Caudal Eyespots As Deterrents Against Fin Predation in the N Eoptropical Cichlid Astronotusocellatus

Copeia, 1990(3), pp. 665-673

Caudal Eyespots as Deterrents against Fin Predation in the N eoptropical Cichlid Astronotusocellatus

KIRK O. WINEMILLER

Astronotus ocellatus and several other large cichlid fishes of South America exhibit bright ocelli, or eyespots, near the base of the caudal fin. Astronotus ocellatus sympatric with fin-nipping piranhas of the genus Serrasalmus shows less extensive fin damage that sympatric cichlids of similar size that lack distinct caudal ocelli. Ecological data and observed interspecific behavioral differences support the hypothesis that eyespots reduce piranha attacks by confounding visual recognition of the prey's caudal region. Dense scalation of the unpaired medial fins, especially the caudal fin, masks visual cues for fin recognition, further enhancing head mimicry by the caudal region. The hypothesis that caudal ocelli in A. ocellatus, and perhaps other large cichlid fishes, function primarily as eye mimics for defense against fin predators is more parsimonious than Zaret's earlier "inhibition of cannibalism" hypothesis.

F ISH fins are a principal food resource for closely (Sazima and Zamprogno, 1985; Nico and many characids of the South American Taphorn, 1988; Winemiller, 1989b). Fishes genera Serrasalmusand Pristobrycon(Machado- captured from lowland aquatic habitats shared Allison and Garcia, 1986; Northcote et al., 1986; by Serrasalmusspp. commonly show extensive Nico and Taphorn, 1988; Winemiller, 1989b). fin damage, particularly around the caudal I:e- Fins provide a seemingly abundant, rapidly region (Northcote et al., 1986; Nico and Tap- newable food resource for these diverse pira- horn, 1988; Winemiller, unpubl.). nhas. Fin-nipping is most prevalent among sub- Fin-nipping must have a serious negative im- adult and small adult size classes of most pact on growth and survival of prey. For ex- Serrasalmus spp. that have been investigated ample, damaged fins of captive fishes are more

@ 1990 bv the American Societv of Ichthvoloilists and Hernetolo.nsts 666 COPEIA, 1990, NO.3 susceptible to bacterial and fungal infection MATERIALS AND METHODS (Reichenbach-Klinke and Elkan, 1965). Fin damage adversely affects swimming perfor- Studysites.-Most specimens used in the present mance, which in turn hinders feeding and es- analysis came from Caiio Maraca, a swamp/ cape efficiency (Sazima and Pombal, 1988). In creek of the Rio Apure drainage of the llanos addition, fin damage impacts the fish's energy in the state of Portuguesa, Venezuela. Full de- budget negatively, since matter and energy scriptions of the Caiio Maraca environment and would be diverted away from gonadal devel- methods of collection are given by Winemiller opment and overall somatic growth, and di- (1987, 1989a, 1989b). Fishes were collected rected toward fin regeneration. Even the dom- monthly during 1984, preserved in 15% for- inance status of individuals is affected by visual malin, and later transferred to alcohol. Speci- cues associated with fin morphology in some mens were collected at less regular intervals species,which in turn can influence mating suc- during 1984 from Caiio Maporal, a broad shal- cess (Baerends and Baerends von Roon, 1950; low creek of the low llanos of Apure state (7~7'OO'N, 69°32'OO'W), and preserved in the Barlow, 1974; Baylis, 1974). Large, slow-moving cichlids with broad me- same manner. Additional collections were made dial fins are conspicuous and potentially vul- at each site during Jan. 1988 and 1989. Both nerable targets for fin-nipping piranhas. This Caiio Maraca and Caiio Maporal exhibit dra- report offers evidence that Astronotus ocellatus matic changes in response to seasonal rainfall and perhaps several other large neotropical (Winemiller, 1989a, 1989b). During the dry cichlids gain a measure of protection against fin season, there is little significant discharge, and predation via caudal mimicry of the head re- water at each site is confined to the main chan- gion. This head mimicry is achieved by three nel. Fish densities are highest at both sites dur- morphological features: 1) the caudal ocellus; ing the dry period (Oct.-April). Heavy rains 2) dense, opaque scalation of the medial fins; from May-Sept. cause extensive flooding of and 3) symmetry of the lateral profile. The cau- broad, low-lying floodplains. Aquatic macro- dal ocellus, or eyespot, of A. ocellatusconsists of phyte and periphyton production is high and a bright orange ring of scales surrounding a many fishes spawn during this period. Caiio black spot at the dorsolateral base of the caudal Maraca has four piranha species:Pygocentrus no- fin. Except for the conspicuous ocellus and or- taus, S. irritans, S. medini, and S. rhombeus(Wine- ange iris of the eyes, the coloration of A. ocellatus miller, 1987, 1989b; see below). Caiio Maporal is drab and cryptic. Depending on the fish's be- has at least eight piranha species:Catoprion men- havioral state and environment, body color- to, Pristobrycon sp;, S. altuvei, S. elongatus, plus ation can quickly change from uniform grey to the four species at Maraca (Nico and Taphorn, a mottled pattern consisting of dark grey and 1988). Specimens were deposited in the TNHC light brown to grey patches. Less than half of and MCNG (institutional abbreviations are as 1% of wild A. ocellatusexhibit two or more ocel- listed in Leviton et al., 1985). li, usually near the base of the dorsal fin or sides of the body (Winemiller, pers. obs., n > 250 Measurements.- Diets of all fish speciescollected over 5 yr period in Venezuelan llanos). These at Callo Maraca were quantified by volumetric multiple ocelli are usually incomplete (some- stomach-contents analysis. Fish prey were iden- times consisting of only a black spot without the tified to species whenever possible. Identifica- orange ring) and apparently anomolous. Except tions were performed using a dissecting micro- for the brilliant caudal ocellus, wild A. ocellatus scope. Fishes were measured (all reported from the Venezuelan llanos are very drab com- lengths are SL) and fins were examined for re- pared with selectively bred color varieties recent piranha nips. Fins damaged by narrow- ferred to as "oscars" in the international aquar- bodied serrasalmin piranhas show distinct char- ium trade. Here I argue that most vertebrates acteristics. First, the distal fin margin within the probably perceive the caudal ocellus as an eye, damage area is usually crescent or rectangular much in the manner suggested for the ocellus in shape. Fins damaged by larger red-belly p.ira- of the eel-mimicking coral reef fish CalloPlesiops nhas, Pygocentrusnotatus, are much broader in altivelis (McCosker, 1977), marine butterfly fish- relation to the depth of the nip than fins nipped es, Chaetodonspp. (Cott, 1957), and eyespots on by Serrasalmusspp. Nips from large P. notatus the wings of certain moths and butterflies (Blest, were extremely rare compared to Serrasalmus 1957). nips. Presumably, large P: notatus nips result WINEMILLER-CICHLID CAUDAL EYESPOT RR7

