ON SEDIMENTATION AND ROULEAUX FORMATION-II.

By
ERIC PONDER. From the Department of Physiology, Edinburgh
University. (With four figures in the text.)
(Received for publication 26th January 1926.)

IN a paper already published (1) it has been shown that the sedimentation
of red cells may be described by certain simple hydrodynamical equa-
tions, provided that the cells are discrete-a condition which precludes
the application of the equations to the fall of cells in plasma or in serum,
in which rouleaux occur. Before we can pursue the study of sedi-
mentation further, we have therefore to consider rouleaux formation,
with which this paper is concerned. The phenomenon is unquestionably
a complicated one, influenced by many factors, each of which may be
considered separately.
I. CHANCE CONTACT.
It will be obvious that two cells cannot cohere to form a rouleau or
part of a rouleau unless they first collide. We have therefore to consider
the chances of contacts taking place between individuals of a large
number of cells, given, to begin with, that no factors save chance enter.
We may imagine that initially the cells are all discrete ; a single cell may
then collide with another to form a group of two, which for convenience
we shall call a rouleau of two. This rouleau of two may then collide
either with a single cell or with another rouleau of two, and thus may
become either a rouleau of three or one of four, and so the process may
go on indefinitely. We wish to calculate the number of single cells,
rouleaux of two, rouleaux of three, and so on, which will be found
after any time t in a volume of fluid containing a very large number of
cells, all discrete when t =0.
To simplify the matter, we may make three assumptions: (1) Every
contact is to result in a cohesion, provided that this contact takes place
in a suitable position as required by the following assumption. (2) The
surface by which a rouleau can add on another rouleau or a single cell
is the same as that by which a single cell can add on another single cell.
(3) Single cells and rouleaux, irrespective of their size, are carried about
in the fluid with equal velocities. The second of these assumptions may
be seen to be true by referring to fig. 1 ; we take it that a contact of the
cell b on the rouleau a will result in the addition to the length of a only
174 Ponder
if b strikes it in the region of its ends; the surface of a, so far as rouleaux
formation is concerned, is then no greater than that of b. The third
assumption will be true if the fluid containing the cells is in turbulent
motion-a condition easily fulfilled in experiment.
If the surfaces of contact and the rate of translation are constant,
whatever the size of the rouleau, we have a problem very similar to that

wt-
1. \)'-
FiGE. M,

Of coagulation, and may deal with 1t1+

it in the manner of SMOLUCHROWSKI ()
In this way we obtain the general expression

m ltM)-
t2 )

Here Wr is the number of rouleaux of r cells, M the number of cells under

consideration, 1 a constant, and t the time after which wT is ascertained.
The form of the expression is identical with that of SMOLUCHOWSKI;
for the mode of derivation, the reader should consult his paper.
The equation presents several interesting points: (1) The constant 1,
under ordinary circumstances, must be small, so that 1tM/2 shall be in
the neighbourhood of unity. Under these circumstances, when t is
small, we have W1> W2>w3, etc.; that is, there are more single cells
than there are rouleaux of two, more rouleaux of two than rouleaux of
three, and so on. (2) When t becomes great, then we approach the
condition W1=W2=w3, etc. ; that is, there are as many single cells as
there are rouleaux of any particular size. This result is rather surprising
at first sight, for one might expect that with increasing time all the single
cells would disappear in favour of the longer rouleaux. A little con-
 On Sedimentation and Rouleaux Formation 175
sideration will show that the result of the equation is in no way at
variance with experience; in a preparation which shows many rouleaux,
it may certainly be difficult to find a single cell or a rouleau of two, but
it may be equally difficult to find a rouleau of exactly twenty-three or
nineteen. (3) The equation has, however, this limitation, that it sets
no limit to the length of rouleaux formed, and, if t becomes large, a
figure may be given for the number of rouleaux of fifty or a hundred
cells. Rouleaux of this length do not occur in practice, and therefore
we must use the expression only up to a certain limit that of rouleaux
of about twenty cells.
In view of this last consideration, it would be desirable to have a
reversible equation which would take account of the fact that, as the
length of rouleaux increases, their tendency to break increases also.
We may take it that the chance of a fracture occurring is directlv
proportional to the length of the rouleau; but even with this assumption
we have failed to develop a reversible equation, and must therefore
content ourselves with the one given above, and with the experience
that, until we come to deal with rouleaux of over fifteen or twenty,
fracture rarely occurs even when the aggregates are carried about by
currents in the fluid.
The validity of the expression may be tested in the following way.
Blood is withdrawn from the arm vein of a subject, and received in a
syringe containing a small amount of oxalate as an anti-coagulant.
This blood is then centrifuged and the plasma removed; the plasma is
again centrifuged to rid it of cells. About 1 c.c. of the clear plasma is
placed in a series of tubes, and kept at whatever temperature is decided
on for the experiment. If the initial stages of rouleaux formation are
to be studied, this temperature should be a low one (150 C.), while if
rapid rouleaux formation is desired it should be higher (370 C.).
As soon as the clear plasma has been obtained, a small quantity of
blood, about 0 05 c.c., is withdrawn from the subject's ear, and mixed
with the plasma. The tube containing the mixture is then shaken to
and fro as one shakes a boiling-tube, for a length of time which is
previously decided upon. During this time the temperature of the
experiment is maintained, and rouleaux formation occurs. Just before
the end of the pre-arranged time, a drop of the mixture is placed on a
clean slide and covered with a coverslip whose edges should be greased
with vaseline. The cells very quickly settle if the vaseline seal is
complete, and may then be placed under a magnification of about 300,
and photographed under conditions of critical lighting, some twenty
or thirty fields being recorded. If good lighting is used, the necessary
exposure is only half a second to a second, and so the necessary number
of plates can be exposed in a couple of minutes. The plates are
developed, and the number of single cells, rouleaux of two, rouleaux
of three, and so on, determined by examination under a low-power
176 Ponder
microscope. Except for the magnification, the procedure is very similar
to that described in another paper (1 1). It is, of course, important that
the fields be taken at random, and not selected for either single cells or
for rouleaux.
The simplest way to record the count of the rouleaux is to count out
1000 peas or beans into a dish, 1000 being a very convenient number to
choose for M, the cells to be observed. If a smaller number is taken,
sampling errors are apt to be large; if a larger number, the counting is

