8 1

ENIIDo. 69(.). 19", ~ . IIH-II60 o I'll by tbe EcoioIjcaI Society of America

SATURNIID AND SPHlNGID CATERPILLARS: TWO WAYS TO EAT LEAVES'

E. A. BERN .... YS DiI'isiQII oJ Biological Control and Departments oj Emom%gical Sciences and Zoology. VI/h'ersily a/Cali/ornia. BerkRfey. Coli/ornia 947]0 USA

AND

O. H . JANZEN Depanmenl a/ Biology, Unil'f'rsily o/Pennsy/I'Qnio., Philadelphia, PennsY/I'onia 19UU USA

Abstract. We demonstrated allometric differences in re lative head mass in different instars of 12 species of Satumiidae and 14 species of . The differences were related to the different ways in which individuals from the two fami lies ate their respective host plants and to the different properties of the hosts that tcnded to be favored by each lepidopteran family, The salurniids tcnded to have various simple cUlling methods, while the sphingids tore and crushed the food, so that in the former, ingested food was in the form of relatively large uniformly sized pieces, and in the latter it was apparently well masticated. Satumiid mandibles were short and simple, while sphingid mandibles were long, toothed, and ridged in a variety of complex ways. The food ofsaturniids tended to consist of old, tough, tannin­ rich leaves, while that ofsphingids was softer, younger, and contained small toxic molecules. The generalists within each group tended to be similar to one another, while the specialists (which occurred more frequcntly among the sphingids) had very characteristic mandibles, each of unique design. One sphingid species feeding on a vine with characteristically very tough leaves had the "satumiid" design of mandibles. The features typical of the two groups of caterpillar are discusscd in relation to feeding strategy, digestion, avoidance of plant "defenses," and rapidity of ingestion. Key WQrds: allometry; digestion; head siu; ingestion; leaf toughness; ; lepidopteran mandibles; Saturniidae: Sphingidae.

INTRODUCTION the contrasting mandibular morphology in these two Caterpillars of the families Satumiidae and Sphin­ fa milies and the potential roles of mandibular gidae are most species rich in the tropics, and where morphology in processing leaves of different types. many species occur together they have few host-plant MATERIALS AND METHODS species in common (e.g., Janzen 1984). Plant species used by Sphingidae lend to be relatively deficient in All catcrpillars except Manduca sexta were collected phenolics but are likely to contain alkaloids and other in the dry forests of Santa Rosa National Park, north­ small tox ic molecules, while Saturniidae use host-plant western Costa Rica (Janzen 1984). They were fixed in species that are rich in phenolics and poor in alkaloids boiling wa ter and preserved in 70% ethanol. (e.g., Janzen and Waterman 1984). There are, however, Relati ve sizes of mandibles, heads, and headless exceptions and complications to this general picture: bodies were measured by taking the dry mass of each satumiid caterpillars often select older leaves and are after removal offood from the gut lumen. species usuall y found in the crowns of adult trees, treelets, or examined and numbers of each are shown i.n Table I . woody vines, while sphingids are less particular about More extensive studies were undertaken with different plant age and commonly feed on younger leaves. They instaTS of the satumiids Orhorene purpurascens and may even eat herbs and other small plants. In addition, Rothschildia lebeau and the sphingids Pachylia saturniidsare re latively more polyphagous than sphin­ and Manduca dilucida. Mandibles of selected species gids (Janzen 1984). A variety of physical and chemical were measured wi th an eyepie<:e micrometer on a Wild features of leaves influences what species of host plant slereomicroscope and drawn at appropriate magnifi­ is fed upon by a species ofcaterpillar . Here we examine cations with a camera lucida. Gut contents of selected species were removed and ' Manuscript received I July 1981; revised S December sampled by random selection from suspensions in water. 1981; accepted 1 December 1981. The samples were mounted on slides for examination, 11 54 E. A. BERNA YS AND D. H. JANZEN Ecolos)', Vol. 69, NO.4

