Birds in the genus Pitohui carry the potent neurotoxin homobatrachotoxin in their skin and feathers. In this study, I tested whether homobatrachotoxin can repel or kill chewing lice (order Phthiraptera). When individual feather lice were offered a choice of two feathers on which to feed or take shelter, the lice preferred nontoxic feathers to feathers of the most toxic pitohui species, Pitohui dichrous. Moreover, the presence of toxic P. dichrous feathers significantly shortened the life span of captive feather lice. These results suggest that homobatrachotoxin repels and kills lice and may thus protect pitohuis against lice infestation. Received 10 April 1998, accepted 8 February 1999.

Department of Ecology and Evolution, The University of Chicago, 1101 East 57th Street, Chicago, Illinois 60637, USA SPECIES IN THE AVIAN GENUS PITOHUI carry a potent alkaloid neurotoxin in their skin and feathers (Dumbacher et al. 1992). Pitohui toxin, known as homobatrachotoxin (homoBTX), is a member of a well-known family of steroidal al- kaloids that depolarize nerve and muscle mem- branes by binding and activating voltage-de- pendent sodium channels (Albuquerque et al. 1971). In some cases, concentrations of toxin are sufficiently high that merely handling an individual Hooded Pitohui (Pitohui dichrous) can irritate buccal membranes and can cause sneezing and burning, watery eyes (Salvadori 1881, Majnep and Bulmer 1977, Dumbacher et al. 1992). Anthropological evidence suggests that the toxin defends pitohuis from human hunters (Majnep and Bulmer 1977; Kocher- Schmid 1991, 1993), and other workers have speculated that it also defends pitohuis against natural predators (Diamond 1992) and arthro- pod ectoparasites (Mouritsen and Madsen 1994, Poulsen 1994). Nevertheless, no studies have directly investigated if the enemies of pi- tohuis are deterred by homoBTX or how homoBTX deters them. Arthropod ectoparasites are natural enemies of pitohuis and potential targets for pitohui de- fensive chemicals. HomoBTX has been shown to affect a wide range of vertebrates and inver- tebrates (Albuquerque et al. 1971, Daly and Present address: Molecular Genetics Laboratory, National Zoological Park, Smithsonian Institution, 3001 Connecticut Avenue NW, Washington, D.C. 20008, USA. E-mail: jdumbacher@nzp. si.edu Spande 1986, Dwivedy 1988). With the notable exceptions of pitohuis and Phyllobates frogs, nearly every animal that contains voltage-de- pendent sodium channels is poisoned by ba- trachotoxins, including distantly related ar- thropods. Also, bird lice can influence host fit- ness in several ways. Lice can affect the ener- getics and survival of hosts (Clayton 1990, Booth et al. 1993, Brown et al. 1995), reduce egg numbers and hatching rates (Derylo 1974, DeVaney 1976), reduce mating success (Ham- ilton and Zuk 1982, Clayton 1990, Loye and Zuk 1991, Clayton and Tompkins 1995), and transmit pathogens (Marshall 1981, Clayton 1990). Therefore, defense against lice might be under selection. Finally, a high proportion of a pitohui's total toxin is concentrated in the skin and feathers. Because lice live and feed on feathers, skin, and subdermal blood supplies, pitohui toxins could constitute a formidable barrier to these ectoparasites. HomoBTXs, and toxins in general, could de- fend birds against lice through several alter- native mechanisms. Toxins could (1) reduce louse fecundity, (2) reduce louse survival, (3) reduce the influence of lice on host fitness (e.g. by delaying maturation, lengthening the life cy- cle, or suppressing appetite), and (4) favorably effect louse transmission rates by reducing im- migration or inducing emigration. Here, I re- port experimental studies that examine wheth- er feather lice exhibit an active choice against naturally toxic pitohui feathers and examine whether the presence of natural levels of TABLE 1. Sources and identification of chewing lice used in petri dish experiments. [Auk, Vol. 116 Avian host species Parasite identification No. parasites Ailuroedus buccoides Alisterus chloropterus Chalcophaps indica Chalcophaps stephani Chlamydera cerviniventris Colluricincla megarhyncha Cuculus saturatus Gallicolumba jobiensis Halcyon torotoro Macropygia amboinensis Melilestesmegarhynchus Pitohui cristatus Pitohuiferrugineus Ptilinopusmagnificus Ptilinopus pulchellus