"I

O.3 1.0 2.0 3.0 4.0 ttm/2

Fic;. 2.

tedious. A distributing box containing ten or twelve divisions is then

taken, the first division labelled 1, the second 2, and so on up to 12.
When a single cell is seen under the microscope, one pea is put into
division I ; when a rouleau of two is seen, two peas are put into division
2; when a rouleau of eight appears, eight peas are put in division 8; and
so on until the 1000 peas are exhausted. The number of peas in each
division is then counted: that in division 1 represents the number of
single cells ; that in division 2, if divided by 2, the number of rouleaux
of two ; that in division 8, when divided by 8, the number of rouleaux
of eight ; and so on. By this method exactly 1000 cells are distributed
withouit difficulty.
We now have to find the value of the constant 1. The easiest way
 On Sedimentation and Rouleaux Formation 177
to do this is to plot a series of curves for the values of w1, W2, W3, W4, and
W5 for values of the factor ltM/2, assigning to this factor successive
values of 0 5, 1, 2, 3, 4, and 5. These values cover the range of the
ordinary experiment. The curves are shown in fig. 2, but are better to
be plotted on a somewhat larger scale. Taking first the number of wl,
the value of the factor ltM/2 which corresponds to this number on the
w, curve is read off and noted down. Next the number of w2 is taken,
and the corresponding value of ltM/2 on the w2 curve is read off.
Similarly, the value of the factor for each of the other observed numbers
is read off the proper curve. The values of the factor are then averaged:
the result gives the best value for ltM/2. Since t, the time, is known,
and since M is 1000, the proper value of the constant I is readily obtained.
Where, as in some of the curves, there are two values for the factor
corresponding to the observed number of rouleaux, that value nearest
the one obtained for the w, curve is taken, for reasons which will be
easily seen as soon as the method is put into practice.
The graph obtained by plotting out the curves as in fig. 2 is an
interesting one; the curves for W2 and rouleaux containing a greater
number of cells than two have a maximum, while that for wL has none.
Further, this former set of curves pass through the origin, indicating
that, when 1 or t is equal to zero, there are no rouleaux present, while
the curve for w, cuts the ordinate at 1000 when t or 1 is zero, indicating
that under these circumstances there are only single cells present.
In order to make the procedure clear, we shall take an actual example.
Cells were allowed to form rouleaux in plasma for 10 minutes, so t=10.
Thereafter the number of rouleaux of each length were counted in the
manner described, and the following figures obtained:

Rouleaux of . 1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

223 110 63 27 15 12 5 3 3 3

Referring to the graph on the w, curve, the value of 1tM/2 corresponding

to 223 is 11. The value corresponding to 110 on the w2 curve is 1 35.
On the W3 curve, 1 0 corresponds to 63; on the W4 curve, 0 85 corresponds
to 27 ; and on the w5 curve, 1.0 corresponds to 15. Adding these
values and averaging, we get the best value of ltM/2 as 1 06. Since t is
10 and M is 1000, the proper value of 1 is 2 12 (10-4). The fact that the
values of ItM/2 obtained from the different curves agree so closely is very
strong evidence of the applicability of the equation.
It will be noticed that the values of w6 and rouleaux longer than
this are omitted in the calculation of the value of ItM/2. The reason
for this is that these values are too liable to experimental error ; if it
178 Ponder
is necessary to investigate the number of rouleaux of ten and above,
graphs showing the relation of ItM/2 to the number of rouleaux must be
constructed in such a way as to cover the experimental range.
If the equation which has been developed is correct, we should be
able to predict the number of rouleaux of any particular size when
t=20: that is, after twice the time considered in this experiment. In
this case, ltM/2=2 12, and the distribution of rouleaux which we should
expect would be:

Boiilealux ()t 1. 2. 83. 4. .,. fi. 7. 8. 9. 10.

102 70 46i :3 I 21 16 10 (i 4 8

The results obtained by experiment were:

Rotlleatix of 1. 2. :. 4. 5. (i. 7. 8. 10.

o.

113 74 47 83222 l 2 183 5 ,5

The correspondence between the calculated and the experimental

results is very good, considering the experimental difficulties, and forms
convincing evidence of the correctness of the equation. The low value
of the constant 1 in these experiments, an indication thalt rouleaux
formation is proceeding very slowly, is principally due to the fact that
the temperature at which the plasma and cell mixture was kept during
the period of shaking was a low one (14° C.), this rendering rouleaux
formation slow.
A number of similar experiments have been carried out. all with a
similar general result. Some of these may now be given as examples,
and arranged in tabular form.
EXPERENIENT
' 1. .\t 15' (. t --20 nilLLutes.

. 2 3. 4. 5. (i. 7. 8. . I 1. 12.

RoLIleauix oW,,erved 117 87 42 :38 2 -t 183 1 :6

( 8
(alculated. hFo, 11 1 1 4 29
': 1. 10 (i 4 82 2 1
It-A 2 =2.
 On Sedimentation and Rouleaux Formation 179

EXPERIMENT 2. At 370 C. ; t=7 minutes.

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11. 12.

RoUleaux observed 40 30 27 20 16 12 12 9 6 6 3 5
Calculated, from 33 27 22 18 14 12 10 8 6 5 4 3
ltM/2=4-5.

