J. Biomedical Science and Engineering, 2010, 3, 677-683 JBiSE
doi:10.4236/jbise.2010.37092 Published Online July 2010 (http://www.SciRP.org/journal/jbise/).
Published Online July 2010 in SciRes. http://www.scirp.org/journal/jbise
A pilot study of a novel pulsatile flow generator using large
collapsible bladder
Ponangi Udaya Prashant1, Nagaraj Balasubramanya2
1Institute G-5, Sneha Sindhu Apartments, Kavalabyrasandra, Bangalore;
2Department of Civil Engineering, M.S. Ramaiah Engineering College, Bangalore, India.
Email: udayaprashant_p@yahoo.co.in; jhrumbaa@gmail.com
Received 30 April 2010; revised 21 May 2010; accepted 22 May 2010.
Background: There are different experimental mod-
els avialable for generating pulsatile flow in labora-
tory and study their heamodynamic effects on blood
vessels. We aim to produce a novel pulsatile flow gen-
erator utilizing a large collapsible rubber bladder
and the phenomenon of fluid structure interactions
occurring in a specially designed flexible tube ar-
rangement. Mehtods: Water enters from a reservoir
above into a large collapsible bladder made of rubber
which opens into ‘U’ shaped tube made of flexible
material and held by non rigid structures. As liquid
starts flowing the distal end of collapsible bladder
collapses under the negative atmospheric pressure
generated inside closing the mouth of ‘U’ shaped tube
and produces pulsatile flow. Resuts: The frequency of
pulsations, pressure fluctuations and velocity profile
resemble that of in vivo blood flow. As the flow en-
tering into collapsible bladder increases the fre-
quency of pulsatile flow decreases and also when
height of the collapsible bladder from the ground was
changed. The whole cycle of alternate collapse/expan-
sion of collapsible bladder with generation of pulsa-
tile flow continue indefinitely as long as there is
enough water in reservoir and vertical gradient to
sustain the flow. Conclusions: The pulsatile flow so
produced has many of the characteristics of physio-
logical blood flow and can be used to study mecha-
nisms of various cardiovascular diseases in labora-
Keywords: Collapsible Tubes; Buckling; Pulsatile Flow
Generator; Unsteady Flow; Fluid Structure Interactions
Cardiovascular diseases are major causes of death throu-
ghout the world, yet the mechanisms of these diseases,
especially atherosclerosis is poorly understood. The most
important aspect of the development and progression of
cardiovascular diseases is the role of pulsatile flow in
hemodynamics of cardiovascular diseases like athero-
sclerosis [1,2]. Pulsatile flow induces complex shear
stress on walls of vessels and is basis for many athero-
sclerotic diseases like aneurysms and stenotic lesions [3].
A number of attempts have previously been made to
simulate pulsatile human arterial flow [4-7].
Hopmann and Liu described a pump that uses an elas-
tic chamber squeezed by a cam, with pressure-actuated
check valves [4]. The use of a gear pump controlled by a
closed-loop servosystem was reported by Issartier et al.
[4,7]. Kiyose et al. have described a mechanical piston
pump system for simulating peripheral arterial flow [4].
Petersen described a system using two gear pumps, one
to generate forward flow and the other to generate re-
verse flow [5]. A microcomputer regulated a pneumatic
control valve so that the total flow at any instant could
be controlled. Errikson et al. used a microcomputer con-
trols a motor which in turn drives a piston pump to gen-
erate the flow [6]. However, the system appears to lack
the flexibility to generate different waveforms.
Many of the existing pulsatile flow generators are
complicated, consume significant power and require
computers to function [4-7]. They are also expensive and
none of them are based on principle of flow limitation
occurring in collapsible tubes. Many interesting phe-
nomena are observed like flow limitation and the pro-
pensity to develop large amplitude self-excited oscilla-
tions during study of flows in collapsible tubes [8].
Flow induce collapse is very common in many phy-
siological situations like regulation of blood vessel cali-
ber, generation of murmurs, snoring sounds, wheezing in
airways, micturation etc [9]. The collapsible tubes buck-
les and collapses at the distal most point of tube as the
internal pressure is least between entry and exit end due
to viscous pressure drop [10]. Buckled vessels are very
flexible and even small changes in fluid pressure can
induce large changes of their cross-sectional area [11].
