36O IJNS - 3eptember ] 9~3

Jeanrenand, B. and Vmmc, M. (1979)J. Clin. (Pfeiffer, E. F., ed.), pp. 281-286, The New 139-150
Invest. 63, 525-531 York Academy of Sciences, New York 47 Wilhelmi, A. E. (1974) in Handbook %1
13 du Vigneaud, V., Ressler, C., Swan, J. M., 41 Skeggs, Jr, L. T., Kahn, J. R. and Shumway, Physiology, Sect. 7, Vol. 4:2, pp. 5~-78,
Roberts, C. W. and Katsoyannis, P. G. (1953)£ N. P. (1956)J. Exp. Med. 103,301-307 American Physiological Society, Washington, DC
Am. Chem. Soc. 75, 4879--4880 42 Spiess, J., Rivier, J., Rivier, C. and Vale, W. 48 Wiamann-Liebold, B. (1982) inMethods in Pro-
14 Erspamer, V. and Melchiorri, P. (1980) Trends (1981) Proc. Natl Acad. Sci. USA 78, rein Sequence Analysis (Elzinga, M.. ed. ), pp
Pharmacol. Sci. 1,391-395 6517-6521 27-63, Humana Press, Clifton, New Jer~y
15 Fischli, W., Gotdstehi, A., Hunkapillex, M. W. 43 Staub, A., Sinn, L. and Behrans, O. K. (1955)J. 49 Yalow, R S. (1978) in Les Przx Nobel pp.
and Hood, L. E. (1982) Proc, Natl Acad. Sci. Biol. Chem. 214,619--632 243-264, Almqvist and Wiksell International,
USA 79, 5435-5437 44 Tatemoto, K. (1982)Proc. Natl Acad. ScL USA Stockholm
16 Grube, D. and Forssmann, W. G. (1979) Horm. 79, 5485-5489
Metab. Res. 11,589--606 45 Terenius, L. and Wahlstr6m, A. (1974) Acta Viktor Mutt is at the Department o f Biochemistry
17 Guillemin, R. (1978) in Les Prix Nobel, pp. Pharmacol. Toxicol. 35, Suppl. 1, 55 II, Karolinska Institute, S-104 O1 Stockholm,
160-193, Almqvist and Wiksell International, 46 Thim, L. and Moody, A. J. ( 1981 ) Regud. Pept. 2, Sweden.
Stockholm
18 Guillemin, R., B ~ u , P., B6hlcn, P., Esch, F.,
Ling, N. and Wehrenberg, W. B. (1982)Science
218,585-586
19 Hammarslr6m, S., Murphy, R. C., Samuelsson,
B., Clark, D. A., Mioskowski, C. and Corey,
E. J. (1979 ) Biochem. Biophys. Res. Comrnun.
91, 1266-1272
20 Herbert, E. and Ub.ler, M. (1982) Cell 30, 1-2 Mobit ms of l-mem
proteins: how are they
21 Hillyard, C. J., Abeyasekera, G., Craig, R. K.,
Myers, C., Stevenson, J. C. and MacIntyre, I.
(1983) Lancet i, 846-848
22 Hughes, J., Smith, T., Morgan, B. and Fothergill,
L. (1975) Life Sci. 16, 1753-1758
23 Kimmet, J. R., Pollock, H. G. and Hazelwood,
mod ed by
R. L. (1968) Endocrinology 83, 1323-1330
24 K6hler, G. and Milstein, C. (1975) Nature
(London) 256, 495--497
cytoskeleton?
25 Lazanas, L. H., DiAugustine, R. P., Johnke, JosephSchlessinger
G. D. and Hernandez, O. (1983) Science 219,
80-81
26 Li, C. H. (1968)Areh. Biol. Med. Exp. 5, 55--61 Measurements o f the lateral diffusion of many membrane proteins show that the
27 Li, C. H., Evans, H. M. and Simpson, M. E. viscosity of the lipid matrix does not control their lateral mobility. Several experiments
(1945)J. Biol. Chem. 159, 353-366 suggest that further interactions with the underlying cytoskeleton limit the mobility of
