Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Habitat partitioning by sparrows along an elevational gradient Repasky, Richard R. 1993

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata


831-ubc_1993_spring_phd_repasky_richard.pdf [ 5.49MB ]
JSON: 831-1.0086388.json
JSON-LD: 831-1.0086388-ld.json
RDF/XML (Pretty): 831-1.0086388-rdf.xml
RDF/JSON: 831-1.0086388-rdf.json
Turtle: 831-1.0086388-turtle.txt
N-Triples: 831-1.0086388-rdf-ntriples.txt
Original Record: 831-1.0086388-source.json
Full Text

Full Text

HABITAT PARTITIONING BY SPARROWSALONG AN ELEVATIONAL GRADIENTbyRICHARD RAYMOND REPASKYB.Sc., North Carolina State University, 1978M.Sc., North Carolina State University, 1984A THESIS SUBMITTED IN PARTIAL FULFILLMENT OFTHE REQUIREMENTS FOR THE DEGREE OFDOCTOR OF PHILOSOPHYinTHE FACULTY OF GRADUATE STUDIESDEPARTMENT OF ZOOLOGYWe accept this thesis as conformingto the required standardTHE UNIVERSITY OF BRITISH COLUMBIADecember, 1992©Richard Raymond Repasky, 1992In presenting this thesis in partial fulfilment of the requirements for an advanceddegree at the University of British Columbia, I agree that the Library shall make itfreely available for reference and study. I further agree that permission for extensivecopying of this thesis for scholarly purposes may be granted by the head of mydepartment or by his or her representatives. It is understood that copying orpublication of this thesis for financial gain shall not be allowed without my writtenpermission.(Signature) Department ofThe University of British ColumbiaVancouver, CanadaDate^2 1 0,2c-to.-1,,^ZDE-6 (2/88)ABSTRACTAlthough species replacements along environmental gradients are commonlyattributed to interspecific competition, predictions of what species distributions wouldbe in the absence of competition are usually lacking. I tested alternative hypothesesthat might account for species distributions including: food, predation, and habitatstructure. These factors were unable to account for the distributions of sparrowswintering along an elevational gradient in the Sonoran Desert of southern California,USA.Sage sparrows (Amphispiza belli), black-throated sparrows (A. bilineata) anddark-eyed juncos (Junco hyemalis) inhabit different vegetation types. Under the foodhypothesis, species occupy different habitats because they eat foods that occur indifferent habitats. Species ate seeds of the same species of plants, and the profitabilityof seeds (seed mass ingested / handling time) ranked similarly among sparrow speciessuggesting that species should have similar distributions. Also, species can apparentlyforage profitably outside of their customary habitats. Several measures of foodavailability suggested that two species are missing from habitats in which food is atleast as available as in occupied habitats. The measures included: food standing crop;food intake rate estimated from seed abundance, seed size, and handling time; andfood intake rates observed in an "introduction" experiment in which individuals ofeach species were placed in an aviary and allowed to forage in each habitat. Observeddifferences in food intake rate between habitats were small suggesting that specieswould be more broadly distributed if food shaped their distributions.Predation could be responsible for habitat distributions if species are safest frompredation in different habitats and predation risk is severe. Alternatively, predationcan not be responsible for habitat partitioning if all species rank habitats similarlyby safety. Data support the latter alternative. All species escape predators by fleeingto woody cover and feed near cover. Hence, they are likely to be safest in the habitatiiiwith the greatest amount of cover. Also, I recorded the rate at which birds scannedthe environment while they foraged in an aviary to test two predictions: (1) if speciesexperience similar changes in risk between habitats and if the same level of vigilanceyields the same probability of detecting predators, species should exhibit similarchanges in vigilance level between habitats; (2) if some species experience increased riskwhen moved from one habitat to another whereas other species experience decreasedrisk, species should exhibit dissimilar changes in vigilance levels. Two species for whichcomparisons were possible exhibited similar changes in vigilance between habitats.I also considered structural features of habitat that might affect foraging abilityor the risk of predation. Foraging microhabitats used by individual species were morewidely distributed than the species themselves, suggesting that species' distributions arenot limited by habitat structure.I addressed the competition hypothesis by testing two conditions necessary forcompetition: species must share limiting resources and they must deplete the amountof food available to one another. Species overlapped in the kinds of foods that theyate and in the types of microhabitats where they foraged. To test the food limitationhypothesis, I carried out a short-term food addition experiment. Bird densitiesincreased as the result of the food addition, supporting the hypothesis of food limitationin the short term. The chronology and magnitude of recruitment to experimentalplots illuminated the existence and dynamics of depletion. Initially less common, yetmobile species (mourning doves and house finches) recruited fastest and in the greatestnumbers to experimental plots, removing up to 72 percent of the total amount of seedsavailable to birds. The initially more common yet least mobile species (white-crownedsparrows, black-throated sparrows) recruited more slowly and in lower numbers. Muchless food was available to these later species than would have been in the absence of theearlier species.ivTABLE OF CONTENTSABSTRACT^  iiTABLE OF CONTENTS^  ivLIST OF TABLES  viiiLIST OF FIGURES^  ixACKNOWLEDGEMENTS  xiGENERAL INTRODUCTION^  1I . HABITAT DISTRIBUTIONS OF WINTERING SPARROWS ALONG AN ELE-VATIONAL GRADIENT: NO RELATION TO FOOD, HABITAT STRUCTUREOR ESCAPE COVER^  5Factors and predictions ^6Food^  7Predation ^7Habitat structure ^8Competition ^9Methods^  9Study site and species  ^9Data Collection^  13Bird census  14Seed standing crop^  14Habitat Structure  15Feeding Habits^  15Seed Handling Times  15Hypothesis testing^  16Results^  24Distributions of Birds^  24VTests of Hypotheses^  24Food^  24Predation  28Habitat structure^  37Competition  37Discussion ^  52Other possibilities ^  52Other evidence  53Future tests ^  57II . FORAGING SUCCESS AND THE HABITAT DISTRIBUTIONS OF WINTER-ING SPARROWS: A TRANSPLANT EXPERIMENT^  59Methods^  60Experimental design^  60Food intake rate  66Peck rate^  67Number of seeds per peck^  67Size of seeds eaten^  67Comparisons^  73Results^  73Food intake rates^  73Habitat comparisons  78Discussion ^  78III . PREDATION, FOOD, VIGILANCE AND THE HABITAT DISTRIBUTIONSOF WINTERING SPARROWS^  83Predation, vigilance and starvation  85Methods^  87Experimental Design^  90viVariables and data collection^  90Analysis^  94Results  96Food supply affects vigilance^  96Comparisons of vigilance rates between habitats^  96Discussion ^  100Vigilance and the tradeoff between food and predation ^ 101Predation and the habitat distributions of species  104IV . INTERSPECIFIC COMPETITION AND THE HABITAT DISTRIBUTIONSOF WINTERING SPARROWS: EVIDENCE FROM A FOODSUPPLEMENTATION EXPERIMENT^  107Possible inferences from a food addition  108Methods^  113Study Area^  113Experimental Manipulation^  113Bird Censuses ^  114Depletion  115Number of Species^  119Statistical Tests  120Results^  120Density  120Depletion^  120Number of Species^  127Discussion ^  132Food limitation^  132Depletion  135Competition among other granivorous birds^  136VIIGENERAL DISCUSSION^  138Summary of results  138Alternative hypotheses^  141BIBLIOGRAPHY^  142viiiLIST OF TABLESTable 1. Principal components analysis of habitat structure^  22Table 2. Sparrow species abundances in three habitats during two winters.^ 25Table 3. Diet overlap among species calculated as proportional similarity in seedspecies composition^  43Table 4. Similarity in relative profitability of seed types among sage sparrows, black-throated sparrows and dark-eyed juncos^  44Table 5. Overlap in use of microhabitats quantified as the rate at which observationswere misclassified by a discriminant function^  49Table 6. Foraging performance in the aviary  74Table 7. The fraction of potential food items that yields edible seeds^ 75Table 8. Levels of statistical significance associated with the relationship betweenvigilance rate and peck rate ^  99Table 9. Response of granivorous birds to a short-term food addition. ^ 123Table 10. Millet removal and extent of food depletion.^  130Table 11. Species richness on control and treatment plots  133ixLIST OF FIGURESFigure 1. Locations of study plots in the lower Coachella Valley, California. ^ 10Figure 2. Diet breadth as a function of number of individual birds sampled. ^ 18Figure 3. Food standing crop for three species of sparrows and sparrow populationdensities in three habitats over 2 winters. ^  26Figure 4. Food intake rates estimated from seed abundance, seed mass, and handlingtime. ^  29Figure 5. Distribution of distances from cover of foraging birds and randomly locatedpoints. ^  31Figure 6. Food density as a function of distance to cover. ^  33Figure 7. Distribution of distances from cover of foraging birds and their food. ^ 35Figure 8. Food standing crop discounted by the risk of predation. ^ 38Figure 9. Foraging locations of species plotted over polygons describing theavailability of microhabitats within habitats. ^  40Figure 10. Profitability of seed foods. ^  45Figure 11. Overlap in the foraging microhabitats of species. ^  47Figure 12. Total sparrow abundance and food abundance in three habitats in twoyears. ^  50Figure 13. Maintenance metabolism as a function of ambient temperature. ^ 54Figure 14. Floor plan of aviary. ^  62Figure 15. One block of the experimental design. ^  64Figure 16. Mean and standard error of seed mass in diet as a function of samplesize.  ^69Figure 17. Differences in foraging success between habitats. ^  76Figure 18. Hypothetical relationships between level of vigilance and the probabilityof starving. ^  88xFigure 19. Schematic representation of one block in experimental design. ^ 91Figure 20. Relationship between vigilance and food abundance for sparrowsconsuming similar seed types in pairs of habitats. ^  97Figure 21. A family of hypothetical starvation curves.  102Figure 22. Food intake rate as a function of seed density for two species differingonly in the time required to handle a single food type. ^  111Figure 23. Density of granivorous birds over the course of the food additionexperiment. ^  121Figure 24. Standing crop of millet over the course of the food addition experiment atSite 2. ^  124Figure 25. Trajectory of millet seed removal rate at Site 2 over the course of the foodaddition experiment. ^  128xiACKNOWLEDGEMENTSThis work was carried out on lands owned by the California Department of Fishand Game, U.S. Bureau of Land Management, U.S. Fish and Wildlife Service, U.S.Forest Service, University of California Land and Natural Water Reserve System, andMrs. Ruth Valeur of Palm Springs, California. Most of these lands are administeredby Deep Canyon Desert Research Center and the Coachella Valley Preserve. DeepCanyon Desert Research Center served as an excellent base of operations. I thank A.Muth and staff at Deep Canyon and C. Barrows of the Coachella Valley Preserve forfacilitating my work, their insights into desert life, and for their hospitality. G. Heintz,B. Boyle (two seasons) and B. Booth spent many long days in the field collectingdata. R. Gullison helped out for a few days. My graduate committee, D. Schluter,W. Neill, A. Sinclair, J. Smith and R. Ydenberg, provided useful guidance and adviceas did many but dwellers on a less formal basis. C. Benkman, P. Feinsinger, A. Frid,T. Price and D. Ward commented on drafts of at least one chapter of the thesis. Myexamining committee, M. Adamson, L. Gass, B. Noon, M. Pitt, and J. Wiens, madeuseful comments on the thesis. S. Welke gave it the final read. I was supported in partby a University of B.C. Graduate Fellowship. The project was supported by NaturalSciences and Engineering Research Council of Canada grants to my supervisor DolphSchluter. I thank Dolph for much more than just his money. His questions and insightswere invaluable, and his patience and encouragement unflagging.1GENERAL INTRODUCTIONSpecies replacements along environmental gradients are striking. A numberof species occur in sequence, sometimes with little or no overlap between theirdistributions (e.g., Terborgh 1971, MacArthur 1972, Wolcott 1973, Adams and Bernard1977). Early explanations of this pattern invoked competition. Darwin (1859) arguedthat competition is responsible for species replacements along elevational gradientsand latitudinal gradients. He reasoned that minor differences between species puteach species at a competitive advantage under different environmental conditionsalong gradients. The first explicit statement of the competitive exclusion principlewas by Grinnell (1904) in a discussion of the habitat distribution of the chestnut-backed chickadee (Lack 1971). Closely related species of temperate birds often dwellin adjacent habitats, and Lack (1944) argued that this pattern of distribution was theresult of interspecific competition. He reasoned that closely related species consumedsimilar foods and that differences in species' morphology were too small to restricttheir distributions. He supported his view with natural experiments in which thedistributions of species were observed in areas in which supposed competitors wereabsent. MacArthur (1972) provided theoretical insight into how competition mightresult in species replacements. The hypothesis that competition is responsible forelevational distributions has been tested frequently (e.g., Terborgh 1971, Terborgh andWeske 1975, Noon 1981, Schluter 1982, Price 1991).The makings of an alternative view also began early. Grinnell (1904) suggestedthat species distributions might be restricted by habitat itself. The distribution ofa species might expand until it reached habitats having characteristics that wereinsurmountable. Gleason (1926) proposed that species are distributed independentlyof one another according to the favorability of the environment to individual species.His hypothesis was supported by the absence of pattern in the distributions of large2numbers of plant species along elevational gradients (Whittaker 1951, 1956). Bowman(1961) suggested that species' food requirements might account for the distributionsof Galapagos ground finches (Geospiza spp.) among islands. He reasoned that eachspecies is adapted to different foods and that species distributions were shaped bythe distributions of their foods among islands rather than by interspecific competitionas Lack (1947) had argued (see Abbott et al. 1977 and Schluter and Grant 1982for a vindication of Lack). The suggestion that species distributions are shapedindependently by environmental factors has been raised (e.g., Sweeney and Vannote1978, Vannote and Sweeney 1980, Wiens and Rotenberry 1981, Strong 1983), but it hasrarely been tested (e.g., Abbott et al. 1977, Schluter 1982, Price 1991). A reasonabletest would compare species' actual distributions with distributions predicted fromenvironmental factors.In this thesis, I ask what factors are responsible for the habitat distributionsof three sparrows along an elevational gradient in the Sonoran Desert of southernCalifornia. Sage sparrows (A mphispiza belli), black-throated sparrows (A. bilineata),and dark-eyed juncos (Junco hyemalis) inhabit different vegetation types along thegradient (Weathers 1983). Present-day competition for food is an obvious factor thatmight be responsible for the distributions of these species. However, I have emphasizedfactors other than competition that might account for the distributions and testedthe competition hypothesis only indirectly. I considered the alternatives that food,predation and habitat structure account for species distributions. In each case, I testthe hypothesis that species specialize on single habitats because each species is so welladapted to a factor in one habitat and so poorly adapted to it in other habitats thatthe factor restricts its distribution. Under the food hypothesis, for example, speciesare suited to different types of foods, and their distributions are determined by thedistribution of food types among habitats. I tested the competition hypothesis byevaluating the conditions that are necessary for competition to occur. Competition is3unlikely to explain species distributions if there is no sign that it operates, whereas itcould shape distributions if the necessary conditions are present.I tested hypotheses using both observational and experimental data. Observationaldata presented in Chapter 1 provide an initial test of each hypothesis. The foodhypothesis is investigated by determining what is food for each species and bycomparing species' distributions with the distributions of their foods. The predationhypothesis is tested by comparing species' use of shrubby vegetative cover that canbe used to escape predators. Predation is unlikely to be responsible for habitatpartitioning if species have similar preferences for cover. The habitat structurehypothesis is investigated by describing the microhabitats that each species uses in itsusual habitat and asking if sites of that type are available in other habitats. Finally, thecompetition hypothesis is tested by determining whether species use similar resourcesor would do so if they occupied the same habitats. Also, I tested the competitionhypothesis by testing one of its major premises: food limits population abundance.Food limitation is evaluated by measuring the relationship between bird abundance andfood supply.Each of the three following chapters (i.e., 2-4) presents experimental evidencedealing with one of the hypotheses. The food hypothesis is pursued in Chapter 2. Anintroduction experiment measured foraging abilities of all three species in each of thethree habitats. These data were used to test whether each species achieves its highestfood intake in the habitat that it typically occupies, and whether significant tradeoffsin foraging ability exist between habitats. The introduction also tested the vegetationstructure hypothesis: any feature of habitat structure that affects foraging abilitystrongly enough to result in habitat specialization should be reflected in food intakerates.In Chapter 3, data from the introduction experiment are analyzed to test thepredation hypothesis. If predation is responsible for habitat specialization, species4ought to perceive different changes in risk between habitats and exhibit dissimilarchanges in vigilance. The predation hypothesis is tested against the alternative thatall species experience similar changes in risk between habitats and the prediction thatspecies exhibit similar changes in vigilance between habitats.The competition hypothesis is treated in Chapter 4. A short-term food additionexperiment was carried out to see if food was indeed limiting. Food must be limiting ifcompetition is to be responsible for species distributions. Additionally, the experimentprovides evidence on whether species are capable of depleting the amounts of foodavailable to one another.Finally in the general discussion, I integrate the results of the individual chapters.5Chapter IHABITAT DISTRIBUTIONS OF WINTERING SPARROWSALONG AN ELEVATIONAL GRADIENT:NO RELATION TO FOOD, HABITAT STRUCTURE OR ESCAPE COVERClosely related bird species often have abutting distributions along environmentalgradients. They probably use similar food, and it is possible that their distributions arerestricted by competition (Lack 1944, Svardson 1949, Terborgh 1971, MacArthur 1972).This reasoning is supported by evidence that species often have broader distributionsin areas where presumed competitors are absent (e.g., Cody 1974, Terborgh andWeske 1975, Noon 1981). Indeed, such evidence indicates that two-thirds of 91 birdspecies along an elevational gradient in the Andes appear to be limited by competition(Terborgh 1971, 1985).Abutting species distributions may be explained in other ways. There may be afactor in the environment whereby a species well suited to one range of values of thefactor is necessarily poorly suited to others (see Levins 1968). If so, there is only alimited potential for current competition to influence distributions. Possible factorsinclude food, predation and abiotic factors. For example, ground finches (Geospizaspp.) differing in size appear to be best suited to eat different sizes of seeds (Grant1986:134), and food, not competition, appears to be the major factor determiningtheir distributions along an elevational gradient (Schluter 1982). Despite the strongpossibility that food determines species distributions, the food hypothesis has beentested directly in only a few studies (e.g., Abbott et al. 1977, Schluter 1982, Schluterand Grant 1982, Price 1991).Hypotheses regarding species distributions can be tested using predictionsstemming from them (e.g., Terborgh 1971, Schluter 1982). Species' distributions canbe compared to distributions predicted from environmental factors (Schluter 1982).6The predictions themselves are based on knowledge of species biology and how thefactors being considered affect species abundances. By testing several factors in thisway, strong inferences (Platt 1964) can be made about the factors that limit speciesdistributions. Although such tests are weaker than direct experimentation, theynevertheless provide important information that is essential for planning experiments.In the present study, I ask what factors shape the habitat distributions of sparrowswintering along an elevational gradient in the Sonoran Desert of California. Sagesparrows (Amphispiza belli), black-throated sparrows (A. bilineata) and dark-eyedjuncos (Junco hyemalis) winter in different vegetation types (Weathers 1983). I testedalternative hypotheses that might determine the distributions, including food, predationand habitat structure and current interspecific competition for food.Factors and predictionsIn testing which factors might limit distributions, I ask whether factors vary alongthe elevational gradient in ways that are consistent with the hypothesis that they limitdistributions. I acknowledge that single factors rarely if ever limit species abundancesto the exclusion of all others and that factors may act simultaneously or interact. Forexample, food and predation are linked through the time budget if foraging placesa bird at greater risk of predation than do other activities (McNamara and Houston1987). My tests eliminate factors that vary along the elevational gradient in ways thatare inconsistent with the hypothesis that they limit species distributions. For example,if two factors vary along a gradient, one becoming more favorable for a species and theother less favorable, only the factor becoming less favorable can restrict the species'distribution. I emphasize factors that might limit distributions in such a way thatthey predict species occur one per habitat. Here, I outline the specific factors andpredictions stemming from them.7FoodFood alone could shape the habitat distributions of species if there is littleoverlap in the type of food different species can eat and species' foods occur indifferent habitats (Schluter 1982). This hypothesis predicts that abundances alongthe elevational gradient should be proportional to food availability and that sharpdiscontinuities in the types of food available occur between habitats.PredationForaging birds avoid areas of high predation risk (e.g., Lima 1990, Watts 1990,Watts 1991) suggesting that predation could influence species' distributions. Predationcould shape habitat distributions if species differ in the habitat in which they are safestfrom predators. If each species is very safe in one habitat and very much at risk inothers, then there is little opportunity for cohabitation. Species differences in the safetyof habitats may be rooted in methods used to escape from predators. Pulliam and Mills(1977) observed three different techniques of escape used by granivorous birds. One setof species foraged close to woody vegetation and fled to it when disturbed. Another setforaged solitarily farther from shrubby cover and crouched when disturbed, apparentlyrelying on crypsis. The third set foraged in flocks at long distances from cover and flewaway when threatened. Species that flush to cover suffer high predation rates awayfrom cover (Watts 1990), and those that tend to feed far from cover and fly off arereluctant to forage close to cover (Lima 1990).The predation hypothesis predicts that each species should be most abundant inthe habitat where it is safest, and less abundant in habitats of greater risk. All three ofthe sparrow species that I studied flush to cover to escape predation, suggesting that allshould be safest and most abundant in the habitat with the most cover. Clearly, thisprediction is discordant with the observation that species live in different habitats. Itested the predation hypothesis by asking whether all three species foraged closer tocover than expected by chance. The predation hypothesis would be supported if some8species prefer to forage close to cover whereas other species avoid cover. It would befalsified if all species forage closer to cover than expected by chance because that resultsuggests that all species are safest from predators while foraging close to cover.Although predation is unlikely to explain habitat partitioning because speciesrespond similarly to avian predators, predation can affect my expectations of therelation between bird distributions, abundance and food supply (Schluter and Repasky1991). For example, the risk of predation could be strong enough