TABLE I. Species analyzed. Unless stated. instars are IV or V. 24 Known food plants Num- eaten b)' indi viduals examined -;; 20 Satumiidae • Ro/hschildia febeau "'" • instar I Zufania guidonia ~ 11 14{10 Casearia corymbosa ~ ,. ~ " III 12 Spendias mombin o IV 12 Exoslellla mexicana V 7 '0= 12 O/horene purpurascellS 1 , 11 3 '"• III 2 Manilkara chide • B IV , ~• V , ~ Hy/esia linea/a ll , : 4 V , Casearia corymbosa % Arsenura arlllida , Bombacopsis quina/um Periphoba areaei 2 Spendias mOlllbin o 2 3 Eades imperialis 3 Spendias mombin Log Body Mass SyssphifIX molina 4 Pilhecellobium saman Cania grandis Flo. I . The: relationship between head mus as a per· centage: of headless bod)' mass and headless bod)' mass for Syssphinx rolfa , Pi,hecellobium saman various species ofSatumiidae: (r .. - 0.95, y " 20.5 - 6.3x). 1 .. l1)desia Iineala; 2 .. Arsenura armida; 3 .. Periphoba AUlOmeris Iugana 4 Annona purpurea areae;: 4 .. Au/omt'ris =ugana; 5 .. Eades imperialis: 6 .. Ci/hmmia fobesis , Spendias mombin AUlOmt'fis io; 7 .. Caio championi: 8 ... Cilheronia lobesis, AU/omeris io 1 Cresa!n/ia alata 9 .. Syssphinx colla; 10 .. SyssphillX molina. - -- the regres­ Caio championi , Bombacopsis quina/um sion line ror Sphingidae.

