2001. The Journal of Arachnology 29:354±366

UNDER THE INFLUENCE: WEBS AND BUILDING BEHAVIOR OF PLESIOMETA ARGYRA (ARANEAE, TETRAGNATHIDAE) WHEN PARASITIZED BY HYMENOEPIMECIS ARGYRAPHAGA (, )

William G. Eberhard: Smithsonian Tropical Research Institute, and Escuela de BiologõÂa, Universidad de Costa Rica, Ciudad Universitaria, Costa Rica. email: [email protected]

ABSTRACT. On the evening that it will kill its host, the orb-weaving spider Plesiometa argyra, the larva of the ichneumonid Hymenoepimecis argyraphaga induces the spider to perform highly ste- reotyped construction behavior and build an otherwise unique ``cocoon web'' that is particularly well- designed to support the wasp larva's cocoon. Cocoon web construction behavior is nearly identical with the early steps in one subroutine of normal orb construction, and is repeated over and over. Usually all other normal orb construction behavior patterns are completely or nearly completely repressed. Experi- mental removal of the larva one or a few hours before cocoon construction would normally occur is sometimes followed by nearly normal cocoon web construction, and sometimes by construction of other highly altered web designs. The mechanism by which the larva induces these changes in the spider's behavior is thus apparently a fast-acting chemical, with effects that are manifested gradually. Partial recovery of orb designs sometimes occurred several days later. Keywords: Parasite, manipulation of host behavior, orb construction behavior, Plesiometa, Hymenoe- pimecis

Manipulation of host behavior by parasites most ®nely directed alteration of host behav- is a widespread phenomenon (Holmes & ior ever attributed to an parasitoid. Bethel 1972; Moore 1984; Barnard & Behnke It has long been known that psychotropic 1990; Toft et al. 1991; Godfray 1994; Mc- substances can modify the forms of orb webs Lachlin 1999; Poulin 2000), but most reports (Witt et al. 1968), but the details of how par- of behavioral modi®cations, especially those ticular steps of the spider's construction be- caused by insect parasitoids in other , havior are affected have never been deter- involve only simple behavior patterns such as mined. Elucidation of which aspects of movement from one habitat to another, adop- behavior are changed can have important con- tion of sleeping postures, or eating more or sequences for the common use of details of less (Wickler 1976; Godfray 1994; McLachlan building behavior as taxonomic characters 1999). Spider behavior is also in¯uenced by (Eberhard 1982; Hormiga et al. 1995; Gris- insect parasitoids (Schlinger 1987). At least wold et al. 1998), as well as how evolutionary some of these changes may be due to rela- transitions may have occurred. It has not al- ways been clear whether or not some variant tively simple mechanisms, such as modi®ca- behavior patterns should be recognized as sep- tion of particular receptors (Jenni et al. 1980). arate traits (Eberhard 1990). If particular be- This report concerns an unusually selective havior patterns can be selectively induced, behavioral modi®cation by the larva of the then the case for their independence from oth- parasitoid wasp Hymenoepimecis argyrapha- er traits, and thus their potential usefulness as ga Gauld (Ichneumonidae), which apparently characters, is strengthened. Clari®cation of the chemically induces expression of the early behavioral effects of this wasp parasite on steps of one subroutine of orb web construc- web construction behavior thus promises to tion in the spider Plesiometa argyra (Wal- improve understanding of the organization of ckenaer 1841) (Tetragnathidae), while sup- behavior within the spider, and of the useful- pressing all the rest of orb construction ness of different behavioral characters in spi- behavior (Eberhard 2000a). It may be the der taxonomy.

