FLOCK

FORMATION TO CYCLICAL

OF TWO PARIDS IN RELATION SEED PRODUCTION IN A

PINYON-JUNIPER

WOODLAND

KIMBERLY A. WITH L3 AND MICHAEL L. MORRISON 2

•Departmentof Biological Sciences, NorthernArizonaUniversity, Flagstaff, Arizona86011USA,and 2Department of ForestryandResource Management, Universityof California, Berkeley,California94720USA

AI•STRACT.--Cyclical productionof seedsin a pinyon-juniper(Pinusedulis-Juniperus monosperma) woodland enabledus to assessthe effectsof resourceabundanceon the winter foraging ecologyand flock formation of two avian seed predators,the Mountain Chickadee(Parus gambeli)and the Plain Titmouse(P. inornatus).During the winter of low seed abundance (1986/1987), chickadeesand titmice convergedin microhabitatuse. Chickadeesexhibited a coarse-grainedresponseby selectivelyforaging in areaswith greater ponderosapine (Pinus

ponderosa) densityduring the winter of high seedproduction,but shiftedto a fine-grained useof microhabitatduring the following yearby randomlyforagingin differenttree species. Titmicedisplayeda fine-grainedresponsein microhabitatuseirrespectiveof seedabundance, but foraged significantlymore in ponderosapine during the winter of low seedproduction (becomingmore like chickadeesin use of tree species).Conversely,foraging behavior was consistentbetween years,and the two specieswere separablebasedupon the use of juniper substrates(e.g. chickadeesforagedmore on juniper needlesthan titmice,titmice foragedmore on the ground beneathjuniper). The convergencein microhabitatuseby chickadeesand titmice during the winter of low seedabundancemay be attributedto the prevalenceof mixed-species flocks.During the year of a mastseedcrop, chickadeesand titmice foragedsingly or in pairs 80%of the time and were never observedtogether. Half of all individuals were observedin flocksthe following year,and two thirds of flockingchickadeesand nearly all (88%)gregarioustitmiceparticipated in mixed flocks.Monospecificflocksof chickadeesused lessjuniper and foraged distinctly from titmice(e.g.probedmore,peckedless,gleanedfrom an inverted position),but converged in these characteristics in the presenceof titmice. Only two titmice ever occurredin mixedspeciesflocks,yet up to six chickadeesformed thesemixed-speciesflocks.Chickadeesapparentlyjoin titmice,assubstantiated by the observedshiftsin foragingecologyby chickadees in the presenceof titmice, and perhapsgain knowledgeof resourcelocationsfrom resident titmice (chickadeesare potentiallyaltitudinal migrantswithin our studyarea).We observed an increasein socialityduring periodsof low seedabundance,which supportsthe proximate role of resourcelevels in promoting flock formation, but does not preclude the possibility that other factorsthat are a consequenceof low resourceabundance(e.g. decreasedtime available for vigilance) provide the primary impetus for flocking behavior. Received18 July 1989,accepted 18 January1990.

PINYON-JUNIPERwoodlands exhibit marked

cyclesin the production of seedsand berries. Massproductionof seedsoccursevery 5-6 years in pinyon pine (Pinusedulis)and every 2-3 years in juniper (luniperusmonosperma), such that in some years no seedsor berries are produced (Baldaand Masters1980). This local synchrony in seed production may be related to escaping

depletion by overwhelming birds and mammals that feed upon seeds("flooding the system" [e.g. Balda 1987]). Two avian seedpredatorsthat winter in the pinyon-juniperwoodlandsof northernArizona are the Mountain Chickadee(Parusgambeli)and the Plain Titmouse (P. inornatus)(Shrout 1977, Balda and Masters 1980). Given the extreme an-

nual variation in resourceproduction,we were interested in the responsesof these two seed 3Presentaddress:Departmentof Biology,Colorado predators in terms of flock formation and forState University, Fort Collins, Colorado 80523 USA. aging ecology(microhabitatuse, foraging be522