Sphingidne and 50 adjacent particles were drawn with a camera Eup)'rrhoglossum sagra 1 Chomelia spinrua lucida. Areas and perimeters were measured using a Eumorpha salelli/a 3 Cissus slcyoides digitizer and IBM PC. Pm:Jl)'liaficus instar II 2 III 2 { Chlorophora /inctoria RESULTS IV 2 Ficus COlini/oIia V 2 The relationshi p between relati ve head mass, ex­ pressed as a percentage: of headless body mass, and headless body mass differs for the satumiid and sphin­ Manduca rustico 2 Amphilophilum panicu· falum gid caterpillars (Figs. 1-3). The slopes of the regressions /tfanducaj1ares/an 1 Cydisla helerophylla for the two groups differ significantly (P < .00 I, t test Mandl/ca lanuginosa 2 Cydis/a helerophylla for parallelism), with the regression lines intersecting Pro/ambufyx s/rigilis 1 Spondias mombin at a body mass of =:: 2 mg. For caterpillars with masses Pachylioides resumens 3 Fors/eronia spicala 1 Aef/opos/adus 1 Genipa americana < 10 mg, the relative: head mass was smaller for sphin­ ocypele 2 Telracera YOlubilis gids than for salumiids. For caterpillars > 101 mg, rel­ Erinnyis ella 1 Sebastiana con/usum ative head mass was greater for sphingids. A certain Manduca ~Xla instar amount of variation in relative head mass wasexpectcd 1 8 11 7 since body mass changes more than does head mass III , Solanum tuberosum during an instar. The variation among sphingids ap­ IV , peared to be greater than that among satumiids (c[ V 8 Figs. I and 3). Manduca dilucida instar For all specics of both families taken togcther, man­ 11 1 III Tabebuia ochracea dible mass was closely correlated with head mass. The IV :{ Sapranthus palanga regressions for the two families were coincident. For V 2 the combined data, r 1 = 0.94, y .. - 1.0 + 1.027x. The Cccytius dupenchel Annona reticulala basic mandible shapes differed in the two families. Perigonia fusca Calycophyllum candi· dissimum Among saturniids all later instars had mandibles that were relati vely short, with a broad base and no obvious teeth. Species that feed on many hosts (Janzen 1984) have the simplest mandibles (e.g., Rothschildia lebeau, August 1988 HOW CA TERPILlA.RS EAT LEAVES 1155 a ' ,. ' 24 ., O . PfJRPfJRASCENS : JC • , R . LEBEAU : • -• • •• , ' • • • • ~ • • ~ 1. • • Il • m • 012 " b • '. , '"• " • : 8 • • • ~ , .'. • .' • .. :-'" -1 0 1 2 L og Body M ass FIG. 2. The relationship between head mass as a percent of headless body mass and headless body mass in the satur­ niids Rothschildia lebeau and O/horene purpurascens. R. le­ beau (r ... - 0.88, y '" 19.2 - 6Ax); O. purpurascens (r " - 0.80. y '" 22.5 - 6.6x ). FIG. 4. Mandibles of (a) ROIhschiidia lebeau inslar V ven­ tral view sligh tly opened out and (b) Eades imperia{is inslar HyleSia lineala, and Eades imperialis; Fig. 4), in which V ventral view separated and in elosed position. Scale line 1 mm. there are no strong grooves or ridges though there may be slight serrations on the very sharp and hard man­ dible edge. The apparently hard edge of one mandible during cutting. This general simplicity was character­ works against a wide region on the inner face of the istic of the Saturniidae. but there were variants. In opposite mandible; this face becomes worn and rough­ Arsenura armida, fo r example, a relative specialist on ened in contrast to the rest o f the mandible. Increased Bombacopsis quinatum, the cutting edges were semi­ wear thus tends to sharpen the edge of the blade. Pre­ circular blades with a short overlap. The outer man­ sumably this rough surface aids purchase of the leaf dible was stopped at a clear-cut ridge on the outer surface of the other, while the inner mandible fitted , , into an irregular groove on the inner surface of the ..• , other (Fig. Sa). In Olhorene purpurascens, which feeds • • , on Manillw.ra chicle, the mandibles were very globular • 12 '. ~*,p ' , ~• • 0 in shape (Fig. 5b), with each having a double edge. 0 " d, 0 m o:' ~ ., , While in all cases there is some asymmetry in sa­ • tumiid mandibles, at closure the left may overlap the '0 ,. ~ , 0 ~ . ' 0 • '\. , right or vice versa, and in the species obscrved closure alternated between the two positions. Either way, the • "$:x/o • 4 • 'j,< mechanisms for cutting appeared similar, with the sharp • , " hard edge of one mandible fitting closely to the inner ~• face of the other. The action appears to be ofa simple x• - 1 0 2 3 snipping device or scissor action, or blade against an Lo, Bod y M ass anvil. However, the first and second instarsofsaturniid caterpillars are usually of the more generalized cater­ FIG. J. The relationship between head mass as a percent of headless body mass and headless body mass in 12 sphingid pillar pattern (Snodgrass 1935) with a simple row of species (r ... - 0.82, y - 11.7 - 2Ax). x ... Manduca dilucida; 4-8 teeth around the curved cutting edge (Fig. 6). o ... Pachylia ficus; * ... Manduca sex/a; I ... Pachylioides Sphingids had very different mandibles from those resumens; 2 '" PrO/(Jmbulyx s/rigilis; J ... Eumorpha sa/elli/a; described above. There was much variation between 4 ... Eupyrrhoglossum sagra; 5 ... Enyorxype/e; 6 " Manduca flores/an; 7 ... Perigonia lusca; 8 ... £rinnyis eflo; 9 ... Aeflopos species but a general panern emerged. Mandibles were fadus; 10 ... Manduca rus/ica; II ... Manduca lanuginosa. longer with narrower bases. Distally there were various --- the regression line for Satumiidae. grooves and teeth, Manduca sexta being the simplest. 11 56 E. A. BERNA YS AND D. H. JANZEN Ecology, Vol. 69, NO.4 '@QP~

~. / Q)0-/

FlO. 6. Mandibles of RQ/hschildia lebeau inSlar I. Scale line I mm.