to reverse thepredicted distributions of species from those expected from food alone. Therefore, I testa joint predation-food hypothesis by comparing species' distributions and abundancesto food abundance discounted by the risk of predation.Habitat structureHabitats can be defined in terms of their structural features, and those featuresare obvious factors that could determine species distributions. This hypothesis is notexclusive of the others, because the effects of habitat should be mediated by food,abiotic factors, competition and/or predation. Yet, structural features of habitatcould influence the foraging abilities of birds or predation rates in ways that are notcaptured by my measures of food availability and safety from predators. For example,soil texture, leaf litter and herbaceous plants may affect a species' ability to search theground for food, and they are not reflected in measures of food standing crop.If habitat structure determines distributions, it should be possible to identifymicrohabitat components of habitat structure that are relevant to foraging and ask ifthe microhabitats used by foraging birds are themselves restricted in distribution. Ifforaging activity is restricted to particular microhabitats and those microhabitats areabsent outside of species' preferred habitat, some feature of habitat structure wouldthen be implicated. Ready availability of suitable foraging microhabitats outside ofpreferred habitats suggests that habitat structure per se is unlikely to account fordistributions.9CompetitionHabitat partitioning is consistent with the hypothesis that interspecific competitionhas shaped species distributions. I explored the competition hypothesis by testing forthe conditions that are necessary for competition. Species compete when they reducethe availability of shared resources that limit population abundance. I asked whetherspecies share resources and whether food might be limiting. Failure to observe eitherof these conditions suggests that competition is unlikely, whereas both conditions musthold true if species compete. First, I asked whether species share food types and foragein similar microhabitats. Second, I tested for food limitation by plotting the abundanceof all sparrows with food density across habitats. A positive relationship is predicted iffood is limiting. Birds in different habitats are different species in my study area, andhence, a positive relationship is likely to represent food limitation rather than simplythe local aggregations of birds that have moved to areas of abundant food (Schluter andRepasky 1991).MethodsStudy site and speciesThe elevational gradient was located in the Sonoran Desert of southern California,USA. It ran from sea level in the Coachella Valley in the vicinity of Palm Desert upthe side of Santa Rosa Mountain to 2660 m. Vegetation varied from very open desertscrub habitat on the valley floor to coniferous forest at the upper elevations (Figure 1see Zabriskie 1979 and Weathers 1983). Between these two habitats lie rocky creosotescrub, a yucca-galleta grass community, pinyon pine-juniper woodland, and chaparral.I studied three common species of sparrows wintering in different habitats alongthe gradient: sage sparrow (Amphispiza belli), black-throated sparrow (Amphispizabilineata) and the dark-eyed junco (Junco hyemalis). Their distributions have beendescribed by Weathers (1983 and unpublished data). Briefly, sage sparrows are largelywinter migrants, common on the valley floor and rare in chaparral. Black-throated10Figure 1. Locations of study plots in the lower Coachella Valley, California. "V" -study plot on valley floor, "A" - study plot on alluvial fan, "P" - study plot in pinyonpine, "S" - study plot used in other portions of the project (Chapters 2 through 4), "E"- area explored as potential study sites or used as a source of birds for captive studies,"W" - study plot of Weathers (1983). Vegetative features taken from topographic mapsare broad scale: small islands of developed land may occur within natural habitat typesand vice versa. Residential/agricultural lands occupy sites that would be either valleyfloor habitat or alluvial fan habitat. Indio, California is marked.1112sparrows are permanent residents occupying the rocky creosote scrub up throughpinyon-juniper woodland. During winter, they are uncommon in pinyon-juniper. Thedark-eyed junco population is a mixture of permanent residents that migrate along theelevation gradient and winter migrants from other areas. Juncos breed in coniferousforest and winter in chaparral and pinyon-juniper. The exact boundaries of species'distributions and the extent to which species distributions overlap are impreciselyknown because Weathers' study sites and my own were located far apart and in theinteriors of habitats (Figure 1). Nevertheless, species distributions are stable: (1)they are consistent with observations made at a variety of sites while exploring forstudy plots (see Figure 1), and (2) fluctuations of species' distributions over 7 winters(Weathers [1983] — 3, this study — 2, subsequent winters [Chapter 2, Chapter 4] —2) were insufficient to place species deep inside of habitats that they do not usuallyoccupy.I worked in three habitat types along the gradient: valley floor, alluvial fan, andpinyon-juniper. The valley floor is bare sand or hard packed sediments vegetated withwidely spaced shrubs (e.g., Larrea tridentata, Atriplex spp.) and patches of herbs(e.g., Schismus barbatus, Cryptantha spp., and Erodium cicutarium). Alluvial fans arelocated in the lower ends of valleys opening into the Coachella Valley. They are rockyand sparsely vegetated with shrubs (e.g., Larrea tridentata, Beloperone californica,Hymenoclea salsola, Ambrosia dumosa, Bebbia juncea), trees (Prosopis glandulosa,Cercidium floridum) and patches of herbs (e.g., Bromus rubens, Schismus barbatus,Cryptantha spp., Plantago insularis). Pinyon-juniper woodland exists on a plateauat about 1200 m elevation. The trees (Pinus monophyla) and large shrubs (up to 2m) (Juniperus californicus, Quercus spp., Rhus ovata) are widely spaced. Commonherbs include Bromus spp., Bouteloua aristidoides, Erioneuron pulchellum, and Stipaspeciosa.13I conducted the study during fall and winter months, after migrant sparrows hadarrived on the study site and before spring seed crops began to set. Sage sparrows anddark-eyed juncos arrive in the study area by early November. Rains occurring fromNovember through January may result in germination and the production of a springseed crop in some years. Seed may ripen as early as late February (R. Repasky personalobservation) or March (Burk 1982). A second rainy period during the summer monthsJuly through September may result in a winter crop of seeds that sets during Decemberand January (Burk 1982), although crops of this type were not observed during eitherof the two winters of the present study or during the following two winters. Thefrequency of years in which summer rains produce winter seed crops is unknown.Data CollectionStudy plots were located along a transect of representative vegetation typesmaintained by Deep Canyon Desert Research Center and the Coachella Valley Preserve(Figure 1). One study plot was located in each habitat type during the first winter(1985-86). A second plot, not less than one mile from the first, was added in eachhabitat type during the second winter (1986-87). Because study plots were few andlimited to a single mountain range, inferences drawn about species distributions arerestricted to the elevational gradient that I studied. Study plots were located awayfrom habitat boundaries to provide a clear test of the hypothesis that species can livein habitats that they do not normally occupy. For example, the availability of food ina habitat should be better characterized by plots located in its interior than by plotsnear its edge. Each plot was a 2-ha rectangle measuring 40 m by 500 in marked withflagging tape. Plots were visited twice during each winter. Visits were made duringDecember, January and February of the first winter. They were advanced to November,December and January of the second winter because a crop of seeds began to ripen latein February of the first winter.14During a visit to a study plot, I censused bird abundance, sampled seedabundance, determined food habits and measured the structural characteristics of thehabitat as well as the sites where birds foraged.Bird census.— I carried out 1 to 4 censuses per visit to each plot during the firstfield season^= 2.5). Estimates of population density were quite variable (standarderror approximately equal to the mean), and so in the second field season I conducted 4censuses per visit to each habitat. A census consisted of a count of the number of birdsfeeding on the study plot during a two-hour period beginning at sunrise. An observerwalked the length of the study plot by advancing 20 m at 5-minute intervals. Onlyactively foraging birds were counted. I recorded every individual and noted whetherit was 0-10 m or 10-20 m from the line of travel.Bird density was estimated from census data using Emlen's (1971) transectmethod. That method adjusts for differences in the probability of observing birds atdifferent distances from the census path. The number of birds in each 10-m band of thecensus plot was enumerated, and the count in the 10-20 m bands was calculated as afraction of that in the inner and then adjusted upward by the reciprocal of this fraction.I made this correction because differences in habitat structure along the gradient mightaffect the probability of observing birds at longer distances from the observer.Seed standing crop .— Seed abundance was estimated during each plot visit bycounting seeds present on 30 quadrats, each 0.125 in2 . Quadrats were randomly chosenfrom a grid coordinate system describing the study plot. Seeds on plants and on theground were counted in situ because regulations at Deep Canyon at the time preventedme from removing soil. Seeds on the ground were counted by systematically pickingthrough the surface soil with a pair of forceps to a maximum depth of 1 cm, dependingon soil hardness. Although this method undoubtedly underestimated the abundancesof small seeds, none of the seeds eaten by birds was too small to be seen in the soil.Whatever bias was introduced should be consistent among habitats, and comparisons15made among habitats are meaningful. I identified seeds by comparing them with aseed collection maintained by Deep Canyon Desert Research Center as well as my ownreference collection.Standing crop was estimated by multiplying seed abundance by mean seed massMasses were determined for seeds in my reference collection by weighing them afterthey had been oven dried for 24 hours.Habitat Structure.— Habitat characteristics were measured at each samplingquadrat. Distance to cover was measured from the quadrat center to the nearest shrubat least 50 cm tall. Shrubs of that size were readily used as cover by startled birds.Other characteristics were estimated visually within a 1-m square plot surroundingeach 0.125 m2 quadrat, including the percentage of the ground surface covered by rock,barren soil and either leaf litter or sprouting herbs. I also estimated by eye the percentvolume of the air column occupied by vegetation at the heights: 0-10 cm, 10-20 cm, 20-40 cm, 40-60 cm, 0.6-1.0 m, 1.0-2.0 m, 2.0-5.0 m, and more than 5.0 m.Feeding Habits. — Observations of foraging birds were made whenever possible.I recorded the location of each bird (ground, plant) and measured the habitatcharacteristics at the spot as described above.To determine diets, I captured birds using mist nets, and I administered to themthe emetic apomorphine hydrochloride (Sigma Chemical Co.; see Schluter 1988b,Esteban 1989). Vomit samples were preserved in alcohol to arrest digestion. In the lab,seeds were identified by comparing them with the reference collection and counted.Seed Handling Times.— Time taken to handle seeds was recorded from captivebirds during the winter of 1988-89. Observations were made after the birds had beenheld in captivity between 1 and 2 months. Birds were housed and observed singly. Ivideotaped them as they ate seeds from a dish, after they had been fasted for an hour.Handling time began when a bird picked up a seed and ended when movement ofthe lower mandible ceased. I measured it by counting the number of frames elapsed16and multiplying by the rate at which frames were recorded. Handling time for eachindividual bird on a seed type was taken to be the median handling time for thatseed type because a few seeds appeared to be handled for inordinately long periods.Handling time on a seed type by a species was taken to be the mean of the medians ofindividual birds. Unfortunately, I was unable to measure handling times of all threebird species on all seeds types eaten by any species. However, my data include theprincipal foods eaten by each species with one exception. Perityle emoryi, a commonfood of black-throated sparrows during the first field season, simply could not be foundwhen I collected seeds.Hypothesis testingHere I outline the methods used to predict distributions of species from thealternative hypotheses.Species' distributions and abundances should be proportional to the availabilityof their foods if distributions are determined by food abundance. I quantified foodavailability in two ways: standing crop and estimated food intake rate. Standing cropis a simple measure that is readily estimated, although it does not necessarily representthe amount of food available to foraging birds. Intake rate is a more realistic measure,but it is more difficult to estimate. I estimated it from its component variables: valueof food items, handling time, and encounter rate.Food itself had to be defined for each species. Not all species were found in allhabitats so I had to decide what a species would eat if it occurred in habitats outsideof its distribution. This problem was tempered by a few vomit samples collectedfrom birds foraging outside of their typical habitats during a transplant experiment(Chapter 2). I defined food for a species as any seed type falling within the rangeof seed morphologies observed in vomit samples. Seed morphology was described interms of the first two principal components of the variables seed mass, seed length(longest dimension), seed depth and seed width (shortest dimension). The two principal17components represented overall seed size and seed shape. Shape described the lengthsof seeds of a given mass. To define food for a species, I plotted all seed types againstthe principal component axes and constructed a convex polygon around those seedtypes consumed by a species. All seed types within the polygon were classified as food.I used rarefaction methods (see Schluter 1988b) to determine whether the numberof birds captured was adequate to characterize species' diets. Diet breadth, measured asarea of diet polygon, was plotted against sample size and inspected for the presence ofan asymptote. Each point represented mean polygon area calculated from 200 randomsubsamples of a given size. Although none of the species exhibited a clear asymptotein polygon size (Figure 2), I concluded that the sample sizes were sufficient becausethe final polygons included the most abundant seed types and further increases in dietbreadth only slightly affected estimated food standing crop. For example, a 20 percentincrease in final polygon area increased estimated food standing crop by more than 1percent for only one species.My estimate of food intake rate was based on Holling's (1959) disc equation,E A i e iR 1^+ E Ai hiwhere R is intake rate in milligrams per second, Ai is encounter rate of seed type iin seeds per second, e i is mass of seed type i in milligrams, and hi is handling timefor seed type i in seconds. Seed masses (ei) were measured from the seed referencecollection. Handling times (hi) were measured in the lab. Only encounter rates (Ai)were unknown. I assumed them to be linearly related to seed abundance: Ai = kdiwhere di is the density of seeds of type i in seeds per square meter and k is a constantrepresenting search rate in square meters per second. Intake rate was calculated as afunction of k. With this formulation, I compared food availability between habitats bycalculating intake rates using a range of reasonable values of k.The three species escape from predators by flying to cover, suggesting that eachis safest in the habitat with the most cover. If this interpretation is correct, the three18Figure 2. Diet breadth as a function of number of individual birds sampled of sagesparrow (  ), black-throated sparrow (^ ), and dark-eyed junco(^). Breadth is the area of the polygon defining food in terms of principalcomponents axes describing seed size and seed shape. The curves were generated byrandomly resampling different sample sizes from the collected sample (see Schluter1988b).0.4 -IIIII0.0 -II4*- -* -A-4 *-*- - 4(4I I I0^10^20^30^40Sample Size20species should forage closer to cover than expected by chance. I compared distancesof feeding birds to cover to distances of randomly located quadrats to cover usingone-tailed comparisons. Wilcoxon 2-sample tests were used because the data werenot normally distributed. To determine whether birds forage closer to cover thanexpected from food availability, I visually compared the distribution of foraging birdswith the distribution of food in each habitat. The relative frequency distribution offoraging birds was described as the probability density curve of foraging distancefrom cover (Becker et al. 1988). The distribution of food was estimated in three steps.First, I estimated the probability density curve of the distance from randomly locatedquadrats to cover. Second, I estimated mean food standing crop as a function ofdistance from cover using non-parametric regression (lowess; Becker et al. 1988), at thedefault parameters. Finally, the distribution of food relative to cover was calculatedby multiplying the probability density curve by food standing crop. The resultingdistribution was standardized by resealing it to have an area underneath it of 1. Thisfrequency distribution of foraging birds in relation to cover was compared to thefrequency distribution of available food.I calculated food abundance discounted by the risk of predation to test thehypothesis that food and predation together account for species distributions. For eachhabitat, I calculated an index ranging from 0 to 1 that described the relative valueof food as a function of distance from cover. The index is based on the assumptionthat predation risk at a given distance from cover is linearly related to the degree ofdisparity between the relative frequency of birds foraging at that distance and therelative frequency of food there. Food value was calculated as °E—E where 0 is theobserved probability density of birds foraging at a distance and E is the probabilitydensity of food at that distance. The largest positive deviation was assumed to occur atthe safest distance from cover and was ascribed the value 1. The most extreme negativedeviation was assumed to occur at the most dangerous distance from cover and was21ascribed relative value 0. Other deviations were linearly scaled between the 0 and 1.Adjusted food standing crop was calculated by multiplying food standing crop at eachsampling quadrat by the index describing relative food value at that distance fromcover.To determine whether species' distributions correspond with the availabilityof their foraging microhabitats, I characterized habitats in terms of the first threeprincipal components of habitat variables. Data from all habitats were combined, andthe percent cover variables were arcsin-square root transformed prior to analysis. Theaxes represented, first, variation in total cover, second, a gradient from shrub coverto tree cover, and, third, variation in rockiness (Table 1). Habitats were described aspolygons plotted against pairs of axes from the principal components analysis. Eachpolygon contained the central 75 percent of the sampling quadrats in the bivariatedistribution. Briefly, the density of points within the neighborhood of each data pointwas calculated using a scatter plot sharpening algorithm (Chambers et al. 1983),and I eliminated points below the 25th percentile of neighborhood density. Finally, Iprojected the habitat structure at birds' foraging sites on to the principal componentaxes characterizing habitats.To test the competition hypothesis, I tested to see whether species share limitingresources. I examined two aspects of overlap in resource use: food itself and foragingmicrohabitat. Overlap in diet composition was calculated in two ways. First, Icalculated proportional similarity (see Hurlbert 1978) in seed species composition ofdiets. This measure provides a limited measure of the extent to which fundamentalniches overlap because not all seed species were found in all habitats. Overlap infundamental niches can be described by comparing the profitabilities of foods amongspecies (Pulliam 1985). Profitability is seed mass divided by the amount of timerequired to handle the seed. Intake rates are maximized by consuming the mostprofitable foods and ignoring others. If foods rank similarly in profitability among22Table 1. Principal components analysis of habitat structure. All variables other thandistance to cover are arcsin square root transformed measures of percent cover.Principal Component Axis1^2^3Eigenvalue 5.90 2.01 1.03Proportion of variance explained 0.49 0.17 0.08Eigenvect ors :Distance to cover -0.26 0.04 -0.22Bare ground -0.30 -0.16 -0.15Rock 0.01 -0.13 0.94Cover (above ground):0-10 cm 0.31 -0.33 -0.1210-20 cm 0.33 -0.34 -0.0920-40 cm 0.36 -0.26 -0.0740-60 cm 0.38 -0.15 -0.0360-100 cm 0.37 0.00 -0.02100-150 cm 0.31 0.29 0.04150-200 cm 0.28 0.37 -0.05200-500 cm 0.21 0.49 -0.01> 500 cm 0.09 0.42 0.0523species, species share preferences for the same foods, and they are likely to compete. Icalculated profitabilities of seeds from seed masses taken from my reference collectionand from handling times measured in the lab. I calculated rank correlations of foodprofitabilities among species and plotted profitabilities to determine whether differencesin profitability were on high or low ranking seed types.Overlap in the use of microhabitats was calculated using a discriminant function inwhich individual birds were classified to species based on the structural characteristicsof foraging sites. This method was used because measures such as proportionalsimilarity are difficult when niches are defined in terms of several variables. The successof a discriminant function at classification increases as the amount of overlap in species'diets declines. Error rates are zero in the absence of overlap between species, and theytend toward . -71- when overlap is complete, where s is the number of species. The errorsrate when overlap is perfect is actually biased below . --1 because of sampling error.sAn index of overlap can be calculated by dividing the observed rate of error by thatexpected by chance if overlap is perfect. Error rates were simply the proportion ofobservations misclassified by the function. Expected rates of error if species overlapperfectly were calculated by randomization. Diet observations were randomly assignedto species, and a discriminant analysis was carried out. This process was repeated 1000times, and the mean error rate among randomized analyses was used as the expectederror rate. Calculations were performed using Procedure DISCRIM (SAS Institute1988). Habitat variables were arcsin-square-root transformed prior to the analysis.I tested the hypothesis that food is limiting by plotting bird abundance againstfood supply. If food is limiting, bird abundance should be positively related to foodsupply. I had six data points that could be considered to be independent: one pointfrom each habitat in each year. Total sparrow standing crop was used because itincorporates differences in body mass among species and because the less common24species within a habitat contribute to resource limitation if it exists. Food was definedto be any seed 1 mg or less in mass.ResultsDistributions of BirdsThe sparrows were distributed nearly one species per habitat with little overlapbetween species in habitat use (Table 2). Sage sparrows were observed only on thevalley floor, and juncos occurred only in pinyon-juniper. Black-throated sparrows weremost common on the alluvial fan and uncommon in pinyon-juniper.These results are consistent with earlier census data from the study areacollected during the three winters 1977-78 through 1979-80 (Weathers 1983, Weathers-unpublished data). Sage sparrows were present exclusively on the valley floor.Black-throated sparrows were present on the alluvial fan and in pinyon-juniper.In Pinyon-juniper, black-throated sparrows were present in low numbers duringautumn (September through November) but absent during winter (December throughFebruary). Dark-eyed juncos were present only in pinyon-juniper.Tests of HypothesesFood.