The Auk 107:522-532.July 1990

July1990]

Flocking Behavior ofTwoParids

havior) during a winter (1985/1986) with a mast crop of both pinyon seedsand juniper berries, and during a winter (1986/1987) with almost no seedand berry production. Specifically,we examined the flocking behavior of chickadees and titmice in relation to seed production during each winter, to determine whether differencesin foragingecologyexistedbetweenthese congeners,and to ascertainwhether foraging ecologychangedbetween winters in response

523

tributed to the increased advantage of heterospecificover monospecificflocksbecausetotal scanningrange of the flock could be increased becauseof the presenceof different speciesthat foraged in different locations. Thus, we made severalpredictionsin regard

to flock formation and foraging ecology of Mountain

Chickadees

and Plain

Titmice

in re-

sponseto changesin annual productivity of seeds:(1) chickadees and titmice will alter their

foraging ecologyin responseto differencesin If one subscribesto the "increasedforaging food abundance between winters; (2) both efficiency" hypothesisas the primary conse- specieswill exhibit increasedsociality during

to differences

in food abundance.

quence of flock formation (Krebs et al. 1972; Caraco1979a,b), then it is expectedthat Moun-

the winter of low seed abundance; and (3) con-

sequently,each specieswould be expectedto divergein foragingcharacteristics, particularly flocksduring the winter of low seedabundance when participating in mixed-speciesflocks as to increasethe likelihood of locatingscarceand comparedwith when foraging in monospecific patchily distributed resources.Formation of flocksor alone.Alternatively,if socialityfacilflockswill bring individuals and speciesinto itatesresourceacquisition,then convergencein closercontact,and competitiveinteractionsmay foraging ecologywould be expectedof species counterpotential benefitsdue to increasedfor- in mixed-speciesflocks. It should be possible, aging efficiency.Speciesthat flock during times therefore, to distinguishbetween the contrastof low resourcelevels ("ecologicalcrunches," ing predictionsof the "competition"hypothesensuWiens 1977)therefore should foragemore sis and "social facilitation" hypothesisregardflock formation. distinctly in mixed-speciesflocks than when ing mixed-species flockingwith conspecifics. Specieswithin flocks METHODS of oakwoodlandbirdsin Arizona were spatially segregatedin microhabitat use. For example, tain

Chickadees

and

Plain

Titmice

will

form

Bridled Titmice (P. wollweberi)altered their foraging ecology (foraging stance,substrateuse, and position in canopy)when in the presence of Bushtits (Psaltriparusminimus)(Austin and Smith 1972). Willow Tits (Parus montanus) avoided trees in which Crested Tits (P. cristatus)

foraged, and shifted their position within the canopy of the tree in the presence of either Crestedor Great (P. major)tits (Alatalo 1981). Conversely,socialfacilitationor "copyingbehavior" predictsincreasedsimilaritiesbetween specieswhen foraging in mixed flocks. Morse (1978) observedseveralunambiguouscasesof copying behavior in Blue Tits (P. caeruleus) attracted to sites previously or concurrently occupied by other foraging individuals (conspecifics as well as other species).Krebs (1973) experimented with mixed-species flocks of Black-capped(P. atricapillus)and Chestnutbacked(P. rufescens) chickadeesin an aviary to demonstrate thatflockmembers(of bothspecies) convergedin foraging behavior in responseto successfulindividuals (of either species).This

led Krebsto proposethat sociallearningplayed an important role in flock behavior and con-

Our study was conductedduring the winters (No-

vember-March) of 1985/1986 and 1986/1987 in a 20-

ha areaca.26 km north of Flagstaff,CoconinoCounty, Arizona. Vegetation on the study area was representative of a transitionbetweenpinyon-juniperwoodlandscharacteristic of lower elevations(1,670m; upper sonoran life zone) and ponderosa pine (Pinus ponderosa) forestscharacteristicof higher elevations (2,100 m; transition life zone) in the southwest. MICROHABITAT