354 EBERHARDÐPARASITE MODIFICATION OF CONSTRUCTION BEHAVIOR 355

The life cycle of H. argyraphaga is the fol- lowing (Eberhard 2000b). The female wasp attacks P. argyra as the spider rests at the hub of its orb, stings it into a temporary (10±15 min.) paralysis, and glues an egg to the spi- der's abdomen. Subsequently the spider resu- mes normal activity, and builds apparently normal orbs to capture prey during the next approximately 7±14 days while the wasp's egg hatches and the larva grows. The larva re- mains attached to the surface of the abdomen, and feeds by sucking hemolymph through small holes it makes in the spider's abdominal cuticle. The second instar larva, on the night that it will kill its host, induces the spider to construct an otherwise unique ``cocoon web'' of dragline silk, molts to the third instar, and then kills and consumes the spider. The next evening the larva spins a cocoon hanging by a line from the cocoon web. The larva (which is barely visible through the thin walls of the cocoon) pupates about 4 days later, and then emerges as an adult wasp after about 7 more days. Figure 1.ÐWeb of an unparasitized adult Ple- METHODS siometa argyra. Scale bar ϭ 3.0 cm. Field observations were made near Parrita, Puntarenas Province, Costa Rica (elev. 10 m) served after the spiders had been taken to San in January and February of 1999 and 2000 in Antonio de Escazu (elev. 1300 m), where they a mature plantation of African oil palm were kept indoors at room temperature for up (Elaeis guineensis L.) where spider popula- to two weeks. On the evening the larva was tions were dense. Web measurements were to be removed, the spider was kept in a small performed in the morning, and thus did not container (6 cm dia.) in which it could not include webs built later in the day (which may spin a web, and then placed on a wire frame have different designsÐEberhard 1988). Con- as soon as the larva was removed between struction of cocoon webs made by spiders car- 2100 and 0200 h. Because the spiders seemed rying wasp larvae was observed indoors near to need air movement to induce web construc- Parrita the night after the spiders were col- tion, they were not kept in cages, but allowed lected and transferred onto silk lines from P. to range freely in rooms. argyra orbs that had been fastened to approx- imately horizontal 0.6 m dia. circular wire RESULTS frames that were hung about 1 m above the Field.ÐThe orbs of spiders carrying wasp ¯oor. Larvae, which would kill their hosts that eggs and larvae were not distinguishable from evening, could be reliably distinguished (15 of the more or less horizontal, moderately open- 15 cases) from others on the morning and af- meshed orbs of unparasitized spiders (Figs. 1, ternoon of the same day, due to their larger 2) (ANCOVA analyses showed no signi®cant size. Voucher specimens of and spiders effects of parasitism by larvae, or by eggs and have been deposited in the U. S. National Mu- larvae (P ϭ 0.91, 0.40). Even parasitized spi- seum of Natural History, the Museum of ders found the morning of the day on which Comparative Zoology at Harvard, and the they would be killed by the wasp larva were Museo de EntomologõÂa of the Universidad de on freshly made, apparently normal orbs. Oth- Costa Rica. er than orbs, the only other webs on which The behavior of spiders from which the lar- unparasitized spiders occurred were small va had been experimentally removed was ob- molting webs (Eberhard et al. 1993). These 356 THE JOURNAL OF ARACHNOLOGY

Figure 2.ÐNumbers of radii and sticky spiral loops (mean number of loops directly above, below, and to the sides of hub) in webs of spiders in the ®eld that were parasitized (®lled symbols) and unparasitized (open symbols). No differences between parasitized and unparasitized individuals were apparent. webs were rare, and several newly molted in- More than 100 cocoon webs were observed dividuals lacked such webs. Despite the dense in the ®eld. They almost always consisted of spider populations, no egg sacs or webs as- a few lines that radiated in a more or less hor- sociated with egg sacs were seen; egg sacs izontal plane from a ``hub'' or central area, may be hidden in leaf litter, as in the closely where the cocoon's suspension line was at- related Leucauge mariana (Keyserling) (Ibar- tached, and were each attached directly to a ra et al. 1991; V. Mendez pers. comm.). support (Eberhard 2000a, Figs. 3±5). Most ra- EBERHARDÐPARASITE MODIFICATION OF CONSTRUCTION BEHAVIOR 357

