LATITUDINAL GRADIENTS IN LARVAL FEEDING SPECIALIZATION OF THE WORLD PAPILIONIDAE (LEPIDOPTERA)*

A major question arising ro.m ecological theory is whether or not there are relatively more specialized species in the lower’ or tropical latitudes than in the higher or more temperate latitudes. There exists a great deal o. speculation concerning this question, but little actual data. This article presents an analysis o such data or one family of Lepidoptera; the Papilionidae. The subject o.t niche breadth and degree o.t competitive overlap has generated tremendous interest in the field of ecology in the last two decades. In terms ot the spatial model ot the niche first pro.posed by Hutchinso.n (I957) and later expanded by Slobodkin (I962), Levins (I968), and MacArthur (I968), niche breadth is the "distance through" a niche hypervolume along a particular resource gradient within a community (Colwell and Futuyma, I97 Whittaker, et al., I973). Although there is presumed to be an ,apparent limit to the similarity and hence the number o.f competing species that can coexist (MacArthur and Levins, 967), the packing of species (MacArthur, 969a) appears to be roughly proportional to the environmental variance, with less overlap, or looser packing, occurring with creased environmental fluctuation or perturbation. May and Yia.cArthur (972) imply that niche overlap may be relatively independent of small to moderate environmental fluctuations, but it remains true that greater packing could still be attained by specialization, and the question concerning what ultimately limits niche width remains unsettled. Also., in a mo.re stable environment ot the tropical latitudes, with larger productivity one might expect that marginally scarce resources would become adequate to. support new .species at the extremes o.f the resource gradient (Odum, et al., I960; C.onnell and Orias, 964). In a more. stable low-latitude tropical environment (MacArthur, I969a), old lakes (Kozhov, I963), the


INTRODUCTION
A major question arising ro.m ecological theory is whether or not there are relatively more specialized species in the lower' or tropical latitudes than in the higher or more temperate latitudes. There exists a great deal o. speculation concerning this question, but little actual data. This article presents an analysis o such data or one family of Lepidoptera; the Papilionidae.
The subject o.t niche breadth and degree o.t competitive overlap has generated tremendous interest in the field of ecology in the last two decades. In terms ot the spatial model ot the niche first pro.posed by Hutchinso.n (I957) and later expanded by Slobodkin (I962), Levins (I968), and MacArthur (I968), niche breadth is the "distance through" a niche hypervolume along a particular resource gradient within a community (Colwell and Futuyma, I97 Whittaker,et al.,I973).
Although there is presumed to be an ,apparent limit to the similarity and hence the number o.f competing species that can coexist (MacArthur and Levins, 967), the packing of species (MacArthur, 969a) appears to be roughly proportional to the environmental variance, with less overlap, or looser packing, occurring with creased environmental fluctuation or perturbation. May and Yia.c-Psyche [December benthos (Sanders,968,969) or coral reefs (Kohn,968) we expect closer packing and greater numbers of competitors per community (Levins,968 MacArt'hur and Levins,967 Vandermeer, I97o) with narrower realized niches and increased ecological differentiation and niche specificity (Whittaker,et al.,954;Klopfer,959 Selander, 965 Millar, 1969).
Although species diversity varies from .one community to. the next, it is often greatest in the lower latitudes.. Sanders (969) incorporates many of the earlier proposed explanations of differences in species diversity (Pianka,966;Paine,966;Connell and Orias,964 MacArthur,965,I969b) in his stability-time hypothesis.
According to this hypothesis, the greater number of species in the tropics implies a greater number of biotic interacti.ons than in the temperate regions. The results of competition would be most obvious in the form of niche divergence and increased specializations over a resource gradient. The corresponding decrease in predictability and stability of environmental c.onditions with an increase in latitude may mean that species in temperate regions are influenced more by the physical environment than by competitive interactions (Slobodkin and Sanders,969 Janzen,I97O). Realized niches may be smaller in the tropical or low latitudes than in higher latitudes, because of the expected difficulties of specializing in an unpredictable' environment and of generalizing in an extremely diverse and competitive environment. The following investigati.on was undertaken in an attempt to determine the correctness of the premise, that realized feeding niches .are relatively more specialized in the tropical and subtropical latitudes than in the higher latitudes.

METHODS
My approach was to choose a single taxonomic group that is cosmopolitan, whose habits are relatively well known and which .all nicely fall into a single 'guild' (Root,I967). The 538 species of Papilionidae (Lepidoptera) of the world, based on Munroe's classification, met these .criteria adequately.