Fig. 1. Caquetia kraussii (left; 170-180 mm) and Astronotusocellatus (right; 170-210 mm) collected on the same date from the same habitat in the Venezuelan llanos. Extensive damage from fin-nipping piranhas is evident on C. kraussii, whereas A. ocellatussuffered only minimal damage to its dorsal and anal fins. Interspecific differences in patterns of caudal ocelli are apparent even in preserved museum specimens. from failed attempts at tearing out pieces of transparent and frequently exhibits abrupt dis- flesh rather than fins alone. However, some evi- continuity with previous soft-ray formations. dence suggests that juvenile P. notatus may at- Fins damaged by other agents, such as intra- tack fins rather than flesh (Machado-Allison and specific aggressive encounters among cichlids Garcia, 1986; Nico and Taphorn, 1988). Some or failed attacks by grasping predators show specimens captured by hook and line were dam- ragged borders rather than smooth edges. aged by P. notatusbefore they could be removed Recent fin nips were counted on all fins of A. from the water (evident from large bleeding ocellatus and Caquetia kraussii, another large wounds). cichlid found at Cano Maraca. Both specieshave Second, Serrasalmus spp. produce clean, a caudal ocellus, but that of A. ocellatus is far abrupt tears at the edge of damaged fins. Aquar- more symmetrical and brightly contrasting with ium observations revealed that three species (S. the fish's background coloration than that ofC. irritans, S. medini, and S. rhombeus)can instantly kraussii (Fig. 1). Specimens smaller than 40.0 clip off pieces of fin tissue with their razor sharp mm were not included in the analysis.Astronotus teeth. Tight occlusion of the teeth apparently ocellatus is widespread throughout lowland facilitates abrupt tearing of fins (some juvenile aquatic habitats of the Orinoco and Amazon Serrasalmusused a jerking motion to dislodge drainage basins. As a result it usually coexists fin fragments). sympatrically with several fin-nipping Serrasal- Regenerating fins retain unambiguous visible mus spp. evidence of the initial extent of fin damage. Caquetiakraussii ranges from the Lake Mara- Newly rejzenerated fin tissue is usually more caibo and coastal drainag:es of Venezuela into 668 COPEIA, 1990, NO.3

undamaged distal tissue on either side of the .c I/) nip. Nips were not measured if bordered by ii: other recently damaged or partially regener- Q; ated tissue. Nips located on the edge of rect- ll- angular fin margins were not measured when l/) Co undamaged distal tissue only remained on one Z side of the nip. Fin nips were also measured on Q) 01 preserved specimens of Crenicichla saxatilis LU Q; (Cichlidae) from Callo Maraca and Cichla ocel- > ct laris (Cichlidae) collected from Callo Maporal. Fin nips were classified as either narrow (depth of nip greater than width), broad (nip width greater than depth), or intermediate (nip depth and width roughly equal, resulting in a semi- circular notch). Based on piranha jaw shapes, broad nips were assumed to correspond to attacks by P. notatusat Callo Maraca. Narrow nips were assumed to correspond to various size classesof S. irritans plus small size classesof S. rhombeus.Intermediate nips were assumed to correspond to S. medini almost exclusively, since