Figures 3±5.ÐTypical cocoon web. 3. Dorsal view. Scale bar ϭ 3.0 cm; 4. Detail of attachments to a leaf. Scale bar ϭ 0.5 cm; 5. Lateral view. Scale bar ϭ 3.0 cm. 358 THE JOURNAL OF ARACHNOLOGY dial lines had many branches near their tips moved directly from one attachment to the and were thus attached at many adjacent next as in normal frame and radius construc- points to the substrate (Fig. 4). They were also tion. sometimes attached at multiple points in the Radial lines were all in nearly the same central area. There were several other indica- plane and were added to the web using two tions, in addition to the planar arrangement of similar, simple behavior patterns (Fig. 8, A radial lines, that the webs from which cocoons and B). The spider began by attaching its were suspended represented modi®ed orbs. dragline at the hub, then walked toward the Some cocoon webs had circular lines similar substrate along a radial line, walked along the to those at the hubs of normal orbs (17% of substrate a short distance and attached the line

41 webs checked for this detail) (Figs. 6, 7), it had laid from the hub (A1,B1). Then it re- though in no case was the central portion of turned to the hub, laying a second dragline as the hub empty, as in normal orbs. Some had it walked along this line or another radial line one or more frame lines connecting the radial that it had laid previously and attached it at lines (29% of 42 webs checked for this detail) the hub (A2,B2). When the substrate was thin (Fig. 6). These frames were typically much (a strand of wire, for instance) the spider usu- shorter and nearer the hub than were the frame ally moved to the opposite side to make the lines of normal orbs (Fig. 1). The most elab- attachment before returning to the hub, as is orate cocoon web had a distinct hub, frame typical of frame construction in orbs (Tilquin lines, and a mesh above and below the hub. 1942; Eberhard 1990). At the opposite extreme, the two simplest co- The two patterns differed in that either the coon webs consisted of a single strong line lines were laid without attachments to previ- with the larva or the cocoon hanging from the ously laid radial lines (A1,A2 in Fig. 8), or central portion. (more often) the spider attached its dragline Cocoon webs spanned smaller spaces than one or more times to radial lines both on the normal orbs of mature females. The distance way out and on the way back to the hub (B1, between the two most distant points of attach- B2 in Fig. 8). Consecutive radial lines were ment of anchor lines was smaller in cocoon always laid in different directions, as in orb webs (mean 36.6 Ϯ 17.2 cm in a sample of construction by other araneoid spiders (Ti- 38) than in orbs (99.6 Ϯ 47.5 cm in a sample lquin 1942; Dugdale 1969; Le Guelte 1966; of 31) (P Ͻ 0.001 with Mann Whitney U- Witt et al. 1968; Eberhard 1982). Each radial Test). These cocoon webs also had fewer an- line was reinforced repeatedly, and the total chor lines (lines directed to the substrate) amount of dragline silk in a cocoon web prob- (mean 3.9 Ϯ 1.5 for cocoon webs, 5.3 Ϯ 1.7 ably represented a major fraction of that in an for the orbs; P Ͻ 0.001 with Mann-Whitney orb. The estimated total numbers of radial U-Test). lines in two ®nished cocoon webs were 36 and Construction behavior.ÐCocoon web 30. Thus the number of radial trips was on the construction behavior, observed in ®ve spiders same order as the typical number of radii (20± captured in the ®eld the same day with larvae 35) in a normal orb (Fig. 1). and a sixth three days after being collected, The behavior of one further individual, col- was very consistent. Early in the evening, the lected four days previously and observed in spider built several lines, repeatedly removing San Antonio de Escazu, was very different. and shifting the points of attachment as typi- The spider descended to the ¯oor about 1.5 cally occurs during the preliminaries of orb below the wire hoop, formed a ``hub'' where construction of many species of orb weavers several lines converged about 1 cm above the (Tilquin 1942; Eberhard 1990). It then re- ¯oor, and then made 5±10 very long radial mained more or less immobile until between excursions (up to 1.3 m each) walking on the 23:30 and 01:00, when construction activity surface of the ¯oor. As it moved away from occurred in bursts. Typically the spider added the hub it walked in a nearly straight line, at- one to several radial lines in quick succession, taching its drag line periodically to the ¯oor, and then spent a minute or more (up to 30 but in some cases it gradually made an arc of min) immobile at the hub before the next burst up to more than 180Њ before it turned back of activity. The spider's movements showed and slowly retraced its path back to the hub. no signs of weakness or vacillation, and it On at least four occasions the spider encoun- EBERHARDÐPARASITE MODIFICATION OF CONSTRUCTION BEHAVIOR 359

