SHORT
COMMUNICATIONS
Formalizing the Avoidance-ImageHypothesis:
Critique of an Earlier Prediction
SIEVERT ROHWER
Department
of ZoologyandWashington
StateMuseum,
Universityof Washington,
Seattle,Washington
98195USA
Two papers on ParasiticJaegers(Stercorarius
parasiticus)that were recently published in this journal
have reported testsof Paulson's(1973) avoidanceimage hypothesis:Arnason (1978) and Furnessand Furness (1980). The authors of these papers, as well as
Paulson(1973:270),fail to distinguishthe processby
which apostaticselection operatesfrom its end result--a frequency-dependentequilibrium between
different morphs.The consequenceof this confusion
of processand result is that both of the contrasting
(and seeminglycontradictory)outcomesof thesejaeger studiescould be predicted from the avoidanceimage hypothesis.Becauseeither of the two possible
but, in the following chapter,Curio reviews another
literature showing that predators preferentially attack odd prey ("odd" apparently meaning unusual
individuals, e.g. color variants, of the samespecies).
Third, so far as I am aware, all of this literature re-
latesto predators
respondingto unusualprey, but the
avoidance-imagehypothesisrelates to prey responding to unusual predators.For prey the problem is
ratherdifferent:to forageefficientlyprey speciesmust
filter stimuli such that they do not madly dash for
coverat every novel stimulus.In short,somediscrimination causingthem more strongly to avoid certain
danger (e.g. hawks of a color that have previously
outcomes of these studies would have been consisattacked) than any novel stimulus could be advantent with the avoidance-imagehypothesis,the stud- tageous,simply becausenovel stimuli must occur
iessayvery little aboutthe validity of the hypothesis. fairly frequently.
Paulson (1973), Arnason (1978), and Furhess and
Paulson (1973) developed and tested the avoidance-imagehypothesisto explain the remarkableinFurness(1980) largely fail to distinguishbetween the
traspecificvariability in coloration in raptors. Almechanismby which a frequency-dependenteffect
thoughmany seemskepticalaboutthe idea,it remains is obtained and the resulting frequency-dependent
the only generallydevelopedexplanationof the re- equilibrium between the morphs. All of these aupeated evolution of color polymorphisms in diurnal
thors assumethat rarity per se is the issue. Conseraptors.Accordingto the avoidance-imagehypothe- quently, all of them predict that the rare morph will
sis, a novel color morph may invade a raptor popu- always enjoy an advantagein encounterswith exlation if its prey have sufficientvisual acuity and inperiencedprey. This predictionis basedsolelyon the
telligence especially to avoid raptors that have
mechanismby which a searchimage (or, by analogy,
attackedthem previously. In short, the formation of
an avoidanceimage) is formed, namely, that a familavoidanceimagesis hypothesizedto causeprey more iarity with objectsmakes them easierto recognize
strongly to avoid hawks similar in color to thosethat
and respond to (see Curio [1976] for a review of the
have attacked them previously than other hawks literature on the formation of search images). Prewhose coloration is novel.
dictingthat the raremorphwill alwaysenjoya higher
The most cogent argument against this idea was prey-capturerate is valid under two conditions:either
put forth by an anonymousreviewer of an earlier (1) both morphsare equally easyfor prey to perceive
version of this paper. He argued that the whole idea (i.e. they are equally noncryptic),or, if they are not,
is nullified by the fact that most vertebratesavoid then (2) the rare morph must be held below the equiany strange stimulus, thus placing a rare invading
librium frequencythat would be achievedsolelyby
color morph at a strong disadvantage.I doubt this apostaticselectionbecauseof somedisadvantageit
argumentfor three reasons.First, it assumesthat spe- suffersoutsideof interactionswith prey.