- the other possibility, adult size classesof S. rhom- 40-79 80-119 120-159 160+ beus, was rare at Callo Maraca (sample n = 1 Length Interval (mm) during spawning period). Fig. 2. Differencesin frequencyof nips to caudal Relative scalation of unpaired medial fins was fins (top) and all other fins (bottom) between Astron- measured on two adult specimens of the follow- DeUSocellatus (n = 76) and Caquetia kraussii (n = 123) ing cichlids collected from the two sites: A. ocel- collected from the Venezuelan llanos with piranhas. latus, C. kraussii, C. ocellaris, Crenicichla saxatilis, Vertical lines represent :t 1 SE. Nip frequencies in C. lugubris (Callo Maporal), GeoPhagussurina- top graph showed significant main effects of species mensis(Callo Maporal), G. jurupari (Callo Ma- (two-way ANOVA; F,.,.. = 47.8; P <0.0001), size poral), Aequidens pulcher (Callo Maraca), and interval (F'.l.' = 28.4; P <0.0001), and interaction effect of species with size (F..,.. = 23.2; P <0.0001). Cichlasomaorinocense (Callo Maraca). Relative fin scalation was defined as the ratio of the length Nip frequencies in bottom graph showed significant main effects of species (F'.l" = 11.0; P <0.01) and of the scaled portion of the longest fin ray di- size interval (F.,l" = 38.2; P <0.0001), but no inter- vided by the total length of the ray (nearest 0.1 action effect of specieswith size (F.".. = 2.3;P =0.073). mm). Long filaments on dorsal and anal fins Differences between species (ANCOV A with size in- were not used for this measure. terval as the covariate) were significant in both comparisons(Fl.'" = 35.7,29.2; P <0.0001). RESULTS AND DISCUSSION Three species of fin-nipping Serrasalmus (S. northern tributaries of the Rio Orinoco in the irritans, S. rhombeus,and S. medini) and the red- western llanos; it is not currently found in the belly piranha, P. notatus, were collected during low llanos (Apure state) and may be a relatively the high-water period at Callo Maraca (June- recent invader of the high llanos, where it is Dec.). Fins were prevalent in stomachs of all sometimes a dominant species of local fish as- three Serrasalmusspp., whereas both whole fish- semblages (D. C. Taphorn, pers. comm.). This es and fish fragments dominated diets of larger cichlid was formerly listed as Petenia kraussii P. notatus (Winemiller, 1989b). Astronotus ocel- (Mago-Leccia, 1980), but was more recently latus and another large cichlid, Caquetiakraussii, considered a separate genus endemic to South were collected year-round at Callo Maraca (1984 total sample n = III and 588, respectively). . America (Kullander, 1982). The depth of fin nips was measured (nearest 0.1 mm) on adult size classesof A. ocellatusand Fin damage.-No fin damage was observed in a C. kraussii (> 120 mm). Nip depth was measured sample of 93 specimens of A. ocellatus and C. as the distance from the most proximal edge of kraussii taken during 4 mo of the harsh dry the damaged area to a line running between season when fin-nipping piranhas were absent WINEMILLER. .CICHLID CAUDAL EYESPOT 669 from Caito Maraca (Winemiller, 1989a). During TABLE 1. PERCENTAGEOF N IPS OF THREE SHAPESDE- the remainder of the year, both the frequency LIVERED TO VARIOUS FINS OF Astronotus ocellatus (180 and magnitude of fin nips was much greater for NIPS) AND Caquetia krausii (530 NIPS). Chi square tests C. kraussii than A. ocellatus(Figs. 1-2). Astronotus for association between cichlid species with frequen- ocellatushad significantly fewer caudal-fin nips cies of different nips, and cichlid species with fre- than C. kraussii for the 40-79 mm size interval quencies of fins nipped were not significant (P > 0.05, (2-tailed t-test = 4.5, df = 31, P < 0.0001), the for all cichlid size classes combined). 120-159 mm interval (t = 4.6, df = 4, P < Inter- 0.025), the 160+ mm interval (t = 4.6, df = 4, Narrow mediate Broad Total P < 0.01), but not the 80-119 mm interval. Except for the 40-79 mm size interval (t = 2.9, Astronotusocellatus df = 31, P < 0.01), lower nip frequencies for Caudal 27.8 22.8 0 50.6 A. ocellatuson fins other than the caudal were Dorsal 9A 12.2 1.1 22.7 Anal 8.3 7.8 0.5 not statistically significant (Fig. 2, lower panel). 