Ptilinopus superbus Tanysiptera danae Brueelia pallida (Piaget, 1880) Neopsittaconirmus circumfasciatus (Piaget, 1880) Columbicola guimaraesi Tendeiro, 1965 Coloceras piriformis (Tendeiro, 1969) Coloceras indicum Tendeiro, 1973 Columbicola sp. Coloceras stephanii (Tendeiro, 1987) Coloceras sp. Myrsidea sp. Philopterus sp. Brueelia sp. Cuculicola latirostris (Burmeister, 1838) Columbicola exilicornis (Piaget, 1880) Alcedoecus sp. Columbicola sp. Coloceras doreyanus (Eichler, 1950) Brueelia sp. Myrsidea sp. Myrsidea sp. or Philopterus sp. Columbicola harrisoni Tendeiro, 1965 Auricotes obscurus Keler, 1939 Hohorstiella sp. Columbicola emersoni Tendeiro, 1960 Columbicola emersoni Tendeiro, 1960 Auricotes pazmartinae Tendeiro, 1976 Myrsidea sp. 1 48 6O 74 5 55 36 2O 7 1 6 1 4O 1 3 18 1 1 7 23 5 1 1 1 5 1 homoBTX affects the captive life span of these lice. Because toxin levels vary across pitohui spe- cies, feathers from different pitohuis provide naturally high and low toxicity treatments, and outgroups provide nontoxic control feathers. Individuals from five pitohui species are known to contain some level of homoBTX (Dumbacher 1997), with P. dichrous containing the highest concentrations. In some popula- tions, P. dichrous feathers contain more than 50 ixg of homoBTX per g of tissue, which is more than 15 times the concentration originally re- ported. Rusty Pitohui (P. ferrugineus) feathers have much lower levels, and Crested Pitohui (P. cristatus) feathers have nearly undetectable lev- els of toxin. METHODS Studies were conducted at the Biological Research Station, Varirata National Park (927'S, 147 21 'E; 840 m elevation), a 400-ha reserve on the Sogeri escarp- ment approximately 40 km east of Port Moresby, southeastern Papua New Guinea. Birds were trapped in mist nets, measured (wing, tarsus, head and bill length, and body mass), banded, visually inspected for parasite loads, and released. Lice were removed from host birds using fumi- gation jars described by Bear (1995) and equipped so industrial-grade CO2 could be pumped constantly into the jar. Feathers were blown and ruffled to de- tach anesthetized lice. The lice revived within 15-30 s of exposure to normal air and then were collected in petri dishes where they were held until they could be placed into experimental arenas, usually within the hour. In the field, lice from each individual host were classified into "types" based on body plan and body size. A type was defined as a group of morphologi- cally similar lice taken from the same host. These louse types were later identified, and they generally corresponded to a particular species and age class (adult or immature) within that species. It was likely that each type within a particular host shared some degree of genetic and/or environmental similarity and thus, louse type provided a natural randomized block for statistical analyses. Table 1 lists all of the avian hosts that provided experimental lice and the identification of these lice to genus or species. Experiments were performed in plastic petri dish- es that contained one or two feathers and one louse. Dishes were kept at 18 to 22C and ambient humidity (>45%) in a darkened room. In general, lice had dif- ficulty walking across the dishes, which prevented them from escaping. In dishes with more than one feather, the second feather was placed above the first, allowing lice to move freely from feather to feather and thereby choose from which feather to feed or rest. All feathers were clean contour feathers in good condition taken from the dorsum just behind the wings. All non-pitohui feathers were presumed to be nontoxic. After returning from the field, toxin con- centrations in pitohui feathers were measured in the Laboratory of Bio-organic Chemistry at the National Institutes of Health using radio-ligand binding as- says. Pitohui dichrous feathers used in this study con- tained an average of 39 p,g of homoBTX per gram of tissue, P. ferrugineus feathers contained an average of 3.5 g of homoBTX per gram of tissue, whereas P. cristatus feathers contained less than 3 g homoBTX per gram of tissue. Feathers collected from a variety of non-pitohui species were used as nontoxic con- trols. Two types of experiments were conducted. The first set of experiments, called choice experiments, tested whether lice showed a preference against toxic pitohui feathers when given a choice of two feathers on which to feed and find shelter. The second set of experiments, called life-span experiments, tested whether pitohui toxins affected the life span of cap- tive lice. Choice experiments.