cial avoidanceis necessaryfor the operation of the
Contrasting with this mechanistic prediction is a
avoidance-image
hypothesis,but this neednot be true. prediction based on the relative fitness of the two
A prey lacking an avoidanceimage for a particular morphswhen they are at their equilibrium frequencolor morph may simply fail to perceivethat hawk as cy. This prediction assumesthat nonforagingaspects
rapidly as it could have done if it had had an avoid- of fitness are unrelated to color. When this is the case
ance image for that color morph. Second,the litera- and when the morphsare near their equilibrium freture on the avoidanceof strange stimuli seemsinquency, then the per-individual foraging successof
consistent. In Chapter 3 of "The ethology of
the morphs should be nearly equal. In the caseof a
predation," Curio (1976) reviews a literature sup- light-dark color polymorphism, such as that of the
porting this idea for predatorsexposedto novel prey jaegers, the assumption of equal visibility of the
("novel" apparentlymeaning new speciesof prey), morphs is almost certainly false:one or the other of
971
972
ShortCommunications
the color morphs must be easierfor prey to perceive.
That morph which is least cryptic should be easierto
avoid and also may be easier to incorporate into an
avoidanceimage (see Girtleman and Harvey, 1980).
In jaegersit is not obviouswhich morph will be less
cryptic:light-phasedjaegersare probablymore cryptic when their victims see them against the sky, but
dark-phased jaegers may be more cryptic when their
victims see them against sea cliffs or the surface of
the sea. To simplify discussionI shall assumefrom
here on that, when the morphs are not equally easy
to perceive,light individuals are morecrypticto prey
than dark, a situation that probably holds for many
open-country Buteohawks. When the dark morph is
more conspicuous,the fundamental problem for the
avoidance-imagehypothesisis to identify conditions
[Auk, Vol. 100
which will occurwhen the frequency of dark-phased
predators is equal to the frequency of light-phased
predators.It is exceedingly unlikely that all of these
assumptions would hold and, therefore, that the
equilibrium frequencyfor the light and dark morphs
would be 1:1. Most iraplausible is that light and dark
forms would be equally easyto perceive, an assumption implicit in each of the preceding three conditional equalities.
Now consider the situation where light predators
are more cryptic. In this situation, predicting the
equilibrium frequency becomesan empirical problem for which
no relevant
data exist. The theoretical
problem is why a conspicuousdark morph should
ever invade a population of lights. Here, I address
the conditions required for the increase of a dark
under which dark individuals
could be favored when
morph when it is rare. As before, I simplify arguthey are rare. I do this below with the consequence mentsby assumingthat prey form avoidanceimages
that new and testable predictions are derived from
with equal easefor either morph, even though the
the avoidance-imagehypothesis.
dark morph is presumed to be more conspicuous.If
The equilibrium frequency between the light and
light predators are more cryptic, this implies that
dark genotypeswill occur when the effectsof color P(LIN) > P(DIN), P(LIEL) > P(DIED ), and P(LIEB) >
on nonforaging, X, and on foraging, F, aspectsof
P(DIEB). When theseinequalities hold, the necessary
fitnessare equal for the light, L, and dark, D, morphs: condition for the invasion of a dark morph is that
f(XL,?L)= f(Xo,?O).
P(DIEL) > P(LIEL ). This is the essenceof the avoid-
ance-imagehypothesis.It is not, however,a sufficient
condition for the invasion of a conspicuous dark
TO simplify arguments, ! shall assumethat XL = XD,
that is that color is only affecting foraging success. morph. The sufficientcondition for invasion occurs
when
Relevant to the avoidance-imagehypothesis are the
P(DIN).n + P(DIEL).I > P(LIN).n + P(LIEL).I (4)
factorsaffecting Fr and FD,which ! shall write as the
probability of capturing an individual prey that has
or, when
or has not had experience with the predator morph
times the frequency of that prey type.
I.[P(D I EL) - P(LI EL)] > n.[P(LIN) - P(D IN)]. (5)
F• = P(LIN)'n + P(LIED).d + P(LIEL).I + P(LIEB)'b
(2)
These inequalities follow from equations (2) and (3)
because,on invasion,d and b are nearly 0; therefore,
n + l • 1. Whether or not (5) will be true depends
F• = P(D[N).n + P(DIEL).I + P(DIED).d + P(DIEB)'b
on the relative frequency of prey that are experi(3)
enced, l, or inexperienced, n, with the light morph.
