TREE VOI. 3, IIO. /‘, .

/u/y 1988

The Evolutionof PartialMigration

in Birds
Per Lundberg
Partial migration, i.e. when one fractiofl of significant differences in the life
fhe populationis migratory and the other histories of individual birds might
sedentary, appears to be a widespread be found-. For such populations it
phenomenon among many animal taxa, seems reasonable to assume that
ranging from insects to higher vertebrates. the decision whether or not to mi-
Partial migration in Girds was first grate is not a completely random
documented for several Holarctic popu- event, but is governed by mechan-
lations many decades ago. The evolution isms shaped by natural selection.
and maititenance of this particular There appears to be no consist-
migratory system have only recently 6edn ent taxonomic pattern to the
more thoroughly examined, but our hnowl- occurrence of partially migratory
edge and understanding of the problem is populations. Since Nice’s’ demon-
still incomplete. Currently, one of the main stration of the phenomenon in song
concerns is the fitness balancing of the two sparrows (Melospiza melodica 1, a
behavioural alternatives, i.e. whether large number of other passerines
migrants and residents within a population have been included in this cat-
are equally fit or if one of the categories is egory. Examples are found both
inferior and making ‘the best of a bad among granivores (e.g. many Euro-
situation’. Closely tied to this question is pean fincheslo and North American
the proximate regulation of the migratory sparrow+’ 1I and insectivores (e.g.
and sedentary habits. II has been thrushes11,13 and North American
suggested that a social dominance system blackbird+\. Partial migration also
might 6e powerful enough to keep this occurs among many Holarctic
migration system going; alternatively, a ducksi4, shorebirds’5 [e.g. lapwings
population might 6e divided into two gen- ( Vanellus vanellus), woodcocks
etically distinct morphs with different pre- (Scolopax rusticola) I and raptors”’
programmed migratory behaviours. (e.g. the Accipiter hawks).
theoretical analysis’“, that partial
migration should be most frequent
in areas with the highest winter
climate variability. Such areas are
often found at intermediate lati-
tudes.
If it is true that intraspecific com-
petition plays a significant role in
these populations, one would pre-
dict that the sedentary fraction of
the population should consist of
individuals somehow superior in
gathering and protecting the scarce
food resources. Such birds might.
for example, be more experienced
(i.e. oldest) or socially dominant
(due to age, sex, body size etc.).
There are indeed indications that,
for example, resident chaffinches
I Fringilla coelebsl, European black-
birds (Turdus merulal, robins
(Erithacus rubecula) and song spar-
rows belong to these categories,
although it is by no means certain
which, if any, of several possible
individual asymmetries are the
most important in determining the
migratory status8.‘5.*‘~-23. As dis-
cussed later, alternative hypoth-
eses to this straightforward
condition-dependent control have
been suggested.

The seasonal migration of birds Conditionsfor partial migration Fitness balancing

is striking in its magnitude and Partially migratory bird popu- Until recently, two related
in the variety of species- and lations seem to be found in most hypotheses have tried to explain
population-specific patterns. Some Holarctic regions, particularly where the evolution and maintenance of
species cover vast distances each the winter conditions are neither the partial migration system (see
year; others only travel some ten extremely harsh, nor especially Ref. 24 for review]. Common to both
kilometres between the breeding benign. Hence, at intermediate lati- is the assumption of a stable sys-
and wintering grounds. In some tudes (in Europe approximately tem with complete fitness equaliz-
species there are also obvious dif- 40-65”Nl in the northern hemi- ation: in the long run, the payoffs to
ferences at the intrapopulation sphere, winter resources are migrants and residents should be
level. This phenomenon is usually apparently sufficient for many exactly the same. It was thought
called differential migration and in- populations, provided that there is that this would be the only way the
creasing attention has been paid to a reduced population density at system could persist.
it during the last decadel-5. that time of the year. Since it is Von Haartman2’ suggested that
Most commonly, differential generally assumed that intra- non-migrants, suffering a relatively
migration means that all indi- specific competition for winter food higher winter mortality at more
viduals of the population migrate, resources is significant in many northern latitudes, would be com-
but the distance travelled varies bird populations in temperate pensated by a higher reproductive
between categories of birds. In this areas1’,‘7,‘“, this seems to be the rate. The reason for this, he
case there is probably no potential main prerequisite for partial assumed, would be that residents
for essential life-history differences migration to occur. One would could initiate breeding earlier in
between categories (P. Lundberg, therefore predict an increasing superior habitats or territories.
Ph.D. thesis University of Ume& proportion of partially migratory Hence, differential mortality would
Sweden, 19851. However, in partially populations with increasing lati- be exactly balanced by differential
migratory populations, i.e. when tude, until we reach areas where reproduction acting in the opposite
one fraction of the population is the winter conditions are so bad direction, making individuals of the
migratory and the other sedentary, that all individuals are forced to two categories equally fit.
leave the breeding areas. This Lack’s26 hypothesis, on the other
seems indeed to be the case, for hand, is based on the assumption
Per Lundberg is at the Department of Wildlife instance, for blackcaps (Sylvia atri- that varying winter conditions
Ecology, Swedish University of Agricultural capilla) in Europe19. It has also would in some years lead to a high-
Sciences, S-901 83 lImei, Sweden been proposed, on the basis of a er mortality for residents than for
172
TREE vol. 3, no. 7, July 1988