Figures 6±7.ÐDorsal view of an unusually elaborate cocoon web with hub loops and a frame line. Scale bars ϭ 2.5 and 1.0 cm respectively. 360 THE JOURNAL OF ARACHNOLOGY

Figure 8.ÐDiagrammatic representations of the sequences of behavior during construction of a cocoon web (A1±A2, and B1±B2) and a frame line in a typical orb (C1±C4). Stippling represents substrate, black spots represent points where the dragline was attached, and dashed lines represent lines laid earlier in the sequence (C1±C4 after Eberhard 1990). Cocoon web construction corresponds to the behavior in C1 and the ®rst part of C2. EBERHARDÐPARASITE MODIFICATION OF CONSTRUCTION BEHAVIOR 361

strate, and more or less converged at the hub (Figs. 10, 11). A fourth spider, which had al- ready begun cocoon web construction when I removed the larva, resumed cocoon web con- struction but did not return consistently to the hub, and made two additional ``hubs''. Sim- pli®ed, ``vestigial'' webs, that had only a few more or less radial lines converging at the point where the spider rested and large masses of silk lines loosely packed together near the central area, were built by 13 spiders (Figs. 12, 13). These radial lines were attached to the substrate at only one or at most a few points. Vestigial webs never had hub loops, temporary spirals or sticky lines, and only sel- dom had recognizable frame lines. One further web was a nearly normal orb, except that the center of the hub was not removed and some Figure 9.ÐDorsal view of a web made by a ma- ture male on the night that it was killed by a wasp portions of the sticky spiral lacked sticky larva. Note multiple attachments to substrate of balls. some radial lines (as in Figs. 4 and 10). Scale bar The construction behavior of spiders from ϭ 3.0 cm. which larvae had been removed was observed for two cocoon webs and three vestigial webs. Cocoon web construction was very similar to tered an object that it could have climbed and that described above for spiders carrying lar- thus have raised its drag line off the ¯oor, but vae, including the frequent pauses between instead it struggled on across the ¯oor. When bursts of construction behavior, except that on I then removed the larva, and replaced the spi- some occasions the spider failed to attach its der on the wire hoop after breaking the lines drag line at the hub when it returned after lay- leading downward toward the ¯oor, the spider ing a radial line. The drag line laid on the next again descended to the ¯oor where it made trip away from the hub was thus not attached another hub. at the hub, but originated part way out the

Wasps generally avoided parasitizing ma- previous radial line (line 2±4 in B2 of Fig. 14). ture males (Eberhard 2000), but two larvae on When this behavior was repeated over and mature males matured and made cocoons in over, the hub gradually expanded and became captivity. One male spider did not make a co- dispersed. The resulting web had large num- coon web (or indeed any supporting structure bers of more or less radial lines attached to whatsoever); the wasp's cocoon hung from a the substrate close to each other, but a diffuse single short strand of spider silk. The second central area (Fig. 15). parasitized mature male spider, however, built During vestigial web construction, the spi- an extensive web that resembled a cocoon der also made radial lines attached to the sub- web in being more or less planar, and having strate just as above. On some return trips to many attachments to the substrate on the night the hub area, however, it broke and removed that it was killed and consumed by the larva these lines, reeling them up and leaving them (Fig. 9). packed loosely together attached to the web. Experimental removal of larvae.ÐLarvae The ®nal product of this process of repeatedly were removed from 22 spiders in captivity on laying and then removing lines was a scanty the evening when the spider was to be killed. array of more or less radial lines, and one or Four spiders built no webs that night. The 18 more large masses of ¯uff (Fig. 13). webs that were built were of three types. None of the 22 experimental spiders that Three were more or less typical cocoon webs, built webs died on the evening the wasp larva with a low number of radial lines which were was removed. In nine cases the spider built a composed of multiple strands that were at- second web on the following night, and the tached at many adjacent points on the sub- second web was of the same type built on the 362 THE JOURNAL OF ARACHNOLOGY