P. U. Prashant et al. / J. Biomedical Science and Engineering 3 (2010) 677-683
Copyright © 2010 SciRes. JBiSE
Many of the physiologically observed phenomena de-
scribed above can be reproduced in laboratory experi-
ments using the ‘Starling Resistor’, shown below [9]
(Figure 1). Inside a pressure chamber, a thin-walled
elastic tube (typically made of latex rubber) is mounted
on two rigid tubes. Fluid (typically air or water) is driven
through the tube, either by applying a controlled pres-
sure drop p-entry and p-exit, between the ends of the
rigid tubes or by controlling the flow rate Q. At suffi-
ciently large Reynolds numbers, the system buckles ax-
isymmetrically and readily produces self-excited oscilla-
tions [9,11].
The present experiment may be considered as extreme
modification of this basic “Starling Resistor”. There are
certain practical applications described for pulsatile flow
produced by rapid flutter of collapsed tube. Microfiltra-
tion performance of Bentonite suspension was greatly
enhanced when a pulsatile flow generated from rapid
flutter of collapsible tube was used [20]. But the pulsa-
tile flow produced by Waxing et al. was different from
the present experimental setup both in mechanism and
the pulsatile waveform generated [20].
Figure 1. The classic “Starling Resistor”. The figure shows
how classical Starling Resistor works. The liquid enters the
collapsible tube at A with velocity V which is enclosed in a
pressurized chamber (pext) and travels distance L and starts
buckling at B. The diameters of entry and exit pipes are same
as that of collapsible tube diameter R0 and all the pipes are
aligned in one straight line.
The cyclical opening and collapse of a large rubber
bladder used here produces large amplitude low fre-
quency pulsatile flow from continuously flowing liquid
unlike previous experiments by a very unique arrange-
ment. The flow very much resembles that occurring in-
side cardiovascular system of living organisms having
characteristics of sinusoidal rise and fall with sharp spike
in between during collapse of bladder.
The experimental model can be demonstrated by con-
structing a simple device using commonly available ma-
terials (Figures 2 and 3). To a source water that has very
negligible head is connected to collapsing rubber bladder.
This rubber bladder is highly elastic and the lower end is
connected to flexible thin garden hosepipe made of syn-
thetic rubber or PVC. This flexible rubber tube hangs
freely after taking initial ‘U’ shaped curve. The distal
end is open to atmosphere and all the connections are
airtight. If the vertical height from the tap to balloon is
less than from ground to the balloon (h1 > ho) and during
certain range of flow rates it exhibits an interesting phe-
Figure 2. Simple sketch of the novel pulsatile flow generator.
R is reservoir; P are connecting pipes from reservoir to col-
lapsing bladder Cb. The collapsible bladder is connected to ‘U’
shaped flexible tube Fp which is in turn connected to long rigid
distal pipe Dp.
P. U. Prashant et al. / J. Biomedical Science and Engineering 3 (2010) 677-683
Copyright © 2010 SciRes. JBiSE
Figure 3. Schematic diagram showing distal flexible ‘U’ shap-
ed tube and its attachements.
nomenon of alternate collapse (buckling) and opening,
along with generation of pulsatile flow of water. The
flow of water or any liquid entering into the collapsible
bladder is regulated by a valve and the flow rate can be
measured by flow meter attached to it. Pressures are re-
corded at two points, one at the entry at the collapsible
bladder and other at the end of ‘U’ tube with help of
Phlips portable MP20 monitor and electronic pressure
transducer system. Average velocity of flow was calcu-
lated by dividing flow rate with area of connecting pipe
to collapsible bladder.
Simultaneous pressure was recorded at the proximal
and distal end of balloon. The pressure at the entry of the
collapsing bladder remains fairly constant at a –ve pres-
sure of -6 to -12 cm of water with slight fluctuations
during the entire cycle.
The distal end of collapsible bladder exhibits rapid
and wide range of fluctuations which can be explained
by very strong fluid structure interactions happening due
to sudden stoppage of flow. This is because the ‘U’
shaped tube is made of flexible plastic pipe and also the
supporting structures are non rigid. An attempt has been
made to graphically record the fluid velocity flow with
help of doppler at the distal end of collapsible bladder
but the flow in this region is very turbulent whereas
other regions it is smooth.