The remaining cells counted were in rouleaux of more than twelve,

and are not recorded in the table.

EXPERIMENT 3.-At 370 C.; t= 3 minutes.

1. 2. 3. 4. 6. 7. 8. 9. 10. 11. 12.

RoUleaUx observe(d 73 55 40 22 19 17 10 10 6 5 5 2
Calculate(d, from 62 47 35 26 20 15 11 8 6 5 3 2
ltM /2 =3.

The remaining cells were in rouleaux of more than twelve.

EXPERIMENT 4.-At 15 '. ; t = 2 minutes.

1. 2. 3. 4. 5. 6.

Rouleaux observed . . 462 120 60 14 10 2

Calculated, from ltM/2=0 5 444 118 50 16 6 2

EXPERIMENT 5.--At 370 C.; t=4 minutes.

These five experiments, together with the previous two, show the
extent to which the equation may be expected to apply. It will be
noticed that results are not given for rouleaux of more than twelve;
this is because the results for such rouleaux are apt to become rather
irregular when only 1000 cells are counted. Moreover, when 1tM/2
becomes greater than about 2, as it does in cases where these long
180 Poncler
rouleaux appear, the sum of the figures obtained by multiplying the
number of rouleaux by the cells in each falls short of 1000, the total
taken for calculation, principally owing to the fact that calculation
gives fractional numbers for the larger rouleaux; we cannot take
account of 0- 5 rouleaux of twelve, for example. This, however, does not
greatly matter, for we are rarely concerned with rouleaux formation so
marked that there are many rouleaux longer than about ten; further,
when we apply the results of this equation to sedimentation, a simpler
method of dealing with the problem presents itself.
To find the mean length of rouleaux present in a suspension of cells
such as we are considering, we multiply the number of each size of
rouleau by the cells in it, and divide the sum of the figures obtained by
the sum of the number of rouleaux. Thus the mean size of rouleau is
-PYWV
or, as 2yw, is M, and 2w, is
M

the mean is

This is a useful result, for the mean can be calculated readily from the
experimental observations, and the value of 1tM/2 can be obtained by
subtracting unity from it. It is, however, to be remembered that the
result will only be correct when 2yw11 =M, that is, when exactly 1000
cells are distributed into rouleaux of various sizes. If Yyw, is less than
M, as in Experiment 2, for instance, where the rouleaux of over twelve
are not recorded, the value of ItM/2 cannot be obtained in this way.
In the first experiment given in full, or in Experiment 4, the condition is
fulfilled, and the value of the factor works out correctly. It would have
done so also in the other experiments had the observations been extensive
enough.
The result is even more useful, however, than appears at first sight,
for it enables us to calculate the value of ItM/2, and therefore of 1, if we
can obtain the mean size of rouleau by any method. Now it happens
that the most convenient way of getting this mean size of rouleau is not
to count 1000 cells and distribute them according to their grouping, but
to allow an unlimited number of cells to sediment in a fluid, and to
calculate from the rate at which they fall the size of the mean rouleau.
The constant I can then be found from the data. The development of
this method, and its applications to the sedimentation of cells in plasma
or serum in which rouleaux formation is proceeding synchronously with
sedimentation, will be deferred for consideration in another paper.
 On Sedimentation and Rouleaux Formation 181
Finally, it will be observed that the rate of rouleaux formation
depends entirely on the value of the constant 1, which expresses the
frequency of chance contact. This constant may vary considerably in
magnitude, although the factor ltM/2 cannot show much variation under
the conditions selected for experiment, since, if it does, rouleaux forma-
tion either becomes too fast to measure or so slow that it is scarcely
detectable. The value of 1 itself depends upon several factors-the
number of cells per unit volume of fluid, the temperature, and such
undefined factors as the charge on the cells and the state of the cell
surface, since anything which may make contacts more or less effective
has its influence on this controlling constant. There is, indeed, Do
reason why 1 should not change in value during the same experiment;
although we have never met with a case in which it was necessary to
assume this, more prolonged experiments, such as those dealing with
sedimentation, may show an instance of this occurrence. In general,
the larger 1 is, the more rapid the rouleaux formation; the smaller 1,
the slower the formation. Its most usual value, the extent to which
this value may vary, and the variation produced by different factors,
have still to be studied.
It is necessary to emphasise the fact that this expression applies only
to cells in a fluid in which movement is turbulent and at the same time
not so vigorous as to cause rouleaux to break up. It does not apply,
without some modification, to rouleaux formation in a fluid with an
orderly flow, nor to rouleaux formation in a fluid which is stationary.
To meet these special cases modifications have to be introduced; these
will be remarked upon as they are required in the course of further work.