--In the choice experiments, feathers from two bird species were placed into petri dishes, one above the other. Each petri dish con- tained either one pitohui feather and one non-pito- hui feather, or one P. dichrous feather and one less- toxic pitohui feather (P. cristatus), matched for size and general shape. Nontoxic control feathers were used from six species: Accipiter poliocephalus (3 trials), Ailuroedus buccoides (25), Chlamydera cerviniventris (8), Colluricincla megarhyncha (49), Dicrurus hottentot- tus (5), and Meliphaga analoga (15). After a feather was used in an experiment, it was stored in a sealed plas- tic bag. Because only a limited number of feathers was available, some P. dichrous feathers were used in a second experiment. Of the 105 trials of P. dichrous feathers versus non-pitohui feathers, 23 (22%) incor- porated previously used pitohui feathers. Reused feathers were always placed in fresh petri dishes and paired with a random unused feather from another species. For the louse to choose between feathers, the de- sign required that one feather had to be placed on top of the other and that each louse had to be placed on a feather to begin the experiment. Because lice may show an overall tendency not to move, or may prefer the bottom feather (possibly due to negative phototaxis; Marshall 1981), combinations of treat- ment effects (which feather the louse was placed on and which feather was on top) were distributed evenly across petri dishes. For each petri dish, feath- ers were randomly chosen from the two species be- ing compared. Lice were randomly distributed to pe- tri dishes and placed on the feather designated by treatment design. Feathers were arranged one above the other such that lice had easy access to both feath- ers. The trial began and the time was recorded when the louse was placed on one of the two feathers. Ev- ery 60 to 90 min between 0600 to 2400 Australian EST, I recorded whether the louse was alive and which feather it was on. If a louse was found alive on the dish but not on a feather, it was returned to the feather on which it had originally been placed. Each louse was monitored until it died. All lice from these experiments were preserved and stored at the Bishop Museum in Honolulu, Hawaii. An experiment was completed when the louse was found dead in the experimental dish. All observa- tions were tallied and a choice was recorded for the feather on which the louse was most frequently ob- served. In the event of a tie or if the louse was found dead in the first observation period, the dish was ex- cluded from subsequent analysis. Choices were tab- ulated in contingency tables and blocked for the two design effects: (1) which feather the louse was placed on, and (2) which feather was on top. Choice exper- iments were conducted during November and De- cember 1995. Life-span experiments.--Only feathers from P. di- chrous, P. ferrugineus, and Colluricincla megarhyncha were used. Colluricincla, the putative sister genus to Pitohui, was chosen as a control to maximize the structural similarity between Pitohui and control feathers. The life-span experiments used the same protocols as the choice experiments. There were three treatments (1) feathers of P. dichrous and C. me- garhyncha, (2) feathers of P. ferrugineus and C. mega- rhyncha, and (3) a feather of C. megarhyncha alone. Lice in treatment 1 were therefore exposed to high natural levels of homoBTX, those in treatment 2 to low natural levels of homoBTX, and those in treat- ment 3 acted as controls that were not exposed to homoBTX. Data were analyzed using analysis of var- iance with life span as the dependent variable and treatment as the independent variable, blocked by louse type. Life-span experiments were conducted February and March 1996. Table 2 shows the proportions of the three most common louse types used in each experiment. The proportion of different louse genera used varied be- tween experiments. RESULTS Effects of toxicity on feather choice.