CANNIBALISM AS THE MAIN FEEDING BEHAVIOUR OF TUCUNARES INTRODUCED IN SOUTHEAST BRAZIL

Individuals of its own genus were the main food item of two species of tucunares (Cichla cf. ocellaris and Cichla monoculus) introduced into the Volta Grande Reservoir. The abundance of adult tucunares may cause intra-specific competition, possibly leading to the high cannibalism rates found.


INTRODUCTION
The Volta Grande Reservoir, on the Rio Grande River, is located in São Paulo and Minas Gerais states (48 o 25' and 47 o 35'W, 19 o 57' and 20 o 10'S).Approximately 25 years ago the dam was closed.The reservoir, with an area of 221.7 km 2 and a water volume of 2,268 km 3 , is used to produce electric energy.The climate is typically tropical with a mean annual temperature of 22°C and mean annual rainfall of 1,635 mm.The two seasonal periods are: warm and rainy from October to April, and cool and dry from May to September (Braga & Gomiero, 1997).The reservoir has a large abundance of tucunares, Cichla spp.Bloch & Schneider, 1801, a fish from the Amazon basin with marked caudal ocellus.
The presence of exotic fishes species presents a serious problem because resources previously available for native fishes are scarce.This leads to competition and predation, mainly among the piscivorous species, and even causes local species extinction (Alves & Vono, 1997).The Cichla species need clear water since they are visually oriented diurnal predators (Winemiller, 2001).
Cannibalism can be controlled poligenetically as it is a complex behavior phenomenon, acting frequently as a control mechanism in fish populations (Chevalier, 1973;Thibault, 1974).It occurs as a response to local condition influencing particular populations, and its frequency and intensity may vary substantially inside and among species (Fox, 1975).Cannibalism intensity is inverse to availability of alternative preys (Polis, 1981), and cannibal fishes consume a high-quality diet, which chemically is very similar to their own constitution (Wootton, 1990).
Fishes providing parental care may feed upon part of their brood.This occurs due to weight loss in periods presenting little foraging opportunity (Halliday, 1988).
The objective of this work was to measure the intensity of cannibalism in two introduced species of tucunaré, Cichla cf.ocellaris Bloch & Schneider, 1801 and Cichla monoculus Spix & Agassiz, 1831, in Volta Grande Reservoir, on the Rio Grande River.

MATERIAL AND METHODS
We sampled the fish in April and June of 1997 and monthly from September 1997 to August 1998, using two types of gear: gill nets and angles.Each sampling period lasted five days.
The gill nets had mesh sizes of 2.0, 3.0, 4.0, 5.0, and 6.0 cm (between adjacent knots).Each net was 150 m long and 1.5 m high, totaling 750 meters for all nets.Natural or artificial baits were used in angling, resulting in the capture of a greater number of specimens.
The caudal ocelli were measured along their greater axis.This length was correlated with total fish length, using the alometric equation y = a x b (Vanzolini, 1993).
Stomach content was identified and preserved in 70% alcohol.When possible we also determined the number, weight, and length of every food item.The preys were analyzed with the following indexes (Matallanas, 1980(Matallanas, , 1982a, b, c, b, c): • numeric index (Ni): proportion of number of preys of one taxonomic group as a percentage of the total number of preys; • frequency index (Fi): proportion of the number of stomachs with preys of a determined taxonomic group as a percentage of total number of stomachs with food; • frequency in weight (Wi): proportion of preys' weight by taxonomic group as a percentage of the total weight of preys.
The feeding coefficient (Q) was calculated with Ni and Wi by the formula: Q = Ni • Wi.This coefficient is classified in three categories: Q more than 200 (preferential prey), Q between 20 and 200 (secondary prey), and Q less than 20 (occasional prey) (Braga & Braga, 1987).
When the Q values of the preys were classified as preferential or secondary, the relative importance index was used: This index indicates the real importance of each taxonomic group in the feeding habit of the fish in question (Hyslop, 1980).
The variation of feeding related to length was analyzed by length class, and for each class the average weight of the most important preys was calculated (AW).

RESULTS
In the period from April 1997 to August 1998, using gill nets and fishhooks we caught 459 C. cf.ocellaris and 170 C. monoculus.
The relation between ocellus length and total fish length was negative allometric in C. cf.ocellaris and isometric in C. monoculus (Table 1 and Fig. 1).
In cases where the food item Cichla was found well preserved, it could be identified by scale counting, and caudal ocelli size.Cannibalism was evident in these cases.
After also separating items representing the species P. squamosissimus and T. rendalli, occurrence frequency revealed the dominance of the food item Cichla sp. in both species and for all length classes (Figs. 4 and 5).
In C. cf.ocellaris the item P. squamosissimus was important in the size class from 42 to 50 cm (Fig. 6).There were no differences among food items and size class for C. cf.ocellaris, according to the twoway ANOVA with single observation, applied for the data.In C. monoculus the item T. rendalli was important in the size class from 34 to 42 cm (Fig. 7).