II. THE ELECTRICAL FORCES.

The researches of NORTHROP and FREUND (3) and of OLIVER and
BARNARD (4) have established that the stability of a red cell suspension
is largely dependent on the difference of potential which exists between
the cells and the suspending medium. If this potential fall below a
certain point, the cells tend to agglutinate; the value of this critical
potential seems to depend on certain factors at present unknown. The
potential difference, amounting to similar charges on the cells, is regarded
as keeping them apart; any force acting in this way and tending to
preserve the discreteness of a cell suspension must clearly be considered
in connection with rouleaux formation.
The potential difference between the red cells and the suspending
medium is ascertained by some method of electrophoresis, the migration
of the cells being measured in ,u per second, and from this the potential
difference calculated by the Lamb-Helmholtz equation. The methods
employed in this type of investigation are principally two-that recom-
mended by MICHAELIs, and used with certain modifications by OLIVER
1 82. Po nder
and BARNARD, and that described by NORTHROP; the nmethods do not
differ in principle, but only in the form of the electrophoretic cell
employed. Of the two methods, we have found the second very much
better, and very consistent results can be obtained by it in most cases,
plrovided that the electrophoretic cell is properly constructed, and that
care is taken to avoid air bubbles or other causes of inequalities in the
cross-section of the channel between the electrodes. In the case of
suspensions such as those of which this paper treats, however, where the
cells are suspended in plasma or serum, and where a considerable degree
of rouleaux formation may be present, certain difficulties arise which
make the method rather unreliable. The principal of these is that the
rouleaux, owing to their size and peculiar shape, tend to fall out of the
plane of focus when the measurements are being made; it takes but a
very short time for the majority of them to collect on the bottom of the
nieasurement chamber, where, of course, they are useless for the purpose
of the experiments. This fact, together with difficulties in the cleaning
of the electrophoretic cell when such solutions containing proteins are
used, has caused us to abandon the microscopic method, and to fall
back on a macroscopic one, which, although it is by no meanis free from
fallacies, has given better results.
The apparatus used consists of a U-tube of about 5 mm. cross-section,
the sides of which are each about 10 Cnm. long. Into the bottom of the
U-tube there is sealed a glass tube of narrower bore, which expands above
into a small reservoir with a stopcock through which fluid can be passed
into the U-tube. Two KCl-agar syphons lead from the limbs of the
U-tube to zinc-sulphate electrodes, which are connected with a 230-volt
circuit through a voltmeter and a reversing switch. A reading micro-
scope with a 3-inch lens is arranged so that it can be focussed on one
of the limbs of the U-tube; the vertical muovement of the microscope
can be read off on a scale, and the finer movements on a mnicrometer
fixed in the eyepiece.
If the reservoir is filled with a suspension of the cells in plasma,
saline, or other fluid, and the U-tube filled with the fluid, but containing
no cells, it is quite easy to allow the suspension to einter at the bottom
of the U-tube, displacing the contents thereof, so that clean interfaces
are formed in each limb about half-way up. To make these interfaces
sharper, the cells may be allowed to sediment for five or ten minutes,
or the current mav be passed for a minute or two in each direction. To
make the reading, the level of one interface is first observed on the
m-icrometer in the eyepiece of the microscope ; the current is then
turned on, and the level of the interface read again after 100 or 200
seconds. Care should be taken to observe the interface which proceeds
away from the electrode on the same side of the U-tube when the
current is passing, and not the one which rises in the limb. The average
of three readings is takeni as the final result, and the movement may be
 On Sedimentation and Rouleaux Formation 183
checked by reversing the current and observing the alteration in level
of the other interface.
The velocity of sedimentation is now to be subtracted from the
velocity of movement of the interface; the result is the electrophoretic
velocity, which is then used for the calculation of the potential difference
between the cells and the fluid surrounding them. The velocity of
sedimentation must be found by measuring
the rate of fall of one of the interfaces
when no current is passing; this sedi-
mentation velocity is usually much smaller
than the electrophoretic. In connection
with it an interesting point arises. The
velocity of sedimentation in such a U-tube
as we describe will always be found to be T
greater than that calculated from the ap- /
propriate equation, and also greater than /
that observed with the same cell suspen-
sion in a straight tube. The explanation
of this fact will be seen by an inspection /
of fig. 3. Cells a and b fall in a given
time to positions a' and b', but at the /
same time a cell c, situated initially in /
the bend of the tube, falls through the /
same distance to c'. This results in clear
fluid being developed in contact with the
upper wall of the bent part of the tube; -c
stich clear fluid never appears in practice, /
however, but is replaced by cell suspension,
while the cell-free fluid passes upwards to FIG. 3.
augment the clear fluid above the interf a ces,
-which appears always greater than the calculated amount in consequence.
This condition of affairs is very similar to that observed in a sloping tube,
and has been discussed at length in (1). The practical difficulty
attendant upon it is that, since the calculated and observed sedimenta-
tion velocities are not the same, there is an error introduced into the
method, and, moreover, an error which one cannot allow for. The
taking of the sedimentation velocity as that observed in the U-tube
when filled ready for an experiment is, however. sufficiently satisfactory
for ordinary purposes.
This method takes no account of the position of the cells relative to
the walls of the tube, but the error introduced in this way is small, as is
that occa.sioned by ignoring the fact that the cells are not all of the same
size. We would not advocate this macroscopic method in place of that
of NORTHROP for routine purposes, as it requires much more deft
handling, but with the suspension media which we use, and with cell
VOL. XVI., NO. 2.-1926. 13
184 Ponder
suspensions which show marked rouleaux formation, we have found it
quick, consistent, and convenient. The ease of cleaning and the small
quantities of material required constitute special advantages: 5 c.c. of
blood gives sufficient plasma and cells for one determination. The
principal difficulty is one which would scarcely be suspected-the
making of the KCl-agar syphons, which must be uniformly filled, clean
on the outside, free from air bells at the ends, and freshly made for each
occasion.
The cell constant is easily found, as the cross-section of the U-tube is
the same at all points. In view of the fact that the continued passage
of a current causes a rise of temperature in the contents of the U-tube, as
well as to avoid diffusion currents due to local temperature alterations,
the apparatus is immersed in a small thermostat, in which may also be
kept the tubes for the necessary viscosity determinations.
The Lamb-Helmholtz equation for the calculation of the potential
difference between the cells and the surrounding fluid is
VId =4vqv
KX'
where V is the potential difference, 1 a measure of the slip, d the distance
between the plates of an air condenser equivalent to that which may be
supposed to exist between the surfaces of the cell and the fluid, K the
specific inductive capacity of the fluid, 'q its viscosity, v the velocity of
migration, and X the strength of the electric field. All measurements
are electrostatic. Of the variables in the equation, d and I are un-
obtainable, and K is difficult to determine; we therefore follow
WINSLOW, FALK, and CAUFIELD (5), to whose excellent critical survey
the reader is referred, and recognise that it is impossible to calculate
the potential difference with accuracy from available data. These
latter writers express their results in migration velocities, after having
given the strength of the field for the series of experiments, a procedure
also followed by EGGERTH (6); but we see no reason why the value of -j
should not be included, as it is an easily determined figure. We shall
therefore give, with a claim to accuracy, the value of v'q/X, v being
observed in ,u per second, and also shall use 14 x (102) times this figure
to indicate what the potential difference in millivolts might be taken
to be if Ild were unity and K were 81.
The values for potential difference between erythrocytes and the
surrounding fluid have hitherto been obtained principally from observa-
tions on cells in isotonic sugar, with or without the addition of electro-
lytes. Some of EGGGERTH'S observations are on cells suspended in
phthalate buffers. OLIVER and BARNARD, observing that M/7 NaCl
lessens the charge of these glucose-suspended cells, go so far as to say
that the saline usually used for agglutination experiments is an un-
fortunately selected medium. With this we cannot agree. The red
 On Sedimentation and Rouleaux Formation 185
cells when in the circulation move in fluid more like M/7 NaCl than
isotonic sugar. Besides this, attention must be drawn to an important
point. If the cells are placed in an electrophoretic field, then electric
transference of ions will take place across their membranes ; whether
this will be great or not is difficult to say, but it will be the greater the
more difference there is between the ionic conditions inside and outside
the cell. This fact makes the measurement of potential difference even
more uncertain, and, looked at from this point of view, isotonic sugar
is a most unfortunate medium for suspension; it is true that it contains
no electrolytes or proteins, but such a condition is more to be avoided
than sought in experiments which are concerned with the cells in their
natural state.
For the purpose of this paper we propose to investigate the migration
of human cells in three media: in 0 85 per cent NaCl, in which rouleaux
formation does not occur; in plasma, in which it occurs to a moderate
extent in the cold; and in plasma heated to 560 C. for fifteen minutes,
in which rouleaux formation occurs very rapidly and extensively
(SELLARDS (7)).
The cells and plasma used were obtained by venous puncture, the
blood being collected over a small quantity of potassium oxalate. The
electrophoretic velocity was measured as described above, the strength
of field being determined in each case-a very necessary precaution-
and the viscosity of the suspending fluid being measured immediately
after the conclusion of the experiment. In all cases, the temperature at
which the U-tube was kept was 150 C., this temperature being selected
because at higher temperatures rouleaux formation is apt to be unduly
rapid. We have been particularly careful to carry out observations
of the electrophoretic velocity as soon after the withdrawal of the blood
as practicable, in view of some of EGGERTH'S findings.
The results of the experiments may be given in tabular form.