--I per- formed 225 choice trials. Eleven trials ended in ties and were excluded from subsequent anal- yses. Lice showed a statistically significant preference for the bottom feather (G-test with Williams' correction, G = 15.52, df = 1, P < 0.01). Data were subsequently split into two TABLE 2. Total number and proportion of louse genera used in choice and life-span experiments. Genus Total no. used % of total Choice experiments (214 lice total) Coloceras 116 54.2 Columbicola 79 36.9 Myrsidea 8 3.7 Other 11 5.1 Life-span experiments (208 lice total) Columbicola 103 49.5 Neopsittacornirmus 48 23.1 Coloceras 35 16.8 Other 22 10.6 blocks: petri dishes in which the louse was placed on the top feather and dishes in which the louse was placed on the bottom feather Within each block, data were compiled into 2 x 2 contingency tables tallying the starting feath- er (toxic or nontoxic) versus the choice (stay on starting feather or move), as shown in Table 3. The two blocks were analyzed jointly using the Mantel-Haenszel one-way X 2 statistic for block- ing tables with one degree of freedom (Mantel and Haenszel 1959, Mantel 1963). Four different treatment combinations were tested (Table 4). Lice avoided P. dichrous feath- ers in favor of nontoxic control feathers (Man- tel-Haenszel statistic = 10.17, df = 1, n = 105 trials, P < 0.0015; Table 4). Even excluding tri- als with previously used P. dichrous feathers, lice avoided P. dichrous feathers in favor of non- toxic feathers (Mantel-Haenszel statistic = 5.84, df = 1, n = 82 trials, P < 0.015), although the level of statistical significance dropped in ac- cordance with losing 22% of the experimental sample. Effects of toxicity on louse life span.--Life span in captivity varied among different types of lice (F = 3.17, df = 28 and 121, P < 0.0001). To avoid confounding effects of treatment with effects of louse type, I used only louse types that were ex- TABLE 3. Contingency table showing the marginal frequency distribution of all choice experiment tri- als, Lice are more likely to move off of a more toxic feather than a less toxic feather. No. of lice No. of lice moving to staying on other Feather on which original feather parasite was placed feather or to dish More-toxic feather 24 82 Less-toxic feather 46 62 posed to all three treatments and performed ANOVA using Type III sums of squares (Shaw and Mitchell-Olds 1993). Life-span data for 29 different louse types were analyzed for a total of 208 trials. Feather treatment (i.e. toxicity) sig- nificantly affected the life span of captive lice (F = 5.30, df = 2 and 56, P = 0.012). Estimates of mean life span for each treat- ment are presented in Table 5. Lice exposed to feathers of P. dichrous or P. ferrugineus died sig- nificantly earlier than those living on feathers of C. megarhyncha (a posteriori Bonferroni t-tests, c = 0.05). Life span did not differ significantly between lice exposed to feathers of P. dichrous versus P. ferrugineus (P > 0.05). DISCUSSION In the choice experiments, lice showed a sta- tistically significant preference against feeding and resting on P. dichrous feathers. The ecolog- ical significance of this choice is unclear be- cause lice rarely have an opportunity to move between hosts except during host mating and nesting periods. For many bird species, mating occurs briefly and lice have little opportunity to discriminate between hosts on the basis of toxicity. Recent evidence suggests that the most toxic pitohui species, P. dichrous, breeds coop- eratively (Legge and Heinsohn 1996), so during TABLE 4. Results of choice experiments. Four different treatment combinations were tested. Sample sizes for each test are given, contingency tables were blocked for effects of top and bottom feathers, and the probability of Mantel-Haenszel test statistics are presented. Toxic feather Less-toxic feather (toxicity) (toxicity) n P Pitohui dichrous (high) Nontoxic control 105 <0.001 Pitohui dichrous (high) Pitohui cristatus (low) 36 0.373 Pitohuiferrugineus (low) Colluricincla megarhyncha (nontoxic) 23 0.771 Pitohui cristatus (low) Colluricincla megarhyncha (nontoxic) 50 0.354 TABLE 5. Mean life span of lice in presence of feath- ers of different toxicity. Mean life Most toxic feather span in petri dish (h) n Pitohui dichrous 35.0 74 Pitohui ferrugineus 41.4 67 Colluricincla megarhyncha 81.5 67 nesting, lice may have a