Below are experiments conducted using above model
which study the relationship rate of fluid flowing and the
frequency of oscillations produced.
3.1. About Collapsible Bladder
The dimensions of the balloon in fully expanded state,
which assumes a geometric shape of ellipsoid are: -
Horizontal diameter is = 11.5 cm
Vertical diameter =19 cm
Volume of fully expanded Balloon is = 1600 cc
The thickness of the rubber used in the rubber is =
0.75 mm
3.2. About Distal ‘U’ Shaped Tube
The distal ‘U’ shaped tube is made of relatively flexible
PVC pipe with following characteristics
The length of ‘U’ tube is = 60 cm
Height of U tube is from collapsible balloon to distal
rigid tube is = 30 cm
Vertical height of distal rigid tube from flexible U
shaped tube is = 140 cm
Inner Diameter of U shaped flexible tube is = 12 mm
Thickness of U shaped tube is = 2 mm
Inner Diameter of distal rigid tube is = 11 m
Thickness of distal rigid tube is = 1.8 mm
The angle of ‘U’ tube which makes to vertical is be-
tween 80°- 70
Similar experiments were conducted in much small
scale using small sized balloons and tubes which worked
very well, though data was not recorded, due to low
volume flows and small sized equipment used.
The dimensions of the pipes and bladder are so cho-
sen to resemble those of small sized human beings car-
diovascular system. The flows used are comparable to
normal physiological range of blood flow. The oscilla-
tions produced by this experiment also are in range of
normal heart beating frequency. The basic experiments
illustrate how simple parameter of flow rate can be var-
ied to alter the frequency of collapsing bladder (Tables
1 and 2). Also at constant flow conditions and when all
other parameters are kept constant of varying height of
the bladder from ground affects the frequency of col-
lapsing bladder (Tables 3 and 4). Thus by altering these
two simple parameters any desired flow condition can
be simulated in laboratory. The effects of pipe’s stiff-
ness and diameters of tubing can be studied and their
flow patterns analyzed when all other parameters are
held constant. Also side branches can be attached to
distal ‘U’ tube or rigid pipe which represent the
branches of main vessels and their flow characteristics
during pulsatile flow can be studied.
Here fall of pressures is sharp unlike the rise which is
observed during pressure recordings inside large arteries
and the pressures recorded here are in negative range
which is reverse of that seen normally inside blood ves-
sels. But the pressure fluctuations resemble closely to
those recorded inside vascular compartment invasively.
P. U. Prashant et al. / J. Biomedical Science and Engineering 3 (2010) 677-683
Copyright © 2010 SciRes. JBiSE
Table 1. Relationship of change in frequency of oscillations with change in flow rate when height of bladder from ground is constant.
Flow rate in
Freq of collapse
per min
Pressure at inlet
in mm Hg
Max pressure at
outlet in mm Hg
Min pressure at
outlet in mm Hg
Range of
fluctuations in
mm Hg
velocity of flow
in m/sec
146 108 -7 -116 -154 37 1.2
200 100 -9 -122 -158 37 1.8
250 74 -10 -146 -196 50 2.2
278 80 -7 -94 -150 56 2.5
303 62 -9 -112 -150 38 2.5
332 52 -7 -100. -127 26 2.6
435 30 -9 -146 -176 30 2.9
Table 2. Similar experiment as above when the vertical height of ‘U’ tube from Ground is changed (h = 90 cm)
Flow rate in
Freq of collapse
per min
Pressure at inlet in
mm Hg
Max outlet pressure
in mm Hg
Min outlet pressure
in mm Hg
Range of pressure
fluctuations in mm Hg
Average velocity of
flow in m/sec
85 128 -5 -50 -70 20 0.75
108 122. -5.5 -65 -75 10 0.95
187 86 -6 -59 -70 11 1.6
188 82 -7 -54 -67 13 1.6
218 75 -5.5 -60 -67 7 1.9
228 72 -7 -34 -56 22 2.0
250 70. -7 -33 -59 26 2.2
286 72 -4.5 -63 -79 6 2.5
298 62 2.5 -67 -72 5 2.6
342 45 -5.5 -60 -75 15 3.0
362 48 -6 -52 -67 15 3.2
Table 3. Variation of frequency with varying vertical height at constant flow rate of = 278 ml/sec average velocity of flow is 2.45
Vertcial height from
ground in cm Freq of collapse per min Pressure at inlet in cm
of ater
Outlet Max pressure in
mm Hg
OutletMin pressure in
mm Hg
Range of pressure
fluctuations in mm Hg
36 76 -7 -63 -100 37
96 66 -8 -68 -108 40
130 58 -7 -116 -154 38
142 70 -7 -94 -130 37
162 57 -6 -2 -4 2
In our study instead of classical models of ‘starling re-
sistor [9,11] we use large balloon such that its diameter
is greater than 3 times the inlet rigid tube diameter. Also
the exit is not along the straight rigid tube but through
the lower side of balloon such that when the balloon
collapses the opposite wall impinges on the exit end
completely occluding it and the flow stops (The path of
fluid in the balloon is not straight as in ‘starling resistor’
but affectively changed to 90’).