Figures 10±13.ÐWebs made by spiders from which the wasp larva was removed on the night when the larva would have normally killed the spider. 10. Cocoon-web type, in which the few radial lines each had multiple attachments to the substrate. Scale bar ϭ 3.0 cm. 11. Close-up of hub of web in Fig. 10. Scale bar ϭ 1.0 cm. 12. ``Vestigial'' type web, in which a few radial lines were attached singly to the substrate (heavy white lines are from previous web of another spider). Scale bar ϭ 3.0 cm. 13. Close-up of the hub of a vestigial web (different web from that in Fig. 12), showing several masses of ¯uff. Scale bar ϭ 1.0 cm; all wire hoops were horizontal.

®rst (two cocoon webs, seven vestigial webs). construction, however, which sometimes oc- Five of the second vestigial webs had at least curs as part of frame construction, the early one hub loop. Due to deaths and emigrations, stages do not involve removing or shifting it was not possible to follow the spiders' be- lines (Fig. 8 C1,C2). Premature termination of havior systematically on subsequent nights. this type of frame construction behavior when Two spiders survived for a week, and gradu- the spider returns to the hub after the ®rst at- ally built webs that were progressively more tachment to the substrate and followed by at- orb-like though still substantially altered (Fig. tachment of the spider's drag line at the cen-

16). tral area (x in Fig. 8 C2), would result in a sequence of operations identical to type A co- DISCUSSION coon web construction (Fig. 8 A). Adding at- Comparison of cocoon web construction tachments to the line already in place on the behavior with the early stages of normal orb way out would result in a sequence similar or construction (Eberhard 1990) indicates that it identical to the second type of cocoon web is probably homologous with the early steps construction behavior (Fig. 8 B). Similar at- of type ``D'' frame construction (Fig. 8 C1± tachments sometimes occur in the closely re- C4). Most anchor line construction in an orb lated L. mariana during frame construction of involves removal of lines already in place, or types ``A'' and ``C'' of Eberhard (1990) but shifting their attachments to each other (Ti- were not seen in conjunction with type D of lquin 1942; Eberhard 1990), but neither of Eberhard (1990) (the same individual often these behavior patterns was ever seen during performed more than one type while building cocoon web construction. In type D anchor a given orb). A further resemblance to attach- EBERHARDÐPARASITE MODIFICATION OF CONSTRUCTION BEHAVIOR 363

Figure 15.ÐCocoon-type web with dispersed hub built by a spider from which the larva was removed on the evening on which it would have Figure 14.ÐDiagrammatic representations of co- normally killed its host. Scale bar ϭ 3.0 cm; wire coon web construction behavior of a spider with a hoop was horizontal. wasp larva (A1,A2) and a spider from which the Figure 16.ÐOrb-like web built by a partially re- wasp larva had been experimentally removed (B1, covered spider. The wasp larva had been removed B2). The experimental spider sometimes omitted the ®ve days earlier, on the evening when it would have ®nal attachment at the hub typical of cocoon web killed the spider, and the spider had spun a typical construction (attachment 3 in A1 and A2; see also vestigial web on that night. Scale bar ϭ 2.0 cm. Fig. 8 A2 and B2); when it moved away from the hub to make the next radial line, the dragline was thus displaced away from the hub (line 2±4 in B ). 2 ther evidence that cocoon webs were homol- Repeated omissions of this attachment resulted in a ogous with orbs is the fact that when these diffuse central area of the web (Fig. 15). webs had more than three radial lines, these were nearly always in approximately the same ments of anchor lines built during orb web plane. In addition, some cocoon webs had construction by other orb weavers (Tilquin frame lines, and a few had hub loops (Figs. 6, 1942: Eberhard 1990) was the attachment of 7). radial lines to thin objects by moving to the The homology of cocoon and orb webs em- opposite side of the object just before attach- phasizes that perhaps the most extraordinary ing. Thus, the spider built the cocoon web by aspect of the wasp larva's effect on the spider apparently repeating the ®rst portions of one was not so much what the spider did, but what type of frame construction over and over. Fur- it did not do. Many aspects of normal orb con- 364 THE JOURNAL OF ARACHNOLOGY struction were completely absent, including design is both unprecedented in P. argyra or both breaking, reeling up and replacing lines any closely related orb weaver, and seems es-