- Distributions of species were unrelated to food supply. Foods of all specieswere most abundant in a single habitat in the first year of study, and they were nearlyequally abundant in all 3 habitats in the second year (Figure 3). In the first year,food standing crop in pinyon-juniper was greater than that in the other two habitats,although the differences could not be tested because of the absence of replication.Differences in food standing crop were slight during the second year and were notstatistically significant (ANOVA - sage sparrow: F = 0.19, df = 2, 3, P = 0.84; black-throated sparrow: F = 0.22, df = 2, 3, P = 0.82; dark-eyed junco: F = 0.22, df = 2, 3,P = 0.81). Clearly, food abundance alone does not explain species distributions.Estimated food intake rates showed a similar pattern. All three species experiencesimilar estimated intake rates in any one habitat, and the habitats rank similarly for all25Table 2. Sparrow species abundances (birds / ha / hr) in three habitats during twowinters. The upper entry for a species in a habitat is mean abundance on a singlecensus plot during the winter of 1985-86. The lower entry is mean abundance on twocensus plots during the winter of 1986-87. Standard errors (in parentheses) listed forthe first year are calculated from individual census on study plots (3 - 8 censuses perplot) because there was only one study plot per habitat. Those for the second year arecalculated from the means of the two study plots in each habitat (8 censuses per plot).A zero indicates that the species was never observed in the habitat during censuses.Bird Density (birds / ha / hr)Valley Floor^Alluvial Fan^Pinyon-juniperDark-eyed junco 0 0 4.28(2.91)0 0 0.57(0.37)Black-throated Sparrow 0 0.65(.49) 0.04(0.05)0 0.49(.02) 0.06(0.06)Sage sparrow 0.10(0.10) 0 00.78(0.23) 0 026Figure 3. Food standing crop (mg /m 2 ) for three species of sparrows and sparrowpopulation densities in three habitats over 2 winters. Circle area reflects food standingcrop, years scaled separately. Actual means and standard errors are listed just beloweach circle. Standard errors for 1985-86 data are based on variation in standing cropwithin study plots because only one study plot existed per habitat. Standard errors forthe second year are based on variation among replicate study plots. Circle color depictsbird population density: black — common, stippled — uncommon, white — absent.110150 ( 20)• 0 0193 ( 79)^179 ( 28)^177 (144)110150 ( 20)^161 (135)19078)0oC=I,a)al2 2_0(I)alm0c)C=2719850 0^129 ( 42)^138 ( 65)^968 (269)0 •^127 ( 42)^138 ( 65)^968 (269)• 0127 ( 42)^139 ( 65)1986968 (269)Valley Floor Alluvial Fan Pinyon-Juniper28three species, regardless of the value of the search rate constant (Figure 4). Pinyon-juniper stood out as having markedly higher predicted intake rates than the otherhabitats in the first year. Predicted intake rates in the second year were higher on thevalley floor than in the other habitats.Predation.— The predation hypothesis predicts that species differ in habitat usebecause they are safest in different habitats. Counter to the hypothesis, the threespecies use the same method to escape predators, suggesting that they should havesimilar habitat preferences. Because the three species escape to shrubby cover, theyshould forage as close to cover as possible. Both sage sparrows and black-throatedsparrows foraged significantly closer to cover than expected by chance (Figure 5; one-tailed Wilcoxon 2-sample tests — sage sparrow: S = 2220, n = 121,50, P < 0.001; black-throated sparrow: S = 1407, n = 117,28, P < 0.001). Dark-eyed juncos did not do soand actually foraged farther from cover than expected by chance (Figure 5; two-tailedWilcoxon 2-sample test, S = 6538, n = 112,64, P = 0.008). Food could be responsiblefor the tendency of dark-eyed juncos to feed farther from cover than expected: foodstanding crop close to cover in pinyon-juniper was approximately half of the standingcrop farther from cover (Figure 6). Indeed, dark-eyed juncos foraged as far from coveras expected from the availability of food when bird distributions are compared to fooddistributions (Figure 7). Sage sparrows and black-throated sparrows foraged slightlycloser to cover than expected from food availability.The similar preferences of sage sparrows and black-throated sparrows for foragingclose to cover suggests that the predation hypothesis alone is unable to account forthe distributions of species along the elevational gradient. Both species should preferpinyon-juniper, the habitat with the greatest amount of cover and the habitat occupiedby dark-eyed juncos. Additionally, if birds avoid foraging far from cover then foodavailability may be inadequately measured by food standing crop. Food in a habitatwith little cover is not equivalent to the same amount of food in a safer habitat. Hence,29Figure 4. Estimated food intake rates of three species of sparrows in three habitats intwo years of study. Intake rates were calculated from the variables mass of food item,handling time, and seed density times the unknown constant search rate for seeds thatdid or were likely to be consumed by species (see text). Predicted intake rates for thefirst year of study are in (a) and those for the second in (b). The species are: sagesparrow (^ ), black-throated sparrow (( ^  )• ) and dark-eyed juncoPinyon-juniper^ a^laWIDIMMONOMOMI.................................. . =1IPAlluvialFanValley FloorII^ I^ I^ I0.006 0.010............ -----------• b300.06 -0.05 -0.04 -0.03 -0.02 -0.01 -in' 0.0 -cm^0.0Ea)csCCa)as 0.20 -t0.15-0.10 -ValleyFloorPinyon-juniper0.002AlluvialFan0.05 -0.0 - I I I0.0 0.002 0.006 0.010Search Rate (m2 / s)31Figure 5. Distribution of distances from cover of foraging birds ( ^ ) andrandomly located points (^ ). Dark-eyed juncos in pinyon-juniper, black-throated sparrows on the alluvial fan, and sage sparrows on the valley floor. Probabilitydensity curves were estimated using the density function of Becker et al. (1988).IIIssts320.8-0.6-0.4-0.2-0.0->, 0.8-. _0CQ)a0.6->, 0.4-13cts..caO 0.2-0_0.0-0.8-0.6—0.4-0.2-0.0-Dark-eyed JuncoIII• Ir\• ts• .^.—2^0^2^4^6^8 10Black-throated SparrowI•■•^I^1 uI I %I\IuiII—2^6^2^.4^.6^8^.10Sage SparrowA,tt sIItII% ^I0 %a r 111—2^0^2^4^6^6^10— oI^II•IMPDistance from Cover (m)33Figure 6. Food density as a function of distance to cover for: dark-eyed juncos inpinyon-juniper, black-throated sparrows on the alluvial fan, and sage sparrows onthe valley floor. The curves in the figures were fit using the scatter plot smoothingalgorithm lowess (Becker et al. 1988) at the default parameter settings.• • •• ••• •••• • • • • • ********* • • •^•^• • •••• • •• ••••••••••.^•• •.^• • • •Valley Floor•• •• •^••••• •••Pinyon-juniper34C\JEC)E Alluvial Fan10001--+10 .2 •0•• • • • • • • * • • • * • •^••••••••Distance to Cover (m)35Figure 7. Distribution of distances from cover of foraging birds (^ ) andfood availability (^). Dark-eyed juncos in pinyon-juniper, black-throatedsparrows on the alluvial fan, and sage sparrows on the valley floor. Distributions offoraging birds are from Figure 4. Food distributions were calculated by multiplying theprobability density curves of the distance of random quadrats to cover (Figure 4) bymean food density (Figure 5) and standardizing so that the area under the curve is 1.0.6-0.4-0.2-0.0-^;0.8-I s • 1I11 1IIi^1itV,1sBlack-throated SparrowA_0.8-Cl)Ca) 0.6-a>.0.4-16—CZ-Ct 0.2-O0_0.0-Dark-eyed Junco36-2^.0^.2^.4^6^6^10-2^0^2^4^6^8^10 Sage SparrowaI%I IIIIIII1 •IIIII0.8-0.6-0.4-0.2-•0.0--2 o.^2^4^6^8^10Distance from Cover (m)37weighting food by distance to cover may yield different predictions about habitatdistributions than those derived from food alone.Adjusting estimates of food standing crop by the distance from cover changed thepredictions of species distributions from food availability (Figure 8). Differences inadjusted standing crop between habitats were marginally significant (ANOVA — sagesparrow: F = 9.91, df = 2, 3, P = 0.048; black-throated sparrow and dark-eyed junco:F = 8.93, df = 2, 3, P = 0.055), and habitats ranked similarly for all species. Adjustedfood standing crop was greatest in pinyon-juniper woodland for all three species in bothyears. Yet, only one species of sparrow is abundant there. Hence, food and predationjointly fail to explain species distributions along the elevational gradient.Habitat structure.— Structural features of habitat might be responsible for speciesdistributions if they influence the foraging abilities of birds in ways that are notcaptured by the simple measures of food and predation that I used. To accommodatethis possibility, I asked if the microhabitats used by foraging birds were themselvesrestricted in distribution and might limit the distributions of species.Foraging microhabitats used by a given species were not restricted to the habitatin which the species is found (Figure 9). The only habitat variable that might limitdistributions was rockiness, the third principal component axis. Sage sparrows andjuncos both live in habitats with few rocks and hence there appears to be relativelylittle foraging habitat for them in alluvial fan habitat where rocky substrate is common(Figure 9b, 8f). However, neither species is restricted to its usual habitat by peculiarhabitat structure, nor is habitat structure likely to exclude sage sparrows or black-throated sparrows from pinyon-juniper where food availability may be highest.Competition.— The competition hypothesis is consistent with the nearly exclusivenature of species distributions. It is also consistent with disparities observed betweenspecies distributions and predictions from the food, predation and habitat structurehypotheses. I pursued the competition hypothesis further by testing for the existence of38Figure 8. Food standing crop discounted by risk of predation for three species ofsparrow and sparrow population density in three habitats over 2 winters. The outerdashed circles represent raw food standing crop, and the inner circles discountedstanding crop. The statistics below the circles are mean discounted standing cropand standard error as in Figure 3. Circle coloration reflects population density as inFigure 3.73 (29) 438 (88)"-'I( 0123 (10)••••23 (10)^73 (29)^438 (88)I23 (10)^74 (29)^438 (88)•II530 ( 5) 62 ( 6) 127 (111)..• — -- ...,...0/^• •^I/ 1 /I I I1^ /^11 / 1‘ /^\ ■al■I30(5)^62( 6) 127 (111)••^%^e^--•• eI / %^/i  0 1 I1^i^I1^I1^II^ I^1 I t1 ■^I^%\e‘^/^\ • /I• 30 ( 5) 72 ( 9) 133 (114)39198500a)rti2 2Ei0cp (ts0Ca)0 2_c0 u)row 0Du:to ct1986Valley Floor Alluvial FanPinyon-Juniper40Figure 9. Foraging locations of species plotted over polygons describing the availabilityof microhabitats within habitats. The first principal component is an axis of increasingtotal cover; the second one represents increasing tree cover and decreasing shrub cover;the third represents increasing rockiness. Habitats are: valley floor (^),alluvial fan ( ), and pinyon-juniper ( ). The polygons enclosethe central 75 percent of each bivariate distribution (see text).000° ••. %••••••d3-2 ,-2--3-x Cx1-x x320%•000 •10-x-11-2--3- .^.^.^.^.^.-2 0^2^4^6^8 -1^6^.^.1^2^3Dark-eyed Juncoe f0%0 ' 0 •••••II0^2^4PCA Axis 1-1^0^1^2^3PCA Axis 30--2-410121Sage Sparrowa^2-1-0-1-2-3b..0 1 2-1 3Black-throated Sparrow42two conditions that must exist if species are to compete: species must share resources,and the resources must be limiting.Species diets overlapped in both the seed species present and in seed morphology.Proportional similarity in seed species present was moderately high between sagesparrows and black-throated sparrows, and it was low between each of these twospecies and dark-eyed juncos (Table 3). Seed profitability was highly correlated amongspecies (Table 4) suggesting that fundamental niches overlap broadly. The rank orderof profitabilities was similar among species with only two exceptions that were dueto more than minor differences in profitability (Figure 10). One of those exceptions(Phacelia spp.) was rare in the environment. Clearly, species share preferences for thesame foods and are potential competitors.The three species also overlap in their use of foraging microhabitats. There wassignificant overlap in the polygons depicting microhabitat use (Figure 11). Errorrates in the classification of foraging observations using a discriminant function ofmicrohabitat characteristics also suggested that there was significant overlap inmicrohabitat use (Table 5). The overall error rate in discriminating species on the basisof microhabitat characteristics was 24 percent or 45 percent of that expected if overlapin microhabitat use was complete. Clearly, these species are potential competitors.Are resources limiting? Bird abundance was positively related to food supplyas expected if food is limiting (Figure 12), but the relationship was due largely to asingle point. There was a large amount of food present in pinyon-juniper during thefirst year of study, and dark-eyed juncos were abundant there that year. If that pointis eliminated, only a small range of food densities is included in the data set. (Therelationship then appears to be negative, but again due largely to a single point.)Hence, my data are inconclusive regarding the hypothesis of food limitation. In thediscussion I consider other evidence suggesting that food is indeed limiting.43Table 3. Diet overlap among species calculated as proportional similarity (see Hurlbert1978) in seed species composition. Standard errors given in parentheses were calculatedusing a bootstrap technique (see Methods).Proportional Similarity in Diet (SE)Black-throated^Dark-eyedSparrow^JuncoSage Sparrow^0.362 (0.101)^0.049 (0.026)Black-throated Sparrow^ 0.059 (0.026)44Table 4. Similarity in relative profitability of seed types among sage sparrows, black-throated sparrows and dark-eyed juncos. Profitability was calculated as seed massdivided by seed handling time. Similarity was calculated as the rank correlationbetween species. n = 11 in each case.Rank Correlation (probability)Black-throated^Dark-eyedSparrow^JuncoSage Sparrow^0.91 (0.0001)^0.90 (0.0002)Black-throated Sparrow^ 0.84 (0.0010)45Figure 10. Profitability (seed mass / handling time) of seed foods to sage sparrows(^), black-throated sparrows (^ ) and dark-eyed juncos(^). Seed types: 1) Bouteloua aristidoides, 2) Aristida adscensionis,3) Schismus barbatus that must be extracted from floret, 4) Festuca octoflora,5) Cryptantha spp. that must be extracted from remnant calyx, 6) Eriogonumfasciculatum, 7) Schismus barbatus bare seed, 8) Cryptantha barbigera, 9) Amaranthusalbus, 10) Cryptantha spp. bare seed, 11) Phacelia spp. either P. crenulata or P.distans.1 .5 -2^4^6^8^10Seed Species47Figure 11. Overlap in the foraging microhabitats of sage sparrows (black-throated sparrows (   ), and dark-eyed juncos ( ^ represent principle components of habitat structure. Polygons enclose the central 75percent of the bivariate distribution (see Methods — Hypothesis testing). PCA 2: Height of CoverC:0^IV^I_L^0 _LK)49Table 5. Overlap in use of microhabitats quantified as the rate at which observationswere misclassified by a discriminant function used to categorize foraging observationsaccording to species on the basis of microhabitat characteristics. Zero error rates areexpected if species do not overlap. Error rates of 53.1 percent are expected if meanmicrohabitat characteristics are identical.Misclassification RateSpecies^ Percent of^Percent of^observations^maximumSage sparrow 40.8 76.8Black-throated sparrow 30.0 56.5Dark-eyed junco 7.9 14.9species combined 23.9 45.050Figure 12. Total sparrow abundance and food abundance in three habitats in two years.Each point represents a single habitat in a single year.*100 -***60 200^400Food Standing Crop (mg/m2 )60052DiscussionDarwin (1859) argued that species replacements along environmental gradientsresult from interspecific competition. Although there are many examples of suchreplacements (e.g., Lack 1944, Terborgh 1971, MacArthur 1972), the hypothesis hasrarely been tested. I tested several hypotheses that might account for abutting habitatdistributions of sparrows, including food, predation, habitat structure and currentcompetition. Food was abundant outside of species' usual habitats suggesting that itis not responsible for species' distributions. Predation is also an unlikely explanation ofthe distributions because all species appear to be safest from predators while foragingclose to cover and they should be safest in the same habitat. The habitat structurehypothesis was rejected: species used foraging microhabitats that were more widelydistributed than the species themselves. I tested for two conditions that are necessaryfor competition. The fundamental niches of species overlap indicating that competitionis indeed possible between them. However, my results regarding food limitation wereinconclusive. Competition is the hypothesis most consistent with my results. Here, Idiscuss alternative hypotheses that might account for the distributions. I also discussother evidence regarding competition and suggest that the competition hypothesis isworth pursuing by more direct tests.Other possibilitiesI considered several biotic factors that might be responsible for habitat partitioningamong species. Factors were considered singly with one exception, food and predationtogether. Species distributions might be explained by other combinations of the factorsthat I considered or by abiotic factors which were not considered. Complex interactionsamong factors are outside the scope of this paper. Here I explain why abiotic factorsare unlikely to explain habitat partitioning. Temperature and water are both relevantto the present study, temperature because the habitats lie along an elevation gradientand water because the lower habitats in my study area are in a desert.53Neither the temperature hypothesis nor the water hypothesis yields the predictionthat species are restricted to different habitats. Rather than forming absolute boundsto species distributions, temperature interacts with food and predation to limitdistributions (Repasky 1991), modulating food requirements and hence the risks ofstarvation and predation (McNamara and Houston 1990). Each species should bemost abundant in the habitat where the least amount of energy is needed to maintainbody temperature. Species actually overlap broadly in the range of temperatureover which energy expenditure is least (Figure 13), and all species should be mostabundant in the same habitat and less abundant elsewhere. Like temperature, thewater hypothesis fails because it predicts that species should share preferences forhabitats. The primary adaptation to aridity by North American passerines appearsto be small body sizes favorable for dissipating heat and conserving water (MacMillen1990). Because passerines lack peculiar adaptations that might restrict desert-dwellingspecies such as sage sparrows and black-throated sparrows to arid areas, all speciesshould be best suited to the presence of water.Other evidenceMy inferences are based on observation rather than experiment, and thus theyare less convincing than experimental results might be. Here, I provide other evidencesuggesting that food, predation and habitat structure are unlikely to shape the habitatdistributions of sparrows and that the competition hypothesis is consistent with thedata.The results of an introduction experiment (Chapter 2) support my conclusionthat food is unlikely to restrict species distributions because food is readily availableoutside of species' typical habitats. Each species was forced to forage in each habitatfor naturally occurring seeds. Only dark-eyed juncos achieved food intake rates in thehabitat that they occupy that were significantly higher than in other habitats. Sagesparrows and black-throated sparrows species experienced smaller tradeoffs in feeding54Figure 13. Maintenance metabolism as a function of ambient temperature: black-throated sparrows (13 g,eyed juncos (19 g, ^ ), sage sparrows (17 g, ^ ), dark-)and for comparison rufous-sided towhees (42 g,^ ). Note that species overlap broadly in the range of temperatures thatare most favorable to them. Estimates were based on allometric relationships in theliterature: basal metabolic rate (the flat portion of each relationship) was calculatedas 3.19M°• 73 where M is body mass in grains (Aschoff and Pohl 1970); the thresholdambient temperature below which additional energy must be spent to maintain bodytemperature was Tb — 11.5M°• 19 where Tb is body temperature are in degrees Celsiusand M is body mass in grams (Weathers and van Riper 1982); the rate at which energyexpenditure increases as ambient temperature declines below the threshold temperatureis 0.28M°•54 in kilojoules per day per degree Celsius where M is body mass (Aschoff1981)._Y)60 -cucCUOal 40 -6220-80 -S..■■•■\•■■,...I^I0^10 20^30Air Temperature (deg. C)56ability between habitats. Those species also achieved food intake rates in the habitatsthat they occupy that were slightly less than in unoccupied habitats. Although thedifferences were small, they suggest that differences in foraging ability among habitatsare minor and that sage sparrows and black-throated sparrows are more narrowlydistributed than they would be if food were responsible for their distributions.The introduction experiment also supported my conclusion that species differencesin predation risk are unlikely to be responsible for shaping the species' habitatdistributions (Chapter 3). Sage sparrows and black-throated sparrows transportedbetween habitats exhibited similar changes in the amount of time that they spentscanning the environment suggesting that they experience similar changes in predationrisk between habitats and that they are likely to be safest in the same habitat.Vigilance data from dark-eyed juncos, unfortunately, could not be compared.Evidence from the introduction experiment also weighs against the habitatstructure hypothesis. The means by which structural features of habitat affect species'distributions are likely to be through their influence on feeding ability and predationrisk. The net effects of these factors were captured in feeding rates and vigilance levelsof birds foraging in the introduction experiment. Those effects were small and appearedto be insufficient to restrict species to single habitats.Do the species actually compete? My data clearly indicate that species share somefoods. Although my observational data on food limitation were inconclusive, two othersources of evidence suggest that food is limiting. First, a short-term food additionexperiment (Chapter 4) showed that birds recruited to food addition plots where foodstanding crop was increased with millet seed. Hence, birds readily switch their feedingto areas of greater food abundance if they are available. Such a response must be seenif food is limiting, although it is not sufficient evidence of food limitation because achange in survival was not demonstrated directly. Nevertheless, the absence of a57response would have suggested that food was not limiting (e.g., Pulliam and Dunning1987).Second, food limitation appears to be a general phenomenon among granivorousbirds in arid lands. Sparrows in the arid grasslands of Arizona appear to be foodlimited in years of low seed production but not in years of higher seed production:there is a positive relationship between sparrow abundance and food abundanceamong years of low food availability (Pulliam and Parker 1979); predictions of speciescomposition that are derived from competition theory tend to be upheld in years ofpoor seed production but not in years of good seed production (Pulliam 1983); andbirds failed to recruit to a food addition plot when the experiment was in a year ofmoderate food availability (Pulliam and Dunning 1987). Also, there is a positivecorrelation between the abundance of finches (broadly defined as small granivoresin several avian families) and food abundance among the arid areas of the worldsuggesting that food limitation is widespread (Schluter and Repasky 1991). Althoughthe worldwide study included data from the present study site, the pattern is clearlypresent without those data.Future testsOn the basis of the evidence above, I suggest that the competition hypothesisis a plausible explanation of habitat partitioning by sparrows. Stronger tests areneeded. Direct experimentation is one possibility. In the absence of direct tests,tests of hypotheses about the mechanism of competition would place the competitionhypothesis at risk of falsification. Three possible mechanisms are: exploitation of food,behavioral interference and avoidance of habitats in which species suffer a competitivedisadvantage.The mechanism of competition among sparrows is more likely to be interference oravoidance than the exploitation of food. Species share food preferences (Figure 10), andthey are expected to have similar distributions. Under these conditions, exploitative58competition would result in habitat partitioning if jointly preferred foods were depletedto the extent that distributions were governed by foods for which species do not sharepreferences. That is clearly not the case in the present study because species sharedpreferences for the foods that were present, and they were not distributed as predictedfrom food availability.59Chapter IIFORAGING SUCCESSAND THE HABITAT DISTRIBUTIONS OF WINTERING SPARROWS:A TRANSPLANT EXPERIMENTTemperate birds are frequently habitat specialists, and adjacent habitats areoften occupied by closely related species. Lack (1944) argued that this pattern ofdistributions results from interspecific competition. He reasoned that morphologicaldifferences between species have only minor consequences on fitness in different habitatsand that species ought to be more widely distributed among habitats in the absenceof congeners. Lack's argument is an attempt to refute the alternative hypothesis thathabitat partitioning results from species differences in feeding ability. Although thecompetition hypothesis has frequently been tested (e.g., Terborgh 1971, Terborgh andWeske 1975, Pulliam 1975, Noon 1981, Garcia 1982, Schluter 1982), the feeding abilityhypothesis has seldom been tested explicitly (Schluter 1982, Price 1991, Chapter 1).Habitat specialization might result from food alone if species are adapted tofeeding conditions in alternate habitats (e.g., Smiley 1978, Futuyma and Wasserman1981, Schluter 1982). The tradeoff in feeding ability between habitats must be largeenough to restrict species' distributions. Hence, there are two testable predictions fromthe foraging success hypothesis. First, a large tradeoff in foraging ability should existbetween habitats. Second, each species should achieve its highest food intake rate in adifferent habitat, and each species should dwell where it is most successful at foraging.I tested the foraging success hypothesis by comparing the feeding rates