16.6 Pectoral 6.7 1.1 0 7.8 The average depth of fin nips was significantly Pelvic greater for C. kraussii (x = 12.6 mm, SD = 5.3) 1.7 0.5 0 2.2 than A. ocellatus (x = 5.4 mm, SD = 2.1; t-test Total 53.9 44.4 1.6 100 = 8.59, P < 0.0001). Caquetia krausii Larger cichlids tended to suffer greater dam- Caudal 34.3 18.7 1.5 54.5 age from fin predators than smaller conspecif- Dorsal 8.5 6.6 0.5 15.6 ics, and interspecific differences in nip frequen- Anal 11.1 4.5 0.5 16.1 cies appeared most pronounced in the largest Pectoral lOA 2.1 0 12.5 size intervals (Fig. 2). Size interval 40-79 mm Pelvic 1.3 OA 0 1.7 A. ocellatushad significantly fewer nips to both Total 65.6 32.3 2.5 100 the caudal and other fins than 80-119 mm (t = 6.1, 5.2; df = 21; P < 0.0001) and 120-159 mm con specifics (t = 2.8, 3.1; df = 13, P < cichlids was 445:251 (1.8:1). The ratio of S. ir- 0.025). The largest A. ocellatussize classactually Titansand small S. rhombeusto S. medini collected averaged fewer caudal nips than intermediate in seine samples at Callo Maraca was n = 126: conspecific size intervals; however, statistics were 68 (1.8:1), which suggests equivalent feeding not significant due to small sample for the 160+ rates of each type of fin-nipping piranha on mm group (n = 5). Size 40-79 mm C. kraussii cichlids. had significantly fewer nips to both caudal and other fins than 120-159 mm (t = 6.3, 5.7; df= Fin predation/head mimicry hypothesis-Based on 28; P < 0.0001) and 160+ mm conspecifics (t observations of piranha feeding behavior (Nico = 7.9, 5.8; df= 13; P < 0.0001). Size 80-119 and Taphorn, 1988; Winemiller, 1989b), cich- mm C. kraussii had significantly fewer caudal lid swimming performance, and cichlid diet data nips than size 120-159 mm (t = 2.3, df = 13, from the field (Winemiller, 1987, 1989b), A. P < 0.05) and 160+ mm con specifics (t = 3.7, ocellatus should be more exposed to predation df = 13, P < 0.01). Size 160+ mm C. kraussii by fin-nipping piranhas than C. kraussii. Astrono- had significantly more nips to fins other than ius ocellatusswims slowly with smooth, seeming- the caudal compared with intervals 80-119 mm ly uninterrupted movement as it examines roots (t = 4.4, df = 12, P < 0.01), 120-159 mm (t = of floating aquatic macrophytes for cryptic in- 4.0, df = 13, P < 0.01), in addition to 40-79 sectsand fishes. This foraging behavior is easily mm conspecifics mentioned above. observed in aquaria containing water hyacinths Approximately half of all fin damage was to or other floating plants. Astronotus ocellatus (n = the caudal fin for both cichlids (Table 1). Paired 176, range 40-210 mm) consumed 19% fishes, fins received little damage relative to unpaired 28% aquatic insects, 30% terrestrial insects, and fins in each case. The two cichlids did not differ 3% crustacea at Callo Maraca. The fishes con- significantly in the proportion of nips to various sumed were all highly cryptic, relatively sed- fins or shapesof fin nips (Table 1). Both species entary catfishes. Three of these small catfishes were attacked most frequently by narrow-snout- (BunocePhalusamaurus, Rineloricaria caracasen- ed piranhas (S. irritans and small S. rhombeus), sis, and Ochmacanthusalternus) were commonly followed by round-snouted S. medini, and only captured from the roots of floating aquatic plants rarely by broad-snouted P. notatus. The ratio of of Callo Maraca. Presumably, the large bulging total narrow nips to total intermediate nips on eves of A. ocellatusoermit flood visual discrim- 670 COPEIA, 1990, NO.3 ination of fine-scale structural features of the incompletely sheathed with tiny transparent environment during foraging for cryptic fishes scales.If the mode of foraging of C. kraussii (i.e., and invertebrates. scanning the environment for midwater fishes) Caquetia krausii, on the other hand, is pri- facilitates visual detection and thus avoidance marily a sit-and-wait, rapid-pursuit predator of of fin predators, then selection may be acting small midwater fishes; it (n = 370, range 40- more weakly on the caudal spot as an eye mimic. 