ABUNDANCE

We quantifiedvegetationabundanceand structure (microhabitat)to characterizethe habitat and to compare bird speciesuse relative to vegetation within this pinyon-juniper habitat. Thirty-five 20-m-radius (0.13 ha) plots were established50 m apart from an arbitrary starting point in the study area in 1987. In each plot we counted all trees 7

m in height(estimated visually).Thesecategories were selectedto separateyoung (reproductivelyimmature) trees(< 2 m) and pinyon and juniper trees(generally 10

m. Variables related to undergrowth were not recordedbecauseshrubswere rare, and grassand forb

[Auk,Vol.107

represent increasinglevels of successby the model in distinguishingbetween the groups.A high percentageof species-centered points(microhabitatuse)

coverwassparse or coveredby snowduringourstudy. classifiedincorrectlyas randomly centeredpoints (microhabitat abundance) would indicate that the

speciesis using microhabitatin proportion to abundance(fine-grainedresponse). Theconceptof "grain" The locationof foraging birds (explainedbelow) isusedtodescribespeciesresponses toenvironmental servedas the centerof a 20-m-radiusplot. Microhab- heterogeneityto provide an understandingof how itat usewasquantifiedasdescribedfor quantification different species"perceive"their environment (AdMICROHABITAT

of microhabitat

USE

abundance.

dicott et al. 1987).

Comparisons betweenspeciesand years.--Two-tailed t-tests were used to evaluate differences

in microhab-

FORAGING BEHAVIOR

itat use between specieswithin and between winters

as well as for both years combined(overall species comparison).The assumptionof homogeneityof variancebetweensampleswasevaluatedby Levene'stest; few comparisons departedfrom this assumptionand a pooled variance estimate subsequentlywas used. Comparisonswere consideredsignificantlydifferent if P < 0.05.

We performeddiscriminantanalysiswith forward stepwiseinclusionof variablesto characterizemicrohabitatuseof the two species.Variableswere entered initially into the equationif the variable significantly (P < 0.1,F-test)improvedthe discriminationbetween groups. We accepteda significancelevel of P < 0.1 for entry of variablesinto discriminantanalysissoas not to exclude variablesthat were nonsignificant(P > 0.05) between speciesin the univariate tests(i.e. t-tests),but that may have provided significantcontributionsin this multivariateanalysisat later steps in the procedure.Multicollinearitybetweenvariables was reducedby eliminating one of any pair of variables with

an r-value

of >0.7.

The variable

retained

for possibleinclusion in discriminant analysiswas the one with the highestF-value for separationbetween groups.

Discriminantfunction equationswere developed for each of the two winters of study and for both winters combined.To determinethe temporalvalidity of these equations,we examined the ability of equations developed from one winter to correctly classifydata (as to one of the two species)collected during the other winter. We validated the discriminant functionsfor eachwinter by randomly dividing the data set for each winter (and the overall data set

for both winters combined),and using this random subsetto develop the discriminant function and the remaining data to determine the accuracyof classification. The assumptionunderlying discriminant analysis of equal variance-covariance matrices between groupswas evaluatedwith Box's-M statistic.

Useversus abundance ofmicrohabitat.--A three-group discriminantanalysiswas run to determine whether each speciesused microhabitatin proportion to microhabitat

abundance.

Because a classification

accu-

racyof 33%is expectedby chancefor eachgroupin a three-group discriminant analysis, percentages > 33%

Foragingactivity increasedafter 0800, peaked between 1000-1300 (85% of all observations),and de-

creasedsharply after 1400. Admittedly, this pattern of diurnal variationin foragingintensitywasamplified further within our data set by the concentration of our efforts during the periods of the day when birdsapparently were mostactive. We systematically traversedthe study area, and when a foraging bird was encountered, we waited 10 s (to reduce bias due to observer disturbance

and towards individuals

for-

aging in conspicuouslocations) and then recorded

activityfor 10-60 s.To describethe foragingactivities and locationsof thesetwo species,we recordedforaging substrate(where foraging motions were directed;e.g. twig, smallbranch,mediumbranch,large branch, and trunk); foraging activity (e.g. glean, probe);distance(m) movedby flying or hovering; distance(m) moved by hopping, using wings only for balance;verticalforagingheight (m); tree height (m); and horizontallocationof bird in canopymeasured as percent distancefrom tree bole (all heights and distanceswere visually estimated as we were trained to make visual estimates;Block et al. 1987).