of sparrowspecies "transplanted" among habitats in the Sonoran Desert of southern California.Sage sparrows (Amphispiza belli), black-throated sparrows (A. bilineata) and dark-eyed juncos (Junco hyemalis) spend the winter in different habitats along an elevationalgradient (Weathers 1983, Chapter 1). Sage sparrows dwell in a creosote bush-saltbush60(Larrea tridentata-Atriplex spp.) shrubland on the floor of the valley at the base of theelevational gradient. Black-throated sparrows occupy the creosote bush shrublands(other common species include brittlebush [Encelia farinosa], burrobush [Ambrosiadumosa], sweetbush [Bebbia juncea], and cactus [Opuntia spp.]) located on rockyalluvial fans at the entrances of small valleys and on the rocky, lower slopes of themountains. Dark-eyed juncos inhabit a woodland of pinyon pine (Pines monophyla)and juniper (Juniperus californi cus), located on a plateau above the other two habitats.I carried out a "transplant" experiment because there were few naturalopportunities to observe foraging success outside of species' typical habitats. Theadvantage of such a manipulation is that observed food intake rates are the net effectof several factors affecting food availability, such as food abundance, vegetation andsubstrate structure and possibly predation risk (Chapter 1, Chapter 3). By measuringforaging success in the aviary, I tested the predictions from the foraging successhypothesis that large differences in foraging ability should exist between habitats andthat species should occupy the habitats in which they forage most successfully. Theresults did not support the foraging success hypothesis: two species' food intake rateswere relatively similar in all three habitats and may actually be slightly higher outsideof the habitats in which each typically occurs.MethodsExperimental designThe experiment was carried out during the winter of 1988-89 in the vicinity ofDeep Canyon Desert Research Center, Palm Desert, California (see Weathers 1983).Study sites in the different habitat types were located on ecological reserve lands alonga transect on the north slope of the Santa Rosa Mountains.Birds used in the study were captured from the wild shortly before the experimentbegan. They were housed individually and maintained on a mix of seeds commerciallyavailable for pet finches, meal worms, water and a vitamin supplement.61Sparrows generally search for seeds on the ground. Food intake rates wereestimated by observing individual birds foraging on seeds naturally occurring on theground inside of an aviary. The aviary measured 4 m x 4 m x 2 m (Figure 14). It wasmade to be portable by constructing it of wooden frames, each 1 m x 2 m, that couldbe bolted together. The roof and 14 of the 16 frames were covered with screening thatcould easily be seen through. The two remaining frames were covered with black fabricto serve as the front of a blind from which observations were made. An apron aroundthe base of the aviary covered gaps between the base and the ground on uneven terrain.Birds entered the aviary from a holding cage attached to a port equipped with a doorthat could be remotely controlled. Rodent burrows were plugged with rocks to preventbirds from seeking asylum or escape.Six individuals of each species were introduced singly into all three habitats. Theywere divided into two lots of three birds each, and lots were tested sequentially. Feedingtrials were scheduled according to a design that experimentally controlled the order inwhich birds experienced habitats because performance might change as experience inthe aviary increases. One individual in each lot of birds experienced each habitat asthe first, second and third habitat in the aviary (Figure 15). Trials could not be carriedout simultaneously in the three habitats, preventing experimental control of seasonalchanges in foraging conditions beyond that afforded by dividing the birds into two lots.Hence, trials did not follow a standard cross-over design (see Mead 1988). Instead,habitat visits were scheduled to minimize the number of times that the aviary wouldbe moved between habitats. For example (Figure 15), the aviary was located in habitat"a" where the first individual was tried. It was then moved to habitat "b" where thefirst and second individuals were tried. All three individuals were tried in habitat "c",and then the aviary was returned to habitats "a" and "b" to complete the trials ofthe second and third individuals. Habitats were randomly assigned to the visitationsequence. Individual birds were randomly designated as the first, second and thirdFigure 14. Floor plan of aviary.62^/BLINDN/ DOORA ACAGEPORT■111- - - -OD.-1 m6364Figure 15. One block of the experimental design. Three individuals of each of threespecies were tried in three habitats. In the upper left box, numbers within rowsrepresent the order in which individuals (regardless of species) were tried in habitats.The average amount of previous experience in the aviary is constant among habitats.Habitats were visited in the order "A", "B", "C", "A", "B". The dashed diagonalline separates trials completed during the first visit to a habitat (above) from thosecompleted during the second visit to a habitat (below). The lower right box is anexample of how depletion was experimentally controlled during a visit to a habitat.Numbers within columns represent the order in which species were tried on any oneday. Each species experienced equal amounts of depletion over a three day visit.1^2^33 • ■ 1^2\■■• ■3DAY1^2^31^2^32^3^13^1^2-J<MCI> 20z1HABITATA^B C6566individuals of their species.Each bird was introduced into the aviary for two 30-minute trials on each ofthree successive days during a visit to a habitat (Figure 15). Each day yielded anindependent observation of an individual's foraging ability in a habitat because theaviary was moved to a new location each day. Birds were fasted for an hour before thefirst trial and for an hour between the two trials.On any one day, one individual of each species was tested in the aviary (Figure 15).To experimentally control for possible depletion of food through the day, all birdswere subjected to the same amount of it. This was accomplished by ensuring that anindividual bird was the first bird in the aviary on one of the three successive days oftrials, the second bird in the aviary on one day and the third bird on the remaining day(Figure 15). The particular sequence (first, second or third) was random. Comparisonsof food intake rates between the first and second foraging trials of individual birds onthe same day indicated that depletion throughout a day was undetectable.Food intake rateI was unable to estimate seed intake rates directly because I could not identifyfrom a distance all of the seeds eaten, especially small ones. Instead, I obtainedestimates of intake rates by recording the rate at which birds pecked at items on theground or on plants, and multiplying this rate by the mean mass of seeds eaten. Thiscalculation is an overestimate because it assumes that each peck yields a seed, whereasthis assumption may not be true. For example, grass seeds are often enclosed in aset of bracts, but a set of bracts picked up from the ground or from a plant may notcontain any seeds. Also, seeds may be spoiled or empty, and subsequently rejected.Hence, I also obtained a lower estimate of intake rate by assuming that birds areunable to determine beforehand whether items picked up are edible. From soil samples,I estimated the proportion of potential food items (seed cases, bracts, etc.) on theground containing edible seeds. This proportion was multiplied by the upper estimate67of intake rate to obtain a lower bound on food intake rate that assumes birds did nobetter than chance at picking up good seeds. Results from the two estimates werenearly identical, and I present only the upper estimates here.Variance in food intake rate was calculated by accumulating the variances of peckrate, probability of obtaining a seed, and mean seed size. The variance of the productof two variables wasV AR(A • C) = V AR(A)V AR(C) + e4 V AR(C) -2c, V AR(A)(Bickel and Doksum 1977), where, for example, A is peck rate and C is size of seedseaten. Standard error was calculated as the square root of variance.Peck rate.— An observer in the blind used a microcomputer programmed as anevent recorder to register the time at which a bird began to forage, each peck at apotential food item made by the bird, and the time at which a bird ceased to forage.Peck rate was calculated as the total number of pecks divided by total time spentforaging during a half-hour trial. Peck rates from the two half-hour trials in a day wereaveraged.Number of seeds per peck.— I estimated the proportion of pecks that might yieldedible seeds. I collected seeds from quadrats placed in patches of high seed density.Seeds were collected from plants, and the surface soil was scraped to a maximum depthof 1 cm from an area of 0.125 m2 (see Chapter 1). Seeds were assumed to be edible ifthey made an audible crack when crushed using a pair of forceps.Size of seeds eaten.— Mean mass of seeds eaten by experimental birds wasestimated by two methods. In the first, data on seeds consumed by birds foraging inthe aviary were combined with prior expectations of seed size based on data from wild-caught birds. Estimates of mean seed size obtained by this method were potentiallyfrail because of small sample sizes. The second method provided a way of judging therobustness of the first. A function describing seed size preference was calculated from68data on wild-caught birds and data on seed abundance, and it was used to predictdiet composition in novel habitats. Mean seed size was calculated from these predicteddiets.A sample of seeds eaten was obtained by administering the emetic apomorphine(see Schluter 1988b) to each bird after its final trial in the aviary. Vomit samples wereobtained only after the final trial to avoid the possibility that birds might develop anaversion to foraging in the aviary. Hence, apomorphine was administered to two birdsof each of the three species in each habitat type. Not all birds vomited with the resultthat n = 12 birds rather than n = 18. Despite these small sample sizes, mean seedsize could still be estimated. Rarefaction analysis of mean seed size as a function of thenumber of wild birds sampled suggests that an unbiased estimate requires sample sizesof 3 to 5 birds (Figure 16).I used a Bayesian estimate of mean seed size in which prior expectations of meanseed size obtained from field data were modified on the basis of samples obtained frombirds foraging in the aviary. The posterior estimate of mean seed size wasit_i_ trii^82 I a2— 1_ _i_ nS I a2(see Stephens and Krebs 1986:77) where p, and u2 are the prior expectations of themean and variance in seed size, 7- and s2 are the sample mean and variance, and n issample size. A and u2 were estimated from vomit specimens obtained from wild-caughtbirds during another portion of the project (Chapter 1). They provide only roughexpectations of the mean and variance of seed size eaten in novel habitats becausethey are obtained from birds foraging only in their preferred habitat. 7 and 32 werecalculated from the vomit specimens obtained from birds after they had foraged inthe aviary. Theoretically, this method should yield an unbiased estimate of mean seedsize. 32 is based on the assumption that each seed eaten by an individual bird is anindependent observation. This assumption was tested using an analysis of variance69Figure 16. Bootstrap estimates of mean (a) and standard error (b) of mean seed sizein the diet as a function of sample size. Sage sparrow (^ ), black-throatedsparrow ( ), and dark-eyed junco (^ ). Each point represents200 repeated samples of vomit specimens from wild-caught birds.5 10 15 200.08 —6)E 0.06J-2 0.04 -coC0.02 -A••*MI llbOb'gm001* 441*011..^ %.*../00 .4 •11....tb 94% WAqp.^4.41,700.15 '4. •^4001360.14 -EN •0 13 - .C.T)"FD3 0.12 -wc.c9 0.11 -a)0.10 -/5^10^15^20Sample Size71of mean seed size eaten by wild-caught birds. Variance in seed size within a singlevomit specimen of a given bird was at least as large as the variance between specimenssuggesting that the assumption is valid.Variance of seed size in the diet was estimated as0-2' = ^1(see Stephens and Krebs 1986:77) where u 2 is the prior expectation of variance, s 2 isthe sample variance, and n is sample size as for te.The above method yielded an unreasonably large estimate of the size of seedseaten by black-throated sparrows on the valley floor. Prior expectations of the mean(it) and variance (o- 2 ) were large because black-throated sparrows eat a wide varietyof seeds on the alluvial fan. Only small seeds were eaten on the valley floor, and asa result the sample mean and variance both were small. The sample variance was sosmall that a sample of at least 250 vomit specimens would have been necessary to bringthe Bayesian estimate of the mean pi to within 130 percent of the sample mean. Seedsof the size it' were extremely rare on the valley floor, and they are large enough that Ishould have been able to observe birds eating them. Hence, I discarded it' and used thesample mean.Two other estimates of mean seed size in the diet could not be obtained by themethod described above because both birds of a species failed to vomit. For sagesparrows foraging on the valley floor, I used the mean seed size in the diets of wild-caught birds because sage sparrows inhabit the valley floor. Observations of sagesparrows foraging in the aviary suggested that they ate mostly Schismus barbatusand Cryptantha spp., just as they do in the wild. Black-throated sparrows failed tovomit after foraging in pinyon-juniper habitat. Field notes made after each foragingtrial suggested that black-throated sparrows foraged similarly to juncos, eating mostlyBouteloua aristidoides, although they occasionally picked up larger seeds, but less1 _L n82 I a272frequently than sage sparrows did. Therefore, I used the mean size of seeds eatenby juncos (7) to estimate it' for black-throated sparrows. This estimate should beconservative because it probably underestimates seed size in the habitat in which intakerates were posited to be highest (see Chapter 1).A second method of estimating the mean size of seeds in the diet was employedbecause of the small sample sizes associated with the Bayesian estimates. Inthe second method, preferences for seeds of a variety of sizes were calculated bydividing percentage seed abundance in the diet by percentage seed abundance in theenvironment. A small constant was added to seed availability to avoid division byzero. The effect of this was that seeds of sizes not available for consumption assumedpreference value zero. Preferences were then multiplied by seed availability to predictdiets in novel habitats. From the predicted diet, the mean size of seeds eaten in ahabitat was the sum of seed mass weighted by numerical abundance in the diet.Distributions of seed sizes in diets and of seed sizes available in habitats weredescribed from data collected during winters prior to the present experiment (seeChapter 1). Briefly, seeds in randomly located quadrats on the ground and in vomitsamples collected from mist-netted birds were identified to species and counted.Frequency distributions of seed mass were generated using a nonparametric probabilitydensity function (Becker et al. 1988). The description of a distribution consists of anarray of coordinate data points in which x is seed size and y is an index of relativeabundance. Abundance values were scaled to sum to 1. Calculations were made byperforming arithmetic operations on abundance values at the same value of seed size.Standard error of mean seed size was calculated by a bootstrap method (Efron1982). Mean seed size was calculated in the manner described above 200 times, eachtime from a random sample of the vomit specimens and a random sample of seedquadrats. The standard deviation among the means was taken as the standard errorof seed mass.73ComparisonsComparisons of food intake rates among habitats were made by calculating 95-percent confidence intervals of the difference in mean food intake rate between habitats.This tack of analysis was taken because peck rate, the proportion of pecks yieldingseeds, and mean seed size were collected from different sampling units. Peck ratesof individual birds were observed in a manner appropriate for a repeated measuresanalysis of variance, whereas estimates of seed size and the proportion of pecks thatyield seeds were unavailable for individual birds. Only one estimate of mean seedsize eaten was available for each species in each habitat, and one estimate of theproportion of pecks yielding seeds was made for each habitat. Hence, there was onlyone independent observation of food intake rate and its variance for each species ineach habitat. Habitats were compared by calculating 95-percent confidence intervals ofthe difference between the means. Differences were considered to deviate significantlyfrom zero if zero fell outside of the confidence intervals. Confidence intervals wereapproximated as the mean food intake rate ± 2 SE.ResultsFood intake ratesSeed intake rates in habitats were determined primarily by the sizes of seedseaten. Peck rate (Table 6) and the fraction of pecks that yield seeds (Table 7) wereconsistently highest on the valley floor, but seed intake rates there were never higherthan in other habitats (Table 1). The proportion of pecks that yielded seeds hadlittle influence on estimated intake rates or comparisons of food intake rates betweenhabitats, and it is not included in the food intake rates presented here (Table 6,Figure 17). Estimates of seed size made by the two different methods (Table 6) werecorrelated (r = 0.72, n = 9, P = 0.03), and conclusions based on the results are thesame regardless of which method is used. Results from both methods are presented inTable 1 and in Figure 17. Hereafter, I focus on results from data collected from birdsTable 6. Performance of captive sparrows foraging in an aviary for naturally occuring seeds. Entries are means (SE).Method 1: seed mass estimated from vomit samples from birds foraging in the aviary. Method 2: seed mass estimated frompreference data. Intake rate is the product of peck rate and seed mass. Entries in bold type are from the habitat that aspecies occupies. See Methods for details.Species HabitatPeck Rate(pecks / s)Seed MassMethod 1^Method 2(mg)^(mg)Seed Intake RateMethod 1^Method 2(mg/s)^(mg/s)Dark-eyed Junco Pinyon-juniper 0.99 (0.03) 0.16 (<0.01) 0.16 (0.01) 0.16 (0.01) 0.15 (0.01)Alluvial Fan 0.52 (0.05) 0.09 (<0.01) 0.13 (0.02) 0.05 (0.01) 0.07 (0.01)Valley Floor 1.08 (0.07) 0.08 (<0.01) 0.10 (0.02) 0.09 (0.01) 0.10 (0.03)Black-throated Sparrow Pinyon-juniper 0.76 (0.04) 0.15 (0.01) 0.11 (0.01) 0.11 (0.01) 0.14 (0.04)Alluvial Fan 0.60 (0.04) 0.15 (0.02) 0.21 (0.08) 0.09 (0.02) 0.13 (0.05)Valley Floor 1.23 (0.07) 0.08 (0.01) 0.12 (0.02) 0.10 (0.01) 0.14 (0.03)Sage Sparrow Pinyon-juniper 0.67 (0.05) 0.22 (0.01) 0.16 (0.02) 0.14 (0.01) 0.11 (0.01)Alluvial Fan 0.70 (0.06) 0.18 (0.01) 0.29 (0.09) 0.13 (0.01) 0.20 (0.07)Valley Floor 1.18 (0.06) 0.09 (0.02) 0.09 (0.01) 0.10 (0.02) 0.11 (0.01)75Table 7. The fraction of potential food items that yields edible seeds, estimated bygathering seeds and squashing them.HabitatEdible Fraction of SeedsP^Standard ErrorPinyon-juniper 0.69 0.02Alluvial Fan 0.61 0.02Valley Floor 0.80 0.0176Figure 17. Differences in foraging success between habitats + 2 SE. Bold, solid symbolsdepict differences between habitats occupied by a given species and unoccupiedhabitats, calculated as (occupied habitat) — (unoccupied habitat). Light, dashedsymbols depict differences between the two habitats not occupied by the given species.Habitat symbols: "PJ" pinyon pine-juniper woodland, "AF" alluvial fan, "VF" valleyfloor. Two estimates of each difference in foraging success are presented, based on thetwo methods of estimating mean seed size (Method 1 and Method 2, respectively).PJ:VFBlack-throated SparrowPJ-AF AF-VFf- 770.15-0.10-0.05-0.0 -0.05--0.10--0.15-Dark-eyed Junco0.2-0.1-0.0-0.1--0.2-T0. Sparrowa)cr)a)CCa)ca)a)a)0 AF-PJ^VF-PJIVF-PJHabitat ComparisonAF-PJVF-AF78foraging in the aviary (method 1) because they are less variable than results from seedpreference data.Habitat comparisonsUnder the foraging success hypothesis, large differences in foraging success arepredicted between habitats. Only dark-eyed juncos experienced a large and significantdifference in foraging success between habitats (Figure 17). Juncos foraging on thealluvial fan ingested seeds at only 31 percent of the rate that they did in pinyon-juniper. On the valley floor they ingested seeds at 56 percent of the rate in pinyon-juniper. However, sage sparrows and black-throated sparrows showed similar differencesin food intake rates between habitats. Sage sparrows on the valley floor consumed seedsat 71 percent of the rate in pinyon-juniper. On the alluvial fan, they consumed seedsat 93 percent of the rate in pinyon-juniper. Black-throated sparrows outside of pinyon-juniper consumed seeds at 82 to 91 percent of the rate that they did in pinyon-juniper.A second prediction of the foraging success hypothesis is that species forage mostsuccessfully in different habitats and dwell in the habitats in which they are mostsuccessful. Only dark-eyed juncos achieved higher food intake rates in the habitat thatthey occupy than in other habitats (Table 6, Figure 17). It was unclear whether sagesparrows and black-throated sparrows are most successful at feeding in their usualhabitats because both species exhibited small differences in seed intake rates betweenhabitats relative to error (Figure 17). However, differences in mean seed intake ratebetween occupied and unoccupied habitats were zero or negative. This suggests thatat best sage sparrows and black-throated sparrows gain only a slight feeding advantageby dwelling in the habitats that they occupy or, at worst, that they might suffer adisadvantage.DiscussionI tested the hypothesis that nonoverlapping habitat distributions of sparrow speciesresult from species' differences in foraging ability. This hypothesis can explain habitat79partitioning only if foraging ability can account for: why species prefer some habitatsover others, why species distributions are restricted, and why each species occurs alone.Here, I summarize why the foraging success hypothesis fails on all three counts.Under the foraging success hypothesis, habitat preferences should result fromdifferences in feeding ability between habitats. Only one of the three species, dark-eyed junco, clearly experienced large significant differences in food intake rate betweenhabitats. Sage sparrows and black-throated sparrows exhibited very similar foragingsuccess in all three habitats. Are these differences in foraging success sufficientto restrict species habitat distributions? Juncos feeding outside of pinyon-juniperexperienced food intake rates that were 30 percent to 60 percent of those inside pinyon-juniper. Such large disadvantages to foraging outside of pinyon-juniper are sufficientto explain why juncos are restricted to this habitat. Sage sparrows and black-throatedsparrows clearly are not doomed to low foraging success outside of their usual habitats,but it is less certain whether such small differences in foraging ability between habitatsmay be sufficient to restrict their habitat distributions.Insights from the theory of habitat selection and comparative data from otherpopulations suggest that the sparrows that I studied would be more broadly distributedif food alone shaped their distributions. Theoretically, species could specialize on singlehabitats despite only slight tradeoffs in fitness between habitats. The difference inforaging success between two habitats at which a population switches from being ahabitat specialist to a generalist depends upon population size and the rate at whichfeeding rate decays as population density increases (Fretwell and Lucas 1970, Fretwell1972). A species might remain a specialist for either of two reasons: foraging rate isindependent of population density or population density is low. However, neither ofthese conditions is likely to be true for sparrows. Feeding rates of sparrows declineas flock size increases (e.g., Caraco 1979) suggesting that foraging success is likely todecline as population density increases. Also, food appears to limit population density80during winter (Chapter 4) suggesting that intraspecific competition for food might besufficient to force species into less than ideal habitats when habitats differ only slightlyin foraging success.Comparative data from other species also suggest that sparrows would be generallydistributed if food shaped their distributions. Abundances of Galapagos ground finches(Geospiza spp.) in habitats are roughly proportional to food abundance along anelevational gradient (Schluter 1982). A habitat having half of the food of another hasapproximately half of the finch density of the other. In my study area, however, species'population densities are near zero outside of the habitats in which they are mostabundant despite the presence of suitable foraging conditions in those habitats (see alsoChapter 1). Hence, differences in foraging success between habitats are unlikely to beresponsible for habitat specialization and habitat partitioning by sparrows.Finally, different species must achieve their highest food intake rates in differenthabitats if foraging ability is to account for habitat partitioning. Because the speciesthat I studied live in different habitats, each species should be most successful foragingin the habitat that it occupies. This prediction was not fulfilled. Although juncosclearly achieved higher feeding rates in the habitat that they occupied than in otherhabitats, sage sparrows and black-throated sparrows actually experienced slightly higherfood intake rates in habitats other