(e.g. Fig. 8 C4), and breaking and then re-at- pecially appropriately designed to increase the taching lines. These two behavior patterns survival of the wasp. Induction of spinning form integral parts of most types of both behavior also occurs in several families of spi- frame and radius construction in normal orbs ders parasitized by acrocerid ¯ies; the spider of this and other species (Tilquin 1942; Eber- spins a thin cell similar to that made just prior hard 1982 1990; Coddington 1986). A single to moulting, and the larva clings to the web failure to repress these behavior patterns could after emerging from the spider (Schlinger be disastrous for the wasp larva, as it would 1952, 1960, 1987). result in the removal of the many-stranded ca- The changes in the behavior of P. argyra ble of radial lines, and its replacement with a are induced chemically rather than by direct much weaker line. This indeed occurred in the physical interference with the spider's nervous vestigial webs built by spiders from which lar- system. The wasp larva contacts only the sur- vae were experimentally removed. Also com- face of the spider's abdomen and limits itself pletely missing were production of the tem- to making small holes through which it im- porary spiral and sticky spiral, and removal of bibes hemolymph (Eberhard 2000a,b). In ad- the central portion of the hub at the end of orb dition, some spiders built normal cocoon webs construction, which again would have resulted after the larva was removed. Some ichneu- in considerable weakening of the support for monids modify host behavior and physiology the wasp's cocoon. These differences between via products injected by the female wasp cocoon webs and normal orbs are appropriate when she oviposits (Gauld 1995). However, to make the cocoon web stronger and less the lack of web modi®cation in the days im- likely to be damaged by falling debris, and mediately following the attack by the wasp, thus a more durable support for the wasp's the sudden abrupt shift in behavior that is co- cocoon than an orb would be. Strong support ordinated with maturation of the larva, and the for the cocoon may be important for the changes in webs produced by removing the wasp's survival, as in the related Hymenoe- larvae, all argue that the larva rather than the pimecis robertsae some pupae died when adult female wasp induced modi®ed web con- heavy rains damaged cocoons (Fincke et al. struction behavior. Secretion of neuromodu- 1990). lators by parasitoid larvae has been implicated The importance of the precision of the be- in behavioral changes produced in some insect havior induced in the spider is also illustrated hosts (Beckage 1997). The variety of web by the effect of occasional omission of one forms and construction behavior observed normal detail, the ®nal attachment at the hub when the larva was removed prematurely sug- after a radial line was built (Fig. 8 A2,B2) that gest a complex, gradual effect rather than an was seen in some spiders from which the lar- abrupt, simple modi®cation. vae were experimentally removed. The result- The ability of Hymenoepimecis argyrapha- ing lack of a clear central point of conver- ga to induce speci®c behavior patterns in spi- gence produced webs that were much less ders indicates that even these ®ne behavioral appropriately designed to support the wasp's details are independent units or modules at cocoon (Fig. 15). It is not clear whether the some level within the spider, and not just ar- aberrant behavior of one spider that laid radial ti®cial constructs. The additional web forms lines on the surface of the ¯oor instead of in produced by experimentally removing larvae the air was something that happens in nature from spiders suggest even further subdivisions (such webs would be missed in the ®eld) or of building behavior. The problem of what was an artifact of captivity. constitutes a biologically realistic behavioral In some cases, claims that modi®cation of unit is crucial in the use of behavior patterns host behavior associated with parasitism rep- as taxonomic characters in orb-weavers Eber- resents an evolved adaptation by the parasite hard 1982; Coddington 1986, 1990; Scharff & to promote its own reproduction have been Coddington 1997; Griswold et al. 1998) as controversial (Toft et al. 1991; Poulin 2000). well as in other (Wenzel 1992). The There can be little doubt on this score with results of this study suggest that it is reason- the species of this study, as the cocoon web able to attempt to use even ®ner behavioral EBERHARDÐPARASITE MODIFICATION OF CONSTRUCTION BEHAVIOR 365 details than those that have been used previ- phaga (Hymenoptera, Ichneumonidae) and its ously in orb weaver taxonomy. The cocoon host, the spider Plesiometa argyra (Araneae, Te- web of P. argyra is similar to the secondarily tragnathidae). Journal of Hymenoptera Research reduced ``asterisk'' web found by Stowe 9:220±240. (1978) in the distantly related araneid Wixia Eberhard, W.G., S. Guzman-Gomez & K.M. Catley. 