180 mm) consumed 43% fishes, 26% aquatic Also, C. kraussii may have a relatively short his- insects, 15% terrestrial insects, and 6% crusta- tory of ecological association with piranhas if, cea. Small, midwater-dwelling characids (As- as postulated, it is a recent invader of the llanos. tyanax bimaculatus,Odontostilbe pulcher, Ctenobry- While data to support the hypothesis of re- con sPilurus, and Hemigrammus sp.) were the cent association are not presently available, the dominant fishes consumed (Winemiller, 1989b). detection/avoidance hypothesis can be tested Aquarium-housed Caquetia kraussii were fre- by comparing fins and ocelli of cichlids of dif- quently observed to ambush small moving prey. ferent sizes.Smaller, deeper-bodied fishesshould The pointed conical teeth and highly protru- have greater abilities for performing evasive sible jaw apparatus of C. kraussii are used to suck darting maneuvers along acute angles or for in and grasp small prey organisms. concealment in dense vegetation. The relative As a sit-and-wait forager, C. kraussii may have degree of caudal-fin scalation is generally great- more opportunities to detect and avoid ap- er for larger llanos cichlids (Fig. 3). Moreover, proaching fin-nipping piranhas than Astronotus highly symmetrical yellow, orange, or red cau- ocellatus,the latter frequently being positioned dal ocelli are found only in four of the longest with its snout amid floating vegetation and its cichlids. As already noted, C. kraussii has a rath- caudal region exposed in the water column be- er weakly developed caudal ocellus, while the neath. As a consequence, the caudal region of smaller GeoPhagusspp., Aequidenspulcher, and A. ocellatusis probably more exposed to fin-nip- Cichlasoma orinocensehave small, dark caudal pers than that of C. kraussii. spots and no ocellus. Two deep-bodied, highly The distinct caudal ocellus of A. ocellatus maneuverable dwarf cichlids (Apistogramma probably functions as an eye mimic (Blest, 1957; hoignei and MicrogeoPhagusramirezi) commonly Wickler, 1968). Ocelli of living specimensclosely collected from dense vegetation in the llanos match both the size and color of the iris of the have small, asymmetrical caudal spots but no true eye. In addition, the position of the ocellus ocellus. The dark caudal spots of small diurnal closely matches the relative height of the eye fishes of numerous other freshwater families in lateral profile (Fig. 1). The combination of may serve a variety of adaptive functions: 1) near complete scalation of the unpaired medial confusion or deflection of predator attacks by fins (producing an opaque, dull gray coloration drawing attention away from the head; and 2) that matches the head and body) with smooth- intraspecific communication (Barlow, 1972; margined, overlapping dorsal, caudal, and anal Lagler et al., 1977; Lowe-McConnell, 1987). fins results in symmetry of the fish's lateral pro- If a simple caudal spot enhances survival by file (Fig. 1). Caudal mimicry of the head region reducing the successrate of predatory strikes, is probably most effective under twilight con- the addition of conspicuous color in the form ditions, a period when fin-nipping piranhas are of an ocellus would only reduce survival by in- particularly active. In addition to enhancing creasing the rate of detection by predators. Al- head mimicry, dense scalation of fins probably ternatively, caudal mimicry of the head region eliminates many visual cues used by fin-nipping could reduce the rate of predation, if more con- piranhas for recognition of appropriate food ventional, whole-fish swallowers sometimes in- resources (e.g., transparency, fin rays, a discrete correctly anticipate the path of escape to be border separating fin from body). taken by a motionless prey organism (Cott, In contrast, the caudal ocellus of C. kraussii 1957). In addition to concealing vulnerable fin is asymmetrical and matches the golden back- tissues, the false head of large piscivorous cich- ground color of the body more closely than the lids would yield the appearance of a predation iris, which can change between gold and crim- threat to small fin-nipping piranhas, thus ills:' son. The most distinctive feature of the caudal couraging many exploratory approaches. Ex- marking of C. kraussii is the large black spot, perimental work on captive fishes could shed rather than the bright ring of scales surround- light on these alternative possibilities. ing the spot in A. ocellatus.The unpaired medial Other possibilities exist for lower incidence fins of C. kraussii are largely unpigmented and of fin-nips on Astronotus ocellatusin addition to WINEMILLER. -CICHLID CAUDAL EYESPOT 671