than those that they occupy. One of two conclusionsfollows from this result: either sage sparrows and black-throated sparrows are absentfrom the habitats in which foraging success is highest, or differences in foraging abilitybetween habitats are minor, suggesting that species might be more widely distributed.Conclusions from other studies of the role of food in shaping species distributionsvary in the importance attributed to food. Food is not responsible for habitatpartitioning by warblers (Phylloscopus spp.) breeding along an elevational gradient(Price 1991). Species specialize on forest habitat types (conifers vs. hardwoods) eventhough their foods are equally abundant in both habitat types. Also, individual birds81in transition areas between forest types readily forage in both types of trees. AmongGalapagos ground finches, however, food explains species' distributions along anelevational gradient (Schluter 1982). Species' distributions corresponded closely withthe distributions of their foods.Obviously, food must play some role in shaping species distributions: species canlive only where the food supply is sufficient for survival. The absence of species fromareas in which food supply is adequate warrants explanation. What other factorsmight explain habitat partitioning by sparrows in the present study? In my work, Ihave addressed these: vegetation and substrate structure, predation, and interspecificcompetition (Chapter 1). Structural features of habitat might affect the ability ofbirds to search for food. Predation risk might affect food intake rates because of atradeoff between feeding and scanning for predators (McNamara and Houston 1986,1987, Chapter 3). However, foraging success should depend on both habitat structureand predation risk and hence both were incorporated in my test of the foraging successhypothesis. The aviary enclosed low vegetation, rocks and leaf litter that mightinfluence foraging success. Also, sparrows in the aviary responded to birds of preyoutside of the aviary, and predators occasionally attempted to attack sparrows insidethe aviary. The failure of the foraging hypothesis suggests that the effects of habitatstructure and predation risk on foraging success are too weak to account for habitatpartitioning.In a separate paper, I test more directly the hypothesis that predation risk itselfcould be responsible for habitat partitioning if species are safest in different habitatsand risk is severe, but consistency in species' vigilance patterns between habitatssuggests that this is unlikely (Chapter 3).Finally, there is the competition hypothesis. My results support Lack's (1944)argument that species' differences in fitness between habitats are generally small andunlikely to be responsible for species distributions. The most conspicuous factor that82might restrict species' distributions is competition from ecologically similar species.Other indirect evidence suggests that current competition is possible among the species.The species share estimated preferences for the same foods, and they forage in similarmicrohabitats (Chapter 1). Finally, food appears to limit population density duringwinter, and the species are capable of depleting the amount of resources available toone another (Chapter 4). In view of this evidence, more direct tests of the competitionhypothesis are desirable.83Chapter IIIPREDATION, FOOD, VIGILANCEAND THE HABITAT DISTRIBUTIONS OF WINTERING SPARROWSInstances in which predation restricts the distributions of individual species arewell known (e.g., Werner et al. 1983, Werner and Gilliam 1984, Aronson 1989). Moreinteresting are situations in which predation affects the distributions of more than oneprey species, shaping community structure. For example, prey species might segregateby habitat if they are safest in different habitats and the risk of predation is severe.A morphological characteristic that renders a species safe in one habitat may placeit at risk in others, and species may occupy different habitats because they are safestin different habitats. This mechanism has received little attention in the literature,although it has been demonstrated among limpets (Mercurio et al. 1985). Limpet shellshaving undulate edges are well suited to irregular rock surfaces and poorly suited toflat rocks where the undulations can be used by predators to dislodge the limpet. Shellswith smooth, flat edges are well suited to smooth flat substrates and poorly suited toirregular rock surfaces. This type of tradeoff in safety between habitats diminishes oreliminates competition between species that use similar resources. Alternatively, severalprey species might all be safest in the same habitat (e.g., Mittelbach 1984, Longlandand Price 1991), and predation could intensify competition if all prey species are forcedinto it to avoid predators (Mittelbach 1984).Some data exist to suggest that predation might shape the habitat distributions ofbirds. Pulliam and Mills (1977) showed that granivorous bird species that use differentmethods of escaping predators dwell in different habitats. Species that forage close tovegetative cover and flee to it are safest there and at greater risk farther from cover(Watts 1990). Species that feed in open habitats, far from cover, and fly far away fromdanger approach vegetative cover reluctantly, suggesting that they may be at greater84risk close to cover than away from it (Lima 1990). Hence, species that escape predatorsby different methods may actually be safest in the habitats that they occupy.Here, I test the hypothesis that tradeoffs in predation risk between habitats areresponsible for habitat partitioning among three species of wintering sparrows alongan elevational gradient in the Sonoran Desert of southern California. Although theylive in different habitats, sage sparrows (Amphispiza belli), black-throated sparrows(A. bilineata), and dark-eyed juncos (Junco hyemalis) respond to predators similarlyby fleeing into woody vegetative cover. All three species forage closer to cover thanpredicted by food availability (Chapter 1). I tested the hypothesis (1) that eachspecies is safest in its preferred habitat against the alternative hypothesis (2) thatall species are safest in the same habitat. I measured rates at which species scannedthe environment, vigilance, and compared qualitative changes in vigilance levels ofindividual birds between habitats using an experimental enclosure. I predicted that ifhypothesis (1) is true species should exhibit differences in vigilance between habitats.Furthermore, some species should increase vigilance when moved from one habitatto another whereas other species should decrease vigilance because predation riskincreases for some species and declines for others. In contrast, hypothesis (2) predictseither similar levels of vigilance, or similar changes in vigilance when species are movedbetween habitats because they experience similar changes in risk between habitats.These predictions are derived in detail in the next section.In the course of testing the above two hypotheses, I also tested a predictionstemming from the theory of vigilance that has not been tested previously in thewild. Vigilance for predators is assumed to occur at the expense of food intake, andthe amount of vigilance that maximizes survival depends upon both the risk of beingkilled by predators and the risk of starving (McNamara and Houston 1986, 1987, Lima1987b). Predictions of the theory of vigilance have been tested by measuring changes invigilance in response to changes in factors that should affect predation risk (e.g., Lima851987a, Lendrem 1983). Predictions of how vigilance should respond to changes in foodabundance have not been tested.My results support the prediction that food abundance affects vigilance. They alsosupport the hypothesis that species perceive similar changes in predation risk betweenhabitats, suggesting that tradeoffs in safety between habitats are not responsible forhabitat partitioning among wintering sparrow species.Predation, vigilance and starvationTo test the hypothesis that predation shapes species distributions, I testpredictions of how predation risk and vigilance should change between habitats.Vigilance levels can be assessed by moving individual birds between habitats andrecording the amount of time that they scan the environment. If species exhibit thesame increase or decrease when moved from one habitat to another, I conclude thatall species rank the two habitats similarly by the threat of predation. In this case,predation can not account for habitat preferences. If on the other hand, some speciesexhibit an increase in vigilance and others a decrease in vigilance when moved from onehabitat to another, I conclude that species rank habitats differently and that predationcould account for species distributions.Here, I briefly describe the components of predation risk and how increased riskof different aspects of predation results in conflicting predictions of vigilance. I alsodescribe how consistency in changes in vigilance between habitats can be used to inferthat species perceive similar changes in the risk of predation between habitats.Factors contributing to predation risk include: the probability of being attacked,the probability of discovering an attack, and the probability of escaping an attackgiven that it has been discovered. The effects of each of these factors on vigilance hasbeen modelled by Lima (1987b). Probability of attack is represented by attack rate.Probability of discovering an attack is represented by the amount of time requiredfor the predator to reach the prey once it has begun to attack: shorter attack times86result in lower probability of detecting an attack for any level of vigilance. Finally, theconditional probability of escaping an attack can be represented by the distance fromcover: a bird that flees to cover is less likely to escape from a predator the farther itis from cover. If only one component of predation risk varies between habitats, effectson vigilance depend upon which component varies. Increased probability of attackshould result in increased vigilance (Figure 3b in Lima 1987b). Decreased probabilityof detecting a predator should result in decreased vigilance (Figure 3a in Lima 1987b).Decreased probability of escaping an attack can result in either increased or decreasedvigilance depending on the probability of being attacked: vigilance increases at highattack rates and declines at low attack rates (Figure 3b in Lima 1987b). All three ofthese factors probably vary between habitats, hampering the prediction of changes invigilance. Hence, it is difficult to use observed changes in vigilance between habitats tomake inferences about how predation risk varies between habitats. However, species canbe expected to exhibit similar changes in vigilance rates between habitats if they aresubjected to similar changes in predation risk. This is the prediction that I test.Both food abundance and predation risk can theoretically affect vigilance levels.Hence, it is necessary to take into account differences in food abundance betweenhabitats if meaningful comparisons of vigilance are to be used to test hypotheses aboutvariation in predation risk between habitats. If vigilance and feeding are exclusiveactivities, the risk of being killed by predators is traded against the risk of starvingto death (McNamara and Houston 1986, 1987, Lima 1987b). A bird that ignores thethreat of predation risks being killed whereas a bird that is constantly vigilant risksstarvation. Both risks can be described as functions of the amount of time spentvigilant, and the optimal level of vigilance is that at which the decrement in predationrisk as vigilance increases is offset by an equal increment in the probability of starving(McNamara and Houston 1987).87How might changes in food abundance affect vigilance? A change in foodabundance alters the probability of starving at any level of vigilance and is representedas a shift in the curve describing the risk of starvation as a function of vigilance(Figure 18). The resulting change in vigilance depends upon how the starvation curve isshifted. No change in vigilance is expected if the new starvation curve is parallel to theold one. A decrease in vigilance is expected if the slope of the starvation curve increaseswhereas an increase in vigilance is expected if the slope decreases (Figure 18).Because food abundance should affect vigilance, changes in food supply betweenhabitats could produce changes in vigilance that are not attributable to differencesin predation risk between habitats. Alternatively, changes in food supply couldmask changes in vigilance resulting from differences in predation risk. One methodof controlling for changes in food abundance between habitats is to describe therelationship between vigilance and food abundance within habitats and then askwhether differences in vigilance between habitats are greater or less than would beexpected from the amount of food available. I accomplished this using analyses ofcovariance. The process of analytically controlling variation in food abundance is itselfa test of the prediction from vigilance theory that food abundance affects vigilancelevel.MethodsThe experiment was carried out in a portable aviary placed over natural vegetationin habitats. The aviary measured 4 m x 4 m x 2 m and was constructed of woodenframes covered with screening that could easily be seen through (further details inChapter 2). Video tape recordings of solitary birds foraging for naturally occurringseeds were made from a blind and used to estimate both vigilance and the foragingprofitability of the patch of ground enclosed by the aviary.The aviary was an appropriate venue for observing vigilance. Although theenclosure might provide foraging birds with a sense of security from attack, birds of88Figure 18. Hypothetical relationships between level of vigilance and the probability ofstarving and the resulting relationship between optimal vigilance level and food supply(after McNamara and Houston 1987). a, c) Family of solid upwardly-sloping curvesrepresents the probability of starving in habitats differing in food supply. The numberto the left of each curve indicates rank of food abundance (1 = low food abundance,5 = high food abundance). Dashed downwardly-sloping curve is the probability of beingkilled by predators. b, d) Relationship between optimal level of vigilance and foodsupply resulting from the tradeoff depicted in the corresponding left panel. The signof the slope depends upon whether starvation curves diverge or converge with increasingvigilance.LowVigilance89 High - b=•alt02a)UCco0)5Low.^.^.Low^High^1 2 3 4 5Vigilance Food Abundance (rank)High • da)UCCl0)5=ast0Low ^High^1 2 3 4 5Food Abundance (rank)90prey attempted to attack sparrows foraging in the aviary, and the sparrows appearedto exhibit normal levels of vigilance. Any effect of the aviary on vigilance shouldbe constant among habitats. Finally, vigilance serves functions other than predatordetection. It might also provide warning of attack by dominant members of a species(e.g., Waite 1987, Roberts 1988) or of attack by other, food-robbing species (e.g.,Thompson and Lendrem 1985). Both of these factors were controlled by observingsolitary individuals free from either threat.Experimental DesignThe experiment was carried out according to a repeated-measures design(Figure 19). Six individuals of each species were caught from the wild 2 to 7 daysbefore the study began. Birds were housed singly in outdoor cages and maintained ona commercially available mix of seeds for pet finches, meal worms, water, and a vitaminsupplement. Each individual was videotaped during one 30-minute foraging trial ineach habitat. This set of trials was a subset of trials conducted to test for differencesin foraging ability between habitats (Chapter 2). Briefly, the experiment was carriedout in two blocks, each containing three birds of each species. The first block was testedbefore the second. The design controlled any effect that previous experience might haveon vigilance by ensuring that equal numbers of birds of equal experience were tested ineach habitat (Figure 19). Each day, the aviary was moved to a new location, and threebirds, one of each species, were tried in random order (Figure 19) beginning one hourafter sunrise.Variables and data collectionVigilance was defined as any time that a foraging bird held its beak such that thelong axis was horizontal to the ground or at an angle above the horizontal. Vigilancerate was calculated as the amount of time spent vigilant divided by net foraging time,defined as total time spent foraging less time spent vigilant. This measure based onnet foraging time ensures that vigilance rate and peck rate (see below) estimated from91Figure 19. Schematic representation of one block in experimental design. One birdof each species was designated as the first, second and third individual of its species.upper) Each bird was introduced into the aviary in each habitat. Numbers within rowsindicate the order in which individuals experienced habitats. Previous experience in theaviary was equalized among habitats. lower) Each day, three birds, one of each species,were tried in a random order. The experiment consisted of two experimental blocks.1^2^33^1^2//r/ I2^3 / II,,^I, I, I,^11392HABITATA^B^C,,,, SPECIES,^,^,, ,/ A / B^CRANDOM ORDER/93the same foraging periods are independent measures and not related to one anothersimply because they are exclusive activities in a fixed time budget. I attempted tomake one estimate of vigilance rate for each bird based on two minutes of foragingactivity, although this was not always possible because of poor visibility or because thebird foraged for only a short period of time. Some estimates were made from severalshort periods of foraging which amounted to 2 minutes of cumulative foraging time.Estimates were made from the first segment(s) of video tape in which a bird could beseen well enough to determine what it was doing. Foraging time in a segment of videotape was measured using a microcomputer programmed as an event recorder. Then,vigilance time was determined by counting the number of frames in which the birdassumed a vigilant posture and dividing the count by the rate at which the camerarecorded frames (30 frames / s).I used the rate at which a bird pecked at either the ground or at plants to pick upseeds as a measure of the profitability of foraging in the patch of ground enclosed bythe aviary. Total number of pecks was divided by net foraging time. Two difficultiesarose with this method: time spent vigilant and time spent actively feeding are onlypartially independent because birds sometimes assume a vigilant posture while handlingseeds, and it was often not possible to determine whether a seed was being handledwhile a bird was vigilant. Hence, patch profitability calculated in this manner assumesthat no handling time is spent vigilant. I made a second calculation of peck rate thatincluded the opposite assumption that all handling time was spent vigilant. Vigilancetime was subtracted from total foraging time as in the first estimate, and handling timewas added back in. Handling time was estimated as the total number of pecks timesthe mean handling time of the seed species being eaten. Mean handling times wereestimated by observing birds consume seeds in the lab (Chapter 1). The results wereunaffected by the method used to estimate patch profitability. For simplicity, I presentthe results calculated under the assumption that all handling time is spent vigilant94because that is probably the more realistic assumption in habitats in which birds eatseeds with long handling times.AnalysisThe goal of the analysis is to determine whether species exhibit consistent changesin vigilance between habitats. I used an analysis of covariance to estimate the differencein vigilance between habitats for each species. The model wasV = ri -I- OX,where V is vigilance rate, X is patch profitability, # is the slope of the relationshipbetween patch profitability and vigilance, and ri is the intercept of habitat i. Thedifference between intercepts reflects the change in vigilance between two habitats.A statistical test for consistency among species was not possible because the analysisresulted in only one independent observation of change for each species: all of theobservations of a species were used to estimate the difference between intercepts.Some degree of confidence in the interpretation of the results can be drawn from thestatistical confidence surrounding individual ANCOVA's. Statistically significantdifferences between habitats strengthen confidence in the rankingsA separate analysis was carried out for each pair of habitats. Each analysis wasrestricted to foraging periods in which seed types common to the two habitats werebeing eaten to ensure that whatever differences in vigilance were observed betweenhabitats resulted from differences in predation risk between habitats rather thanfrom differences in seed types. Seeds of different size could result in different rates ofvigilance because they offer different opportunities to scan for predators without cost.Small seeds that require little handling time offer little opportunity for free vigilancewhereas large seeds that require longer handling times offer more opportunity forvigilance. Sparrows usually consumed small seeds of the grass Schismus barbatus onthe valley floor and larger seeds of Bouteloua aristidoides in pinyon-juniper habitat.95They consumed a mix of both of these grass species and a third large species Aristidaadscensionis on the alluvial fan. Therefore, I analyzed data from the valley floor withdata from the alluvial fan when birds there were feeding on Schismus barbatus, and Ianalyzed separately data from pinyon-juniper and data from the alluvial fan when birdswere feeding on large seeds.Analysis of the data was complicated by missing observations. Not all individualscould be observed eating both large and small seeds while foraging on the alluvial fan.One outlier was from a bird that was lethargic on one day, and it was removed. In afew other instances, good video tape footage was unavailable.I used generalized least squares regression to analyze the results (see Rawlings1988). The main advantage of generalized least squares was that it could incorporaterepeat observations on individuals without estimating large numbers of parameters.Generalized least squares is a form of least squares regression in which the assumptionof independent observations can be relaxed. This is accomplished by including a weightmatrix in the least squares equation that describes the variance-covariance structureof the data. The matrix effectively adjusts the error mean square according to thedegree of independence in the data set. I relaxed the assumption of independence toincorporate repeat observations on the same individual. This was reflected in a weightmatrix in which l's lay down the diagonal and off-diagonal elements representing pairedobservations on individual animals were set to the correlation coefficient between theresiduals of repeat observations. The average correlation between paired residuals fromseparate analyses was —0.53. It may result from changes in vigilance associated withprevious experience in the aviary: a bird might be nervous when tried the first time,be calm for the second trial, and be frustrated with captivity by the third trial. Someindividuals were reluctant to forage by their final trial. I used the mean correlationamong analyses rather than using separate estimates for each analysis for two reasons.First, a single parameter estimate should be applicable to all of the analyses because96the same set of individuals is used in all analyses. Second, generalized least squares issensitive to error in the estimation of weights, and a combined estimate of correlationshould be more robust than individual estimates.ResultsFood supply affects vigilancePlots of vigilance against food supply suggest that vigilance varies with foodsupply (Figure 20). Vigilance declined as foraging patch profitability increased in 4 ofthe 6 analyses, significantly so in three (Table 8). A positive relationship was observedin one analysis, although the pooled slope was not statistically significant (Table 8).In the sixth analysis (Figure 20a), no case could be made for a relationship betweenvigilance and food supply on the alluvial fan or valley floor where juncos do not usuallyOMIT.A decline in vigilance with increasing foraging patch profitability is counter to theintuitive reasoning that increased patch profitability lessens the threat of starvationand allows more time to be spent vigilant. A negative relationship between vigilanceand patch profitability could be an artifact if both vigilance rate and peck rate arecalculated from total time spent foraging because time is constrained and the numberof pecks must decline as vigilance increases. This is an unsatisfactory explanation of thepresent results because I calculated vigilance rate and peck rate from net foraging time.The negative relationship is completely consistent with theory if the slope of the curvedescribing the risk of starvation increases with increased food abundance (Figure 18a).Comparisons of vigilance rates between habitatsEvaluation of similarities in species' responses to changing habitats depends uponmaking habitat comparisons for each of the species. These comparisons were possibleonly between the valley floor and the alluvial fan. Comparisons were not feasiblebetween alluvial fan habitat and pinyon-juniper habitat for several reasons. First, therewere insufficient data on sage sparrows to make the comparison reliably. Second, the97Figure 20. Relationship between vigilance and food abundance for sparrows consumingsimilar seed types in pairs of habitats. Comparisons between valley floor habitat andalluvial fan habitat are in the left set of panels, and comparisons between alluvial fanhabitat and pinyon-juniper habitat are in the right set of panels. Lines spanning onlythe range of data values are least-squares regression lines fit in a single habitat. Valleyfloor: 64v 77 , ( ) ; alluvial fan "a", (   ) ; pinyon-juniper: 64 97p  ,(^) . Solid parallel lines spanning the entire range of the abscissa were fitby generalized least squares analysis of covariance (see Methods). Results of statisticalinferences are in Table 8.1.