b little variation (Fig. 9). In any individual on a single host plant the small standard deviation in particle size was notable while the overall pattern of size change with instar was quite consistent (Table 2). Also the pieces were of simple shape (Fig. 10), which resulted in relatively small perimeters. On the other hand, par­ ticles in sphingid guts were very small, were extremely variable and irregular in shape, and had relatively larg­ Flo. 5. Mandibles of (a) Arsenura armida instar V and (b) er perimeters (Fig. 10; Table 3). In spite of the different Othorene purpurascel1S instar V. Scale lines I mm in each food plants and the many species examined, the pat­ ~~. terns found in the two fa milies were consistently quite The most distal region of the mandible had 2- 3 irreg­ different. ular rows of sharp-edged projections, while the inner face had a series of irregular ridges and grooves (Fig. DISCUSSION 7). Other species displayed variations on this theme, The two groups of caterpillars processed leaf blades with heavy and broad-based teeth forming a spiked in two quite different ways. Satumiids simply snipped club at the distal region and the inner faces being var­ iously grooved and ridged. Pachylia ficus was one of the most extreme, and, like most, there was a dorsal .'m ;'. ";r region with a fine serrated edge (Fig. 8). , ", \? As with the saturniids, the closure could occur with {} left over right or right over left, usually alternating with ([j • b successive bites. In either case the teeth covering the distal region of one mandible fi t tightly into grooves on the inner face of the other. The action appeared 10 be one of crushing the blade fragment after it was cut or tom from the leaf. While sphingid mandibles had a basic si milarity, the development of the parts varied: no two spedes had identical mandibles. Perhaps each species has a slightly different style oftearingand crush­ ing related to the nature of its host leaves. Of the species examined, Enyo ocypete was the most elltreme (Fig. 7), with the teeth and grooves so reduced as to more resemble the saturniid type. The width of the mandibles (the distance across the ~ base of the right mandible) in a number of different caterpillars of different species in different instars, was ~ compared with the area ofa sample of foliage particles FlO. 7. Left mandibles of five sphingid species from var­ ious anlero-ventl1l1 angles: (a) Manduco lanuginoso instar V, taken from the gut of the same individual. Among (b) Cocylius duponchel instar V, (c) Pachylioides resumel1S salUrniids the mandible width was directl y related to instar V, (d) Protambu/yx Slrigilis. and (e) Enyo ocypele. Scale the particle size bitten off and swallowed, with very line I mm. August ]988 HOW CATERP]LLARS EAT LEAVES 11 57 off piecesorthe blade. This produced panicle sizes that .6 were closely correlated with mandible width and rel­ atively invariant. Different sptties or satumiid cater­ • pillars of the same instar snipped off pieces of leaves • , • o r about the same size (even though they were eating • "E , different species orJeaves). Ir a single sptties of satur­ E •• .. , • ni id caterpillar, such as Rothschi/dia lebeau or Eacles , ~ imperialis, fed on a numberofdifferent species ofieaves, - , ' • .!I.E",,, the same consistency among instars occurred. We fee l , "~ • that these should be viewed as havi ng snipping • • • rather than chewing mouthpans, since there appears " .2 " , ~ • to be no further mechanical processing of the food after "' • • it is bitten of[ Such simple snipping behavior suggests • , " u .' "- , 3 !.. , ~ a simple control mechanism; the entrance of the leaf 2 ~ ,~ , to a certain point in the mouth causes a biting response, -" ~ , ~ L ~• .!!.~ rollowed by swallowing. , ~ ~ 4. i~~! The newly eclosed first-instar satumiids examined •• .8 1.2 .6 in this study eat the same tough and mature leaves as Mandible Width (mm) do the later instar larvae (e.g., Janzen 1984). Funher­ Flo. 9. Relationship between sizeorchewed rood pankles more thei r hosts are almost all trees, many of which and mandible widlh. Numbers - salumiids; letters wi lh un­ are evergreens with exceptionally thick and tough leaves derlinin8 - sphingids. Sa\llmiids (r - 0.88, y - 0.07 + 0.38x): (e.g., Hymenaea cQurbaril, Manilkara chicle, Quercus I - Olhor,.n,. purpurasc,.ns on Manilkara chide; 2 - Roth­ oleoides). As with the grnss-feeding saturniids in the schildia I,.beau on Casearia corymbosa: 3 - Rothschildia 1,.­ beau on Zuelania guidonia; 4 - ROfhschiidia l,.beau on SPQn­ southwestern United States, the necessary power for dias mombirr, 5 - Syssphinx mo/inaon Pithecellobium soman; biting through tough leaf tissue must come from large 6 - Arsenura armida on Bombaeopsis quinalum; 7 - Hylesia mandibular adductor muscles and heavily sclerotized lineala on Casearia corymbosa: 8 - lIyl,.s;a lineala (food not recorded): 9 - ROlhschildia I,.beau on £xQSI,.ma m,.xicanum. Sphingidae: A - £up}'rrhog/ossum sagra on Chom,./ia spi· nosa; B - Manduca rustica on Amphilophi/um panieulawm; C - Manduca flor,.stan on Cydista h,.t,.rophylla; D - Man· auca lanuginosa on Cydista heterophylla: E - Piich),lioides r,.sumens on ForSI,.ronia spica/a; F - p,.rigonia fusca o n Ca· • Iycophyllum candidis:rimum: G --Pachylia ficw on Chloro­ phoro linclOria; H - Manduca71i1ueida on Tabebuia ochracea; I - Cocytius duPonch,.1 on Annona r,.tieulala; J - Erinnyis ello on Sebastiana con/wum; K - Manduca difUcida on Sa· , pranthus palanga; L - Pach}·ffQ ficus on Ficus cotini[o/ia; M - Eumorpha sat,.lIhaon Cissus sicyoid,.s: N - Manduca S,.xta on Solanum tuberO$um; P - Poehylia 7icus (food not re­ corded); Q - Pach}1ioid,.s-r,.sum,.ns (food not recorded); .!! - Enyo ocype/e on Tetracera l"OIubilis.