28 Luben, R. A., Bmzeau, P., B6hlen, P. and membrane proteins. By analogy to the role of the actin-spectrin network :m
Guillemin, R. (1982) Science 218, 887-889
29 MacCannell, K. L. and Lederis, K. (1983) Fed.
erythrocytes, it is possible that the recently discovered 'spectrin-like' molecules in
Proc. Fed. Am. Soc. Exp. BioL 42, 91-95 nucleated cells regulate the mobility of membrane proteins in these cells as well. Two
30 McKusiek, V. A. (1980)J. Here& 71,370-391 new approaches provide important information on membrane-cytosketeton
31 Montecucchi, P. C., Anastasi, A., de Castiglione, interactions. Measurements of the rotational diffusion of membrane proteins probe
R. and Erspamer, V. (1980)Int. J. Peptide Pro- their local constraints and aggregation state. Microinjection of fluorescently labelled
tein Res. 16, 191-199 cytosketetal elements into living cells allows the investigation of dynamic processes
32 Montecucchi, P. C., de Castiglione, R., Piani, S.,
in situ.
Gozzini, L. and Erspamer, V. (1981 ) Int. J. Pept.
Protein Res. 17, 275-283
33 Morris, C. J. O. R. and Morris, P. (eds) (1976) It is now well established that the plasma components x~,a°. In all the versions of FPR,
Separation Methods in Biochemistry, Pimaan membrane of the animal cell is an important the lateral mobility of molecules on the cell
Publishing, Bath cellular organelle. Many cellular processes surface is determined by measuring the rate
34 Mutt, V. (1982) in Vitamins and Hormones are initiated as a consequence of molecular of local changes in the concentration of
(Munson, P. L., Glover, J., Diczfainsy, E. and interactions which occur on the plasma fluorescent markers of cell-surface
Olson, R.E., eds), Vol. 39, pp. 231-427. membrane. Some of these processes molecules. A small region on the cell sur-
Academic Press, New York
depend on the dynamic properties of the face which is labelled with fluore,.w,em
35 Nakanishi, S., Inoue, A., Kita, T.. Nakamura,
M., Chang, A. C. Y., Cohen, S. N. and Nutria, molecular constituents of the plasma mem- molecules is illuminated with a focused
S. (1979)Nature (London) 278, 423--427 brane; namely, on their ability to translate laser beam. A brief and intense pulse of
36 Porath, J., Carlsson, J., Olsson, I. and Belfrage, and rotate within the fluid lipid matrix. light causes an irreversible photobleaching
G. (1975)Nature (London) 258, 598-599 of scxae of the f l ~ in the ilhm'fin-
37 Rivier, J., Spiess, J.. Thorner, M. and Vale. W. D e t m n h m t i o n of the lateral mobility of ated region. The laser is then turned down
(1982) Nature (London) 300, 276-278
ceH-sm'faee molecules to monitor the fluorescence intensity within
38 Schally, A. V. (1978) in Les Prix Nobel, pp.
201-234, Almqvist and Wiksell International, Several approaches have been used to the spot without causing further bleaching.
Stockholm analyse the motion of membrane As fresh fluorophores diffuse into the
39 Seidah, N. G., Gossard, F.. Crine, P., molecules 17.=°. The most common method bleached region the intensity increases, and
Gianodakis, C., Routhier, R. and Chr~tien, M. is called fluorescence photobleaching the diffusion coefficients of the labelled
(1980) in Precursor Processing in the Biosyn- recovery (FPR) (for review see Refs 17, molecules may be calculated from the rate
thesis o f Proteins (TAmmennan, M., Mumford. 20). Different versions of FPR were exten- of this process 7-t7.2o.
R. A. and Steiner, D. F., eds), Vol 343, pp.
sively used by various laboratories in the
443-447, The New York Academy of Sciences,
New York last six years ~°. These studies provided a General pattern of the latend mobility
40 Siedel, V. W. and Seh6~e, H. H. (1969) in wealth of interesting results concerning the The general pattern which emerges from
Handbuch des Diabetes Mellitus. Band l dynamic properties of many membrane the investigation of the lateral mobility of
~) 1983,ElsevierSciencePublishersB.V., Amsterdam 0378- 5912/83/$O1.00
T I N S - September 1983 361

many membrane molecules is as follows. between band 3 and the cytoskeletal web, Dynamic aspects-of
First, lipid molecules diffuse rapidly with a thus reducing its apparent diffusion coeffi- membrane--cytoskeleton interactions
diffusion coefficient (D) o f - 10-s cm2 s -1 in cient; or, alternatively, a simple 'friction- In order to investigate dynamic aspects
a continuous membrane matrix over macro- like' interaction between the inner portion of membrane--cytoskeleton interactions, it
scopic distances without being partitioned of band 3 protein and the dense cytoskeletal is important to employ methods which can
into closed regions. The mobility of the networkXl,19. be used on intact living cells. Unfortu-
lipid probe seems to be limited by the nately, most studies concerning mem-
viscosity of the lipid matrix ~.17.2o. Spectrin-like molecules in non-erythroid brane-cytoskeleton interactions utilize
Second, most membrane proteins diffuse cells fixed cells and therefore cannot provide
about fifty times slower than lipid For many years it was thought that spec- direct information on the dynamic properties
molecules [D = 2 - 6 × 10-1° cm 2 s-X], trin existed only in erythroid cells. How- of these structures.