P. U. Prashant et al. / J. Biomedical Science and Engineering 3 (2010) 677-683
Copyright © 2010 SciRes. JBiSE
Table 4. Variation of frequency with change in height at constant flow rate of 240 ml/sec & average velocity of flow is 2.14 m/sec.
Vertical height from
ground in cm
Frequency of col-
lapse per min
Pressure at inlet in
mm Hg
Max Pressure at
outlet in mm Hg
Min pressure at outlet
in mm Hg
Range of pressure
fluctuations in mm Hg
30 96 -2 -3 -4 1
50 92. -4 -1 -21 20
96 91 -4 -41 -47 27
130 86. -5 -60 -70 30
150 98 -5 -82 -96 20
165 58 -4 -80 -90 14
There is another major deviation from classical ap-
proach the distal end is not exactly ‘rigid straight tube’
but a ‘U’ shaped flexible tube with mobile supports. The
purpose of it was to make the distal end flexible, so that
during the collapse, it executes rotational motion and
also swings forward. The axial and radial motion of dis-
tal ‘U’ shaped pipe due fluid solid coupling can be de-
scribed by series of differential equations [13]. As the
supports holding the distal ‘U’ tube are elastic and the
‘U’ tube are itself flexible, due to strong fluid structure
interactions occurring inside the system, the ’U’ tube
gets displaced [14] and the collapsed rubber bladder
opens up and fills with liquid flowing continuously from
reservoir above. Also some doubts about the flow re-
sembling certain kinds of slug flow have to be dismissed
totally as it is purely a monophasic flow and experimen-
tal observations revealed that the presence of air or any
gas inside the tubes does not produce pulsatile flow. All
the connections have to be airtight and very little air is
sucked inside the tubes at the lower end of distal rigid
tube (7 of Figure 2).
The basic mechanism is that once flow of water starts
the positive hydrostatic pressure acting inside collapsible
bladder proportional to height ho (ho*d*g) becomes
negative lateral pressure when flow is fully established
flow and this negative pressure (-h1*d*g) is proportional
to height h1 [17]. Also this negative pressure is the com-
pressing transmural pressure acting on the collapsible
bladder which causes it to buckle axisymmetrically at its
distal end [12]. This sudden collapse occludes the mouth
of ‘U’ shaped tube and stops the flow instantly creating a
large –Ve waterhammer wave distally with strong junc-
tion coupling [15,18].
The pressure wave form so produced can be analyzed
by combined effects of pressure surges and pipe motion
with help of method of characteristics and finite element
methods respectively [16,18]. This strong junction cou-
pling occurring at U shaped tube has sufficient force to
overcome the collapsing force of the bladder at the
mouth of ‘U’ tube {(h1- h0)*d*g*A} and causes the
Pressure recorded distal to collapsible bladder when all conditions constant
13 2
23 5
Time in sec
Pressure in mm Hg
Typical pressure fluctuations recorded at the mouth of ‘U’ tube when
all the flow conditions are held constant. Here flow rate was = 280
ml/sec, vertical height of ‘U’ tube from ground is 140 cm and fre-
quency of collapse is = 84/min
Figure 4. Graph showing pressure readings with time.
bladder to open up with re-establishment of flow, thus
producing pulsatile flow [18]. The total strain energy
which has developed due to the fluid structure interac-
tions can be calculated knowing bulk modulus of flow-
ing liquid and elastic modulus of pipe as well as its mass
density per unit length [17].