In equation (2) P(L,N) is the probability of a lightIf l is very small, then invasion by the dark morph
phasedpredatorcapturinga naive prey (naive prey cannot occur unless P(L IN) - P(D IN) • 0. From this
hold avoidanceimages for neither morph); P(L,ED) relationship it follows that a shortage of naive prey
and P(L,EL) are the probabilities of a light predator is required for the invasion of a dark morph when
capturing prey that are experiencedwith dark and dark is more conspicuousthan light.
light predators,respectively;and, P(LIEB) is the
These argumentsshould make it clear that the two
probability of a light predator capturing a prey havrequirementsfor the invasion of a conspicuouscolor
ing avoidanceimagesfor both light and dark pred- morph are, first, that the conspicuousmorph is more
ators.The probability termsin (3) are defined anal- successfulthan the cryptic morph at capturing prey
ogously.Frequenciesof prey that are totally naive or that have formed an avoidance image for the cryptic
are experienced with dark, light, or both predator morph, and, second,that there is a shortageof naive
morphsare symbolizedby n, d, l, and b,respectively, prey. The requirement of a shortageof naive prey
andn+d+l+b=l.
suggeststhat, among speciesof predators,thosethat
are most successfulon a per attack basiswill be less
If light and dark predatorswere equally easy for
prey to perceive,then P(LIN)= P(DIN), P(LIEL) =
likely to evolve color polymorphismsthan thosethat
P(DIED ), and P(LIEB) = P(DIEB). If it is alsotrue that
often missprey. Could it be that the PeregrineFalcon
avoidanceimagesare equally as easyto learn for light
(Falcoperegrinus),a speciesthat preys almost excluand dark morphs,then the foragingaspectsof fitness sively on birds, is not color polymorphic becauseit
for light and dark morphs will be equal when has an unusually high ratio of capturesper chase?
P(LIED).d = P(DIEL).I. This will occur when d = l,
A shortage of naive prey or victims seems quite
October 1983]
ShortCommunications
plausible for hawks of temperate areasin winter and
for jaegershunting around breeding seabirdcolonies. Naive prey accumulate either by new recruits
entering the population of prey or by experienced
individuals forgetting previouslyheld avoidanceimages. Becausemost polychromatic raptors prey on
higher vertebrates(Paulson 1973), memory may be
long enoughto contributesubstantiallyto a shortage
of naive prey and thus facilitatethe invasionof rare
andpresumably
lesscrypticdark-colormorphs.If the
invasion of a conspicuous,dark morph is permitted
by a shortageof naive prey, then the fitnessof dark
individuals will decline astheir frequencyincreases,
becausean increasing proportion of the prey will
form an avoidanceimage for the dark morph. This
frequencydependenceshould be sufficientto producea stableequilibrium between the light and dark
morphs.
At the risk of redundancylet me restatehow these
arguments affect the prediction drawn by Paulson
(1973), Arnason (1978), and Furness and Furness
(1980), namely that the rare morph should have a
higher per encounterrate of prey capture.This prediction is basedsolely on the mechanicsof avoidance-image formation and makes the implicit assumption either that the morphs are equally visible
or that the rare morph suffersa disadvantagein activities unrelated to foraging. If either of these assumptionsis true, the prediction should hold, becausefewer prey will hold an avoidanceimage for
the rare than for the commonmorph. Note that the
assumptionthat fewer prey will hold an avoidance
image for the rare morph includesthe assumption
that many attacks,even by the cryptic and presumablymorecommonmorph,areunsuccessful
and,thus,
that many "educated" prey exist.
In contrast,it may be assumedthat success
in prey
capturelargely or entirely determinesthe equilibrium frequency, and that most populations are near
equilibrium. In this case,neither morph will have an
advantage.This is so becauseselectionshould have
balancedthe opposingadvantages(1) of surpriseenjoyed by the more cryptic morph and (2) of a frequent lack of specialavoidanceon the part of experiencedprey that is enjoyedby the more conspicuous,
but less abundant, morph when it attacks.