1. CELLS SUSPENDED IN 085 NaCi.

Experiment. v?7/X. Approximate P.D.

1 0 0052 7-3
2 0-0048 6.7
3 0-0062 8-7
4 0-0046 6-5
5 0 0060 8-4
6 0-0055 7-7
7 00063 8-8
8 0-0052 7-3
9 0-0049 6-8
10 0-0065 9-1
186 Ponder
2. CELLS SUSPENDED IN PLASMA.

Experiment. 1ij X. Approximate P.D.

1 0 0076; 10-6
2 00047 66
-
3 0-0080 11-2
4 0-0070 9-8
00-0063 8.8
6 00057 8-0

:3. CELLS SUSPENDED IN HEATED PLASXJA.

Experimi-en-t. njII X. Approxinmate P.D.

0 007A)5 10.5
0-0085 11.9
3 0 0060 8-4
4 oo00070 9i8
00063- 9' 1
6 0-0081 11-3

The average P.D. of the cells in saline is 7.7 millivolts, that of the cells
in plasma 9 2 millivolts, and that of the cells in heated plasma 102-'
millivolts. The difference between the two last figures may not be
significant, for the experimental error might be made to account for a
difference of about 1 millivolt, but we are inclined to look upon the
difference between that P.D. observed for the cells in saline and that
found for cells in plasma as a real one. It may be mentioned that the
difference is not striking in the actual experiment, the rate of migration
being about the same in each case ; introduction of the proper values for
the viscosity, however, at once shows the higher charge in the cells
suspended in plasma. It is, of course, to be remembered that these
figures are correct only on the assumption that K =81 and lld = 1,
conditions which may not be fulfilled, and which may differ in the cases
of plasma and saline respectively. It will also be noticed that the P.D.
between cells and 0O85) NaCl is considerably lower than that between
cells and isotonic sugar, a result which might be expected froii some
of the results of WINSLOW, FALK, and CAUFIELD with bacterial cells,
and of other workers with varied material.
Allowing for all sources of error, we have to conclude from these
results that (1) rouleaux formation will occur even if there is a consider-
able potential difference between the cells and their surrounding medium;
that (2) the failure of the cells to form rouleaux in saline is not due to
the repulsive forces between them being unusually great ; and that (3)
 On Sedimentation and Rouleaux Formation 187
the increased tendency to form rouleaux in heated plasma is not due to
the repulsive forces between the cells being diminished. The results
of the investigation are thus negative.

The Charge on a Single Cell.