the head-mimicry hypothesis. For example, the ~. AnoICaudal fins of A. ocellatus may contain distasteful sub- Clchlasoms orinocense(112} m Donal

stances that discourage attacks following initial Geophagua }urupari(150)

sampling and learning by piranhas. In addition, Geophagua surinamenala(I'.} the abundant scales on fins may increase the . C..nlclchla saxall11S(175}

piranha's energetic foraging cost by hindering Caque/la k..uasl/(223) removal of fin fragments. Because some indi- .C..nlclchla lugubris(253} vidual A. ocellatus had multiple nips and nips were generally as abrupt and smooth-edged as nips on other llanos cichlids, these explanations 0.0 0.2 0.4 0.6 0.8 1.0 seem unlikely. Alternatively, A. ocellatus could RelatlvaFIn Scalatlon be more adept than Caquetia kraussii at pro- Fig. 3. Relative scalation of anal, caudal, and dor- tecting their caudal regions. As described pre- sal fins for cichlid fishes collected from two sites in viously, foraging behavior of captive cichlids the Venezuelan llanos inhabited by fin-nipping piranhas. Relative fin scalation was measured as the ratio suggestsjust the opposite for situations in which of the length of the scaled portion divided by the total piranhas remain undetected. Upon detection of length of the longest fin ray. Asterisks denote species an approaching piranha, both cichlid species with a distinct, brightly colored caudal ocellus. Values appeared capable of effectively repelling attacks in parenthesis are maximum standard lengths among with frontal displays (see description from ex- material examined. perimental trial below). Additionally, A. ocellatus might detect fin- predators more effectively than C. kraussii via that indiscriminate predation on conspecifics social cooperation. Both cichlids form pair bonds reduces lifetime fitness. This assumption relies during reproductive bouts and brood protec- on the additional assumption that offspring re- tion that may last several weeks (Winemiller, main within the home range of their parents 1989c). A dense school consisting of over 300 following the period of brood defense. More- juvenile A. ocellatus measuring 35-55 mm was over, inhibition of cannibalism implies that re- captured at Cano Maraca during Aug. 1984. duced mortality from callibalism more than off- The relatively large size of these schooling ju- sets the increase in mortality caused by higher veniles, which were assumed to be siblings, in- rates of detection by other diurnal piscivores dicates a 2-3 mo period of parental care. By caused by the conspicuous ocellus. This is un- comparison, C. kraussii pairs were never cap- likely in native environments in the Amazon tured with offspring larger than 15 mm. Captive and Orinoco river basins, where other species groups of A. ocellatusas large as 130 mm exhib- of diurnal piscivores far outnumber C. ocellaris ited tight schooling behavior when disturbed (Goulding, 1980; Taphorn and Lilyestrom, by human observers. Yet, the same individuals 1985; Winemiller, 1987). showed little hesitation in confronting solitary In addition to the questionable validity of these S. irritans with frontal displays. Although social assumptions, the primary assumption that C. facilitation remains a viable possibility for de- ocellaris never exhibits cannibalism is based al- fense against fin-predators, it appears somewhat most entirely on data from two populations lo- unlikely pending further evidence. cated outside of its native range. 1 know of no other large, freshwater piscivore for which a Other neotroPicalcichlids with caudal ocelli.- The similar scenario has been proposed. Zaret (1977) head-mimicry hypothesis probably applies to the presented data that indicated a lower integrity large piscivore Cichla ocellarisand Crenicichlaspp. of the caudal ocellus pattern in C. ocellaris from as well. Based primarily on data taken from a an exotic Panamanian population compared recently established population outside the nat- with native Amazonian populations. The dif- ural range of the species(Zaret and Paine, 1973), ference was hypothesized to be a result of re- Zaret (1977; Schroder and Zaret, 1979) pos- laxed selection on inhibition of cannibalism tulated that cannibalism in Cichla ocellaris is in- during the initial phase of explosive population hibited via the ocellus, which functions as a sig- expansion (Zaret, 1977). Following the fin-pre- nal for species recognition. This evolutionary dation hypothesis, the same data can be attrib- hypothesis relies on the assumption that C. ocel- uted to relaxed selection on the ocellus as an laris encounters its own offspring as potential eye mimic in the absenceof fin-nipping piranhas prey at frequencies sufficiently high in relation in the exotic setting. to encounters with unrelated individuals. such Zaret (1977) theorized that A. ocellatusand 672 COPEIA, 1990, NO.3