At Arnason's(1978) study site, jaegersof the rare
light phase were significantly more successfulas
kleptoparasitesthan were dark-phased individuals.
Furness and Furness (1980) employed the victim's
speedof reactionto light- and dark-phasedpursuers
asan inversemeasureof hunting success.
They found
no difference in responsetimes to attacksby lightand dark-phasedjaegersat a different locality where,
again,light-phasedindividuals were rare. By the mechanics of apostatic selection and its attendant assumptions,Arnason'sresultssupportthe avoidanceimage hypothesisand those of Furnessand Furness
do not. In contrast,by the equilibrium frequencyat-
973
gument and its attendant assumptions,the resultsof
Furness and Furness support the avoidance-image
hypothesisand thoseof Arnasondo not. Clearly,more
and better data are required for an adequate intrapopulationaltest of this hypothesis.
I shall end by making two suggestionsfor future
fieldwork in this area, both of which follow from the
preceding considerations.First, simple observations
on unmanipulated populationsare unlikely to affect
the credibility of the avoidance-imagehypothesis
strongly unless naive and experiencedvictims can
be distinguished.Such might be possibleif attacks
by jaegersof the different color morphswere separately tallied for adult and for immature seabirds.If
naive and experienced victims could somehow be
distinguished,then two predictionsfollow: (1) the
more common(more cryptic) morph should be more
successfulin attackson naive prey, but (2) the rarer
(less cryptic, but less often avoided) morph should
be more successfulin attackson experiencedprey.
This testmight be particularlyfeasibleon Eleonora's
Falcon (Falco eleonorae),because at certain colonies
mostprey (fall migrant birds) may be assumednever
to have been attackedpreviously and becauseseveral
falcons of either or both color morphs often repeatedly attack a single prey (Walter 1979). Thus, the relative success
of light and dark falconson firststoops
would measure their relative crypticity and the relative success
of attacksby like color morphspreceded
by like and oppositecolor morphs precededby opposite would estimate the advantage a novel morph
would experiencein attackson experiencedprey.
Second, an experimental field test of the avoidance-imagehypothesis could be achieved by comparing the averagehunting success
of individualsof
different color morphs before and after their relative
frequencieswere alteredby removals.BreedingParasiticJaegersthat foragearound alcid or larid colonies
should be ideal for such a test because their numbers
are sufficiently low that their relative frequencies
could be altered easily by removals.After a substantial proportion of the individuals of one morph had
been removed, the remaining individuals of that
morph should enjoy a gradual increase in foraging
success.
The changemight be slow if mostpotential
victims had already been attackedby individuals of
both morphs, becausesome time would be required
for previously establishedavoidance images to be
forgotten. Over time, however, the average success
of the remaining individuals of the reduced morph
should increase, while that of the unmanipulated
morph should decrease.A hidden weakness of this
test,as well as the precedingone (if applied to jaegers),is the assumptionthat the color polymorphism
is relevant
to the time
over which
the observations
are made. BecauseEleonora'sFalconpreys primarily
on insectsin its nonbreedingseasonbut exclusively
on birds in its breeding season,this hidden testing
assumptioncertainly holds for this species.
974
Short
Communications
I thank Lev Ginzburg, Dolph Schluter, Craig Sargent, and an anonymousreviewer for their helpful
commentaries on this problem. I was supported in
part by a Harry Frank Guggenheimgrant during my
work on this project.
[Auk,Vol.100
FURNESS,
B. L., & R. W. FURNESS.1980. Apostatic
selection and kleptoparasitism in the Parasitic
Jaeger:a comment. Auk 97: 832-836.
GITTLEMAN,
J. L., & P. H. HARVEY.1980. Why are
distastefulprey not cryptic?Nature 286:149-150.