mandibles (Bernays 1986). The consequence is that b newly hatched larvae have an enormous relative head mass; while still in the egg, the head capsule appears to take up most of the egg volume. This may be why satumiids lay eggs that generally have 2- 3 times the volume of the eggs of sphingjds with the same adult body mass. Not only is the head mass relatively large, but the first-instarcaterpill ars are themselves relatively large (e.g., first-instar E. impen'alis caterpillars prior to feeding are 5-7 mm long, with heads upto 2 mm wide). The minimall y processed si mple leaf discs that are swall owed by satumiid caterpillars appear to pose a digesti ve challenge. The onl y plant tissue readily avail­ able fo r rapid digestion or re moval of nutrients is that around marginsofthe leaf piece. A caterpillar has noth­ ing analogous to the gizzard ofa bird, and passage rates FlO. 8. Mandibles of Pachylia ficus inS\3.r V showing (a) anlem-venlra] view. (b) inner faces. and (c) mode of closure. are measured in hours, which does not leave time for Scale li ne I. S mm. macrodegrndation by microflora. The only tissue that 11 58 E. A. BERNAYS AND D. H. JANZE N Ecology. Vol. 69. No. 4 locking jagged surfaces crunch the pieces into smaller pieces and punClUre the cuticle. Because of their shape the mandibles produce something that is much closer to true chewing (mastication). However, we do not know if a sphingid bites more than o nce on any given leaf disc. Observation of the feeding process suggests a single bite per disc. The particles in the sphingid caterpillar gut are ell­ tremely varied in size and shape, including some fi ­ brous particles that have been tom off the leaf. Whole mounts of gut material show onl y a small proportion of the original leaf blade to be intact. Sphingid cater­ pillar fecal pellets are also a packed mass of extremely small and unrecognizable mushy tissue, and are easil y b distinguished from the wads of leaf discs defecated by satumiid caterpillars. The striking contrasts in vari­ ability ofleaf particle area (Table 2) were probably even underestimated in this study; the sphingid guts contain a slurry of fine cellular plant material that was not measured or included in the particle-size analys is be­ cause it graded into the indeterminately minute. The mashed and pulveri zed nature of the sphingid gut con­ tents could be partly created by digesti ve kneading,

FIG. 10. Drawings of food particles from the midgut of (a) a satumiid. ROlhschifdia /rlJeau instar V (scale line 2 mm), TABLE 2. Represen tati ve enmpJes of individua l ca terpil­ and (b) a sphingid, Pach}'lia ficus instar V (scale line I mm). lar's rood panicle areas and the coefficient or variation (cv) of those areas (Sa tumiidae). was conspicuously removed from the discs was that Par- around the disc margins. Indeed, the fecal pellets of ticle satumi.id larvae are simply tightly packed wads of al- .~. A~ Insect and instar Food pla nt (m m') most morphologicall y intactleafdiscs. This means that "" the larger the pieces snipped off, the lower the pro- fly/esia lineara II and Cascaria CQ- . 10 .19 V rymbosa .49 .40 portion of the food that is in a form such that its nu- A,ynura armida V /J()moocops/s .55 .70 trients are rapidly available to the larva. This places a quinalum .S) .78 constraint on the size of the piece to be bitten off, which RQ/hschifdia /ebeau Cos('O.r/a c0- in our data is reR ected in the relati vely small heads of II rymbosa .109 .46 III . t43 .35 large satumiid larvae; during caterpillar development, IV .18 .23 the relative head mass changes from :::;25% of body V .42 .36 mass to :::; I % of body mass. We predict that satumiid V .'0 .57 caterpillars will be found to spend proportionately large II Spondias mOIll- .06 .4, III bin .18 .38 amounts of time cutting off and swallowing large IV .22 .31 amounts of leaf tissue. Whether this will also lead to V .4S .52 disproportionate increase in gUI transit time will de- li EXQslema mex- . 14 .28 pend on the relati ve yields fro m a small a mount of /cono .39 .23 IV .39 .51 processing of much tissue vs. a large a mount of non- I'"V .38 .38 mechanical processing of less tissue. Many other fam - V .56 .67 ilies of caterpillars also feed on tough leaves. A prelim- II Zulania gui- .08 .40 inary survey (E. A. Demays, personal observation) donia .13 .34 '"IV .20 .37 suggests that these, including grass specialists. also ap- V .38 .30 pear to have the simple mandibular shape, snipping Olhorene pllrpurasce1lS Monilkara chi- action, and constant leaf-disc size descri bed here fo r I d, .037 .10 satum iids. II .081 .07 .101 .07 The sphingid caterpillar mandi ble shapes are very '"IV .179 .0' different fro m those of the satumiids. The varied and V .4' . 11 complex array of mandibular teeth and ridges grasp Syssphinx molina V Pilhecellobium .50 .44 the sphingid's somewhat softer food and ro ughly tear saman V it away (rather than cleanl y snip it away). The inter- Cassia grandis .4S .33 August 1988 HOW CATERPILLARS EAT LEAVES 1159