and a substantial fraction of membrane pro- ever, several spectrin-like molecules have New methods were recently introduced
teins are immobile on the time scale of FPR recently been discovered in non-erythroid to cell biology which allow the investiga-
experiments 7.17.2°. Thus, it seems that the cells, including proteins immunologically tion of dynamic processes on living cells. It
viscosity of the lipid matrix does not control cross-reactive with chicken erythrocyte is now possible to determine the rotational
the lateral mobility of many membrane c~-spectrin in avian and mammalian diffusion of cellular molecules using the
proteins 7.17.~s.20. neurons, Schwann cells, lens cells, muscle method of time-resolved phosphorescent
Several experiments suggest that inter- cells, and in endothelial and epithelial emission and anisotropy t. In contrast to
actions between plasma-membrane proteins cells ~4, and proteins called fodrin from the translational diffusion coefficient, for
and cytoskeletal elements provide cons- brain lb and TW 260/240 from the terminal example of a membrane protein, the rota-
traints limiting the lateral mobility of cer- web of the intestinal microvillis. Like ery- tional diffusion of the molecule provides a
tain membrane proteins. It was reported throcyte spectrin, these proteins have rod- sensitive measure for changes in its size TM.
that the cross-linking of ceil-membrane pro- like shapes, bind calmodulin, and have A comprehensive analysis of the rotational
teins by concanavalin-A initiates a global been localized to the cytoplasmic side of the and lateral diffusion of the membrane
inhibition of mobility of other cell-surface plasma membrane 14. Peptide maps suggest receptor of epidermal growth factor (EGF)
molecules even when the latter do not that ankyrin, fodrin and spectrin are struc- was recently reported 1o.23.
directly interact with the lectin, and drugs turally related. Similarly, immunoreactive EGF binds to randomly distributed
that disrupt microtubules partially release ankyrins were also found in non-erythroid membrane receptors which rapidly cluster
the constraint 6'7. cells14. The widespread abundance of spec- mainly on coated pits and become internal-
Furthermore, the lateral diffusion coeffi- trins and ankyrins in nucleated cells is very ized via receptor-mediated endocytosis~'2.2.
cient of membrane proteins is 20-50 fold striking. By analogy to the role proposed At 4°C the receptors remain dispersed and
larger in membrane blebs (D - 10 -g cm ~ for the actin-spectrin network on erythro- they diffuse with a rotational correlation
s -~) than in intact cells21. Similarly mem- cytes (see Fig. 1), it is possible to extrapo- time of 20-50 #s 23. This corresponds to a
brane proteins incorporated into artificial late a similar role for this network in many free rotation of a single receptor molecule in
lipid bilayers diffuse with D ~ 5 × 10-9 nucleated cells; namely, that transmem- a lipid matrix with apparent viscosity in the
cm 2 s 1 (for review see Ref. 17). Hence, in hrane proteins will interact with the range of 1-5 P. However, the lateral diffu-
the absence of an underlying cytoskeletal actin-spectrin web which will, in turn, reg- sionofEGFreceptorat 4°C (D = 3 × 10 lo
network, receptor mobility approaches the ulate their mobility and function. Indeed, it cm 2 s -t) indicates that the viscosity of the
maximum theoretical prediction for a large was reported that fodrin is concentrated in lipid matrix is not the major limit of the lat-
protein in a viscous lipid bilaye#. ~7.ts. the cortical cytoplasm of neurons and eral diffusion of the receptoP °. Hence, the
moves down the axons by axonal trans- rotational diffusion of EGF receptor seems
The erythrocyte as a model system for port TM. Moreover, the capping of cell- to be controlled by the viscosity of the lipid
investigating membrane--cytoskeleton surface molecules in lymphocytes and cul- matrix, whereas the lateral mobility of the
interactions tured fibroblasts is accompanied by the receptor molecule seems to be limited by
The erythrocyte membrane provides an redistribution of fodrin molecules in these interactions in addition to viscous drag 1°23.