DISCUSSION
The ocelli of both fishes grew differently, being allometric negative to C. cf.ocellaris and isometric to C. monoculus.The allometric negative growth of ocelli in C. cf.ocellaris was relatively less than the total length of the fish, however the isometric growth of ocelli of C. monoculus was relatively proportional to fish length.The adaptive meaning of the ocellus to Cichla monoculus must be stronger than that of Cichla cf.ocellaris.Zaret (1977) and Winemiller (1990) hypothesized that the ocellus could help in peer recognition during pair formation or as a defense against predation: the ocellus in the tail makes the head hard to distinguish According to these authors, Zaret's first hypothesis would not be valid, because with fish size increase the ocellus tends to lose its shape and even to disappear, although reproductive activity continues normally.The second hypothesis appears to lack validity because fishes with this adaptation generally have a strip obscuring their real eyes (Barlow, 1972), which does not occur in tucunare.A third hypothesis, presented by Zaret (1977), is that coloration and the ocellus developed in the tail by juvenile tucunares (Cichla ocellaris) just after the end of parental care, is an important sign for intraspecific recognition and prevents cannibalism.Two other hypotheses were also proposed.Arcifa & Meschiatti (1993) found in tucunare stomachs only specimens that did not yet present caudal ocelli, and Lowe-McConnell (1969) suggested that the ocellus in Cichla would serve for brood orientation during parental care.Zaret (1977), in analyzing 300 stomachs of Cichla ocellaris from Gatun Lake, found no evidence of cannibalism.This author suggested that the following conditions prevent cannibalism: 1) the territoriality of a given couple, 2) juveniles of the same area are offspring of a single couple, and 3) abundance of prey.In addition to this, when juveniles become independent, shelter is found in the aquatic vegetation.Young stay in these shelters until reaching 18 to 20 centimeters, when predation pressure is lower.The caudal ocelli serve as interand intraspecific signs inhibiting predation and thus would serve as another factor reducing cannibalism.As reported, in environments lacking submerged vegetation, cannibalism occurs (Arcifa & Meschiatti, 1993;Fox, 1975;Santos et al., 2001).
Our data indicated that Cichla sp. was the most important item in the diet of both species considered.In the sampling site, a high index of predation was found upon the genus itself (Cichla), particularly following reproduction peaks.This result disagrees with the ocellus function proposed by Zaret (1977).Better explanations for this cannibalism are: 1) the great number of adult specimens that are not predated, 2) low abundance of prey coupled with a great number of young specimens of the same genus, and 3) lack of aquatic vegetation to shelter the juveniles.Cannibalism reduces the number of intraspecific competitors.In populations where cannibalism is frequent, greater stability exists and extinction probabilities are low.However, there is a cost.If fishes prey upon their own brood, by eliminating potential partners they reduces their own reproductive chances (Chevalier, 1973;Fitzgerald, 1992;Fox, 1975;Polis, 1981).
In environments where tucunares have been introduced, high cannibalism rates can occur (Durães et al., 2000;Santos et al., 1994).They also occur, but at lower rates, in natural habitats of this genus (Jepsen et al., 1997;Winemiller, 2001).Bedarf et al. (2001) compared two lakes; into the first one was recently introduced a predatory cichlid whereas, in the second, this species was native.In the first lake, population growth was exponential and there was little cannibalism due to prey abundance.In the second, the species found few preys and practiced large-scale cannibalism.According to the authors, in the first lake as the preys became rare and intra-and interspecific competition increased, a tendency to cannibalism developed.In Guri Reservoir (Venezuela), into which the tucunare had previous been introduced, cannibalism was evident in samples from 1985 to 1989, when fishes of this genus were very abundant (Gil et al., 1996;Lasso et al., 1990;Novoa, 1996;Novoa et al., 1990).However, fish caught from 1993 to 1994 did not present cannibalism, probably because of reduced tucunare abundance (Williams et al., 1998).

T. rendalli
Average w eight (g) . value of the intercept; B = angular coefficient value; a = transformed value of the intercept; S b = b deflection; p 0.05 = probability of b being larger or smaller than 1; R = correlation coefficient; N = number of data pairs; s 2 yx = regression variance; F 0.05 = result of variance homogeneity test.

Fig. 1 -Fig. 2 -
Fig. 1 -Linear regression referring to the ocellus length compared to the total length of the Cichla cf.ocellaris and Cichla monoculus.