TABLE 2. PREY ITEMS TAKEN FROM STOMACHS OF = 4.8 nips per fin; all fins = 8.0 nips per fish), Cichla ocellaris FROM THE Low LLANOS OF ESTADO the average depth of 50 fin nips on 14 Apure ApURE, VENEZUELA (N = 10). specimens was 5.9 mm (SD = 3.0), a valuelower - - than that obtained for C. kraussii (t-test = 7.86, Volume Proportion P < 0.0001) but not A. ocellatus(t-test = 0.94, Prey N (ml) (Vol.) P > 0.05). Crenicichla saxatilis (Cich. Comparisons of fin damage on other llanos lidae) * 1 11.0 0.30 fishes suggest that cichlids are more vulnerable Pellona sp. (Clupeidae) 2 7.5 0.21 Poptella orbicularis to fin predation than native ostariophysan fishes (Northcote et al., 1986; Sazima and Pombal, (Characidae) 2 3.3 0.09 Geophagussurinamensis 1988; Winemiller, unpubl.). To test this hypothesis, an experiment was conducted for 3 d (Cichlidae) 3.3 0.09 PygocentTUs notatus using four specimens of S. irritans (30-160 mm), (Characidae) 2 3.0 0.08 each housed in a separate aquarium with a mix- Curimatella inmaculata ture of 16 characiform, siluriform (catfishes), (Curimatidae) 2.4 0.07 and cichlid prey (>25.0 mm). Cichlids were at- Anostomus cf trimaculatus tacked most frequently (11 nips recorded on (Anostomidae) 2.0 0.05 three Aequidenspulcher; no nips on one Creni- Cichlasoma festivum (Cich- cichla geayi with a bright caudal eyespot), fol- lidae) 1.7 0.05 lowed by catfishes (14 nips on 17 individuals MicrogeoPhagus ramirezi representing eight species), and characins (18 (Cichlidae) 0.8 0.02 nips on 42 individuals representing seven Eigenmannia virescens species). Predation attempts by the piranhas (Sternopygidae) 0.7 0.02 were most frequent during twilight conditions Unidentified fish 0.4 0.01 in the aquaria. Megalamphodus sweglesi To test the head-mimicry hypothesis, exper- (Characidae) 0.15 <0.01 iments using S. irritans and manipulatuions of . Species having distinct caudal ocellus. ocellus markings on cichlids were set up in large concrete pools during the 1989 llanos dry season. The experiment was terminated prema- turely after the grease paint used for masking Crenicichla spp. have evolved ocelli as mimics of and enhancing Astronatus ocellatusand Caquetia the intraspecific recognition signal of Cichla ocel- kraussii ocelli wore off. During the trial, both laris as a means to avoid predation by the latter. cichlids frequently were observed to turn and Sixteen specimens of C. ocellaris from the low repell piranhas with frontal displays (i.e., mouth llanos were examined for stomach contents (Ta- open, opercula and branchiostegal membranes ble 2). A single, 85 mm Crenicichla saxatilis with flared) when the latter approached or hovered a bright caudal ocellus was the most important near their caudal regions. Persistent piranhas prey consumed by volume in this small sample were sometimes driven off with charges and bites (stomachs of six individuals were empty). The by both cichlids. Experiments of similar design single occurrence of C. saxatilis with a bright using models or another marking technique eyespot within the stomach of one of 10 spec- could test the fin-predation/head-mimicry hy- imens of Cichla ocellaris offers no support for pothesis. Zaret's hypothesis, whereas ingestion of two ju- Fin damage inflicted by piranhas is undoubt- venile piranhas (Pygocentrusnotatus) illustrates a edly a major selective pressure on large, slow- predation threat to fin-nippers posed by the an- moving fishes that inhabit lowland aquatic hab- terior region of the large piscivore. itats of South America. Conspicuous ocelli in A. Both Crenicichla saxatilis and Cichla ocellaris ocellatusand certain other large cichlids prob- appear to suffer comparatively low levels of fin ably serve as false eyes for caudal-head mimicry predation. Twenty-three specimens of Creni- that partially deters fin-nipping. Although not cichla saxatilis collected with piranhas as Callo mutually exclusive, the fin-predation/head- Maraca averaged only 0.26 nips per fish (all mimicry hypothesis explains the function of caudal). Although the average number of nips caudal ocelli on neotropical cichlids more par- per individual of Cichla ocellaris was not signif- simoniously than Zaret's inhibition of canni- icantly lower than the value obtained for Ca- balism hypothesis, because it invokes fewer in- quetia kraussii in the earlier comparison (caudal terrelated assumptions. WINEMILLER-CICHLID CAUDAL EYESPOT 673