This may be found in an approximate manner from the migration
rate by the use of a modification of the Stokes Equation, in which the
force moving the cell and the resistance keeping it back are balanced
against one another. We have two important cases: (1) where the cell
is a sphere, as it is in the Goughian form, and (2) where the cell is in its
normal form of a biconcave disc.
(1) When the cell is a sphere, we have the resistance to its movement
as 6-f-r. The velocity in proper units is v/X, and so the force acting on
it is, in e.s. units, e =6-z-rv/X. Here r is the radius of the cell, while
v, X, and -q have the same meaning as above.
The charge on a single cell works out as a very small quantity.
Taking a cell of radius 3,u, and assuming that it moves in a fluid of
viscosity 0.01 with a migration rate of v/X=1, the charge is
approximatelv 1 7(10-6) e.s. units. This corresponds to a potential
difference of about 14 millivolts, as observed by electrophoresis.
(2) If the cell is in the discoid form it will meet with a different
resistance as it passes through the fluid, which it does broadside on.
The resistance is that met with by a sphere of radius 0 85c, where c is
the radius of the disc, and so we have the force, e =-6ir(0. 85c)v/X. This
gives approximately 2 1(10-6) e.s. units for a cell under the same con-
ditions as above.
The Repulsive Forces between Cells.
It would be of great interest to be able to determine the magnitude
of the repulsive forces which exist between two similarly charged cells.
Unfortunately it is almost impossible to estimate these forces with any
accuracy. The reason for the difficulty is that, although we are able to
express the charge on a single cell in electrostatic units, this charge is
due to a double layer at the cell surface-indeed, as the cell is a balloon-
like body, to two double layers, one at each surface of the membrane
and as such cannot be treated as if it were a fixed charge, such as resides
on a pith ball. The repulsive forces, if we may judge from analogous
cases of double layers round colloid particles, only manifest themselves
when two cells approach to within a very small distance of each other,
and occur owing to a redistribution of the charges as a result of the
opposition. The repulsive forces differ in another respect from those
due to fixed charges, for they do not vary with the square of the distance
of the charges from one another, but with a higher power, so that when
188 Ponder
the cells are more than a small distance from each other the repulsive
forces are negligible.
It is, however, quite instructive to treat the charges on the cells as if
they were fixed charges, always bearing in mind that no arguments can
be based on such results alone. In the case of spherical cells we can
locate the charges at the centres, in which case we have the repulsive
force between two cells each charged with 1 7 e.s. units, and whose
centres are 10, apart, as 3 6(10-8) dynes-a very small quantity.
When it is remembered that the erythrocyte is normally in the form
of a biconcave disc, the difficulties of forming an estimate of the repulsive
forces becomes even greater, for two cells can approach one another in a
number of ways, either broadside on, edge on, or in any intermediate
position. For each position there will be a corresponding alteration
of the distribution of the charge on the cell, and a corresponding repulsive
force will develop. Any attempt to study these forces is beyond the
scope of this paper; if, however, we treat the charges as fixed electro-
static charges, some interesting points appear.
Since the surface of the cell in its normal form approaches closely to
the form of an equipotential sur-
face to a charged ring, we may
2 z locate the charge borne by the
cell on this hypothetical ring, whose
X3 diameter is about five-sixths of that
of the cell. In order to facilitate
computation, we may further re-
place this ring by the inscribed
2
2 hexagon, and locate one-sixth of
the total charge of the ring at each
s point of the figure. The charge
.borne by the cell is thus supposed
14 to reside at the points of the hexa-
gon inscribed in the ring to which
the surface of the cell constitutes
3. an approximately equipotential sur-
face (13).
We can now consider the three
principal positions in which two cells
FIG. 4. can lie with respect to one another;
1. Equatorial planes coplanar. they may lie with their equipotential
2. Equatorial planes at right angles. planes in one plane, with their equi-
3. Equatorial planes parallel,
potential planes at right angles, or
with their equatorial planes parallel (fig. 4).
1. Equatorial Plane8 Coplanar.-For the purposes of computation,
take two hexagons inscribed in circles of 1 unit radius, whose centres
are 4 units apart. If the hexagons lie so that their sides are parallel, the
S
 On Sedimentation and Rouleaux Formation 189
sides nearest to one another will be 2 units apart. Let each point of the
hexagon bear unit charge. The repulsive force between the hexagons
is then approximately found by joining point 1 of the first hexagon
to all the points of the second in rotation, measuring the lengths of the
lines so formed, and summing the reciprocals of their squares ; thereafter
point 2 of the first hexagon is joined to all the points of the second, and
the sum of the reciprocals of the squares of these lengths obtained.
The procedure is continued till all the points of the first figure have
been joined to all the points of the second; the sum of the reciprocals
of the thirty-six squared distances gives the repulsive force in arbitrary
units. In the case of hexagons whose equatorial planes are coplanar,
this force is 2 8 units.
2. Equatorial Planes at Right Angles.-A similar summation, this
time in three dimensions, gives for hexagons of the same size as in 1,
whose centres are 3 units apart and whose edges are therefore 2 units
apart, a repulsive force of 3 2 units.
3. Equatorial Planes Parallel.-If the planes are 2 units from each
other, and the hexagons the same as in 1, the repulsive force is 6 3 units.
Treating the charges in this way, the repulsive forces are greatest
when the cells are broadside on, and least when they are edge on, the
distance between two opposing edges being the same in each case. If
the two cells were to move together to form a rouleau, the greatest work
would be done in moving them in the former position.
Slightly different results are obtained if the hexagons are taken as
lying with the points of one opposed to the sides of the other; but the
general nature of the result is not affected. Nor is the general result
altered by assuming that the repulsive force varies with the reciprocal
of the cube, the fourth power, etc., of the distance instead of with the
square, as is most likely in actual fact. Whether it would be obtained
from a rigidly correct treatment of the charges due to the double layer,
we are not in a position to say at present, but preliminary work indicates
that in this case too there is more work done in causing two cells to
approach broadside on than is done if the cells approach edge to edge.
It also indicates that the proper solution is one which involves very
great mathematical difficulties. The point is perhaps not of much
importance, for the currents which cause the cells to collide are quite
capable of overcoming the small electrical repulsions, and, moreover,
these currents tend to orient the cells broadside on.