▪ 20:6.0.8 1.00.5-0.4-.0.8 1.0 1:20.3-ap\0.8-0.6-0.4-0.2-0:6a p0.2-0.1-0.30^0.35^0.4098Dark-eyed Juncoa.^.^.0.30^0.35^0.40Black-throated Sparrowc d0.6-0.4-0.2-.^.^.0.30^0.35^0.40Sage SparrowePatch Profitability (pecks / s)99Table 8. Levels of statistical significance associated with tests of Analysis ofCovariance models describing the relationship between vigilance and peck rate in pairsof habitats (Figure 20). The alternative hypotheses tested were: either slope differsfrom zero or mean vigilance rate differs between habitats (Overall ANCOVA), pooledslope differs from zero, the intercepts of the parallel lines are unequal, and the slopes ofindependent linear regression lines fit to each habitat are unequal.POverall^Pooled^Unequal^UnequalSpecies^ ANCOVA^Slope^Intercepts^SlopesAlluvial Fan — Pinyon-juniper comparisonDark-eyed Junco 0.03 0.03 0.81 0.68Black-throated Sparrow 0.11 0.50 0.19 0.33Sage Sparrow 0.17Valley Floor — Alluvial Fan comparisonDark-eyed Junco 0.77 0.54 0.96 0.81Black-throated Sparrow 0.02 0.01 0.03 0.40Sage Sparrow 0.02 0.01 0.02 0.89100relationship between vigilance and food supply available to dark-eyed juncos was notamenable to simple comparisons between habitats because the lines did not appearto be parallel, suggesting that the difference in vigilance observed between habitats isdependent on the quantity of food available (Figure 19b).Comparisons of vigilance rates between habitats were made to test the hypothesisthat species exhibit similar changes in vigilance between habitats. Relative to foodabundance, both sage sparrows and black-throated sparrows were more vigilant on thevalley floor than on the alluvial fan. This result suggests that both species perceivesimilar changes in predation risk between habitats, and that they are both likely to besafest in the same habitat and at greater risk in the other. Hence, it is unlikely thatspecies differences in the tradeoff in predation risk between habitats is responsible forhabitat partitioning between these two species.DiscussionBehavioral ecology has offered insight into how populations are affected bydecisions made by individual animals. For example, the habitat distribution of aspecies reflects choices made by individuals weighing the tradeoff between the riskof predation and foraging profitabilities of habitats (e.g., Milinski and Heller 1978,Werner and Gilliam 1984, Gilliam and Fraser 1987, Todd and Cowie 1990). I useda defense behavior, vigilance, to detect perceived changes in predation risk betweenhabitats and test two hypotheses of how predation might influence the distributionsof wintering sparrow species. The first hypothesis was that sparrow species aredistributed one-per-habitat because the species are safest from predators in differenthabitats. The alternative was that the species are all safest in the same habitat andthat predation can not be responsible for habitat partitioning among them. Theanalysis was complicated because defense behavior itself is subject to a tradeoff betweenthe predation and starvation. It was necessary to compare levels of defense betweenhabitats after adjusting for variation in food abundance.101Vigilance and the tradeoff between food and predationThis study supports the previously untested prediction from vigilance theorythat vigilance level is affected by food abundance (McNamara and Houston 1987,Lima 1987b, Figure 18). Vigilance declined in two species as food patch profitabilityincreased, but the third species exhibited a positive although statistically insignificantrelationship between vigilance and food supply (Figure 20).A negative relationship between vigilance level and food patch profitability isconsistent with theory (McNamara and Houston 1987, Figure 18), but it is inconsistentwith the intuitive reasoning that increased food supply alleviates the threat ofstarvation, allowing more time for vigilance. The relationship begs the question: isa family of starvation curves that results in such a relationship (Figure 18a) likely tooccur in nature? The answer is yes. McNamara and Houston's (1987) simple modelof starvation results in a family of starvation curves that could yield either a positiveor a negative relationship between vigilance and foraging patch profitability. In themodel, food intake was assumed to be normally distributed with a mean and a variance,and the probability of starvation was the probability that food intake fell below somethreshold. Increased vigilance reduced mean food intake and thereby increased theprobability of starving. If habitats are assumed to differ in mean food intake, a familyof starvation curves results that both diverges and converges (Figure 21), regardlessof whether increasing vigilance reduces food intake linearly or curvilinearly, andregardless of whether variance in food intake is assumed to be constant or allowed tobe proportional to net food intake. Solutions to the tradeoff between vigilance andstarvation that lie in the area where curves diverge yield a positive relationship betweenvigilance and food abundance whereas those in the area where curves converge yield anegative relationship as in Figure 18.An intuitive explanation of a negative relationship between vigilance level andfood patch profitability follows from the family of curves in Figure 21. Starvation102Figure 21. A family of hypothetical starvation curves derived from the model ofMcNamara and Houston (1987). Food intake is assumed to be normally distributed,and the probability of starving is the probability of obtaining less than some thresholdof food. Mean food intake in the absence of vigilance differs between the curves.Adjacent curves diverge with increasing vigilance to the left of the dashed vertical lines,and they converge to the right.High -c01.,ascj(r)0loas_c)0a_Low -Low^HighVigilance104curves converge when the probability of starving is high. Under this condition, thebenefit of increased food availability is so great that survival is maximized by sacrificingvigilance time to take advantage of it. This interpretation might be used to concludethat sparrows in nature are subject to high starvation risk, but that conclusion isunwarranted because birds used in this study were fasted prior to trials in the aviaryand may have behaved as if the risk of starvation was inordinately high.The family of starvation curves in Figure 21 also suggests a plausible explanationof why black-throated sparrows exhibited a switch from a negative relationship betweenvigilance and foraging patch profitability while feeding on small seeds to a positiverelationship while feeding on large seeds. The sparrows may have been closer tostarvation while feeding on small seeds than they were while feeding on large seeds. Ifso, they may have been subject to convergent starvation curves while feeding on smallseeds and to divergent curves while feeding on large seeds. Why would this be true forblack-throated sparrows but not for the other two species? Black-throated sparrows aresmaller than the other two species (mean body mass 13 g vs. 17 g and 19 g for sagesparrows and dark-eyed juncos, respectively), and they should be less likely to starve atany food intake rate because they have smaller energy requirements.Predation and the habitat distributions of speciesSage sparrows and black-throated sparrows exhibited similar changes in vigilancebetween habitats. This result suggests that species rank the danger of predation inhabitats similarly and that predation is unlikely to shape their habitat distributions.If predation strongly affected species' distributions, species would have similar habitatdistributions.This conclusion is consistent with previous work on these three species in thestudy area. The species use similar methods to escape from avian predators, and theyforage closer to woody vegetation than expected from the distribution of available food,suggesting that the species should be safest in the habitat with the greatest amount105of cover (Chapter 1). Interestingly, a similar conclusion was recently drawn regardingthe risk of predation and microhabitat use by desert rodents (Longland and Price1991). Heteromyid rodents tend to forage farther from cover than other species, anddespite their remarkable adaptations for feeding far from cover (bipedal locomotion andlarge auditory bullae for hearing), these rodents are safest close to cover. Perhaps thepredation hypothesis will explain the habitat distributions of a group of species differingmarkedly in their response to predators such as species that use different methodsof escaping from predators like those observed by Pulliam and Mills (1977) or Lima(1990).What might account for the habitat distributions of these sparrows if a tradeoff inpredation risk between habitats is not responsible for them? There is another way inwhich predation might restrict species to different habitats. Species might segregate ifthey experience greater predation rates when they occur together than when they occurseparately (Holt 1977, 1984, Schmitt 1987). This possibility remains to be tested. Foodis an unlikely candidate. First, species are not distributed among habitats as predictedfrom food availability: some species are absent from habitats in which food is at leastas abundant as those in which they occur (Chapter 1). Also, transplant experimentssuggest that only small differences in foraging ability exist between habitats and thatspecies might be predicted to have broader habitat distributions than they do if foodgoverned distributions (Chapter 2). Habitat partitioning could result from interspecificcompetition (Lack 1944, Svirdson 1949, Noon 1981, Price 1991). Indirect evidencesupports this hypothesis. Profitabilities of seeds in the different habitats rank similarlyamong species suggesting that species share preferences for the same resources, andindeed they consume some of the same seed species even though they occupy differenthabitats (Chapter 3). Food appears to limit population abundance, and sparrows arecapable of depleting the amount of food available to one another (Chapter 4).106Hence, interspecific competition, rather than predation, is likely responsible for habitatdistributions. More direct tests of the competition hypothesis remain to be done.107Chapter IVINTERSPECIFIC COMPETITIONAND THE HABITAT DISTRIBUTIONS OF WINTERING SPARROWS:EVIDENCE FROM A FOOD SUPPLEMENTATION EXPERIMENTDirect experimental evidence of interspecific competition among birds is rare(see Schoener 1983, Wiens 1989b) because bird populations operate on a largespatial scale that defies experimentation (e.g., Wiens et al. 1986). Most evidenceof competition between bird species is based on observational studies of patterns inmorphology, distribution and abundance (see Wiens 1989a, 1989b). Patterns alonecan be misleading, and it is desirable to challenge the competition hypothesis withfurther tests (e.g., Dayton 1973). An indirect test of the competition hypothesis canbe achieved by measuring the requisite conditions: species should overlap in their useof limiting resources, and each species should deplete resources available to others. Ifeither condition is absent, species do not compete for resources; hence verifying theconditions is a strong test of the competition hypothesis.Here, I present the results of a food addition experiment carried out on granivorousbirds wintering in the Sonoran Desert of California. The three common species ofsparrows in the area are sage sparrows (Amphispiza belli), black-throated sparrows(Amphispiza bilineata) and dark-eyed juncos (Junco hyemalis). There is evidencethat they partition habitats along an elevational gradient (Chapter 1). The speciesoverlap extensively in the types of seeds they eat. Their distributions differ from whatmight be predicted from the availability of winter food alone: sage sparrows and black-throated sparrows are absent from the habitat in which their food, the seed of herbs, ismost abundant. These distributions could result from interspecific competition, eitherinterference competition occurring presently or evolutionary specialization on habitatsresulting from competition in the past. I undertook the present experiment to test the108competition hypothesis. I sought evidence of food limitation and evidence that resourcedepletion could occur among sparrows if they were to occupy the same habitat. Theexperiment also afforded an opportunity to test a prediction that finches (Fringillidaeand carduelinae: Emberizidae) use richer, more patchily distributed resources thansparrows (emberizinae: Emberizidae) (Benkman and Pulliam 1988). My results wereconsistent with the hypothesis that interspecific competition occurs among sparrows,and supported the prediction that sparrows and finches differ in the types of resourcesthey use.Possible inferences from a food additionFood additions have been criticized recently (Wiens 1989b), so I shall first clarifywhat inferences about food limitation and depletion can and can not be drawn fromtheir results.Food is limiting when individual survival depends upon its availability. Morefood means higher food intake rates and thus lower chances of death from starvationor predation (McNamara and Houston 1987). One way that individual birds inthese circumstances can enhance their food intake rates is to locate themselves inplaces where intake rate is high, provided that the cost of relocating is not exorbitant(Bernstein et al. 1991). Hence, the recruitment of individuals to a food addition plotwould support the hypothesis that food is limiting to individuals. Note that thisargument does not depend on the assumption that birds strictly maximize food intakerate in accordance with optimality theory; it assumes only that foraging birds aresensitive to spatial variation in food abundance and that they increase food intake rateover what it would be if spatial variation in food availability were ignored. Indeed,birds are sensitive to spatial variation in food abundance even if they do not strictlymaximize food intake rates (Krebs et al. 1974, Smith and Sweetman 1974, Cowie1977, Ydenberg 1984, Hanson and Green 1989, Kacelnik and Todd 1992). Behavioralevidence of food limitation is indirect, because a change in survival itself is not109demonstrated. Nevertheless, if resources are not limiting, then a response to a foodaddition is not expected (e.g., Pulliam and Dunning 1987). Therefore, a food additionexperiment offers an opportunity to falsify the hypothesis that resources are limiting.Depletion occurs when one species reduces the rate at which food may be gatheredby another. To demonstrate depletion, resource availability in the presence of apotential competitor must be less than it would have been in the absence of thatcompetitor. A food addition experiment provides an opportunity to examine theabilities of species to deplete resources. Supplying a single food profitable to severalspecies mimics a pulse of food produced at the beginning, during or at the end ofa growing season (e.g., Grant and Grant 1980, Schluter 1982, Repasky personalobservation). As species deplete shared resources, diet overlap often declines (see alsoSchoener 1982). The dynamics of depletion can be observed easily after a food additionbecause resource production is zero. Depletion can be described as the percentagedifference between the amount of food that was available to a species and the amountthat would have been available in the absence of a potential competitor. Depletion canbe calculated if food standing crop and removal rates are known.Food additions also illuminate overlap in the use of resources. Only species thatcan share a resource, that is species having overlapping "fundamental niches," canrespond to a food addition. Although these species are potential competitors, theycan compete only in the presence of foods profitable to all. In the absence of commonfoods, they do not compete for food, and ecological differences between them shouldbe manifested in the way they respond to the food addition. For example, sparrowsand finches have been predicted to use different resources based on the rates at whichthey can handle seeds (Benkman and Pulliam 1988). In this view, a finch of a given sizeeats larger seeds than a sparrow of the same size because it can more profitably handle(seed mass ingested / handling time) large seeds than the sparrow. Finches, therefore,can obtain more of their daily energetic needs per seed, and they require fewer seeds110per day for the same metabolic expense. Because they need fewer seeds, finches canspend more time and energy searching out patches of particularly high seed abundance.Sparrows, on the other hand, must spend most of their time searching out and eatingmany small seeds, and they are less able to make use of very rich but scattered patchesof food. Hence, finches should be more mobile and better able than sparrows to exploitrich, widely scattered patches of food.How might differences in mobility between sparrows and finches be illuminated bya food addition? A food addition plot is effectively an extremely rich patch of foodprofitable to both finches and sparrows. Finches should recruit to a food additionplot faster than sparrows if indeed they are more mobile than sparrows, as predictedby Benkman and Pulliam (1988). Differences between species in the availability ofnatural food should be reflected by differences in the standing crop of seeds presentwhen species depart a food addition. Birds are expected to remain in a patch offood (food addition plot) only while food intake rates there exceed the average inthe environment (Krebs et al. 1974, Charnov 1976, Cowie 1977). If finches abandona food addition plot earlier at a higher seed density than sparrows, it suggests thatthey experience higher intake rates in the environment than sparrows for the followingreason. Finches foraging on a large seed such as millet can achieve a higher intake ratethan similarly-sized sparrows at any food density because they handle seeds faster thansparrows (Figure 22). As seed abundance declines, finches should depart earlier thansparrows only if their foraging success off of the food addition plot is greater than thatof sparrows (Figure 22).Finally, competition can limit the number of species present in a guild. A foodaddition might relax food limitation and hence competition temporarily, allowingspecies number to increase. Hence, an increase in the number of species following afood addition would also support the competition hypothesis.111Figure 22. Food intake rate as a function of seed density for two species differing onlyin the time required to handle a single food type. The function is Holling's (1959)Ae hdisc equation: R = 1+A where R is intake rate , A is the rate at which seeds areencountered, e is seed mass, and h is handling time. Food density is expressed as aconstant which when divided by the shorter handling time gives encounter rate; intakerate is the proportion of the maximum intake rate attainable by the species havingthe faster handling time (^). The species having the lower intake rate(^) handles seeds 70 percent slower than the species with the higher intakerate, corresponding to the rate black-throated sparrows handle millet seed relative tohouse finches (Repasky unpublished data). The horizontal dashed line is a hypotheticalaverage food intake rate by the slow species and the intake rate at which it abandonsa food addition plot. The fast species departs earlier than the slow species only if itsaverage intake rate in the environment is greater than the intake rate that it wouldhave when the slow species departs (arrow).to -0.8 -0co-„— 0.6 --lizi4--a)0.4 -ma)as4E' 0.2 - ea.---------------do, oe e e o. e..• 4., ••do o0.0 -111111^I^I^I^I^I^I^III^I^I^I^1^11111^10.1 1 10^20Seed Density (encounter rate constant)113MethodsStudy AreaI carried out the study in the Sonoran Desert of southern California, USA.Experimental and control plots were located in alluvial fan habitat in the CoachellaValley near Deep Canyon Desert Research Center. The rocky creosote-scrub vegetationhas been described in detail by Zabriskie (1979). Briefly, an alluvial fan is a fan-shapedincline formed by the flow of rock and soil out of a narrow valley. The intermittentstreams that feed an alluvial fan change course through time and dissipate as theydescend. The resulting surface is a series of terraces and washes. Prominent plants onterraces are creosote bush (Larrea tridentata), brittlebush (Encelia farinosa), burrobush(Ambrosia dumosa), sweetbush (Bebbia juncea) and the chollas (Opuntia acanthocarpa,0. echinocarpa) and the prominent plants of the washes are palo verde (Cercidiumfloridum), chuparosa (Beloperone californica), desert lavender (Hyptis emoryi), andcheesebush (Hymenoclea salsola).Granivorous birds of the alluvial fan habitat include Gambel's quail (Callipeplagambelii), mourning dove (Zenaida macroura), house finch (Carpodacus mexicanus),black-throated sparrow and white-crowned sparrow (Zonotrichia leucophrys) (Weathers1983). The white-crowned sparrow is a winter resident, whereas the other speciesare year-round residents (Weathers 1983). Sparrows consume the seeds of grasses(e.g., Schismus barbatus, Aristida adscensionis, Bouteloua aristidoides) and forbs(e.g., Cryptantha sp., Perityle emoryi, Camissonia sp.) whereas house finches tendto consume the seeds of shrubs such as creosote bush and desert lavender (Repaskyunpublished data). Quail consume seeds of both shrubs and herbs (Goldstein and Nagy1985). The natural diets of dove in the area are not known.Experimental ManipulationOne control plot and one treatment plot were located on each of two alluvialfans. Deep Canyon alluvial fan and Devil Canyon alluvial fan (hereafter sites 1 and 2,114respectively) were approximately 11 km apart and separated by three mountains. Plotsat each site were at least 1.6 km apart. At site 2, a coin toss was used to assign plots totreatment and control groups. Plot assignment at site 1 was dictated by ownership ofthe properties. The plot on the property of Deep Canyon Desert Research Center wasdesignated as the control plot because seed additions were not permissible. Each plotwas a 8.4-ha rectangle measuring 600 m by 140 m. Plots were marked with flaggingtape.I used white proso millet (6 mg/seed, Union Pacific Mills, San Bernardino, CA)in the experiment because it was preferred by captive black-throated sparrows overother seeds in a mix. Seeds were applied once at a density of 2.71 g/m 2 . The dosewas approximately 58 times the estimated standing crop of foods observed during theprevious winter (46.8+ SE 6.2 mg/m 2 ). Such a large dose was warranted because birdsresponded relatively slowly to the added seeds and recruited only after most of it hadbeen removed by other granivores. The standing crop of native seeds, sampled duringthe experiment (see below), was relatively low (17.4+ SE 2.9 mg/m2 ) because of lowrainfall during the previous growing season.Seeds were spread by two researchers casting seeds as evenly as possible by handwhile repeatedly walking the width of a plot. Although no placebo was spread oncontrol plots, we paced about the plots to disturb them in a manner similar to that ontreatment plots. The seed used in the study had been fumigated with methyl bromideto minimize germination.Bird CensusesPlots were censused six times over a six-day period prior to the seed additionand 10 times over a 31-day period afterward. Each plot was visited 4 or 5 times priorto the treatment and censused once or twice on each of those days. Post-treatmentcensusing began 12 days after the seed addition so that birds would have time to recruitundisturbed to the study plots. Each plot was then visited at 4-day intervals, and two115censuses, one in the morning and the other in the afternoon, were carried out each visit.Each census lasted 2 hours. Morning censuses began at sunrise, and afternoon censusesended at sunset. Over the course of the experiment, equal numbers of censuses wereconducted in each time period of the day and by each of two observers.Censuses were carried out in a 2-ha strip (500 m x 40 m) centered within each 8.4-ha study plot. An observer walked the length of the strip down the center at a rate of20 m every 5 min, recording the birds within the 20-m segment in front of him. Everyeffort was made to avoid double-counting individual birds that remained within thecensus strip, moving forward along with the observer, or that crossed the census stripturned and reentered it.DepletionSeed standing crop was estimated repeatedly after the seed addition to determinedepletion rates. Fifty sampling points were randomly chosen in each census strip. Ateach point, a combined area of 0.125 m 2 was sampled from 8 square plots of equalsize. Subplots were positioned at one-meter intervals along a straight line centered atthe sampling point and oriented parallel to the width of the census strip. Seven or 8samples were taken from each food addition plot on each of the five visits followingthe seed addition. Seed standing crop on control plots was sampled less systematicallythrough time.All seeds on plants were counted in situ. Surface soil was collected by scrapingit into a bag using a toothbrush. The depth of the collection depended on surfacehardness and was limited to 1 cm in soft soil in order to restrict the sample to seedsavailable to birds. Soil from all subplots at a sampling point was combined and waslater spread over a large tray where seeds were counted visually.Seed disappearance alone overestimates seed removal by birds because