; McCauley, M. E. Motion sickness incidence induced
is pointless. by complex periodic waveforms. Aviat. Space Environ. Med. 53:554 –
The present modelling of motion sickness greatly relies on the 563; 1982.
data of O’Hanlon and McCauley [11,12]. Because most of these 10. Mayne, R. A systems concept of the vestibular organs. In: Kornhuber,
H. H., ed. Handbook of sensory physiology IV/2: vestibular system.
data pertain to frequencies above 0.16 Hz, the symmetric depen-
Berlin: Springer-Verlag; 1974:493–580.
dency of the MSI envelope on frequency is questionable. It is 11. McCauley, M. E.; Royal, J. W.; Wylie, C. D.; O’Hanlon, J. F.; Mackie,
reasonable in sofar that at zero frequency there is no motion and R. R. Motion sickness incidence: exploratory studies of habituation,
people (by definition) then do not get motion sick. Nevertheless, pitch and roll, and the refinement of a mathematical model. Human
the very low frequency range of MSI should be explored more in Factors Research Inc. Technical Report 1733-2; 1976.
the experimentation. 12. O’Hanlon, J. F.; McCauley, M. E. Motion sickness incidence as a
It is a challenge to expand the model to allow all possible function of the frequency and acceleration of vertical sinusoidal mo-
three-dimensional movements. Then not only the magnitude of the tion. Aerospace Med. 45:366 –369; 1974.
verticals will be involved but also their direction, and it may follow 13. Oman, C. M. A heuristic mathematical model for the dynamics of
sensory conflict and motion sickness. Acta Otolaryngol. Suppl. 392;
that seasickness not only originates from vertical (heave) motions 1982.
but also from (in itself subthreshold) cumulated crosscoupled 14. Reason, J. T.; Brand, J. J. Motion sickness. London: Academic Press;
Coriolis stimulation due to pitch and roll. It is the power of the 1975.
present model that it predicts motion sickness based on vestibular 15. Wertheim, A. H.; Bos, J. E.; Bles, W. Contributions of roll and pitch
characteristics and therefore gives the MSI as a function of what- to sea sickness. Brain Res. Bull. 47:517–524; 1999.

5
Numerical analysis reveals an MSI optimum versus K. At this optimum dMSI/dK 5 (. . .)dE/dK 5 0. With E given by Eq. 6 it then follows that
dE
dK
52
a
2 S t 2v 2 2 sK t 1
K2
v2
11 D S
23/ 2
22 t 1 2
K
D
v2
50

and thus v 2 5 K/ t at the optimum. Substituting this result in E it follows that E 5 a. Because now E is independent of K, also the MSI according to
Eq. 5 at the optimum is independent of K itself.

You might also like