choice of multiple adult and nestling hosts. Therefore, if lice can move to less-toxic individuals in the nest en- vironment, birds without toxins may bear an unusually high lice load, and selection for an- tiparasite toxins could occur. During nonbreeding seasons of avian hosts, lice may have little or no opportunity to move between hosts. Lice may still be repelled by homoBTX and drop off the host because many lice chose the dish over the toxic feather. A sec- ond possibility is that lice may be transmitted via phoresis, that is, attached to another more- mobile parasite such as a hippoboscid fly. Al- though phoresis is believed to be uncommon (Marshall 1981), it may facilitate transmission away from toxic hosts because hippoboscid flies were common on pitohuis (Dumbacher 1997). Alternatively, lice may migrate to areas of the bird's body that contain lower concentra- tions of toxin. If lice are repelled from toxic ar- eas on an individual host, selection may favor a toxin distribution that protects regions of the body that are the most likely to be parasitized, or that pose the highest threat to host fitness. HomoBTX profoundly increased louse mor- tality in the life-span experiments. Lice placed in dishes with P. dichrous feathers died sooner than those in dishes with C. megarhyncha feath- ers, which have undetectable levels of toxin. The life span of lice exposed to P. ferrugineus feathers decreased significantly compared with those exposed to C. megarhyncha feathers, even though toxin levels in P. ferrugineus feathers are about 10 times lower than those in P. dichrous feathers. Thus, toxin concentrations in P. ferru- gineus and P. cristatus may be too low to repel lice but still may increase louse mortality. Reducing the life span of lice can have three profound consequences for host-parasite inter- actions. First, any given louse will feed for a re- duced period of time, reducing its individual effect on its host. Second, on average, the entire population of lice will be smaller at any given time. Smaller populations may irritate the host less, or may be less visible to a host's potential mates. Third, if the life span of lice is reduced enough, the probability of survival to mating decreases, and subsequent generations of lice will be reduced in number In addition, lice that lived the longest ap- peared to be feeding on feathers of C. mega- rhyncha. Black powder, assumed to be fecal pel- lets or discarded feather bits, accumulated on the petri dishes beneath C. megarhyncha feath- ers. After 4 to 5 days, these lice had damaged a noticeable fraction of the feather. In dishes with P. dichrous feathers, however, lice rarely showed evidence of eating either of the feath- ers. In dishes with P. dichrous feathers, lice also became immobile and inactive. This may allow pitohuis to remove lice more easily during preening or even while flying. Given that many lice were sluggish and did not feed on pitohui feathers, pitohui toxin may also lower feeding rates and thus reduce each louse's effect. These hypotheses deserve future testing. Life span varied significantly among louse types (F = 3.17, df = 28 and 121, P < 0.0001). For the extreme examples, Brueelia lived an av- erage 145 h in captivity, whereas Coloceras lived an average of 25 h. This difference may be due to many attributes of the different louse types, including overall differences in life span, feed- ing needs, previous condition, or the response to stresses of captivity. Seven lice collected from a single Pitohui ferrugineus host lived on average 194 h, which is the longest average cap- tive life span of any louse type in these exper- iments. This may be due to homoBTX insensi- tivity because these lice have coevolved with toxic pitohuis. However, even these lice showed a marked difference in life span across treat- ments; those on P. dichrous feathers lived an av- erage of 36.6 h (n = 3), those on P. ferrugineus feathers lived an average of 129.8 h (n = 2), and those on C. megarhyncha feathers lived an av- erage of 417 h (n = 2). Also, the ANOVA re- vealed a significant interaction of treatment by louse type (F = 1.63, df = 56 and 121, P = 0.013), which suggests that different types of lice react differently to the effects of homoBTX. For example, Neopsittaconirmus circumfasciatus showed no overall