/
BODY WEIGHT
migrants, and in other years the
?everse. In the long run, this would
In average give both categories
equal payoffs. Lack did not make
jny assumptions about the repro-
ductive success of the two cat-
20
-gories. Thus, only differential mor- z MIGRANT
:alities would account for the
2 1
s:oexistence of migrants and resi-
dents.
These two hypotheses contain
<assumptions that still seem to be
appropriate, e.g. differential mor-
rality and reproduction, but they
:tre not able to predict either the
proportion of the two categories, or
the decision rule(s) governing Fig. I. Body-weight development, molt, and migratory restlessness IZugunruhel of two representative
tvhether to migrate or not. individual robins lErithacus rubeculal measured in the laboratory during their first autumn. The birds,
originating from the same population, were hand-raised and kept in identical simulated natural
individual asymmetries day-length conditions. Only the ‘migrant’ type shows the typical body-weight development of a migrant
bird, coinciding with the period of migratory restlessness These and similar results with other species
Since the time of von Haartman
suggest genetic dimorphism as the mechanism for partial migration From Rel 117with permission
and Lack, it has become clear that
Lndividual asymmetries should be
(onsidered. This has led to the vary from year to year depending with blackcaps and E:uropean
(:onclusion that partial migration on the initial proportions of the robins, that the determination of
could be regarded as a pure individual conditions. partial migration has a strong gen-
ctrategy with two condition- etic component, i.e. a partially
dependent tactics: ‘migrant’ and Geneticdimorphism migratory population is genetically
’.esident’7,23. Individual conditions Some years ago, a German re- dimorphic with respect to rnigratory
such as age or sex would then de- search group concluded, on the behaviour29-s’. This was not an en-
termine the migratory status. De- basis of results from experiments tirely new idea. Lack15.Lh suggested
pending on the relative benefits
znd costs to either category for
t-leing migrant or resident, the Rd(67) R 9 (56) tvt@ (22) M9 (28)
categoi-y’s behaviour would be
// \ i \ li \ / ’
\
selected for accordingly. For in-
stance, if juveniles had low chances R M R M M R M R
cf reproductive success during 100

I
t ieir first breeding season
irrespective of migratory habit, it
would pay them to migrate if the
casts of winter survival were higher
50
at the breeding ground than in
M,intering quarters elsewhere lin-
eluding the costs of migration). 0 !l
Adults, on the other hand, might
benefit from early breeding in high
quality territories. Thus, being resi- R ad. (80) R juv. (43) Mad. (29) M juv. (21)
dent might ensure an adult a good / \ / \ i ‘\ i \
breeding territory, which could out-
R M R M M R M R
v,eigh the increased mortality risks 100
during winter at the breeding