excellent model system for the investiga- cells TM. The differential constraints on the transla-
tion of specific interactions between mem-
brane components and the cytoskeletal
network. It has been proposed that the band 3
cytoskeletal network of erythrocytes is
composed of spectrin, actin and band 4.1
lipid lipid
protein (see Fig. 1)5. This three-
dimensional network interacts with band 3 matrix matrix
protein (an integral membran~ protein
which functions as an ion channel) through
a protein called ankyrin (see Fig. 1)s. The
lateral diffusion coefficient of band 3 pro-
tein in intact red blood cells is D = 4.5 ×
10- ~ cm 2 s-L However, in spectrin-
deficient erythrocytes it is D = 2.5 × 10 -9
4,1 ""~" "'
cm 2 s -1 (Ref. 11).
Two different models could account for Fig. 1. A schematic description o f the cytoskeleton o f erythrocytes and its interaction with a transmembrane
the decrease in the mobility of band 3 pro- ' protein (according to Re[. 4). The cytoskeletal network is composed ofspectrin rods attached to each other by
tein in intact red blood cells: either a kinetic oligomerfc actin and by band 4.1 protein. The transmembrane protein band 3 interacts with the
model which assumes a certain affinity cytoskeletal network via the protein ankyrin.
 Fig. 3. (Above) bluorescence reeovery after photobleaching ~j N-uctm m ~Ue.~:~
fibers. Stress fibe~ o f living microinjected cells (A ~were bleached by u ]bcu,~ed &~scr
beam (1.5 lain radius'). The photobleached line i~ mdtcated hy #re arrows and the
double line in (B). The video.intensified image of the cell was photographed from the
screen o f the television monitor prior to the bleaching (AL and 2 rain (B/, / 0 mtn (f)
and 35 rain (D)after bleaching. Bar 10 iJ,rn. (Taken from Ret" ! ~

Fig. 2. (Above, left) ( omparison o f the distribution o/cellular acun labelled with fluoreseein-phalloidin and with microinjccted R-atria. ( 7iicken gtzzard cells were
mh'roinjected with R-actin and fixed either 30 rain (A and B) or 24 h (C and D) later, permeabilized and labelled with fluorescein-phalloidin which labels &e total
cellular aetin. Panels A and C show the distribution oJ R-actin, and panels B and D show the distribution o f fluorescein-phalloidin in the same cells. Notice the
closely related patterns o f R-actin and flm~rescein-phalloidin in injected cells and the absence o f R-actin from the neighboring, fluorescein.labelled celL~. Bar ,
I0/am. (From Ref. 13. )

tion versus rotation of the receptor demons- in Fig. 2. The virtually overlapping pattern The conclusion from this analysis is
trate an important distinction between local of the two markers indicate that R-actin is a that actin in cytoskeletal structures can
diffusion (around the axis of the molecule) valid probe of the cellular actin. exchange with its soluble mobile pool with
and lateral motion over macroscopic dis- Two main conclusions were drawn from a relaxation time of 2-10 min. Thus, stable
tances .l. If we assume that the receptor .these experiments. First, R-actin inside the cytoskeletal structures exchange their con-
spans the plasma membran e , it is possible cytoplasm of living cells can be divided into stituents by dissociation/association reac-
to envisage a model in which the receptor two classes with respect to its mobility: tions with the surrounding soluble pool.
interacts with a dense cytoskeletal network actin molecules within stress fibers, focal This process could be regulated by many
which on the one hand impedes macro- contacts and filamentous structures are factors including other cytoskeletal ele-
scopic motion, but on the other hand does 'immobile' on the time scale of FPR experi- ments or local concentrations of ions. The
not affect the rotational diffusion of the ments; and actin within the interfibrillary regulated polymerization reaction could, in
receptor molecule around its axis **. space is mobile with D = 2 - 3 x 10 -9 cm ~ turn, affect many cellular processes which
A new approach for the investigation of s ~) (Ref. 13). Second, slow fluorescence are influenced by the cytoskeleton.