Data for substrateuseand foragingmode were converted to percent use for each individual to standardize observationperiodsof different lengths. Once we recorded data on an individual, the bird

was not followed. We recordeddata on only one individualper flockto minimizebiasdue to potentially correlatedactivities of flocking individuals. The observer resumed traversing the study area in the direction opposite to that in which the bird or flock disappeared.This minimized encounteringthe same individual or flockagain.Encounterratesof birdsby observerswere calculatedfor each species,and statisticalcomparisons (t-tests;comparisons significantly different if P < 0.05)were madebetweenspeciesfor eachwinter and betweenwintersfor eachspeciesto assess the distribution

and abundance

of individuals

(or flocks,becausedata on only one individual per flockwere recorded)throughoutthe studyarea. Comparisons betweenspecies and years.--Differences between speciesin foraging behavior within each winter and for bothwinterscombined(overallspecies comparison)were evaluated using two-tailed t-tests

July1990]

Flocking Behavior ofTwoVarids

525

as for analysisof microhabitatuse.Similarly, differencesbetween yearsin foraging behavior for each specieswasassessed with a two-tailed t-test.Discriminant analysiswasemployedto characterizeforaging behavior of the specieswith the same criteria and validation proceduresas describedpreviously for analysisof microhabitatuse. FLOCKING BEHAVIOR

When we encountereda foragingbird, we recorded whether the bird wasforagingsingly,asa pair, or in a flock (>-3 birds),and the speciescompositionand number of individuals if foraging in a flock. Percentageof time observedforaging singly, in pairs, or in monospecificor heterospecific flockswasassessed for eachspeciesand betweenwinters.We usedt-tests to test for differences

in mean

flock size between

speciesand betweenyears(comparisonssignificantly different if P < 0.05). Differences in microhabitat use

and foragingbehaviorbetweenthe two specieswere assessed separatelyfor monospecificand heterospecific flocks using Mann-Whitney tests. To examine

ßwhether chickadeesand titmice foraged differently in mixed-species flocksthan when in the presenceof conspecifics, the foragingecologyof eachspecieswas contrastedbetween these different assemblages.

RESULTS

MICROHABITAT

ABUNDANCE

Theunderstorycomprisedmostly(95%)small (7-m tall were ponderosapines(10 stems/ha). Live foliagewasconcentratedmainly (63%)below 3 m.

MICROHABITAT

USE

Comparisons betweenspecies andyears.--Mountain Chickadeesforagedin areasof significantly (P < 0.05) denser ponderosapine than Plain Titmice, and in areaswith greater foliage cover between 3 and 7 m in height during winter 1985/1986 (Appendix 1). During winter 1986/ 1987, chickadeesused significantly denserju-

niper in the midcanopy(2-7 m) than titmice. Chickadeesdifferedbetweenyearsonly by for-

eqV

V

526

WrrHANDMOP, RISON

[Auk,Vol. 107

TABLE2. Percentsuccess,as classifiedfrom discriminant analysisof Mountain Chickadees(MOCH) and Plain

Titmice (PLTI) microhabitatuse, and use of microhabitatvs. abundance,during winters 1985/1986 and 1986/1987 in northern Arizona. Values in parenthesesare resultsof subgroupclassification(validation) usinga randomsubsetof 50%of that year'sobservations; NA = not applicable. Predictedgroup Winter