ground.
A conditional strategy, however,
rarely produces tactics with equal
fitnessz7. Rather, the inferior tactic
uill persist by doing the ‘best of a
bad iob’28. Hence, if there is a
sl.rong conditional element
vlllved in the decision whether to
migrate or not, we would expect
that (I I the fitnesses of the two
categories would not necessarily be
equal, and (21 the proportion
n-igrants in the population should
I
50
0
in-
Fig. 2. The change of migratory behaviour in successive years according to sex I upper panell and age
(lower panel) in mid-European blackbirds (Turdus merula). Above each pair of bars is assigned the
status of the birds (M=migrant: R=resident) during a preceding winter, and the arrows indicate which
status the birds had in a following winter. Numbers indicate the number of birds in which the status had
been determined in two successive years. There was a significant tendency for migrants to become
of residents the following season rather than the reverse (~2 test; KO.001 1. This might indicate that age
(or some correlated trait) is an important condition for residency From Ref I ? with permission.

173
 TREE vol. 3, no. 7, July 1988

the genetic dimorphism hypothesis riumj’. However, individual mixed

is that the fitnesses of the two ESSs seem to be rare in nature,
categories should be equal; this partly because of the strong
question has not yet been com- tendency for condition depen-
pletely resolved. Unfortunately, the dence27,28, which makes this
arguments of these authors for the alternative less probable in partial-
evolution of partial migration ly migratory populations.
assume group selection without Alternatively, partial migration is
i I ( --I stating how this might be the expression of two coexisting
9* 1 q
operating3”m3’. Moreover, a bal- strategies in the population, each
Fig. 3. The inclusive fitness (NGFI of bet hedging anced genetic dimorphism will only of them specifying a single tactic:
parents as a tunction of q, the proportion of migrants ‘migrant’ or ‘resident’. This would
result if the fitness of the genotype
in the brood. 7 denotes the periodicity of harsh winters
(here ‘migrant’ or ‘resident’) is mean that the two behavioural
at the breeding ground IO=always mild; I=always
harsh; 0.5=randomly mild and harsh winters). if related to its frequency32,33, alternatives are genetically deter-
T = I it would pay the parents to produce migrant which was not assumed by mined and coexisting in an ESSt. If
offspring only If T i I, some offspring should become Berthold3’, Berthold and Querner2” so, one would predict that the indi-
migrants and some residents. q* indicates the equilib-
and BiebachjO. viduals’ behaviours should not vary
rium proportion of migrants in the brood if 7 =O 5. For
any -r < I, the equilibrium proportion of migrants in the between years. A migrant should
brood does not equal the equilibrium proportion of Frequency-dependent selection always be a migrant and a resident
migrants in the population when selection acts directly Most hypotheses so far have should stay in the breeding area
on individuals, not the brood optimum. Therefore q’ is usually implicitly, a
assumed, every winter, irrespective of con-
usually not a stable equilibrium. From Ref. 7 with
permission.
strong density dependence for the ditions and environmental irregu-
evolution of partial migration7,17J3. larities. Data from mid-European
However, a most neglected aspect blackbirds, however, seem to show
is the obvious basis for frequency that this is not necessarily true”
a genetic dimorphism, but only dependence as well. Swingland2’ (Fig. 21. Indeed, individuals do