the distribution and dynamic properties of recovery occurs also on photobleached Acknowledgements
cytoskeletal elements involves the micro- areas of either stress fibers or focal contact I would like to acknowledge fruitful discus-
injection of fluorescently labelled cyto- areas. The time required for complete sions with my colleagues E. L. Elson, G. M.
skeletal elements into cultured cells. We recovery of fluorescence in these areas is in Hillman, Y. Yarden, T. Kreis, B. Geiger, T.
have recently applied several methods to the range of 2-10 min. Fig. 3 shows Jovin, R. Zidovetzki and M. Schwartz. This
analyse the intracellular dynamics of fluorescence recovery into a photobleached research was supported by grants from NIH
(CA-25820), Stiftung Volkswagenwerk and the
rhodamine-actin (R-actin), which was micro- line across a parallel array o f stress fibers
US-Israel Binational Foundation.
injected into living cells (for review see within a living cultured cell. The fluores-
Ref. 12). Image-intensification fluores- cence recovery into the bleached line was Re.~in~ list
cence microscopy was used to tbllow the recorded by a video-intensified time-lapse 1 Austin, R. H., Chan, S. S. and Jovin, T. M.
fate of the microinjected actin, and FPR camera. The dark photobleached line (see (1979) Proc. Natl Acad. Sci. USA 76.
experiments were used to measure the Fig. 3B) was clearly apparent 2 min after 5650-5654
mobility of R-actin within the living cells *~. the photobleaching. However, it gradually 2 Barak, S. L. and Webb, W. W. (1982)£ Cell.
A comparison between the distribution of Biol. 95,846--852
disappeared (Fig. 3C) and was not detect-
3 Bennet, V. and Stenbuck, P. (1979) J. Biol.
the microinjected R-actin and the total cel- able 35 rain later (Fig. 3D). Interestingly, Chem. 254, 2533--2541
lular filamentous actin visualized by label- the time required for this slow recovery is 4 Branton, D, (1982) Cold Spring Harbor Syrup:
ling with fluorescein-phalloidin (a toxin independent of the size of the photo- Quant. Biol. 46, 1-5
which binds to filamentous actin) is shown bleached area TM. 5 Branton, D., Cohen, C. M. and Tyler, J. (1981)
T I N S - September 1 983 363

Cell 24, 24-32 12 Kreis, T. E. and Birchmeier, W. (1982) Int. Rev. (Ehemstein, G. and Lecar, H., eds), Vol. 20, pp.
6 Edelman, G. M. (1976)Science 192,218-226 Cytol. 75,209~227 197-226, Academic Press, New York
7 Elson, E. L. and Schlessinger, J. (1978) in The 13 Kreis, T. E., Geiger, B. and Schlessinger, J. 21 Schlessinger, J., Elson, E. L., Webb, W. W.,
Neuroscience Fourth Study Program (Schmitt, (1982) Cell 29,835-845 Yahara, l., Rutishaaser, U. and Edelman, G. M.
F. D. and Worden, F. G., eds), pp. 691-701, MIT 14 Lazarides, E. and Nelson, W. J. (1982) Cell 31, (1977) Proc. Natl Acad. Sci. USA 74,
Press, Cambridge, MA 505-508 ll00.-lll4
8 Glenney, J. R., Jr, Glenney, P. and Weber, K. 15 Levine, J. and Willard, M. ( 1981 )J. Cell Biol. 90, 22 Schlessinger, J., Shechter, Y., Willingham, M. C.
(1982) Proc. Natl Acad. ScL USA 79, 631-643 and Pastan, I. (1978) Proc. Natl Acad. Sci. USA
4OO2-4OO5 16 Levine, J. and Willard, M. (1983) Proc. Natl 75,265%2663
9 Haigler, H. T., McKanna, J. A. and Cohen, S. Acad. Sci. USA 80, 191-195 23 Zidovetzki, R., Yarden, Y., Schlessinger, J. and
(1979)J. ('ell Biol. 81,382-395 17 Peters, R. (1981) Cell Biol. Int. Rep. 5,735-760 Jovin, T. M. ( 1981 ) Proc. Natl A cad. Sci. USA
10 Hillman, G. M. and Schlessinger, J. (1982) 18 Saffman, P. G. and Delbruck, M. (1975) Proc. 78, 6981~5985
Biochemistry 2 I, 1667-1672 Natl Acad. Sci. USA 72, 3111-3113
11 Koppel, D. E., Sheetz, M. P. and Sehindier, M. 19 Schlessinger, J. (1983)Biopolymers22,347-353 Joseph Schlessinger is at the Department o f
(1981) Proc. Natl Acad. Sci. USA 78, 20 Schlessinger, J. and EIson, E. L. (1982) in Chemical Immunology, The Weizmann Institute
3576-35811 Methods o f Experimental Physics: Biophysics o f Science, Rehovot 76100, Israel.