ACKNOWLEDGMENTS MAGO-LECCIA, F. 1980. Lista de los peces de Vene- zuela. Ministerio de Agricultura y Cria, Caracas, D. C. Taphorn, L. G. Nico, A. Barbarino, L. Venezuela. K. Winemiller, P. Rodriguez, and E. R. Pianka MCCOSKER, J. E. 1977. Fright posture of the ple- provided technical assistance in the field. Spe- siopid fish CalloPlesiops altivelis: an example of cial thanks go to D. C. Taphorn and la Univer- Batesian mimicry. Science 197:400-401. sidad de los Llanos Occidentales Ezquiel Za- NICO, L. G., AND D. C. T APHORN. 1988. Food habits mora for providing a base of operations in of piranhas in the low llanos of Venezuela. Biotro- Venezuela. A collecting permit was obtained pica 20:311-1321. NORTHCOTE, T. G., R. G. NORTHCOTE, AND M. S. from the Direccion Administraccion y Desarro- ARCIFA. 1986. Differential cropping of the caudal llo Pesquero of the Republic of Venezuela. I fin lobes of prey fishes by the piranha, Serrasalmus thank E. R. Pianka, L. K. Winemiller, M. Mor- sPiloPleura Kner. Hydrobilogia 141:199-205. ris, and C. Hubbs for helpful comments on the REICHENBACH-KLINKE, H., AND E. ELKAN. 1965. The manuscript. I thank R. L. Beach for assistance principal diseases of lower vertebrates. Academic in preparation of the manuscript. Field work Press, New York, New York. was funded by the National Geographic Society SAZIMA, I., ANDJ. P. POMBAL,JR. 1988. Mutilacao and the Denton A. Cooley Professorship (E. R. de nadadeiras em acaras, Geophagus brasiliensis, por piranhas, Serrasalmus spilopleura. Rev. Brasil. BioI. Pianka). 48:477-483. -, AND C. ZAMPROGNO. 1985. Use of water hy- LITERATURE CITED acinths as shelter, foraging place, and transport by young piranhas, Serrasalmus spilopleura. Environ. BAERENDS,G. P., AND J. M. BAERENDSVAN ROON. BioI. Fish. 9:87-101. 1950. An introduction to the study of the ethology SCHRODER,S. L., AND T. M. ZARET. 1979. The adap- of cichlid fishes. Behaviour suppl. 1: 1-243. tive significance of color patterns in Cichla ocellaris. BARLOW,G. W. 1972. The attitude offish eye-lines Copeia 1979:43-47. in relation to body shape and to stripes and bars. TAPHORN,D.C.,ANDC.LILYESTROM. 1985. Lospeces Copeia 1972:4-12. del modulo 'Fernando Corales'. Resultados ictiolo- -. 1974. Contrasts in social behavior between gicos del proyecto del investigacion del CONICIT- Central American cichlid fishes and coral-reef sur- PIMA 18. Rev. Cie. Tecn., UNELLEZ 3:55-85. geon fishes. Am~r. Zool. 14:9-34. WICKLER, W. 1968. Mimicry in plants and animals. BAYLIS,J. R. 1974. The behavior and ecology of McGraw-Hill Book Co., Inc., New York, New York. HerotilaPia multisPinosa(Pisces, Cichlidae). Z. Tier- WINEMILLER, K. O. 1987. Tests of ecomorphological psychol.34:115-146. and community level convergence among neotrop- BLEST,A. D. 1957. The function of eyespot patterns ical fish assemblages. Unpubl. Ph.D. dissert., Uni- in the Lepidoptera. Behaviour 11 :209-255. versity of Texas, Austin, Texas. COTT, H. B. 1957. Adaptive coloration in animals. -. 1989a. Development of dermal lip protu- Methuen and Co. Ltd., London, England. berances for aquatic surface respiration in South GOULDING,M. 1980. The fishes and the forest. Uni- American characid fishes. Copeia 1989:382-390. versity of California Press, Berkeley, California. -. 1989b. Ontogenetic diet shifts and resource KULLANDER,S. O. 1982. A revision of the South partitioning among piscivorous fishes in the Ven- American cichlid genus Cichlasoma(Teleostei: Cich- ezuelan llanos. Env. BioI. Fish. 26:177-199. lidae). Swedish Museum of Natural History, Stock- -. 1989c. Patterns, of variation in life history holm, Sweden. among South American fishes in seasonal environ- LAGLER,K. F.,J. E. BARDACH,R. R. MILLER AND D. ments. Oecologia 81:225-241. R. PASSINO.1977. Ichthyology, 2nded.John Wiley ZARET, T. M. 1977. Inhibition of cannibalism in Cich- & Sons, New York, New York. la ocellaris and hypothesis of predator mimicry LEVITON, A. E., R. H. GIBBS,JR., E. HEAL AND C. E. among South American fishes. Evolution 31:421- DAWSON.1985. Standards in herpetology and ich- 437. thyology: part. I. Standard symbolic codes for in- -, AND R. T. PAINE. 1973. Species introduction stitutional resource collections in herpetology and in a tropical lake. Science 182:449-455. ichthyology. Copeia 1985:802-832. LOWE-McCONNELL,R. H. 1987. Ecological studies DEPARTMENT OF ZOOLOGY AND TEXAS MEM- in tropical fish communities. Cambridge University ORIAL MUSEUM, UNIVERSITY OF TEXAS, Press, Cambridge, England. AUSTIN, TEXAS 78712. Accepted 24 July MACHADO-ALLISON,A., AND C. GARCIA. 1986. Food habits and morphological changes during ontogeny 1989. in three serrasalmin fish speciesof the Venezuelan floodplains. Copeia 1986:193-195.