more recently were the critical tests was the first to suggest explicitly change behaviours from year to
performed29-3’. Although this work that frequency dependent selec- year and tend to do so in a particu-
showed that offspring bred and tion might be operating, an idea lar direction; former migrants more
tested in the laboratory indeed dif- that was recently developed in often become residents than the
fered in, for instance, body-weight more detai17. One can distinguish at reverse (cf. also Ref. 91.
development and migratory rest- least three possible mechanisms
lessness (Fig. I I. such results do not for such selection: Parentalbet hedging
unambiguously lead to the rejec- (II Partial migration is a The basic argument for the gen-
tion of other hypotheses2’. Never- mixed (stochastic1 evolution- etic dimorphism hypothesis, as
theless, the focal prediction from arily stable strategy IESSI Biebach<O and BertholdS’ put it, is
with two phenotypic tactics that if the parents were able to
with equal fitness payoffs. produce both migrant and resident
(21 Partial migration is a bal- offspring, it would ensure them at
W, (dam.)
~~
anced genetic dimorphism, least some surviving young,
i.e. a mixed ESS at the irrespective of winter conditions or
population level where the migratory costs. A rather rigid gen-
two morphs coexist at an evol- etic control would therefore be
W, idom., sub.1
utionarily stable state I ESSt I safer for the population than an
Li W, (rub.)
with equal fitnesses. Again, unreliable environmental control ot
i ,” this is indeed what Biebach~O the fraction of migrants. Ignoring
; ~~-- and Bertholdj’ implicitly have the obvious group selectionist
assumed, although the proxi- argument, Lundberg7 was able to
mate mechanism they suggest show that parental bet hedging
Fig. 4. Conditional elements included in the (parental bet hedging1 actual- probably cannot be an evolution-
frequency-dependent choice The fitness of migrants ly will have a different evol- arily stable system (Fig. 31. More-
(W,,I and of residents (W,I is illustrated as a function of utionary outcome7. over, Dominey17 has pointed out
the proportion of migrants in the population (PI In
13) Partial migration has that experimental results revealing
this example, conditions are determined by position in
the social dominance system ldom.=dominant, evolved as a conditional ‘genetic’ differences between indi-
sub.=subordinatel Dominant birds should in this ex- strategy with frequency- viduals do not necessarily mean
ample always be residents lW, ldom 1‘ , W, (dam t for dependent choice. The that the population is genetically
all pI Subordinate birds should always be migrants IW, fitness balancing in the latter dimorphic with respect to that par-
lsub 1< W, (sub I for all values of pi. Hence the fitness
of dominants is always greater than the fitness ot
case is still unclear. ticular trait. This is certainly true ii
subordinates, i.e subordinates are making the ‘best of In a mixed strategy at the indi- there is also a conditional element
a bad iob’ However, if the fitness functions cross at any vidual level, there should be no involved. ‘Even individuals ex-
value of p, indicating equal fitness for the two cat- individual asymmetries and the pressing the same condition-
egories, the situation becomes less predictable. Thus,
choice purely stochastic, ultimately dependent tactics may differ gen-
frequency dependence tends to equalize fitnesses,
whereas condition dependence tends not to. From Ref so adjusted that the two tactics etically as to the precise conditions
7 with permission have equal fitnesses at equilib- which will elicit particular tactics