In order to analyze a niche breadth it was necessary to choose a criterion that could be compared over a range of latitudes, and not one that varied between .communities. Therfore, I investigated larval host plant relationships arbitrarily designating those, species feeding on more than one taxo.nomic family of plants as "generalists," or  Because larva.e generally remain upon the host plant throughout most o their existence, the ood-plant is especially likely to be the key actor in th.e niche o most butterfly species since, to a large extent, it is also the shelter, substrate, or habitat upon which allelochemical co-evolution (Ehrlich and Raven, 1965;Whittaker and Feeny, 1971) o the entire 'component community' (Root, 1973) takes place. In this sense it is essentially difficult to differentiate between 'niche' and 'habitat' (Whittaker,Levin,and Root,I973), and their term 'ecotope' might be more appropriate since it entails the inter-community as well as the intra-community variables. I have compiled range maps o each o the 538 species o Papilionidae o the world (Munroe,196o) using the techniques o Slansky (I972). I used Goode's Atlas (Espenshade, I96O) and locality data r.ecorded in the literature, primarily in Rothschild (I895), Rothschild and Jordan (I9O6), Aurivillts (I9O8-I9IO), Bollow (929), Fruhstorfer (9o8), Jordan (9o7, 9o8-x9o9), von Rosen (929), Seitz (9o6), Stichel 9o6), Shirozu 96o), D'Almeida (I966), and Common and Waterhouse (I972).
The number of species known to occur in each ten degree latitude belt was recorded. The worldwide data were divided into the three geographical sections in figures , 2, and 3 to simplify the analysis..
The components of each of t'hese figures are shown in tabular form (Table ).
The species of known larval food-plants were compiled from the literature, although care was used in interpreting recorded host plant records (see Ehrlich and Raven,965;and Shields,Emmel,and Breedlove, 969). The plant classification follows Willis (I973).
A supplementary table with each species of Papilionidae with its latitudinal range, food-plant families fed upon, and citations for these food plant records is a.vailable from the author and at the following libraries: Cornell University (Entomology), Harvard (M.C.Z.), and the U.S.N.M.
With those species eeding on more than one taxonomic family of plants as my "generalists" or wide-niche species, a species was. counted in all parts of its range as being a wide-niche species, or generalist, if it eeds on more than one plant family .anywhere in its range.
Latitudinal gradients in niche breadth were prepared by enumerating the species for each ten degree belt of latitude, and calculating the percentage o.f species exhibiting a "general" feeding strategy.
Although the ;oodplants of the Papilionidae are probably as well known as those of any other taxon of similar magnitude', there are still gap,s in the record. In the case. of a species .o.f unknown o.odplant preference I have assumed that the host plant family range utilized is, in general, similar to those o. closely related species of the group. There are, for example, many tropical Troidini for which the precise species of .dristolochia utiliz.ed are not known.
Longitudinal and altitudinal variations in species richness are' an important and interesting part o, distributional patterns, but were: not considered in this study.  Figure 4 s'how:s the overall pattern of latitudinal gradients in the number of species of Papilionidae, corrected for distributional overlap of certain species that occurred in more than one of the earlier three figures. Also the number of species for each tribe of Papilionidae feeding on two or more plant families and the overall percentage of generalized feeders for each latitudinal belt are, shown in relation to this pattern. Note that there is a considerable decrease in the percentage of generalists toward the lower tropical latitudes ( Figure 5).

DISCUSSION
Air, hough the absolute number of generalists is not strikingly different, it is readily apparent that the percentages are considerably less in the lower latitudes. This fact appears to support the premise that there are relatively fewer generalists in the tropics and relatively more in the higher latitudes. There is, however, no sharp division, but instead a gradient of increasing relative degree of specialization from the extremes of latitude to the equatorial regions. It is not known whether increased competition could be the cause or the result of fhe increasing diversity gradients toward the lower latitudes, nor is it certain whether the observed gradients in feeding niche breadth are due to ecological release from competition in higher latitudes or just the inability to specialize in their unpredictable and unstable environmental conditions. It was assumed throughout this study that the niche breadth as measured here is that of the 'realized' ni.che as opposed to the 'fundamental' (Odum,97x). This study of feeding niche breadths involves a cross section of both 'evolutionary' and 'ecological' time (Slobodkin,x962) and the differentiation at both intraand interpopulation levels. That is, there are .probably areas of active, competition where w.e are not viewing only the neat results of evolutionary processes that have already and permanently narrowed the realized hostplant ranges from those plant families which are. physiologically exploitable, but there are also the currently unfolding consequences of ecological interactions that someday may or may not result in more specialized herbivores. It should be noted that the relative merits of either specializing or generalizing may involve 'qualitative' and 'quantitative' chemical defense of the host plants (Feeny, 974) as well as other ecological factors besides competition alone.