1993. Correlation between spermathecal mor- ectypa (Walckenaer). Whether or not this evo- phology and mating systems in spiders. Biolog- lutionary transition involved chemical chang- ical Journal of the Linnean Society 50:197±209. es similar to those produced by H. argyra- Fincke, O., L. Higgins & E. Rojas. 1990. Parasitism phaga remains to be determined. of Nephila clavipes (Araneae, Tetragnathidae) by an ichneumonid (Hymenoptera, Polysphinctini) ACKNOWLEDGEMENTS in Panama. Journal of Arachnology 18:321±329. I thank I.D. Gauld and H.W. Levi for iden- Gauld, I.D. 2000. The re-de®nition of pimpline ge- tifying the wasp and the spider respectively. nus Hymenoepimecis (Hymenoptera: Ichneu- This research was ®nanced by the Smithson- monidae) with a description of putatively ple- ian Tropical Research Institute and the Vicer- siomorphic new Costa Rican species. Journal of rectorõÂa de InvestigacioÂn of the Universidad Hymenoptera Research 9:213±219. de Costa Rica. Gauld, I.E., J. Ugalde & P.E. Hanson. 1998. Guia de los de Costa Rica. Revista de Biol- LITERATURE CITED ogia Tropical. 46 (Supl. 1):1±189. Barnard, C.J. & J.M. Behnke. 1990. Parasitism and Godfray, H.C. J. 1994. Parasitoids Behavioral and Host Behaviour. Taylor & Francis, London. Evolutionary Ecology (Princeton University Beckage, N.E. 1997. New insights: how parasites Press, Princeton, New Jersey. and pathogens alter the endocrine physiology and Griswold, C. E., J.A. Coddington, G. Hormiga & development of insect hosts. Pp. 3±36. In Para- N. Scharff. 1998. Phylogeny of the orb-web sites and Pathogens, Effects on Host Hormones building spiders (Araneae, Orbiculariae: Deino- and Behavior (N.E. Beckage, ed). Chapman and poidea, Araneoidea). Zoological Journal of the Hall, New York, New York. Linnean Society 123:1±99. Coddington, J.A. 1986. Monophyletic origin of orb Hormiga, G., W.G. Eberhard & J.A. Coddington. webs. Pp. 319±363, In Spiders, Webs, Behavior 1995. Web construction behavior in Australian and Evolution (W. Shear, ed). Stanford Univer- Phonognatha and the phylogeny of nephiline and sity Press. tetragnathid spiders (Araneae, Tetragnathidae). Coddington, J.A. 1990. Cladistics and spider clas- Australian Journal of Zoology 43:313±364. si®cation: araneomorph phylogeny and the Ibarra, G., O. Jaramillo M. & T. de Castro C. 1991. monophyly of orbweavers (Araneae: Araneo- Observaciones sobre el comportamiento reprod- morphae; Orbiculariae). Acta Zoologica Fennica uctor de las aranÄas del geÂnero Leucauge (Ara- 190:75±87. neae: Tetragnathidae) del Soconusco, Chiapas, Dugdale, B.E. 1969. The weaving of an engineering Mexico. Pp 65±66. Memorias del 26 Congreso masterpiece, a spider's orb, done at Fryson Nacional de Entomologia, Veracruz, Mexico. Lakes, N. J., August 8, 1942, as observed and Jenni, L., D.H. Molyneux, J.L. Livesley & R. Gal- reported by B. E. Dugdale, structural engineer. un. 1980. Feeding behaviour of tsetse ¯ies in- Natural History 78(3):36±41. fected with salivarian trypanosomes. Nature 283: Eberhard, W.G. 1982. Behavioral characters for the 383±385. higher classi®cation of orb-weaving spiders. Le Guelte, L. 1966. Structure de la toile de Zygiella Evolution 36:1067±1095. x-notata Cl. (AraigneÂes, Argiopidae) et facteurs Eberhard, W.G. 1988. Behavioral ¯exibility in orb qui reÂgissent le comportemente de l'araigneÂe web construction: effects of silk supply in dif- pendant la construction de la toile. TheÂse Publi- ferent glands. Journal of Arachnology 16:303± cations de l'Universite Nancy Faculte des Sci- 320. ences. Eberhard, W.G. 1990. Early stages of orb construc- McLachlan, A. 1999. Parasites promote mating suc- tion by Philoponella vicina, Leucauge mariana, cess: the case of a midge and a mite. and Nephila clavipes (Araneae: Uloboridae and Behaviour 57:1199±1205. Tetragnathidae), and their phylogenetic implica- Poulin, R. 1994. Meta-analysis of parasite-induced tions. Journal of Arachnology 18:205±234. behavioural changes. Animal Behaviour 48:137± Eberhard, W.G. 2000a. Manipulation of host spider 146. web construction behavior by a wasp larva. Na- Poulin, R. 2000. Manipulation of host behaviour by ture 406:255±256. parasites: a weakening paradigm? Proceedings of Eberhard, W.G. 2000b. The natural history and be- the Royal Society of London, Series B 267:787± havior of a parasitoid, Hymenoepimecis argyra- 792. 366 THE JOURNAL OF ARACHNOLOGY