III. THE COHESIVE FORCES.

Some of the most important observations in connection with rouleaux
formation are those of SELLARDS (7). This investigator records the
interesting facts that (1) normal human serum, when heated to 600 C.
for fifteen minutes, develops the property of causing human erythrocytes
190 Pon(ler
to collect in rouleaux; (2) this property is due to an increase of a normal
property of the serum ; (3) the property of causing rouleaux formation
is rapidly lost at room temperature, and less rapidly in the cold, the loss
being more marked in unheated than in heated serum; and that (4)
rouleaux break up spontaneously after some hours, the necessary length
of time being dependent on the temperature. He also records the fact
that rouleaux formation is dependent on the number of cells present in
the suspension, being more marked when many cells are present, as
might be expected from the foregoing consideration of chance contact.
We have carried out the experiments of SELLARDS, using both serum
and plasma, and employing various modifications: we can confirm his
results without exception. Especially noticeable is the effect of heating
the serum or plasma; we have used this method of producing marked
rouleaux formation in several experiments in other parts of this paper.
One of the most important conclusions is that serum possesses a
normal property of producing rouleaux formation. Plasma, according
to our finding, has also this normal property. This is in keeping with
the fact that red cells will form rouleaux in the vessels of an animal if
the circulation is arrested for a few seconds, and may be seen well in the
mesenteric vessels of the mouse. Further, the cells move along the
wider vessels broadside on, as we should expect, and frequently in
columns these columns may be broken up as the cells enter the
narrower capillaries, but the cells in circulating blood are arranged, if
not in rouleaux in the usual sense, in a manner very similar. The
formation of the typical rouleaux is by no means a phenomenon to be
observed in shed blood only, as is sometimes stated or implied.
WVe have shown above that the difference between the power of
unheated and heated plasma to form rouleaux is not to be explained by
the cells in the former fluid having a higher potential difference than those
in the latter. Wte are inclined to consider that the effect of heating the
suspending medium is to cause a change in the surface stickiness of the
cells suspended in it, or, in other words, to cause a change in what
NORTHROP, OLIVER, and BARNARD, and others, term the ' cohesive
forces" between the cells. It is known that the stickiness of the
surface of the red cell can be altered by the addition to the suspending
fluid of various substances (FENN (8)), and both NORTHROP and OLIVER
and BARN-ARD employ such a conception to explain the effects of NaCl,
BaCl2, and CaC12 on the cells.
The more sticky the surface of the cell, the greater will be the
number of collisions which result in permanent contacts. MIoreover,
such a conception goes to explain why, in typical rouleaux formation,
end-contacts are permanent and side-contacts not permanent, for the
former type of contact involves large surfaces over which these cohesive
forces act, while in the latter type of contact only small surfaces, and
accordingly small cohesive forces, are involved. We are, indeed,
 On Sedimentation and Rouleaux Formation 191
inclined to think that the stickiness of the cell surface forms the key
to the whole problem, and that variations in the potential difference
between the cells and the suspending fluid play a comparatively small
part.
Unfortunately there is at present no satisfactory method of investi-
gating these cohesive forces, and a definite statement on the point is
therefore impossible.
Regarding SELLARD'S point that rouleaux spontaneously disintegrate,
we may add one observation. After some hours, cells in plasma not only
break apart so that no rouleaux are present, but undergo an interesting
change of form. They lose their discoid form, and become spherical.
Measurements before and after the change show that the volume of the
spherical form is the same as that of the discoid form, and that the
change is therefore not unlike that described by GOUGH (9), except that
it is not reversible. The fact is important in connection with the cause
of the maintenance of the normal discoid shape; it appears that after
some time in plasma the component of the cell membrane which is
responsible for the typical shape gives way, and that the cell then
assumes the form which presents minimum surface for volume. This
change is not dissimilar to that observed when cells are haemolysed by
saponin and the bile salts (10), or by guinidine (12).

IV. THE SURFACE FORCES.

Quite apart from the question of the stickiness of the surfaces of
cells, surface forces play an important part in the formation and
stabilisation of rouleaux.
When two cells come into contact, a certain reduction in surface
takes place; if they are pulled apart again, there must be produced an
enlargement of surface which entails the expenditure of work, quite
apart from the overcoming of the cohesive force which causes the
surfaces to stick to one another. This fact tends to stabilise rouleaux
and maintain the contact of cells; it will be clear that the more reduction
which takes place in the surface as the result of a particular position of
contact, the more stable will that contact be.
When two cells collide with each other broadside on, the resulting
reduction of surface is as great as can be, and the collision results in a
very stable contact, for the separation of the cells involves an increase
of surface energy of the order of 10-6 ergs. Cells which collide with
rouleaux at other points than in the neighbourhood of their ends,
however, can be easily caused to break away again, for as the result of
such a collision there is little diminution of surface, and therefore little
stability associated with such a position of contact. Thus it comes
about that collisions in the region of the ends of rouleaux result in
permanent contacts being formed, while collisions in other regions are
192 Ponder
practically ineffective-an assumption which has been made in an earlier
part of this paper. In the former case, there is a resistance to separation
of the cells, for such separation means an increase of surface energy;
in the latter case there is no such resistance, and separation takes place
under any slight applied force, whether it be that of the currents in the
fluid or that of the electrical repulsion which is associated with the cells.
The existence of these surface forces is shown in an interesting way
in the deformation of the cells which build up a rouleau. Since the
surfaces of these components will be deformed so as to produce a
minimum surface, provided that they are deformable, the cells of a
rouleau are pressed together, considerable surface reduction being
effected in this way. Further, the whole rouleau is compressed, for the
surface forces tend to shorten it until its height is equal to the diameter
of its base-the condition for a cylinder of the minimum surface for its
volume. Such a complete compression is not possible, for the cells are
not sufficiently deformable under the forces acting on them, but an
approach to it is seen in the decreased thickness of the cells composing
the rouleau, and also in their increased diameter. This may be shown
by an experiment such as the following:
A preparation of cells, suspended in their own plasma, is made in the
manner indicated earlier in the paper, the conditions being arranged so
that a moderate degree of rouleaux formation is present. A series of
microphotographs are taken of the preparation, with all the precautions
set forth in (11); the fields are selected so that some show a number
of single cells while others show rouleaux of moderate length. The
diameters of the cells are then measured from the plates-first the
diameters of the single cells, and then those of the cells in rouleaux.
The following table shows typical results for five experiments of this
sort with human cells, each experiment being based on the blood of a
different individual. Mean diameters are shown.