174
‘TREE vol. 3, no. 7, Julv 1988

f.or instance, genetic variants might will reveal interesting aspects of I2 Lack, D. (19541 The NaturalRegulation of
Animal Numbers, Oxford University Press
use or emphasize different cues in the evolution and coexistence of
I3 Schwabl, H. (1983) 1. Ornithol 124,
switching from one tactic to different life-history solutions with- 101-l I6
another, or might switch at different in populations. As Greenberg6 has I4 Ogilvie. M A. (19751 Ducks ofBritaIn and
points along a gradient of a single shown, the survival and reproduc- Europe, T. and A.D. Poyser
c Je’27. tive components of fitness are 15 Lack, D. (19441 Br. Birds 37. 122--l 30;
143-150
heavily influenced by the migratory
I6 Newton, I. (I9791 Population Ecology of
C,)nditional strategy with frequency- strategy of an individual bird. Raptors, T. and A.D. Poyser
dependent choice However, these problems are not I7 Cox, G.W. II9681 Evolution 22, 180-192
Obviously, a condition-depen- necessarily elucidated by partially I8 Pulliam, H.R. and Milikan. C.C ( 19821 in
.4vian Biology IVol. 61 IFarner, D.S , King, J.R.
dent element in partial bird mi- migratory populations only, but in
and Parkes, K.C., edsl. pp 169-197
g!ation cannot be ignored. Albeit fact by most populations with dif- Academic Press
u Iproved, frequency dependence ferential migration. 19 Berthold, P. 119781 Experientia 34, 1451
n-ight also be operating. If so, a 20 Cohen,D.ll967)Am.Nat lOl.5--I7
problem arises since a conditional 21 Gauthreaux, S.A., fr. 11978) in Perspectives
References In Ethology(Vol. 3) (Bateson, P.P.G. and
strategy rarely gives the tactics
I Cauthreaux, S.A., Ir. C19821 in Avian Biology Klopfer. P.H., edsl, pp. 17-54, Plenum Press
equal fitness, whereas frequency- (Vol. 6) (Farner, D.S., King, J.R. and Parkes. 22 Dolbeer, R.A. II9821 1. Field Omitho/ 53,
dependent selection tends to do K. C., eds), pp. 93-168, Academic Press 28-46
SOAR. Fig. 4 illustrates a situation 2 Ketterson, E.D. and Nolan, V., Jr. 119761 23 Swingland, I.R. 1I9831 in The Ecology of
Ecology 57,679-693 Anima/Movement(Swingland, I.R. <and
w lere individual asymmetries are
3 Ketterson, E.D. and Nolan, V.. Ir ( I9791 Auk Greenwood, P. I., edsl, pp 102-I Ii
created by a social dominance 96,532-5X Clarendon Press
hierarchy. Similarly, of course, other 4 Ketterson, E.D. and Nolan, V., Ir. C19821 Auk 24 Baker, R.R. (1978) The Evolutionary
kinds of asymmetries could be 99,243-259 Eco/ogyofAnima/Migration Holmes and
analysed. In this particular example 5 Ketterson, E.D. and Nolan, V., Jr. II9831 in Meier
Current Ornitho/ogy(Vol. II Ifohnston, R.F, 25 von Haartman. L. II9861 Omis Fenn 45.
the outcome is rather simple. It
ed.), pp. 357-402, Plenum Press l-6
would always pay dominant indi- 6 Greenberg, R. 11980) in Migrant Birds in 26 Lack, D. f 19681 Oikos 19, I-G
vijuals to be residents since their the Neotropics IKeast, A. and Morton, E.S., 27 Dominey, W I. II9841 Am. Zoo/ 24
fitness function IW, (dom.11 is edsj, pp. 493-504, Smithsonian Institute 385-396
Press 28 Dawkins, R. II9801 in Beyond Nature/
greater than any alternative for all
7 Lundberg, P. I I9871 /. Theor. Biol. 125, Nurture? (Barlow, G.W. and Silverberg, I
values of p (the proportion of mi- 35 I-360 eds), pp. 33 l-367, Westview Press
gr,3nts in the population). The sub- 8 Lundberg, P. ( 19851 Behav. Ecol. Sociobiol 29 Berthold, P and Querner. U f lY821
ordinates, on the other hand, 17, 185-189 Experientia 38,805
9 Nice, M. ( 19371 Trans. Linn Sot. N.Y. 4. 30 Biebach, H. t I9831 Auk 100, 606-610
should be migrants, since their
6 l-247 31 Berthold, P. I 1984 I The Ring IO. 25 3-265
fitness function, if migrating IW, 32 Fisher, R.A. ( 19301 The Genetical Theory
IO Newton, I. II9721 Finches. Collins
(sub.)], is greater than the fitness II Fretwell, S.D. (19801 in MigrantBirdsin ofNatura/Se/ection. Clarendon Press
fuiiction for residency I W, (sub.) I for the Neotropics (Keast, A. and Morton, E S., 33 Maynard Smith, I. ( 19821 Evolution and
all values of p. If, however, the eds), pp. 517-527. Smithsonian Institute the TheoryofCames. Cambridge University
Press Press
fit less functions cross each other at
any value of p, the situation be-
comes equivoca17.

Concludingremarks
rhe idea of frequency-depen- In the next issue of TREE:
dent choice has not yet been
further explored. In fact, the rela-
tive fitness payoffs of the two tac-
tic; (or strategies) have not yet * Gaps in fossil teeth, J.S. Jones
been satisfactorily established,
alt lough measurements of repro-
du :tive variables have been Rodent dynamics and community processes, L. Hansson
made13. Nevertheless, we are at and H. Hentonnen
present left with two main hypoth-
eses: genetic dimorphism and
a conditional strategy, possibly Differential foraging for resources and structural plasticity
with frequency-dependent choice. in plants, M.J. Hutchings
There is empirical support for
bo! h of them, although no crucial
evidence has been presented in How do migrating birds cross the Sahara? F. Bairlein
terns of fitness calculations or
delerminations of frequency
dependencies. Paleoecological feedback and the Vendian-Cambrian tran-
l-respective of the final result, sition, M.A.S. McMenamin
however, a deeper understanding
of the partial bird migration system