From these basic experiments it can be concluded that
as the flow rate increases the frequency of oscillations
progressively reduce and the average velocity of flow
increases. The pressure at the inlet of balloon remains
relatively constant with minor fluctuations per collapse
of bladder. The outlet pressure fluctuates widely and the
range is 30 – 40 mm Hg which can be comparable to
pulse pressure of human cardiovascular system.
Definite relationship between pressure and flow rate
could not be established one reason being relatively in-
sensitive instruments used to record a very dynamic and
P. U. Prashant et al. / J. Biomedical Science and Engineering 3 (2010) 677-683
Copyright © 2010 SciRes. JBiSE
unstable pressure tracing. The minimum flow where it
functions is where the flow inside the pipe work be-
comes closed channel laminar flow [17].
That means the whole of pipe area is in contact with
flowing water. At flow rates below this it is open channel
flow. Also the velocity of flow is 0.5 -2 m/sec which is
very low. At flow velocity > 3 m/sec or at higher Rey-
nolds numbers the efficiency markedly falls and little
higher flow velocity it doesn’t behave as pulsatile bal-
loon but it simply expands and remains in that position
and water starts flowing passively. When diameter of U
and distal pipe is increased the lower and upper flow rate
limit where these pulsations are observed both are in-
creased and also simultaneously the range at which op-
erates efficiently are increased correspondingly.
Yannick et al. performed in a specially designed mo-
ckup simulating the heart’s behavior (the Dual Activa-
tion Simulator) to understand pulsatile flow inside heart
like chambers in terms of mechanical behavior to under-
stand cardiovascular diseases in laboratory without re-
sorting to animal models or patients [19].
The gear pumps used by Issartier et al., Petersen et al.,
etc produce cavitation and damage to suspended parti-
cles [4-5] whereas the present model the occurrence of
cavitation is very less and can be easily prevented by
altering flow conditions. There is also no damage to
blood particles as it does not involve rotating metal
Peristaltic pumps which were developed later to pre-
vent particle damage of gear pumps suffer the drawback
of production of only limited subset of waveforms even
if they are computer controlled and the systems lack
flexibility to operate under wide flow conditions [4,6,7].
Also generating continuous flow from peristaltic pumps
is difficult. The present model operates under wide range
of flow conditions which can be produced by simply
opening the inflow valve or raising the height of col-
lapsible bladder from ground. Besides pulsatile flow,
continuous can be easily produced simply by halting the
oscillations of ‘U’ shaped tube and no further additional
arrangement is required.
Microcomputer controlled piston pumps are very
complicated, difficult to set up and cumbersome to oper-
ate [4-7]. The present apparatus offers an easier solution
of producing pulsatile flow in laboratory by using read-
ily available cost effective materials and can save lot of
expenses and power.
The collapse of collapsible bladder resembles some-
what superficially of “beating heart” and waveforms
produced are very close to invivo flow including sudden
surges of pressures seen during valve closure. The pulsa-
tile flow so produced can be studied for progression of
diseases like atherosclerosis, effects of vessel wall stiff-
ness and other properties for development of vascular
diseases like hypertension and atherosclerosis.
Initial preliminary studies the apparatus works even
when glycerol water having viscosity similar to blood
are used as flowing liquid instead of water. However this
is a pilot study and needs to be validated by further re-
search and all the variables which affect the flow have to
be studied more exhaustively with better instrumentation
and advanced mathematical approach, before real useful
conclusions drawn.
A novel and cost-effective pulsatile flow generator can
be constructed using a large collapsible bladder con-
nected to a flexible ‘U’ shaped tubing system and the
pulsatile flow so produced can be used as simulation
model for studying in vivo blood flow.
PUP developed the experimental setup from beginning,
conducted the experiments and drafted the manuscript.
NBS provided technical support for conducting experi-
ments in MSRIT lab and suggested the utility of using it
as simulation model for cardiovascular system.
I thank MSRIT institute for providing me laboratory support for my
research especially Dept of Civil engineering and the Principal of
[1] Afshin, A.-B., Mohammad, T.-S., Nasser, F., et al. (2008)
A new system to analyze pulsatile flow characteristics in
elastic tubes for hemodynamic applications. American
Journal of Applied Sciences, 5(12), 1730-1736.
[2] Chatzizisis, Y.S. and Giannoglou, G.D. (2006) Pulsatile
flow: A critical modulator of the natural history of ath-
erosclerosis. Electronic Publication, 67(2), 338-340.