Experiment. Single cells, Cells in rouleaux.

diameter in Mi. diameter in ,u.

1 8 72 8 96
2 8 80 9.11
3 8-68 905
4 8 75 9 10
5 8-81 924

In all cases the diameter of the cells in rouleaux was greater than
that of the cells which were discrete. Setting aside the scarcely possible
explanation that the rouleaux were formed principally from the larger
cells of the preparation, we are left with the conclusion that the cell
diameter is increased in the rouleaux-evidence of the compressive
 On Sedimentation and Rouleaux Formation 193
force which we should expect on theoretical grounds. The practical
result of this finding is that it is not permissible to include cells in
rouleaux when obtaining the mean diameter of the cells in a sample of
blood-a fact which we have insisted upon in a previous paper (11).
Similar experiments to the above show a fact which might be fore-
seen, that in rouleaux whose height is equal to their base, and in rouleaux
shorter than this, these compressing forces are not in evidence, and there
is no measurable increase in the diameter of the component cells. Under
certain circumstances, the cells in such rouleaux form an exception to
the rule just laid down, and may be included in determinations of the
mean diameter of a sample. If this is done, as is sometimes convenient
and occasionally necessary, the measurements of the cells in rouleaux
should be kept separate from the measurements of the single cells of the
sample, and included with them only if suitable statistical tests show
that there is no significant difference in the diameters of the cells from
the two sources. We have frequently demonstrated that this is a safe
procedure, if employed with caution.

V. SUMMARY.
From this short study of rouleaux formation several points have
emerged.
The first condition for rouleaux formation, and one which is fulfilled
under all but unique circumstances, is that there shall be collisions
among the cells. The frequency of these collisions plainly depends on
such factors as the movements in the fluid, the number of cells present
per unit volume, the viscosity of the fluid, and the size of the cells. The
second condition is that the collisions shall result in permanent cohesions.
Such cohesions will not occur if the surfaces of the cells are not sticky,
if the repulsive forces between the cells are very great, or if the cells
collide with one another so that the surfaces which come into contact
are small. If, on the other hand, the collisions bring large surfaces into
contact, permanent cohesions are likely to occur, for not only are there
large surfaces over which the cohesive forces between the cells can act,
but separation of these surfaces entails a considerable amount of work.
The consideration of these points indicates why cells collect in
rouleaux, and not in aggregates of roughly spherical form, for all contacts
except those which occur at the ends of the shorter rouleaux are not
permanent, whereas all contacts in which the cells are broadside on
result in lasting cohesions. To this must be added the fact that discoid
cells moved by a fluid orient themselves broadside on, and therefore
collisions in this position are encouraged.
From the mass of work on the potential difference between the red
cells and the suspending fluid, it appears that, if the potential difference
be reduced below a certain point, agglutination of the cells occurs, all
194 On Sedimentation and Rouleaux Formation
contacts being permanent, in whatever position they occur. From this
may be drawn the deduction that the repulsive forces accompanying
this potential difference are sufficient to prevent the permanence of all
contacts except those which involve a large extent of surface in con-
nection with which considerable surface and cohesive forces are called
into play. Before this explanation is regarded as the correct one,
however, it must be shown that those electrolytes, and other substances
which reduce the potential difference, do not also increase the stickiness
of the cell surfaces. The cases of CaCl2 and MgCl2, which reduce the
potential difference and yet do not cause agglutination, and the instance
of heated plasma, which increases rouleaux formation without decreasing
the potential difference, show how necessary it is to take account of
this latter possibility, the importance of which has been clearly recog-
nised by such investigators as OLIVER and BARNARD and NORTHROP.
As regards the kinetics of rouleaux formation, we have found that a
simple equation containing one easily evaluated constant is sufficient to
describe the course of the phenomenon under such conditions as are
extremely suitable for experimental purposes.
This research was carried out unider a grant from the Royal Society.

REFERENCES.
(1) PONDER, Quart. Journ. Exper. Physiol., 1925, xv. 235.
(2) SMOLUCHOWSKI, Zeitschr. f. physik. Chem., 1917, xcix. 129.
(3) NORTHROP and FREUND, Journ. General Physiol., 1924, vi. 603.
(4) OLIVER and BARNARD, Journ. General Physiol., 1924, vii. 99.
(5) WINSLOW, FALK, and CAUFIELD, Journ. General Physiol., 1924, vi. 177.
(6) EGGERTH, Journ. General Physiol., 1924, vi. 587.
(7) SELLARDS, Johns Hopkins Hospital Bulletin, 1908, xix. 271.
(8) FENN, Journ. Exper. Medicine, 1922, xxxv. 271.
(9) GOUGH, Biochem. Journ., 1924, xviii. 202.
(10) PONDER, Quart. Journ. Exper. Physiol., 1924, xiv. 334.
(11) PONDER and MILLAR, Quart. Journ. Exper. Physiol., 1924, xiv. 67.
(12) SECKER, Journ. Physiol., 1925, Ix. 286.
(13) PONDER, Journ. General Physiol., 1925, ix. 197.
The paper by WINSLOW, FALK. and CAUFIELD (5) contains an extensive
bibliography of the work on potential difference.