Year / actual

group

1985/1986

Winter 1986/1987

MOCH

PLTI

Abundance

69 (82) 32 (53) 3 (0) 72 (91) 23 (53)

22 (0) 26 (12) 11 (6)

9 (18) 42 (35) 86 (94)

MOCH

PLTI

Abundance

1985/1986

MOCH PLTI Abundance MOCH PLTI

NA

28 (9) 77 (47)

37 48

63 52

23 (36) 28 (47) 11 (50) 56 (64) 81 (73)

40 (0) 52 (13) 3 (0) 44 (36) 19 (27)

1986/1987 MOCH PLTI Abundance

MOCH PLTI

NA

70 24

30 76

37 (64) 21 (40) 86 (50)

aging in areaswith lessdensefoliage coverbe-

tain Chickadeeswere more specificin their use

tween 3 and 5 m during the second winter.

of microhabitatthan Plain Titmice during the

Titmicesignificantlyreduceduseof juniper (but juniper was still used in proportion to its abundance) and increaseduse of 2- to 7-m tall ponderosapines between years. Discriminant analysis identified four variables that significantly separatedmicrohabitat usebetween speciesduring the first winter (Table 1); three variablesrelated to use of ponderosapine and areaswith foliage cover between 3 and 7 m in height were previously identified by univariate tests and were discussedabove. The discriminant function classifiedcorrectly

first winter.

The discriminant

function

distin-

guished between chickadee microhabitat use and microhabitat

abundance data, but was un-

able to successfullyclassifytitmice (Table 2). Validation proceduresusing a subsetof the data from

the first winter

resulted

in about

three

quarters of the chickadees (70%) and titmice

(76%)being identified correctly.

ca. 75% of individuals based upon microhabitat use data from the first winter (Table 2). Vali-

During the second winter, the discriminant function was less successfulin discriminating the speciesand randomly centered points, although microhabitat abundancedata remained a distinct group (86% classificationsuccess; Table 2). These relationships were considerably

dation of the discriminantfunction usinga sub-

weakened

set of the data resulted

validate the model as speciesmicrohabitat use

in 50% of the titmice

being incorrectlyclassifiedas chickadeeswith respectto microhabitatuse.The model derived from the secondwinter and applied to habitat use data from the first winter correctly identified •nany (70-76%) of the chickadeesand tit-

and

when

microhabitat

a subset of data was used to abundance

data

became

in-

separable. FORAGING BEHAVIOR

mice.

Comparisons betweenspecies andyears.--We enDiscriminantanalysisseparatedspeciesby use countered individuals of either speciesat 35of 2- to 7-m tall juniper during the secondwin- min intervals (SD = 18.9) over both winters ter (Table 1). Classificationwas poor, however, (1985/1986:27 = 34 + 18.9 min; 1986/1987:27 = as 81% of titmice and 44% of chickadees were 36 + 19.0 min; P > 0.05, t-test). No significant misclassified.

The discriminant

function

from

differences

in encounter

rates were

found

be-

the first year misclassified60% of chickadees tween species for either winter (1985/1986: Mountain Chickadee 27= 36 + 28.2 min, Plain and approximately50% of titmice (Table 2). Use versusabundance of microhabitat.--Moun- Titmouse 27= 41 + 14.7 min; 1986/1987: Moun-

July1990]

Flocking Behavior of TwoVarids

527

TABLE3. Discriminantanalysisof Mountain Chickadeeand Plain Titmouseforagingbehaviorduring two winters in northernArizona.Variablesretainedfor discriminantanalysisresultedfrom the stepwiseinclusionof variablesthat significantly(P < 0.1, F-test)improvedthe discriminationbetweengroups(see text for details). Winter 1985/1986

Eigenvalue Canonical

correlation

Chi-square (df)

1.229 0.743

19.328 (5)

P

Variables • and correlation

Winter 1986/1987

0.383 0.527

47.702 (9)

0.002

JUTW JUSE JUHOP JUNE JUSBMB

Box's-M (P)