[3] Finol, E.A. and Amon, C.H. (2001) Blood flow in ab-
dominal aortic aneurysms: Pulsatile flow hemodynamics.
Journal of Biomechanical Engineering, 123(5), 474-484.
[4] Law, Y.F., Cobbold, R.S.C., et al. (1987) Computer-con-
trolled pulsatile pump system for physiological flow si-
mulation. Medical & Biological Engineering & Comput-
ing, 25(5), 590-595.
[5] Petersen, J.N. (1984) Digitally controlled system for re-
producing blood flow waveforms in vitro. Medical and
Biological Engineering and Computing, 22(3), 277-280.
[6] Eriksson, A., Persson, H.W. and Lindstrom, K. (2000) A
computer-controlled arbitrary flow wave form generator
for physiological studies. Review of Scientific Instru-
ments, 71(1), 235-242.
[7] Holdswoth, D.W., Rickey, D.W., et al. (1991) “Com-
puter-controlled positive pump for physiological flow
P. U. Prashant et al. / J. Biomedical Science and Engineering 3 (2010) 677-683
Copyright © 2010 SciRes. JBiSE
simulation,” Medical & Biological Engineering & Com-
puting, 29(4), 565-570.
[8] Kamm, R.D. and Shapiro, A.H. (1979) Unsteady flow in
a collapsible tube subjected to external pressure or body
forces, Journal of Fluid Mechanics, 95(1), 1-78.
[9] Heil, M. and Jensen, O.E. (2003) Flows in deformable
tubes and channels -- Theoretical models and biological
applications. Chapter 2 of: flow in collapsible tubes and
past other highly compliant boundaries. Pedley, T.J. and
Carpenter, P.W., Eds., Kluwer, Dordrecht, Netherlands
Heil, 15-50.
[10] Heil, M. (1996) The stability of cylindrical shells con-
veying viscous flow. Journal of Fluids and Structures,
10(2), 173-196.
[11] Heil, M. and Pedley, T.J. (1996) Large post-buckling
deformations of cylindrical shells conveying viscous flow.
Journal of Fluids and Structures, 10(6), 565-599.
[12] Heil, M. (1998) Stokes flow in an elastic tube -- A large-
displacement fluid-structure interaction problem. The In-
ternational Journal for Numerical Methods in Fluids,
28(2), 243-265.
[13] Tijsseling, A. (2007) Water hammer with fluid-structure
interaction in thick-walled pipes. Computers and Struc-
tures, 85(11-14), 844-851.
[14] Heinsbroek, A.G.T.J. and Tijsseling, A.S. (1994) The
influence of support rigidity on waterhammer pressures
and pipe stresses. Proceedings of the Second Interna-
tional Conference on Water Pipeline System and BHR
Group, Edinburgh, 17-30.
[15] Wiggert, D.C. and Tijsseling, A. S. (2001) Fluid tran-
sients and fluid-structure interaction in flexible liquid
filled piping. ASME, 455-481.
[16] Tijsseling, A.S. and Heinsbroek. A.G.T.J. (1999) The
influence of bend motion on waterhammer pressures and
pipe stresses. Proceedings of the 3rd ASME & JSME
Joint Fluids Engineering Conference, Symposium S-290
Water Hammer (Editor JCP Liou), San Francisco, July
1999, ASME-FED, 248, Paper FEDSM99-6907.
[17] Kumar, D.S. (2010) Fluid mechanics and fluid power
engineering. S. K. Kataria & Sons Publishers and Dis-
tributors, New Delhi, Chapter 9.
[18] Ahmad A. and Ali R.K. (2008) Investigation of the junc-
tion coupling due to various types of the discrete points
in a piping system. The 12th International Conference of
International Association for Computer Methods and
Advances in Geomechanics (IACMAG), Goa, India.
[19] Knapp, Y., Bertrand, E. and Mouret, F. (2003) 2D-PIV
measurements of the pulsatile flow in a left heart simu-
lator. Proceedings of PSFVIP-4, Chamonix, F4082.
[20] Wang, W.X. and Christopher D.B. (2007) Effects of col-
lapsible-tube-induced pulsation vigour on membrane fil-
tration performance. Journal of Membrane Science, 288
(1-2), 298-306.