TAlIU 3. Examples of individual caterpillar's food panicle areas and the coefficient of variation (cv) of those areas (Sphin­ gidac).

Panicle: area Insect and instar Food plant (m ml) Area cv Manduca sexla V Solanum IUberosum .08 1 1.03 Protambulyx sirigilis IV Spondias mombin .053 1.19 EupyrrhogfOS$um sagra V Cholt/elia spinosa .044 1.58 Eumorpha satellila IV Cissus sicyoides .118 0.737 Enyo ocypete- V Telracera vofubilis .341 0.559 Manduca floreSlan V Cydisla heterophyffa .102 1.870 Manduca ruslica V Amphifophifum panicuialum .165 1.0 Manduca lanuginosa V Cyriisla heterophyffa .06' 1.06 .OS7 1.90 lIfanduca difucida V Tabebuia ochracea .04 1.3 IV .0' 0.' HI Sapramhus palanga .0' 0.76 Pachylioides reSU II!ens IV ForSleronia splcala .04' 0.98 V .044 1.21 V .049 1.1 7 V .039 1.57 Perigonia lusca IV CalycophyllulI! candidissimum .082 0.97 Pachyliaficus V Chlorophora linctoria .012 1.33 IV Ficus cotinifolia .0 11 1.23 HI .010 1.12 Cocytius duponchel lV Annona reticulata .06' 1.36 Erinnyis e/lo IV Sebasliana con/usum .076 1.1 2 • This species has "satumiid type" mandibles. since once a leaf disc has been ripped and broken, fTUm (tough evergreen leaf blades), and Ficusspp. (thick turbulent digestive movements can mechanically break but fragile and nearly evergreen leaf blades), has the it down further (as opposed to the small impact of such most massively destructive mandibles. movemenison intactleafdiscs in a satumiid caterpillar It is possible that the species differences in mandible gut). shape simply indicate different ways to maximize the The large sphingid caterpillars have relati ve head rate at which food can be ingested on the particular masses almost double those of large satumiid cater­ foliage utilized. Specialized mandible shapes may thus pillars. Sphingid digestion should not be hampered by be more obvious in species with narrow host range, as increasing the initial bite size as the larva increases in appears to be the case. The virtue of rapid ingestion size. We found that sphingid caterpillars with large rate depends on the ye t unknown selection for reduced heads produced particlesjusl as small as did those with time spent feeding, a pressure sometimes postulated to small heads. Pachyliajicus. the sphingid with the most be imposed by visuall y hunting predators. complell mandibular teeth, had the smallest food par­ The sphingid method of feeding may represent a ticles in its gul, yel it had the largest relati ve head mass quite different method ofcircumventing plant chemical of any spe<:ies. defenses than that which is used by the satumiid cat­ Sphingid host leaves range from extremely flim sy to erpillars. The sphingid way of processing leaves creates relatively tough (Janzen 1984). However, many, ifnot a soup in the gut, one in which the nutrients and the all, of the first instars of the sphingids examined here other chemicals are potentially in direct contact with fed on very new leaves, leaves that were delicate and each other and the caterpillar gut tissues (and free­ thin. Almost all sphingid hosts in the study area were ranging gut nora). The satumiid host plants are re­ deciduous and had relatively flim sy leaves. Finall y. nowned for having foliage rich in phenolics including many species of sphingids feed on relatively herba­ tannins (Janzen 1984. Janzen and Watennan 1984) ceous plants, which also have very flimsy leaves. Even and not conspicuous in the phytochemistry literature the muscle mass in a very small head capsule can drive as producers of directly toxic small molecules. Sphingid sphingid macelike mandibles to triturate such leaf host plants (e.g., Rubiaceae. Apocynaceae, Euphorbi­ blades. As sphingid larvae become