difference in life span due to feather toxicity, whereas Brueelia showed pro- found differences. Although I collected few lice from pitohuis for this study, infestation rates in pitohuis do not appear to differ significantly from those of other muscicapids (Dumbacher 1997, R. Elbel pers. comm.). Three caveats should be mentioned concern- ing my experiments. First, the experimental dishes were cooler, drier, darker, and more sed- entary than a live bird's plumage. The stress of captivity may intensify effects of pitohui toxins on lice. Second, the experimental dishes may expose lice to unnaturally low concentrations of toxin. In the dishes, lice have a choice of two feathers, one of which is nontoxic. Also, be- cause feathers contain much lower concentra- tions of toxin than skin (Dumbacher et al. 1992), exposure to feathers alone may under- estimate the effect of natural levels of homo- BTX on lice. Third, many experimental lice were collected from bird species distantly re- lated to pitohuis. Although pitohui toxin pro- foundly affected these lice, the toxin may have different effects on lice that coevolved with pi- tohuis because parasites often evolve resistance to host defenses. However, pitohui toxins have profound effects on many species of lice, sug- gesting that these toxins had profound effects on pitohui lice during the evolution of toxicity in pitohuis. It has been suggested that chewing lice have little or no influence on the fitness of their avian hosts (Rothschild and Clay 1952, Ash 1960, Marshall 1981), although recent studies sug- gest otherwise. Lice can reduce egg number and hatching success in chickens (Derylo 1974, DeVaney 1976), and studies of mate choice in Rock Doves (Columba livia) have shown that fe- males discriminate against males with high louse loads (Clayton 1990). In addition, lice can damage plumage to the extent that thermal conductance is increased, which would in- crease the metabolic costs of temperature reg- ulation (Booth et al. 1993). Even in low num- bers, lice can transmit diseases that profoundly affect the fitness of their avian hosts (Clayton 1990). Pitohui toxin may also affect other arthropod ectoparasites. Pitohui toxin attacks sodium channels composed of highly conserved pro- teins. These sodium channels are found in all other arthropod ectoparasites including feath- er mites, sucking mites, hippoboscid flies, soft and hard ticks, and ephemeral ectoparasites such as chiggers, mosquitoes, and leeches. Many of these ectoparasites are believed to be more detrimental to host fitness than are feath- er lice. Additional studies of these other ecto- parasites are needed. This study clearly demonstrates the mecha- nisms by which homobatrachotoxin can defend wild pitohuis against ectoparasites. However, because homoBTX affects predators and per- haps other parasites as well as lice, it is difficult to assess the relative role of louse defense in the evolution and maintenance of pitohui toxins. Naturally occurring levels of homoBTX affect New Guinea predators such as green tree py- thons (Chondropython viridis) and brown tree snakes (Boiga irregularis) and are known to de- ter human hunters (Majnep and Bulmer 1977, Kocher-Schmid 1991). The bright orange-and- black plumage pattern of the two most toxic pi- tohuis, P. dichrous and P. kirhocephalus, probably serves as an aposematic signal for visual pred- ators such as hawks and thus would not affect lice. Consequently, homobatrachotoxin repre- sents a single evolutionary innovation that may simultaneously influence a broad spectrum of pitohui enemies. ACKNOWLEDGMENTS I thank the Papua New Guinea Department of En- vironment and Conservation and the PNG Research and Conservation Foundation for providing visa as- sistance and permission to conduct this study, and the personnel at Varirata National Park for logistical support, especially P. Ainie and Bisikau Iova. I also thank Bulisa A. Iova for helping capture birds and lice, and S. O'Steen and R. M. Gaylord for assisting with experimental design and field work. S. Arnold, J. Bergelson, D. Clayton, R. Elbel, M. Kreitman, R. Page, S. Pruett-Jones, R. Price, and M. Wade provided insightful comments on the manuscript. Chemical analyses were directed by J. W. Daly and supported and carried out in the National Institutes of Health Laboratory of Bio-organic Chemistry. R. Price pro- vided invaluable assistance in identifying lice to ge- nus or species; however, I accept responsibility for any misidentifications. Funding for field work was provided by the University of Chicago Hinds Fund and National Geographic Society Grant 5082-93. The author was supported by a William Rainey Harper Fellowship and GAANN Ecology Fellowship from the University of Chicago. LITERATURE CITED ALBUQUERQUE, E. X., J. W. DAL3/4, AND B. WITKOP. 