granivorousants and rodents also took seeds. Therefore, I estimated seed removal by two methods.116First, I estimated the amount of seed removed by birds as the difference betweentotal removal and an estimate of the seeds removed by other granivores (mainly antsand rodents). To accomplish this, I assumed that non-avian granivores removed seedat a constant rate, estimated as the rate at which seeds disappeared between the timethe millet was spread and the time that birds began to recruit. Once birds began torecruit, the rate at which they removed seed was calculated as the difference betweenthe removal rate observed then and the rate at which food was removed before theyarrived. Removal rate was calculated as the instantaneous rate of seed disappearanced^where p is the proportion of seed present at the beginning of a time intervalthat remained at the end of the time interval, and t is time elapsed (days). Removalwas calculated for each period between visits to a plot and summed to give totalremoval by birds and other granivores. This method probably underestimated seedremoval by ants and rodents and, hence, overestimated removal by birds because itassumed that ants and rodents began removing seeds at full speed immediately whereastheir rate of removal undoubtedly accelerated as the numerical response of rodents wascompleted. (Functional responses probably were immediate, and a numerical responseby ants from colonies adjacent to the site was probably limited because ants are central-place foragers that must commute from nest colonies to the study plot.)Second, I estimated seed removal by birds more directly using census dataalong with estimates of metabolic requirements of birds taken from the literature.Seed removal on each census day was calculated as the product of bird density andindividual consumption rate. Removal on days between census days was interpolatedby calculating the change in daily removal between two census days and distributing itevenly among the intervening days.Consumption rates were obtained from known allometric relationships betweenenergy expenditure and body mass. The consumption rate of an individual free-living117bird was estimated asER -= AC (1 )where R is the amount of seed removed by an individual bird (g/bird/d), E is totaldaily energy expenditure (kJ/bird/d) of free living birds, A is the assimilation efficiencyand C is the energy content of seeds (kJ/g). E was calculated as 0.949W° 749 forpasserines and 0.703W" for non-passerines, where W is body mass in grams (Nagy1987). Assimilation efficiency A was estimated to be 0.81, the mean of the valuesreported by Kendeigh et al. (1977:Table 5.7) for cardinals (Cardinalis cardinalis) andsong sparrows (Melospiza melodia) feeding on seeds of hemp (Cannabis sativa) andfoxtail (Setaria faberii). C, the energy content of seeds, was assumed to be 19.1 ± 0.8(SD) kJ/g, which is the mean of the values presented by Kendeigh and West (1965) forgrass seeds.Seed removal calculated from census data underestimates true removal and soprovides a lower bound of seed removal by birds. This is because both the number ofbirds removing seeds from the plots and the amount of seed removed by individualbirds were probably larger than estimated for three reasons. First, peak densities ofbirds may have been missed. After a food addition, bird abundance increases andthen declines. Censuses conducted at 4-day intervals are likely to miss the maximumabundance if the response is quick and short-lived as it was in this study. Second,on any day, more birds probably used a food addition plot than was assumed in thecalculations. I assumed that the average number of birds observed during the fourhours of census each day was the total number of birds using the plot and that thosebirds obtained all of their food from the plot on that day. Undoubtedly, more birdsused the plot, and each bird obtained some fraction of its daily need from the plot.Under these circumstances, the amount of seed removed by the population would beRp = x • R • Alpo^(2)118where Rp is seed removal by the population of birds (g/d), x is the number of birdsobserved on the census, R is the minimum daily food requirement of an individualbird (g/d) from eq. 1, pr is the proportion of the daily requirements that a birdactually obtains from the study plot, and Po is the proportion of birds using the plotthat are actually observed on the census. If p r is equal to Po , eq. 2 reduces to theproduct of the number of birds observed and the daily requirements of individualbirds, the estimate I used. Otherwise, my estimates should be biased according tothe ratio Nip,. That ratio was probably greater than one. Birds visiting the foodaddition plots may have obtained most of their daily food requirements there. Theenergy requirements of sparrows and finches, calculated as given above, could be metwith fewer than 100 millet seeds per day. That amount of seed might theoreticallybe obtained in only 8 minutes by a house finch feeding non-stop from a dish of milletand in 13 minutes by a black-throated sparrow in the same circumstances (Repaskyunpublished data). If either species took as long as 30 seconds to find a seed inthe field, daily food requirements could be fulfilled in about an hour. Hence, birdsvisiting the plot probably obtained nearly all of their daily food requirements there.The proportion of birds actually observed on censuses, Po , was certainly less than1 because we counted only actively foraging birds and because I completely missedbirds visiting the food addition plots outside the census periods. Finally, minimumdaily energy requirements were probably greater than estimated from allometricrelationships because energetic expenditure increases as ambient temperature declines.The relationships used describe energy expenditure during the breeding season whereasthe study was carried out during the winter, when daily temperatures averaged 12°C.This bias may be ameliorated somewhat because thermoregulatory needs are partiallycompensated by energy spent in day-time activities (Webster and Weathers 1990).Variances of seed removal rates based on energetics were calculated bycompounding the variances of the component variables in eq. 1. Variance of the119estimate of daily energy expenditure, E, for an individual of a species was calculatedusing the equations given by Nagy (1987). Variances of assimilation efficiency, A, andenergy content, C, were calculated from data in the original sources. Note that for eachof these three variables I used the estimate of the population variance rather than theestimate of the variance of the population mean because I was interested in predictingvalues for individual species. To compound variances, the variance of the product oftwo estimates was calculated asV AR(A • C) = V AR(A)V AR(C) + 2A V AR(C) + .2c. V AR(A)(Bickel and Doksum 1977), where, for example, A is assimilation efficiency and C isseed energy content in the denominator of eq. 1. Variances of more complex functionsof a variable such as the reciprocal or exponentiation were calculated using the deltamethod (Efron 1982). The variance of the sum of two estimates was calculated as thesum of the variances (Snedecor and Cochran 1967:190).Number of SpeciesMeasures of species richness tend to exhibit sample size biases (Hurlbert 1971,Simberloff 1972): on average, larger samples include more species. Therefore, I usedrarefaction to standardize species richnesses by sample size before making comparisons.I used the flock as my resampling unit, defined as a group of birds encountered togetherat the same time and place during a census, because individual birds often forage ingroups and are not independent observations.Species richness on each plot was estimated for the entire period following the seedaddition. Species richness was taken to be the actual number of species observed on theplot on which the fewest number of flocks (N1) was observed. Species richness on eachof the other plots was adjusted to the same number of flocks by resampling. Adjustedrichness was calculated as the mean number of species in random samples of size Nfusing the program NICHE (Schluter 1988b).120Statistical TestsI treated plots on the same alluvial fan as paired samples because they might notprovide independent observations of species abundances. Changes in bird abundancesand species richness were, therefore, tested using one-tailed paired-sample t-tests(Snedecor and Cochran 1967) with n = 2 pairs. One-tailed tests were used because,a priori, abundance and diversity are expected to increase when food is added.ResultsDensityThe density of granivorous birds increased 24 fold before it began to subside(Figure 23). Densities were higher on post-addition treatment plots than on controlplots or on the treatment plots before food was added to them. The response wasshort lived. At site 2, densities began to increase as censuses resumed 13 days after thefood addition, and they declined to pre-treatment/control levels by 31 days after theaddition. The recruitment phase at site 1 began before censusing resumed, and only thedecline phase of the response was observed.Eight of nine species were more abundant on the treatment plots than on controlplots (Table 9). The chance probability of an increase in that many species wasapproximately 0.018 by a one-tailed binomial test. Only the responses by house finchesand white-crowned sparrows were statistically significant as judged by paired t-tests(Table 9).DepletionMillet disappeared quickly from the treatment plots (Figure 24). Ants beganremoving seed as we spread it, often within a few minutes of when it landed on theground. They also altered the distribution of millet. By the end of the study, most ofthe millet on both study plots was in refuse piles of the desert harvester ant (Messorpergandei), and the remaining sparrows travelled between ant mounds and scratchedthrough piles. The standing crop at site 1 declined 99.0 percent (2.71 g/m 2 to 0.03121Figure 23. Density of granivorous birds over the course of the food addition experimenton control plots (x ) and treatment plots (<>^ ). Vertical barsrepresent standard errors.12250macmal■10 _.am5masite 1maonT+ 1 -(15U)16-1'. 50(7")ca)0^_Diij 105IMdEr1 am-5^0^5^10^15^20^25^30Days Since Food AdditionI^s^I^g^g^1^II^I-5^0^5^10^15^20^25^30i •I.- •MEMAI^e^g^I^•^I^f^Isite 2123Table 9. Response of granivorous birds to a short-term food addition. Maximumdensity for each plot after the seed addition and the mean of the differences betweenthe maxima of paired treatment and control plots (d). Differences significantly greaterthan zero by a one-tailed one-sample t-test (one degree of freedom) are marked.bird density (birds/ha/hr)site 1Treatment Controlsite 2Treatment Control d11.2 0.0 32.8 0.0 22.023.8 0.0 30.0 1.2 26.2*9.5 0.0 8.8 1.2 8.5*1.0 0.5 3.0 0.2 1.60.8 0.0 1.5 1.0 0.60.5 0.0 0.5 0.0 0.51.2 1.0 0.0 0.0 0.10.2 0.0 0.0 0.0 0.10.0 0.0 0.0 0.5 -0.231.5 0.5 39.0 2.0 34.0*Speciesmourning dovehouse finchwhite-crowned sparrowblack-throated sparrowBrewer's sparrow'lark sparrow 2Gambel's quaildark-eyed junco3lesser goldfinch4species combined*P < 0.051 Spizella breweri2 Chondestes grammacus3 Junco hyemalis4 Carduelis psaltria124Figure 24. Standing crop of millet over the course of the food addition experimentat Site 2. Standing crop was estimated by sampling seeds on the ground. Means andstandard errors are connected by the dashed line. Data for Site 1 are not presentedbecause millet standing crop had declined to levels lower than could be detected by thetime sampling commenced.5.00 :.1.1. ..,.. Oft ... ..... Ow .11b ... ...•II• , •I• ,••■• ..... , ..... I.11, ..., ......11.■■0.01 -0^5^10^15^20^25^30Days Since Food Addition126g/m2 ) during the 12 days between the food addition and the resumption of censuses.It was too low to be detected (below 6.9 mg/m 2 , 99.7 percent removal) by day 16.The decline in standing crop at site 2 occurred in three phases (Figure 24). Firstwas a period of relatively slow decline during the first 13 days after food was addedand before birds discovered the site. 50.5 percent of the millet disappeared duringthis period. Second was a period of rapid decline between days 13 and 17 whenbirds discovered the plot and 48.3 percent of the millet disappeared. Finally milletabundance declined slowly after day 17. It dropped below the minimum detectablestanding crop by day 26.Birds removed between an estimated 11±0.3 (SE) percent and 46±28 percent ofthe millet at site 2. Total removal by birds at site 1 was not estimated because mostof the seed had been removed and bird abundance was declining by the time censusesresumed. The difference between the lower and upper estimates of seed removal at site2 reflects differences in the assumptions that the methods make about the cause of thedramatic decline in millet abundance between days 13 and 17. Most of the decline isassumed to be by ants and rodents in the lower estimate whereas it is attributed tobirds in the upper estimate. The lower estimate, based on bird abundance and foodconsumption, assumed that bird abundance between days 13 and 21 was never higherthan it was on day 17. If bird abundance actually peaked at about day 17, it failedto increase enough to account for the dramatic decline in food abundance. The upperestimate, on the other hand, assumed that the dramatic decline was due largely tobirds. Removal by ants and rodents was assumed to be constant at the average ratethat had been established in the 13 days between the time food was added and censuseswere resumed. Fluctuations in total removal around that rate, as happened betweendays 13 and 17, were attributed to birds. The actual amount of millet taken by birdsshould lie between these extreme values.The pattern of recruitment by different species suggests that bird species depleted127the amount of food available to one another. Species recruited in sequence (Figure 25),and early arriving species reduced the amount of food available to later arrivingspecies. At site 2, mourning doves and house finches peaked in abundance firstfollowed by white-crowned sparrows and then black-throated sparrows. The patternof species abundances was similar at site 1 where most of the seed was depleted beforecensusing began: mourning doves and house finches were at their peak abundanceswhen censusing resumed whereas white-crowned sparrows peaked later and black-throated sparrows lacked a distinct peak in abundance. Indeed, species that peakedin abundance early recruited in greater numbers and removed more seed than thosethat peaked later (Table 10), and they depleted the amount of seed available to laterarriving species.The extent of depletion is the amount of seed that would have been availableto species recruiting more slowly had species recruiting quickly failed to appear. Tomake the calculations it was necessary to assume that seed removal by birds did notaffect the amount of seed removed by ants and rodents. The effects of removal on foodavailability were considerable (Table 10). At site 2, for example, the amount of foodavailable to white-crowned sparrows would have been 3.2 times greater than actualvalues, had mourning doves and house finches not recruited. Food availability to black-throated sparrows would have been 23 times greater in the absence of all earlier speciesand 6.4 times greater in the absence of white-crowned sparrows. Hence, the more slowlyrecruiting species might have recruited in larger numbers than they did had the quicklyrecruiting species been absent.Number of SpeciesSupplemental food did not increase the number of species present on plotssignificantly. Although more species were observed on the treatment plots than oncontrol plots, all of the newly recruited species occur in alluvial fan habitat irregularlyor in low numbers (Weathers 1983). The numbers of species were similar on controlall bird species(R ^), mourning doves (o^ ), house finch), and black-throated sparrow(*), white-crowned sparrow (x128Figure 25. Trajectory of millet seed removal rate at Site 2 over the course of thefood addition experiment: all species including ants and rodents (^),(o —^). Total removal rate was estimated from a regression of standing cropagainst time, and removal rates for birds were estimated from energetic allometricrelationships described in the text.10010.15---,d+ 0.14 -c.iE 0.13 -cg)2(15 0.12 -cc 0.10 -15^20^25^30Days Since Food Addition130Table 10. Millet removal and extent of food depletion among predominant birdspecies. Each entry is the percentage increase in the amount of millet seed availableto more slowly arriving species that would have occurred in the absence of the targetspecies or of the target species and all species recruiting more quickly. Species are listedin descending order by recruitment rate.Increase in foodavailability (percent)in the absence of a ...Time of peakabundance^Millet removed^Target^Target(days since^(percent of total^species^species and allTarget species^food addition) removed by birds)^only^faster speciesSite 1Mourning doves 12 24.4±0.2 34±0.4 34±0.4House finchl 12 27.4±0.5 62±1.5 117±2.2White-crowned sparrow 16 40.0±0.4 976±25.8 2239±56.0Black-throated sparrow _2 4.1±0.1Site 2Mourning dove 17 50.5±2.7 105±10.5 105±10.5House finch 17 21.1±1.9 79±14.3 268±43.2White-crowned sparrow 21 22.5±0.7 536±54.8 2240±1251Black-throated sparrow 25 4.2±2.3continued ...131Table 10 continued.1 Depletion was underestimated because this species recruited to the food addition plotsbefore censuses were resumed.2 No distinct peak occurred.132plots and treatment plots after being adjusted for sample size differences (Table 11),and the paired t-test was not significant (d = 0.442, t = 0.777, P > .05). Althoughthe power of the test was low (P(reject when ea,/ = 3 species) = 0.149), the observeddifference in species richness was small. Hence, richness itself was not increased; rather,rare species increased in local abundance and were more likely to be observed on thefood addition plots. A true increase in species number is unlikely to occur in a suchshort-term experiment because species must be present in low numbers if they are todiscover a food addition plot at all.DiscussionI carried out an indirect test for interspecific competition by testing the conditionsnecessary for competition. In doing so, I put to a test the hypothesis that present-daycompetition is responsible for habitat partitioning among three species of sparrows. Myresults supported the competition hypothesis. Here, I clarify this interpretation andspeculate on how the results bear on the interactions between other avian granivores.Food limitationBirds recruited to the food addition plots while food abundance remained elevated.This response is consistent with the hypothesis that food limits population abundance,although it is indirect support because a change in survival was not demonstrated.The strength of the support depends on whether bird abundance would be positivelycorrelated with food abundance in the absence of food limitation. Here, I give tworeasons why food is likely to be limiting when birds accumulate in areas of high foodabundance. First, the only existing models that predict a positive correlation betweenconsumer abundance and prey abundance are rate maximizing models describingsystems in which the addition of food is likely to increase survival. Second, it isunlikely that consumers would exhibit behavior that enhances food intake rate if it wereuncoupled from survival.Food abundance probably affects survival of consumers that enhance food intake133Table 11. Species richness on control and treatment plots as observed and as adjustedfor differences in sample sizes. The adjusted richness of each plot was calculated as themean of a set of samples, each the size of the smallest sample, drawn from the originaldata. See text for details.Plot^Species richness (No. species ± S.E.)unadjusted^adjustedSite 1control^2^2.0treatment 8^3.0Site 2control^5^2.6treatment 6^2.4134rate by choosing where to forage. The models of Bernstein et al. (1988, 1991) describethe distribution of consumers under circumstances similar to those of granivorous birdsduring winter, consumers searching for immobile prey during the non-breeding season.Consumers maximize intake rate by abandoning areas (patches) in search of betterareas if the intake rate in them is less than the average intake rate that they experiencein the environment. Increased food abundance should increase intake rate by decreasingboth the number of poor patches and the amount of time consumers spend in them.Would consumers enhance intake rate if it were independent of survival? Probablynot, a consumer doing so would gain no benefit and might incur a cost. A likely costis the risk of predation taken while needlessly searching out rich patches of food.Indeed, predation risk strongly influences the foraging behaviors of birds (e.g., Caracoet al. 1980, Valone and Lima 1987, Lima 1987, 1990). Hence, food is likely to belimiting in systems in which consumers actively distribute themselves according to preyabundance, as happens after a food addition.Support for the food limitation hypothesis is consistent with a growing body ofevidence that food limits the abundance of granivorous birds during the non-breedingseason. First, finch abundance (broadly defined as the abundance of small, seed-eatingpasserines) is positively related to seed abundance across sites over four continents,suggesting that food abundance influences population abundance in a consistentmanner in many places (Schluter and Repasky 1991). The present study sites in theSonoran Desert were included in that analysis, and the additional evidence presentedhere of food limitation at the study sites further supports the hypothesis that thepositive relationship between seed abundance and food abundance among continentsresults from food limitation.Second, the present results are consistent with the conclusion of Pulliam andhis colleagues that food limits sparrow abundance only during years of poor seedproduction. In their study site in southern Arizona, sparrow abundance is correlated135with food abundance during poor years but not in good years (Pulliam and Parker1979). The species composition of the guild of sparrows is predictable from resourcedistributions in poor years but not in better years (Pulliam 1983). Third, grasslandsparrows in southern Arizona failed to respond when food was supplemented in ayear of moderate food abundance (Pulliam and Dunning 1987). In the present study,birds recruited to food addition plots in a year when rainfall, a surrogate measure forseed production, was relatively low (26th percentile of the long-term distribution).The positive response during a year of low productivity in the present experiment isconsistent with the lack of response during a year of moderate productivity in Pulliamand Dunning's experiment.DepletionBirds are small players among desert granivores (Parmenter et al. 1984). Iestimated that most of the seed removed during the experiment was taken by othergranivores, probably ants and rodents. So much seed was taken by these othergranivores that it might have been impossible to effect an increase in the amount offood available to birds without saturating ants and rodents with a large quantity ofseed. The presence of abundant, competitively superior species that can drasticallyreduce food supplies could also make it difficult to detect interactions among lesscommon species (e.g., Hairston 1980, 1981, Grubb 1986). That ants and rodentscan severely deplete the availability of food to birds supports the hypothesis thatcompetitors in non-avian taxa limit the morphological diversity of granivorous birdson continents (Schluter 1988a, 1988b).Can sparrow species deplete the amount of food available to one another? Becausebirds removed less seed than ants and rodents removed, it might be argued that thereis little potential for depletion. To the contrary, my results suggest that birds cansignificantly deplete the availability of resources to one another. For example, white-crowned sparrows and black-throated sparrows were the slowest species to respond136to the food addition and removed the least amount of seed. Yet, had white-crownedsparrows not recruited to plots, the amount of food available to black-throated sparrowswould have been 6 to 11 times the actual amount (Table 10). Hence, species that occurtogether are capable of affecting one another's abundances through food depletion.Competition among other granivorous birdsA number of species responded to the food addition, more than the three specieshypothesized to partition habitats, suggesting that their fundamental niches overlap.In nature, these species might also overlap in diet at times when resources are availablethat are profitable to all. They would compete if resources were limiting and were beingdepleted at those times.The extent to which species shared diets and may have been competing for naturalfoods at the time of the experiment should be reflected by similarities in the timesthat they arrived at the food addition plot and departed from it. Arrival time reflectsmobility, and departure time the mean intake rate a species experiences outside of afood addition. Species that use the same resources and value them similarly shouldhave similar intake rates and mobilities. More mobile species should recruit faster thanless mobile species, and species with high intake rates are likely to depart sooner thanthose with low intake rates.House finches differed markedly from sparrows in both arrival times and departuretimes, suggesting that house finches and sparrows were not competing for naturalseeds during the experiment. House finches recruited quickly to the food additionand abandoned it early, while sparrows were still recruiting. High mobility and highintake rates of finches are consistent with the prediction that finches use profitableresources that are clumped into rich patches whereas sparrows use poorer, more evenlydistributed