larger, they incor­ aceae. Solanaceae, Bignoniaceae, Asclepiadaceae, porate both old and new leaves in their diets. It is Moraceae, Sapotaceae, Lauraceae). however, can easily striking that the sphingid larva that eats the toughest be characterized as rich in toxic small molecules and leaf blades, Enyo ocypete feeding on Tetracera \'olubilis are nOt famous for production of tannins (Janzen 1984). (a nearly evergreen vine), has the most satumiid-like We hypothesize that the sphingid caterpillar feeding mandibles and leaf fragments in its gut. On the other on a particular species of plant is explicitl y resistant to hand, the sphingid that feeds on the greatest variety of the toxic chemicals in that plant, and therefore can ieaftypes, Pachylia ficus feeding on Chlorophora tine­ thoroughl y triturate the leafso as to get the maximum toria (leaf blades like tissue paper), Brosimum alicas- amount of nutrient from it. If there are also phenolics 1160 E. A. BERNA YS AND D. H. JANZEN Ecology. Vol. 69, No.4 in the foliage, these will be present in such low amounts cussed here (D. H. Janzen. personal observation). but that they do not interfere with this mechanism, and preliminary results strongly suggest that a sphingid cat­ may even contribute to the detoxification process by erpillar can accumulate dry mass almost twice as fast binding with toxic molecules. as can a satumiid caterpillar of the same size. Such a digestive mechanism implies that sphingids will be largely host specific, to one or a few closely ACKNOWLEDGMENTS related plants, which they are (Janzen 1984). By being This study was supported by NSF grants BSR 8516813 to E. A. Bemays: NSF BSR 83-07887. BSR 84-03531 , BSR 83- lightly host specific, there is the opportunity for the 08388, and DEB 80-1 1558 to D. H. Janzen: and by the Ser­ evolution of colors, morphologies, and behaviors that vicio de Parques Nacionales de Costa Rica. Caterpillar col­ are themselves extremely protective vis-a-vis the spe­ lection was aided by W. Hallwachs, R. Espinosa, A. Espinosa, cific host plant. Satumiid caterpillars, on the other hand, C. Chapman. F. Joyce, R. Evans, M. Johnston, L. Kellogg, and I. Gauld. The manuscript was constructively reviewed being able to feed on plants with more varied or less by W. Hallwachs. species-specific "defenses" (as long as the defenses largely stay put within the untriturated leaf discs). find LITERA T1JRE CITED themselves on a variety of backgrounds. This leads to Bemays, E. A. 1986. Diet-induced head allometry among a selective regime favoring individual defenses, such foliage-chewing inseets and its importance for gramini· as urtication and its mimicry, that function in a wide vores. Science 231 :495-497. variety of circumstances (Janzen 1984). Janzen, D. H. 1984. Two ways to be a tropical big moth: Santa Rosa satumiids and sphingids. Oxford Surveys in Thorough trituration of leaf fragments before the Evolutionary Biology 1:85-139. digestive process should lead to a greater amount of Janzen, D. H., and P. Waterman. 1984. A seasonal census nutrient removal by the caterpillar per amount of leaf of phenolics, fibre and alkaloids in foliage of forest trees in consumed, as compared with the satumiid digestive Costa Rica: some factors influencing their distribution and relation to host selection by Sphingidae and Satumiidae. process with intact leaf discs as substrate. This may Biological Journal of the Linnean Society 21 :439-454. lead to either less leaf consumption or faster growth Snodgrass. R. E. 1935. Principles of inseet morphology. by the sphingid than by the satumiid caterpillar. Both McGraw-Hill. New York, New York, USA. topics are under examination with the species dis-