1971. Batrachotoxin: Chemistry and pharmacol- ogy. Science 172:995-1002. AsH, J. S. 1960.,A study of the Mallophaga of birds with particular reference to their ecology. Ibis 102:93-110. BEAR, g. 1995. An improved method for collecting bird ectoparasites. Journal of Field Ornithology 66:212-214. BOOTH, D. T., D. H. CLAYTON, ANDB. A. BLOCK. 1993. Experimental demonstration of the energetic cost of parasitism in free-ranging hosts. Pro- ceedings of the Royal Society of London Series B 253:125-129. BROWN, C. R., M. B. BROWN, AND B. RANNALA. 1995. Ectoparasites reduce long-term survival of their avian host. Proceedings of the Royal Society of London Series B 262:313-319. CLAYTON, D. H. 1990. Mate choice in experimentally parasitized Rock Doves: Lousy males lose. American Zoologist 30:251-262. CLAYTON, D. H., AND D. M. TOMPKINS. 1995. Corn- parative effects of mites and lice on the repro- ductive success of Rock Doves (Columba livia). Parasitology 110:195-206. DALY, J. W., AND t. F. SPANDE. 1986. Amphibian al- kaloids: Chemistry, pharmacology, and biology. Pages 1-274 in Alkaloids: Chemical and biolog- ical perspectives (S. W. Pelletier, Ed.). John Wiley and Sons, New York. DERYLO, A. 1974. Studies on the economic harmful- ness of the Mallophaga. II. Influence of lice in- festation on egg laying and hatching of hens. Medycyna Weterynaryjna 30:406-410 [In Pol- ish]. DEVANEY, J. g. 1976. Effects of the chicken body louse, Menacanthus stramineus, on caged layers. Poultry Science 55:430-435. DIAMOND, J. M. 1992. Rubbish birds are poisonous. Nature 360:19-20. DUMBACHER, J.P. 1997. The ecology and evolution of chemical defense in the avian genus Pitohui. Ph.D. dissertation, University of Chicago, Chi- cago. DUMBACHER, J. ?., B. M. BEEHLER, t. E SPANDE, H. M. GARRAFFO, AND J. W. DALY. 1992. Homobatrach- otoxin in the genus Pitohui: Chemical defense in birds? Science 258:799-801. DWIVEDY, A. K. 1988. Alkaloid neurotoxins-depen- dent sodium transport in insect synaptic nerve- ending particles. Comparative Biochemistry and Physiology 91C:349-354. HAMILTON, W. D., AND M. ZUK. 1982. Heritable true fitness and bright birds: A role for parasites? Sci- ence 218:384-387. KOCHER-SCHMIO, C. 1991. Of people and plants: A botanical ethnography of Nokopo Village, Ma- dang and Morobe Provinces, Papua New Guin- ea. Ethnologisches Seminar der Universitit und Museum fiir V/51kerkunde, in Kommission bei Wepf und Co., Basel, Switzerland. KOCHER-SCHMID, C. 1993. Birds of Nokopo. Muruk 6: 1-15. LEGGE, S., AND R. HEINSOHN. 1996. Cooperative breeding in Hooded Pitohuis Pitohui dichrous. Emu 96:139-140. LOYE, J. E., AND M. ZUK. 1991. Bird-parasite inter- actions: Ecology, evolution, and behaviour. Ox- ford University Press, Oxford. MAJNEP, I. S., AND R. BULMER. 1977. Birds of my Ka- lam Country. Aukland University Press, Auk- land, New Zealand. MANTEL, N. 1963. Chi-square tests with one degree of freedom: Extensions of the Mantel-Haenszel procedure. American Statistical Association Journal 58:690-700. MANTEL, IN[., AND W. HAENSZEL. 1959. Statistical as- pects of the analysis of data from retrospective studies of disease. Journal of the National Cancer Institute 22:719-748. MARSHALL, g. G. 1981. The ecology of ectoparasitic insects. Academic Press, London. MOURITSEN, K. N., AND J. MADSEN. 1994. Toxic birds: Defence against parasites? Oikos 69:357-358. POULSEN, B. O. 1994. Poison in Pitohui birds: Against predators or ectoparasites? Ernu 94:128-129. ROTHSCHILO, M., AND t. CLAY. 1952. Fleas, flukes, and cuckoos: A study of bird parasites. Collins, London. SALVADORI, t. 1881. Ornitologia della Papuasia e del- le Molucche. Socio Residente della R. Accadernia delle Scienze di Torino, Parte Seconda. SHAW, R. G., AND T. MITCHELL-OLDS. 1993. ANOVA for unbalanced data: An overview. Ecology 74: 1638-1645. Associate Editor: A. J. Baker