resources (Benkman and Pulliam 1988). House finches are sufficientlymobile to specialize on clumped resources because they handle large seeds moreefficiently than sparrows and require fewer seeds per day. Indeed, house finches eat137larger seeds than sparrows during mid winter (Repasky unpublished data).Unlike the contrast between sparrows and finches, differences in arrival anddeparture times between other taxa lack easily identifiable causes. Here, I speculateon the causes and their implications for interspecific competition. Mourning dovesresponded much like house finches; they recruited early and departed early. They, too,probably use rich, spatially clumped resources similar to those used by house finchesand are more likely to compete with house finches than with sparrows. Among thesparrows, white-crowned sparrows recruited more quickly than black-throated sparrows,and they began to leave the plot before black-throated sparrows. This difference intiming might be attributable to a difference between the diets of the two species.Although the species share many of the same seed species, white-crowned sparrows tendmore toward herbivory than do black-throated sparrows (Repasky unpublished data).White-crowned sparrows eat mostly fresh vegetation, and they are abundant onlyduring winters when sprouting plants are abundant (Repasky personal observation).If white-crowned sparrows are dependent upon fresh vegetation, competition betweenthem and black-throated sparrows might be restricted to wetter winters in which thereis a sufficient amount of fresh vegetation for white-crowned sparrows.138GENERAL DISCUSSIONMy census data and earlier census data (Weathers 1983, Weathers unpublisheddata) suggest that species are distributed one species per habitat. Censuses wereconducted in the interiors of habitats, away from habitat boundaries. Hence, sagesparrows, black-throated sparrows and dark-eyed juncos enjoy sole occupancy of thecenter portions of habitats, although it is unclear how much species distributionsoverlap at habitat boundaries and how closely distribution boundaries of species matchhabitat boundaries.I tested alternative hypotheses that might explain why three sparrows aredistributed one-species-per-habitat along an elevational gradient. The alternativesincluded: food, predation, structural features of habitat, and interspecific competition.Here, I summarize the evidence against the food, predation and vegetation structurehypotheses and show how the data are consistent with the competition hypothesis. Idiscuss alternative explanations of the results briefly and consider how the competitionhypothesis might be pursued in the future.Summary of resultsFood is unlikely to shape species' habitat distributions. Several lines of evidenceargue against the prediction from the food hypothesis that species specialize onalternate foods that occur in different habitat. First, species' foods are not restrictedin distribution. Foods are at least as abundant outside the habitats that speciesoccupy as they are inside. Second, species do not specialize on different types of food.Species in two different habitats eat seeds common to both habitats. Also, the foragingprofitabilities (mass of seed ingested / handling time) of seed species rank similarlyamong the sparrow species. Finally, only dark-eyed juncos experienced significanttradeoffs in feeding ability between habitats when species were experimentallyintroduced into each habitat and allowed to forage. Differences between habitats in139the foraging abilities of sage sparrows and black-throated sparrows were small. Feedingrates outside of their typical habitats were at least as large as those inside, suggestingthat these two species would be more broadly distributed than at present if food wereresponsible for their distributions.Species differences in predation risk between habitats could account for use ofdifferent habitats by species if species are safest from predators in different habitats.This condition is unlikely to be true for the sparrows in my study area. Two resultssuggest that species perceive similar changes in predation risk between habitats ratherthan opposite changes in risk as predicted by the predation hypothesis. First, the threespecies rely on shrubby cover to escape from predation, and they forage closer to coverthan predicted from the availability of food. This result suggests that the species aresafest close to cover and that they are all safest in the habitat with the greatest amountof woody cover. Second, sage sparrows and black-throated sparrows both increased theamount of time they scanned the environment, relative to food abundance, when theywere moved from the alluvial fan to the valley floor, suggesting that at least these twospecies perceive similar changes in predation risk between habitats. Hence, if predationwere responsible for habitat distributions, the species should have similar distributionsrather than specialize on different vegetation types.Finally, species are unlikely to be adapted to structural features of their habitatsthat restrict their distributions. Foraging microhabitats used by individual species aremore broadly distributed than the species themselves. The effects of habitat structureon foraging ability and predation risk appear to be minimal. Only small differences inforaging ability existed between habitats, whereas large changes would be predictedif habitat structure strongly influenced foraging ability. Also, changes in vigilancebetween habitats were consistent among species whereas disparate vigilance patternswould be predicted if species perceived the relative safety of habitats differently becauseof structural habitat characteristics.140Interspecific competition may be responsible for habitat specialization. The mostconspicuous difference between habitats that might limit species' distributions is thepresence of other species of sparrows. Species are capable of eating the same species ofseeds, and they do so to the extent that seed species occur in more than one habitat.Data on seed profitability suggest that species would share preferences for the sameseeds if they were to feed in the same habitats. Species also forage in similar microhabitats. Clearly, these species are potential competitors.Competition is contingent upon resource limitation. Although observationaldata on sparrow abundance and food abundance were equivocal, the food additionexperiment provided evidence that food limits population abundance. Sparrowsrecruited to short-term food addition plots indicating that individual birds are on thelookout for rich food sources as they ought to be if food is limiting. No such responseis predicted in the absence of food limitation because individual birds would pay thecost of relocating without benefit. The experiment was carried out over the course ofa month. Food limitation over such a short time period is sufficient to suggest thatcompetition shapes species distributions along the elevational gradient because thepattern of distribution exists only during the winter. Sage sparrows migrate to breedin the Great Basin of the western United States, and dark-eyed juncos migrate uphill tobreed in pine forest near the tops of the mountains (Weathers 1983).Results of the food addition experiment also suggest that sparrows are capable ofdepleting the amount of food available to one another when they occur together. In theexperiment, which was carried out in alluvial fan habitat, both white-crowned sparrowsand black-throated sparrows recruited to the food addition. White-crowned sparrowsrecruited before black-throated sparrows and also departed before them. White-crowned sparrows reduced the amount of food available to black-throated sparrows.Hence, competition is a plausible explanation for habitat partitioning among thespecies. If competition is indeed responsible for habitat partitioning, it is likely to occur141by one of two mechanisms. First, species might avoid habitats in which they would losein competition. Second, species might segregate because of aggression among species.A third possible mechanism, exploitative competition, is unlikely to be responsible forsparrow distributions. Under exploitative competition species should be distributed aspredicted from the availability of their foods. Species might occupy different habitatsbecause food in preferred habitats has already been depleted to the extent that foragingis no longer profitable in them. These hypotheses are worth testing because they placethe competition hypothesis at risk of falsification.Alternative hypothesesHypotheses other than those considered here that might explain habitatpartitioning by sparrows fall into two broad categories. The first is a set of hypothesesin which at least two factors account for all species' distribution boundaries. Onefactor might explain some boundaries, and other factors other boundaries. Forexample, competition might limit some species' distributions and predation others toproduce the pattern one-species-per-habitat. The second set of alternatives consistsof reformulations of the hypotheses that I have considered after some assumptionsunderlying the tests have been relaxed or refined. For example, I concluded that speciesshould share preferences for seed species because the handling times of different seedsrank similarly among species. This conclusion rests on the assumption that seed typesare equally digestible by different species, an assumption that may not be true. Manyother simplifying assumptions have been made in this thesis. However, I believe thatthe present results are clear enough to suggest that the next logical step is to pursuethe competition hypothesis by more direct tests. Should those tests fail, it will beworthwhile to sift through more complex hypotheses and the assumptions of the presentanalyses for avenues to pursue.142BIBLIOGRAPHYAbbott, I., I.K. Abbott, and P.R. Grant. 1977. Comparative ecology of Galapagosground finches (Geospiza Gould): evaluation of the importance of floristicdiversity and interspecific competition. Ecological Monographs 47:151-184.Adams, M.J., and G.I. Bernard. 1977. Pronophiline butterflies (Satyridae) of the SierraNevada de Santa Marta, Columbia. Systematic Entomology 2:263-281.Aronson, R.B. 1989. Brittlestar beds: low-predation anachronisms in the British Isles.Ecology 70:856-865.Aschoff, A. 1981. Thermal conductance in mammals and birds: its dependence on bodysize and circadian phase. Comparative Biochemistry and Physiology 69A:611-619.Aschoff, A., and H. Pohl. 1970. Der Ruheumsatz von VOgeln als Funktion der Tageszeitand der KOrpergrae. Journal Fir Ornithologie 111:38-47.Becker, R.A., J.M. Chambers, and A.R. Wilks. 1988. The new S language. Wadsworth& Brooks/Cole Advanced Books & Software, Pacific Grove, California, USA.Benkman, C.W., and H.R. Pulliam. 1988. The comparative feeding rates of NorthAmerican sparrows and finches. Ecology 69:1195-1199.Bernstein, C., A. Kacelnik, and J.R. Krebs. 1988. Individual decisions and thedistributions of predators in a patchy environment. Journal of Animal Ecology57:1007-1026.Bernstein, C., A. Kacelnik, and J.R. Krebs. 1991. Individual decisions and thedistribution of predators in a patchy environment. II. The influence of travelcosts and structure of the environment. Journal of Animal Ecology 60:205-225.Bickel, P.J., and K.A. Doksum. 1977. Mathematical statistics: basic ideas and selectedtopics. Holden-Day, SanFrancisco, USA.Bowman, R.I. 1961. Morphological differentiation and adaptation in the Galapagosfinches. University of California Publications in Zoology 58:1-302.Burk, J.H. 1982. Phenology, germination, and survival of desert ephemerals in DeepCanyon, Riverside County, California. Madrono 29:154-163.Caraco, T. 1979. Time budgeting and group size: a test of theory. Ecology 60:618-627.Caraco, T., S. Martindale, and H.R. Pulliam. 1980. Avian time budgets and distancefrom cover. Auk 97:872-875.Chambers, J.M., W.S. Cleveland, B. Kleiner, and P.A. Tukey. 1983. Graphical methodsfor data analysis. Wadsworth International Group, Belmont, California, USA.Charnov, E.L. 1976. Optimal foraging theory: the marginal value theorem. TheoreticalPopulation Biology 9:129-136.143Cody, M.L. 1974. Competition and the structure of bird communities. PrincetonUniversity Press, Princeton, New Jersey, USA.Cowie, R.J. 1977. Optimal foraging in great tits (Parus major). Nature 268:137-139.Darwin, C.R. 1859. On the origin of species by means of natural selection. JohnMurray, London, UK.Dayton, P.K. 1973. Two cases of resource partitioning in an intertidal community:making the right prediction for the wrong reason. American Naturalist 107:662-670.Efron, B. 1982. The jackknife, the bootstrap and other resampling plans. Society forIndustrial and Applied Mathematics, Philadelphia, Pennsylvania, USA.Emlen, J.T. 1971. Population densities of birds derived from transect counts. Auk88:323-342.Esteban, M.D. 1989. Eficacia de un emetic() (apomorfina) para el estudio de las dietasde paseriformes granivoros. Ardeola 36:185-191.Fretwell, S.D. 1972. Populations in a seasonal environment. Princeton University Press,Princeton, New Jersey, USA.Fretwell, S.D., and L.H. Lucas. 1970. On territorial behaviour and other factorsaffecting habitat distribution in birds. Acta Biotheoretica 19:16-36.Futuyma, D.J., and S.S. Wasserman. 1981. Food plant specialization and feedingefficiency in the tent caterpillars, Malacosoma dissitra and M. americanum.Entomologia Experimentalis et Applicata 30:106-110.Garcia, E.F.J. 1983. An experimental test of competition for space between blackcapsSylvia atricapilla and garden warblers Sylvia borin in the breeding season.Journal of Animal Ecology 52:795-805.Gleason, H.A. 1926. The individualistic concept of the plant association. Bulletin of theTorrey Botanical Club 53:7-26.Goldstein, D.L., and K.A. Nagy. 1985. Resource utilization by desert quail: time andenergy, food and water. Ecology 66:378-387.Grant, P.R. 1986. Ecology and evolution of Darwin's finches. Princeton UniversityPress, Princeton, New Jersey, USA.Grant, P.R., and B.R. Grant. 1980. Annual variation in finch numbers, foraging andfood supply on Isla Daphne Major, Galapagos. Oecologia 46:55-62.Grinnell, J. 1904. The origin and distribution of the chestnut-backed chickadee. Auk21:364-382.Grubb, P.J. 1986. Problems posed by sparse and patchily distributed species in species-rich plant communities. pages 207-225 in J. Diamond and T.J. Case (eds.).Community ecology. Harper & Row, New York, New York, USA.144Hairston, N.G. 1980. The experimental test of an analysis of field distributions:competition in terrestrial salamanders. Ecology 61:817-826.Hairston, N.G. 1981. An experimental test of a guild: salamander competition. Ecology62:65-72.Hanson, J., and L. Green. 1992. Foraging decisions: patch choice and exploitation bypigeons. Animal Behaviour 37:968-986.Holling, C.S. 1959. Some characteristics of simple types of predation and parasitism.Canadian Entomologist 91:385-398.Holt, R.D. 1977. Predation, apparent competition, and the structure of preycommunities. Theoretical Population Biology 12:197-229.Holt, R.D. 1984. Spatial heterogeneity, indirect interactions, and the coexistence ofprey species. American Naturalist 124:377-406.Hurlbert, S.H. 1971. The nonconcept of species diversity: a critique and alternativeparameters. Ecology 52:577-586.Hurlbert, S.H. 1978. The measurement of niche overlap and some relatives. Ecology59:67-77.Kacelnik, A., and I.A. Todd. 1992. Psychological mechanisms and the marginal valuetheorem: effect of variability in travel time on patch exploitation. AnimalBehaviour 43:313-322.Kendeigh, S.C., and C.C. West. 1965. Caloric values of plant seeds eaten by birds.Ecology 46:553-555.Kendeigh, S.C., V.R. Dol'nik, and V.M. Gavrilov. 1977. Avian energetics. pages127-204 in J. Pinowski and S.C. Kendeigh (eds.). Granivorous birds inecosystems. International Biological Programme 12. Cambridge UniversityPress, Cambridge, UK.Krebs, J.R., J.C. Ryan, and E.L. Charnov. 1974. Hunting by expectation or optimalforaging? A study of patch use by chickadees. Animal Behaviour 22:953-964.Lack, D. 1944. Ecological aspects of species formation in passerine birds. Ibis 86:260-286.Lack, D. 1947. Darwin's finches. Cambridge University Press, Cambridge, UK.Lack, D. 1971. Ecological isolation in birds. Harvard University Press, Cambridge,Massachusetts, USA.Lendrem, D.W. 1983. Predation risk and vigilance in the blue tit (Parus caeruleus).Behavioral Ecology and Sociobiology 14:9-13.Levins, R. 1968. Evolution in changing environments. Princeton University Press,Princeton, New Jersey, USA.145Lima, S.L. 1987a. Distance to cover, visual obstructions, and vigilance in housesparrows. Behaviour 102:231-238.Lima, S.L. 1987b. Vigilance while feeding and its relation to the risk of predation.Journal of Theoretical Biology 124:303-316.Lima, S.L. 1988a. Vigilance and diet selection: a simple example in the dark-eyedjunco. Canadian Journal of Zoology 66:593-596.Lima, S.L. 1988b. Vigilance and diet selection: the classical diet model reconsidered.Journal of Theoretical Biology 132:127-143.Lima, S.L. 1990. Protective cover and the use of space: different strategies in finches.Oikos 58:151-158.Longland, W.S., and M.V. Price. 1991. Direct observations of owls and heteromyidrodents: can predation risk explain microhabitat use? Ecology 72:2261-2273.MacArthur, R.H. 1972. Geographical ecology. Princeton University Press, Princeton,New Jersey, USA.MacMillen, R.E. 1990. Water economy of granivorous birds: a predictive model.Condor 92:379-392.McNamara, J.M., and A.I. Houston. 1986. The common currency for behaviouraldecisions. American Naturalist 127:358-378.McNamara, J.M., and A.I. Houston. 1987. Starvation and predation as factors limitingpopulation size. Ecology 68:1515-1519.McNamara, J.M., and A.I. Houston. 1990. The value of fat reserves and the tradeoffbetween starvation and predation. Acta Biotheoretica 38:37-61.Mead, R. 1988. The design of experiments. Cambridge University Press, Cambridge,UK.Mercurio, K.S., A.R. Palmer, and R.B. Lowell. 1985. Predator-mediated microhabitatpartitioning by two species of visually cryptic, intertidal limpets. Ecology66:1417-1425.Milinski, M., and R. Heller. 1978. The influence of a predator on the optimal foragingbehavior of sticklebacks (Gasterosteus aculeatus L.). Nature 275:642-644.Mittelbach, G.G. 1984. Predation and resource partitioning in two sunfishes(Centrarchidae). Ecology 65:499-513.Nagy, K.A. 1987. Field metabolic rate and food requirement scaling in mammals andbirds. Ecological Monographs 57:111-128.Noon, B.R. 1981. The distribution of an avian guild along a temperate elevationalgradient: the importance and expression of competition. EcologicalMonographs 51:105-124.146Parmenter, R.R., J.A. MacMahon, and S.B. Vander Wall. 1984. The measurementof granivory by desert rodents, birds and ants: a comparison of an energeticsapproach and a seed-dish technique. Journal of Arid Environments 7:75-92.Platt, J.R. 1964. Strong inference. Science 146:347-353.Price, T.D. 1991. Morphology and ecology of breeding warblers along an altitudinalgradient in Kashmir, India. Journal of Animal Ecology 60:643-664.Pulliam, H.R. 1975. Coexistence of sparrows: a test of community theory. Science184:474-476.Pulliam, H.R. 1983. Ecological community theory and the coexistence of sparrows.Ecology 64:45-52.Pulliam, H.R. 1985. Foraging efficiency, resource partitioning, and the coexistence ofsparrow species. Ecology 66:1829-1836.Pulliam, H.R., and J.B. Dunning, Jr. 1987. The influence of food supply on localdensity and diversity of sparrows. Ecology 68:1009-1014.Pulliam, H.R., and G.S. Mills. 1977. The use of space by wintering sparrows. Ecology58:1393-1399.Pulliam, H.R., and T.A. Parker, III. 1979. Population regulation of sparrows.Forschritte der Zoologie 25:137-147.Rawlings, J.O. 1988. Applied regression analysis: a research tool. Wadsworth &Brooks/Cole Advanced Books & Software, Pacific Grove, California, USA.Repasky, R.R. 1991. Temperature and the northern distributions of wintering birds.Ecology 72:2274-2285.Roberts, S.C. 1988. Social influences on vigilance in rabbits. Animal Behaviour 36:905-913.SAS Institute. 1988. SAS/STAT user's guide. Release 6.03. SAS Institute, Cary, NorthCarolina, USA.Schluter, D. 1982. Distributions of Galapagos ground finches along an altitudinalgradient: the importance of food supply. Ecology 63:1504-1517.Schluter, D. 1988a. Character displacement and the adaptive divergence of finches onislands and continents. American Naturalist 131:799-824.Schluter, D. 1988b. The evolution of finch communities on islands and continents:Kenya vs. Galapagos. Ecological Monographs 58:229-249.Schluter, D., and P.R. Grant. 1982. The distribution of Geospiza difficilis on Galapagosislands: tests of three hypotheses. Evolution 36:1213-1226.Schluter, D., and R.R. Repasky. 1991. Worldwide limitation of finch densities by foodand other factors. Ecology 72:1763-1774.147Schmitt, R.J. 1987. Indirect interactions between prey: apparent competition, predatoraggregation, and habitat segregation. Ecology 68:1887-1897.Schoener, T.W. 1982. The controversy over interspecific competition. AmericanScientist 70:586-595.Schoener, T.W. 1983. Field experiments on interspecific competition. AmericanNaturalist 122:240-285.Simberloff, D. 1972. Properties of the rarefaction diversity measurement. AmericanNaturalist 106:414-418.Smiley, J. 1978. Plant chemistry and the evolution of host specificity: new evidencefrom Heliconias and Passiflora. Science 201:745-747.Smith, J.N.M., and H.P.A. Sweatman. 1974. Food-searching behavior of titmice inpatchy environments. Ecology 55:1216-1232.Snedecor, G.W., and W.G. Cochran. 1967. Statistical methods. Sixth Edition.University of Iowa Press, Ames, Iowa, USA.Stephens, D.W., and J.R. Krebs. 1986. Foraging theory. Princeton University Press,Princeton, New Jersey, USA.Strong, D.R. 1983. Natural variability and the manifold mechanism of ecologicalcommunities. American Naturalist 122:636-660.Svardson, G. 1949. Competition and habitat selection in birds. Oikos 1:157-174.Sweeney, B.W., and R.L. Vannote. 1978. Size variation and the distribution ofhemimetabolous aquatic insects: two thermal hypotheses. Science 200:444-446.Terborgh, J. 1971. Distribution on environmental gradients: theory and preliminaryinterpretation of distributional patterns in the avifauna of the CordilleraVilcabamba, Peru. Ecology 52:23-40.Terborgh, J. 1985. The role of ecotones in the distribution of Andean birds. Ecology66:1237-1246.Terborgh, J., and J.S. Weske. 1975. The role of competition in the distribution ofAndean birds. Ecology 56:562-576.Thompson, D.B.A., and D.W. Lendrem. 1985. Gulls and plovers: host vigilance,kleptoparasite success and a model of kleptoparasite detection. AnimalBehaviour 33:1318-1324.Todd, I.A., and R.J. Cowie. 1990. Measuring the risk of predation in an energycurrency: field experiments with foraging blue tits, Parus caeruleus. AnimalBehaviour 40:112-117.Valone, T.J., and S.L. Lima. 1987. Carrying food items to cover for consumption: thebehavior of ten bird species feeding under the risk of predation. Oecologia71:286-294.148Vannote, R.L., and B.W. Sweeney. 1980. Geographic analysis of thermal equilibria:a conceptual model for evaluating the effect of natural and modified thermalregimes on aquatic insect communities. American Naturalist 115:667-695.Waite, T.A. 1987. Vigilance in the white-breasted nuthatch: effects of dominance andsociality. Auk 104:429-434.Watts, B.D. 1990. Cover use and predator-related mortality in song and savannahsparrows. Auk 107:775-778.Watts, B.D. 1991. Effects of predation risk on distribution within and between habitatsin savannah sparrows. Ecology 72:1515-1519.Weathers, W.W. 1983. Birds of southern California's Deep Canyon. University ofCalifornia Press, Berkeley, California, USA.Weathers, W.W., and C. van Riper, III. 1982. Temperature regulation in twoendangered Hawaiian honeycreepers: the palila (Psittirostra bailleui) and theLaysan finch (Psittirostra cantans). Auk 99:667-674.Webster, M.D., and W.W. Weathers. 1990. Heat produced as by-product of foragingactivity contributes to thermoregulation by verdins, Auriparus flaviceps.Physiological Zoology 63:777-794.Werner, E.E., and J.F. Gilliam. 1984. The ontegenetic niche and species interactions insize-structured populations. Annual Review of Ecology and Systematics 15:393-425.Werner, E.E., J.F. Gilliam, D.J. Hall, and G.G. Mittelbach. 1983. An experimental testof the effects of predation risk on habitat use in fish. Ecology 64:1540-1548.Whittaker, R.H. 1951. A criticism of the plant association and climatic climaxconcepts. Northwest Science 25:17-31.Whittaker, R.H. 1956. Vegetation of the Great Smoky Mountains. EcologicalMonographs 26:1-80.Wiens, J.A. 1989a. The ecology of bird communities. Volume 1. Foundations andpatterns. Cambridge University Press, Cambridge, UK.Wiens, J.A. 1989b. The ecology of bird communities. Volume 2. Processes andvariations. Cambridge University Press, Cambridge, UK.Wiens, J.A., and J.T. Rotenberry. 1981. Habitat associations and community structureof birds in shrubsteppe environments. Ecological Monographs 51:21-41.Wiens, J.A., J.T. Rotenberry, and B. VanHorne. 1986. A lesson in the limitations offield experiments: shrubsteppe birds and habitat alterations. Ecology 67:365-376.Wolcott, T.G. 1973. Physiological ecology and intertidal zonation in limpets (Acmaea):A critical look at "limiting factors". Biological Bulletin 145:389-422.149Ydenberg, R.C. 1984. Great tits and giving-up times: decision rules for leaving patches.Behaviour 90:1-24.Zabriskie, J.G. 1979. Plants of Deep Canyon and the central Coachella Valley,California. Phillip L. Boyd Deep Canyon Desert Research Center, Palm Desert,California, USA.


Citation Scheme:


Citations by CSL (citeproc-js)

Usage Statistics



Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            async >
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:


Related Items