Emmel and Emmel (969) have shown that fo,r Patilio indra (Reakirt) and Papilo rudkini (Comstock) direct competition of sympatric populations is avoided most of the time by temporal iso,lation and utilization of mutually exclusive food-plant families. In one instance however they found lhat food did become the limiting resource and the ,separate host preferences broke down when the favored food became scarce. In this situation both Umbelliferae and Rutaceae were then utilized by the, insects of these populations. that permit the coexistence of several species of butterflies should probably be sought in the larval rather than the adult stages has been obtained by Young (972). The critical dimension of the niche (or ecotope) may in fact be something other than the range o.f larval hostplants utilized; however, as well as being one that is relatively easy to measure, it is assumed here to be a crucial parameter in most cases.
In .a relatively recently speciated group of swallowtails in the U.S., the polyphagous Papilio glaucus (L.) is apparently entirely excluded geographicall.y by three species which are. s.ympatric and vhich have subdivided the range of plant families utilized by P. glaucus (Brower,I958a;959). Besides their specialized eeding habits, isolating mechanisms .such as seasonality, coloration, and mating preferences appear to have been involved in the Pleistocene speciation. There is no clear cut distinction between .ecological and evolutionary time, and the importance of trophic specialization is uncertain 19:r3] Scriber PaIilionidae 363 in this and similar cases such as the i:our species ot the P. demodocus (Esper) group ot swallowtails that have evolved on Madagascar, or the three species of the P. nireus (L.) group in West Arica that are presumed to have speciated during the Tertiary when the climate and vegetation were rapidly changing (Owen,I97 ). Furthermore, because the author is in this study to a large degree uncertain o.t l"egional food plant preferences that might distinguish a polymorphic population of specialized individuals from a monomorphic population of generalized individuals, no attempt has been made to differentiate any 'components o. niche breadth as described by Roughgarden 972). It is probably fair to assume that neither the latitudinal gradients in diversity nor those o.r trophic niche breadth remain constant for any long periods of time. The phylogenetic histories o the Papilionidae are certainly intertwined with the zoogeographical movement and changes in the flora of major continents (Hovanitz,958;Carcasson,964;and Kostrowicki,969), and this paper is merely .an attempt to look at the present state of these continually changing phylogenies.
Besides regional preferences there ,can be tempo,ral changes in foodplant preferences and t,hese are requently accentuated by human agriculture. Owen (97) points out that several species o Arican papilionids have recently expanded their ranges and seasonal abundance by switching from their wild Rutaceous oodplants onto cultivated varieties of Citrus which, unlike the wild plants, maintain their leaves throughout the dry season and .are spreading rapidly with agricultural development. The extent to which members of t:he Papilionidae have become secondarily polyphagous on .completely new plant families or else by revitalized usage upon re-exposure to. ancestral hostplant families is uncertain. It is very likely that besides the general ecological pressures such as competition with more specialized herbivol"es, there may well be both 'fixed' and 'variable' metabolic 'cost' associa.ted with any extension o.f the normal range o,f ood plants utilized by a species (Feeny,974  land (Australia), the larvae are unsuccessful and eventually die when left to feed upon leaves of this relatively recent introduction (Edwards, 1948;. It appears thus that in at least some populations metabolic specializations or adjustments may have coevolved along with the regional restriction in food-plant choice such that Citrus is perhaps in some way no. longer physiologically suitable for those individuals. It should be pointed out that individuals from some Australian populations can and do successfully utilize the native rutaceous Microcitrus australis (D'abrera, I971 ), and those in New South Wales can be reared successfully upon Citrus (see Straatman,I96:z).
it is interesting that in Africa Papilio demodocus (Esper), a species which is very closely related to and once considered a race o P. demoleus, is primarily a Rutaceae-feeder over most of its range.
It: is however .polymorphic (Clarke,et al. I963).and also utilizes the Umbelliferae in South Africa where the richness of this plant family is relatively high (Dethier, 1941). Similarly, the P. machaon group of Umbellifer feeding species in Sardinia, East Asia, and North America are also close to P. demoleus in ancestry (Munroe,196o).