We welcome comments from our readers. Short communications

Letter to the Editor stand the best chance of publication. The Editor reserves the right
to take extracts from the longer ones.

Treating animal behaviour as a human and animal decision-making precisely. The way this is done depends on
game processes, but that there is a clear analogy whether the problem is best regarded as a
SIR: between the process of a human player game in which animals interact in pairs, as
In a recent T I N S article2 Calvin Harley and choosing between alternative strategies and occurs in contests over resources, or as a
John Maynard Smith presented a stimula- natural selection choosing between alterna- game between an individual and a field of
ting analysis of animal learning mechan- tive genetically determined behavioural other contestants, as would be appropriate
isms based on evolutionary game theory. strategies over evolutionary time. Thus, in for male display strategies in species in
Since the basis of their approach may be an evolutionary game the strategies are which each female chooses one mate from a
unfamiliar to many T I N S readers, this note behavioural phenotypes, the payoffs are group of displaying males. Here we will
introduces the elements of evolutionary changes in Darwinian fitness (the product consider only the former case. In addition,
game theory, a relatively new but increas- of reproductive rate and survival), and the some specific assumptions about the nature
ingly important technique for analysing criterion for selecting the winning strategy of the population and the way in which
behavioural evolution4. is evolutionary stability 4. The last of these strategies are inherited are necessary. For
As originally developed by economists a, concepts is central to the logic of evolutio- analytical simplicity an infinitely large,
game theory is a branch of mathematics nary game theory. random-mixing population and asexual
intended to analyse human decisions in Broadly defined, the evolutionarily reproduction are usually assumed.
situations where there is a conflict of inter- stable strategy (ESS) for a game is that Table I shows the general form of an
est between two or more parties and to strategy which when adopted by most evolutionary game in which individuals
which the following conditions apply: members of a population provides a higher interact in pairs and there are just two alter-
(a) each player can choose among a clearly payoff than any alternative strategy within native strategies, X and Y. The table is a
defined set of strategies, where each some specified set. The ESS may be either payoff matrix in which the cell values (a, b,
strategy is a specification of what he would one of the pure strategies from the strategy c and d) give the payoffs to the row strategy
do in any situation in which he found him- set or a mixed strategy of the form 'play when played against the column strategy,
self during the game; pure strategy X with probability P and pure Thus a is the payoff to strategy X played
(b) the payoffs (gains or losses) for adopt- strategy Y with probability 1 - P'. To against itself, b the payoff to X played
ing any given strategy against any given apply the ESS concept to an evolutionary against Y, and so on. The key to finding the
opponent's strategy are measurable along a problem it is neccessary to define it more ESS for this, and similar, games is to corn-
single value scale for all strategy pairs;
TABLE 1. The general form of a two-player, two-strategy, symmetric evolutionary game
(c) the strategy giving the highest payoff to
a player depends on the strategy of his (a) Payoff matrix
opponent(s); and Player l's slrategy Payoff to player 1 when player 2 plays strategy:
(d) each player chooses the strategy giving X Y
him the highest expected payoff given cer- X a b
tain specific assumptions about how his Y c d
opponent will play. (b) Conditions for four ESSs
Condition ESS
The solution to a game (i.e. the strategies
1. a~> c a n d b > d
adopted by each player) is highly depen- OR
dent on the nature of these assumptions. a>candb=d Pure X
While it is evident that many social 2. a < c a n d b ~ < d
interactions between animals involve con- OR
flicts of interest, similar to those of humans, a=candb<d t:hlre Y
3. a < c a n d b > d Mixed: play X with probability P, and Y with
it is not initially obvious why a theory
probability 1 - P where
incorporating rationality should be relevant
P = ( b - d)/(b + c - a - d)
to animal behaviour. The reason is not that 4. a > c a n d b < d Either pure X or pure Y
there is necessarily any similarity between