Scharff, N. & J.A. Coddington. 1997. A phyloge- erus cordatus and Wixia ectypa (Araneae: Ara- netic analysis of the orb-weaving spider family neidae). Journal of Arachnology 6:141±146. Araneidae (Arachnida, Araneae). Zoological Tilquin, A. 1942. La Toile GeÂometrique des Araig- Journal of the Linnean Society 120:355±434. neÂes. Presses Univ. France, Paris. Schlinger, E.I. 1952. The emergence, feeding hab- Toft, C.A., A. Aeschlimann & L. Bolis. 1991. Par- its, and host of Opsebius diligens Osten Sacken. asite±Host Associations. Oxford University Pan-Paci®c Entomology 28:7±12. Press, New York. Schlinger, E.I. 1960. A revision of the genus Ogco- Wenzel, J.W. 1992. Behavioral homology and phy- des Latrielle with particular reference to species logeny. Annual Review of Ecology and System- of the Western Hemisphere. Proceedings of the atics 22:361±381. United States National Museum 111:227±336. Wickler, W. 1976. Evolution-oriented ethology, kin Schlinger, E.I. 1987. The biology of Acroceridae selection, and altruistic parasites. Zeitschrift fuÈr (Diptera): true endoparasitoids of spiders. Pp. Tierpsychologie 42:200±214. Witt, P.N., C. Reed & D.B. Peakall. 1968. A Spi- 319±327, In Ecophysiology of Spiders (W. Nen- der's Web Springer, New York. twig, ed.). Springer-Verlag, New York. Stowe, M.K. 1978. Observations of two nocturnal Manuscript received 15 December 2000, revised 1 orbweavers that build specialized webs: Scolod- May 2001.