Further speciation has apparentl.y taken place in those regions where Umbelliterae are most diverse. Several ot fhese species, however, are still able to utilize rutaceous oodplants..
Feeding patterns ot the Papilionidae and phytophagous insects in general, range ro.m strict monophagy (stenophagy), in which a single species o( (oodplant is utilized, to wide polyphagy (euryphagy) in which many species, genera, or tamilies may be utilized (Brues,92o; Dethier, 954). The exploitation o a particular hostplant involves many adjustments on the part o the insect to. the plant's microcommunity. This includes host-specific predators (Brower,I958b) and parasites (Read et al.,97o). Facultatlve poiyphagy in some Papilio larvae, conveying the ability to take advantage ot ovipositional mistakes ot the adult, may allow some escape rom the intense mortality rates that can result trom host-specific parasitism (Stride and Straatman,962). Polymorphisms in larval color patterns can also be involved (Clarke et al.,I963), but perhaps, the most important ]actor in this coevolving community is the metabolic adjustments on the part ot the insect to the secondary chemistry ot the plant (Fraenkel,959 Jermy, I966; and Thorsteinson, I96O). As pointed out by Ehrlich and Raven (965), the plasticity o t'he chemoreceptive response and the potential tor physi.ological adjustments to secondary plant chemicals may be very important tactors in determining the potential tor evolutionary radiation in a phytophagous species. Since the possibilit.y exists that taxonomically generalized teeders may in act be allelochemical specialists that cue in on one key set ot chemicals in the plants, eeding niche breadth may well take on new meaning.
Increased specialization by a species is generally presumed to increaser the competitive ability o a species largely because the species utilizes its resources more efficiently (Emlin,I966;MacArthur and Levins,964;MacArthur and Pianka, Willson,969). I have employed a somewhat different measure here, and by altering the number o amilies or genera used as a criterion or 'generalist', the amplitude ot: the observed latitudinal gradient would vary but probably not the overall pattern. Other studies have described a 'generalist' in somewhat different terms .as those species able to maintain populations over a broad range o1: environmental types or substrates (McNaughton and Wol1:, 972), or those exhibiting a greater variance or breadth o1: morphology or eeding behavior (Levins, z968; Roughgarten, 972; Schoener, 97). These also may or may not reflect actual niche breadth as such, but latitudinal comparisons within a taxocene would probably be possible, to prepare in these terms as well.
It is interesting that many species o1: Papilionidae are relatively specialized on certain tropical and sub-tropical plant 1:amilies (Slan-sky, 972). In the higher latitudes where fewer species of most major amilies of host plants occur there has probably been some breakdown of the phylogenetic larval host plant preferences that tie some species to these major evolutionary.host plant gr.oups, perhaps avoring polyphagy at the plant amily level. The Patilio troilus L. and P. glaucus L. groups, for instance, have expanded their diets considerably (rom the Lauraceae and Magnoliaceae they are thought to have originally ed upon. Thus., besides 'ecological release' and the inability to specialize in unpredictable and unstable environments as possible explanations o.i: the observed gradients in feeding specialization, a third possibility exists. Perhaps there was a. necessary alteration ot: those allelochemical specializations deve.loped over a long period of evolutionary time, either as species of Papilionidae extended their ranges into the higher latitudes where their usual host plant families never existed, or as they remained in higher latitudes where their ancestral hosts had since been eliminated, perhaps by climatic changes.
The precise reason for Papilionidae being more specialized in their eeding habits in tropical latitudes is uncertain and will probably remain largely as such. One .contributing factor or this is. that no one knows whether polyphagy in general is the. ancestral condition from which more specialized feeding habits arose or not. The act exists, however, that there are greater percentages of eeding specialists in the tropical latitudes, and that this percentage decreases, as one moves toward the higher latitudes. These data do not suggest that species are more specialized in the tropical than temperate latitudes, but that more species are specialized.

SUMMARY
Latitudinal gradients in ood plant specialization are examined for the swallowtail butterflies of the world (Papilionidae). It is found that the absolute number o. generalized feeders existing in each.ten degree belt of latitude is fairly constant, but the relative, percentage of generalized species is significantly higher in the higher latitudes. This fact is discussed in terms of trophic niche breadth and species diversity of the Papilionidae in tropical and temperate regions.