PAULSON,
D. R. 1973. Predator polymorphism and
apostatic selection. Evolution 27: 269-277.
LITERATURE CITED
WALTER,
H. 1979. Eleonora'sFalcon.Chicago,Univ.
ARNASON,
E. 1978. Apostaticselectionand kleptoparasitismin the Parasitic Jaeger.Auk 95: 377-
Chicago Press.
Received6 July 1982, accepted22 March 1983.
381.
CURIO,E. 1976. The ethology of predation. Berlin,
Springer-Verlag.
Kirtland's Warbler, Victim of Its Own Rarity?
HAROLD
F. MAYFIELD
9235 River Road, Waterville, Ohio 43566 USA
The recent history of the population of the Kirtland'sWarbler (Dendroicakirtlandii)presentssomebaffling questions.The population declined from about
500 singing males in 1961 (Mayfield 1962) to about
200 in 1971 (Mayfield 1972) under intolerable pressure from nest parasitismby the Brown-headedCowbird (Molothrusater), which has expanded its breeding range to include that of the warbler only in the
last 100 yr (Mayfield 1960). The cowbird has continued to increase.In the late 1960's it parasitized 70%
or more of all Kirtland's Warbler nestsand depressed
the production of young to lessthan one fledgling
per pair of adults per year (Ryel 1981). Measures to
control
the cowbird
started
in 1972 and are continu-
ing with remarkable success,reducing parasitismto
negligible levels (Shake and Mattson 1975). The production of young warblers immediately rose to rates
never seen before and has remained at these high
levels (Walkinshaw 1977, Nicholas L. Cuthbert un-
publ.), setting to rest any doubts about the fecundity
of the species.The decline in the population was
arrested, but the number of adults has not increased
as expected. The number of singing males has remained nearly level from 1971 to 1982, fluctuating
between 167 and 243 (Ryel 1982).
What factorshold the population down in spite of
excellent nesting successand survival of adults?
Plainly, some combination of factorshas been operating selectivelyagainstthe young with specialforce
in the last two decades.I will consider several possibilities, beginning with problems in the breeding
season.
Intuitively, most biologists turn first to the breeding habitat. Indeed, this is restrictedand specialized,
consistingof extensivehomogeneousstandsof young
jack pines (Pinusbanksiana)on poor sandy soil in
northern lower Michigan. This habitat is a transitory
phasein the forestsuccession
occurringnaturally after forest fire and in a modified form after cutting
and planting. We suspectthe habitat was most extensive in historical times during the 1880's and 1890's
in Michigan and perhapsalso in Ontario, Wisconsin,
and Minnesota at the height of lurebering in the region, when forest fires were frequent and unretarded. At that time
Kirtland's
Warblers
were
found
on
their wintering grounds in the BahamaIslands much
more frequently than today (Mayfield 1960). Jerome
Weinrich (unpubl.) has calculated that suitable hab-
itat alsodeclined40%from 1961 to 1971 along with
the recent decline in warblers. Thus, a probable maximum and a recent minimum in population have correspondedwith grosstrends in the extent of suitable
habitat. Without doubt, at some level of population,
the available habitat would limit the number of birds,
and, therefore, efforts to increase the habitat would
be a wise conservationmeasure,but field study has
not turned up evidence that the present habitat is
inadequate.
Nesting successis good. Unmated individuals are
rare. In nesting "colonies"the area usedby eachpair
always seemsmuch smaller than the spaceavailable
to them.
Measured
territories
have a mean
area less
than 4 ha (Mayfield 1960), but the area available on
occupied tracts usually amounts to more than 12 ha
per pair. When mapped, the territories are rarely
contiguouson all sidesas though crowded. In addition, many tractsof young pines that appear suitable
to human eyesare not usedby warblers.An example
is a large cutover tract in Oscoda County that held
the following numbersof singing malesin the years
beginning in 1975: 0, 6, 0, 1, 1, 10, 14, 22. If it was
suitablein 1976when it held six males,it was surely
not less so in the next three years when it had virtually none. In 1980 it became more productive.