Study area - Samples were taken at six sites located along the Caracu and the São Pedro streams (Fig. 1). These streams, affluents of the Rio Paraná eastern side, are second and third order respectively according to the classification by Jeffries and Mills (1990). Their mouths are about two km apart. The Caracu is about 5.2 km long, springs at the 320 m contour line, and has degraded shoreline forest limited to a few sites. The São Pedro is about 19.2 km long, springs at the 340 m contour line, and has poorly developed riparian forest.
The Rio Paraná water level, the seasons' characterization (based on water level data), and the pluviometric data for the region during the study period are shown in Fig. 2.
The spatial and temporal variation of some limnological parameters of the streams has been described by Pavanelli et al. (1997). They found that there was a reduction of variability (increase in the buffering capacity) along the upper-downstream direction. They also suggested that the total-P was the variable with the highest coefficients of variation.
Caracu. - The sampling station at the headwater was located at the place where the stream has been artificially widened to about 10 m and 0.95 m mean depth, creating a lentic situation. The sieves were passed longitudinally for 20 m at this site. The intermediary site was a semi-lotic environment, also artificially widened to about 10 m wide, with a mean depth of 0.70 m. There, the sieves were passed for about 15 m. The mouth, with a mean width of 8 m and mean depth of 1 m, was sieved for about 30 m. This was a lotic environment where water level fluctuated highly, influenced by the Rio Paraná level mainly during rising and high water seasons.
São Pedro. - The site in the headwater had a mean width of 2 m and depth of about 0.20 m and was sieved for 130 m. The intermediary site, about 1.60 m deep and 3 m wide, had fast current and high banks. Its physiography did not allow the use of sieves at that sampling site. The mouth, about 25 m wide and 2.00 m in mean depth, had fast flowing water and was sieved for 30 m.
Sampling and data collection - Samples were taken monthly from March 1991 through February 1992. Each stream was sampled at three sites, at its headwater, its mouth, and at an intermediary site near the mouth (Fig. 1). Due to physiographic differences on the sites, Caracu was sampled only by sieves and São Pedro by sieves at the headwaters, by gill nets at the intermediary site and at the mouth by both of the fishing gears. Two collectors used two sieves at the same time, and they were passed in the shallow waters, mainly under the aquatic macrophytes for 40 minutes. The standard lengths in millimeters of the specimens captured with sieves were measured, after species identification. At the São Pedro intermediary site and at the mouth, gill nets, 10 m long and mesh sizes of 3, 4, 5, and 6 cm were set up. The first two gill nets were 1.20 m high, and the last two were 1.50 m high. At the mouth, an additional trammel net, 10 m long and 1.50 m high with 6 cm mesh was also set up. The nets were placed at sunset and checked twice, at night and at dawn. Nupélia/Finep (1989) collected data on the occurrence and constancy of the species in the Rio Paraná. These data were obtained by different methods and at different times than the present study, and are included here only for qualitative comparison purposes. The specimens taken with nets, after identification, had their standard length measured in millimeters and total weight in grams, and were dissected to identify their sex and gonadal stage. Pavanelli (1994) gives a more detailed analysis of the reproductive biology of the most abundant species. Pavanelli and Caramaschi (1997) provided a taxonomic list and discussion of the ichthyofauna composition of both streams. Regarding this taxonomic list, in the present paper Astyanax bimaculatus (Linnaeus, 1758) = A. altiparanae Garutti and Britski, 2000; Aphyocharax nasutus Ahl, 1936 = A. cf. anisitsi Eigenmann and Kennedy, 1903; Cheirodon notomelas Eigenmann, 1915 = Serrapinnus notomelas (Eigenmann, 1915); Cheirodon sp. = Serrapinnus sp.; Hyphessobrycon callistus (Boulenger, 1900) = H. eques (Steindachner, 1882); Jobertina sp. = Characidium sp.; Myloplus levis (Eigenmann and McAtee, 1907) = M. tiete Eigenmann and Norris, 1900; Parodon tortuosus Eigenmann and Norris, 1900 = P. nasus Kner, 1859; Auchenipterus nuchalis (Spix, 1829) = A. osteomystax (Ribeiro, 1918); Hypostomus aff. derbyi (Haseman, 1911) = H. ancistroides (Ihering, 1911); and Rhamphichthys rostratus (Linnaeus, 1766) = R. hahni (Meinken, 1937).
Data analysis - The abundance of each species was determined by the CPE (catch per effort). Sieved fishes are expressed in number of individuals caught by sieve in 6 hours, and the netted ones are expressed in number of individual caught by 1,000 m2 of gill nets in 12 h. The species whose relative abundance was higher than five percent were considered abundant.
The constancy was calculated from the frequency with which each species was recorded in the samples, constancy (c) being the relationship expressed as the percent c=(p x 100)/P, where p was the number of samples containing the species in question and P was the total number of collections (Dajoz, 1983). A species was considered constant when this percentage exceeded 50%, accessory between 25 and 50%, and occasional below 25%.
The cumulative frequency was obtained from the cumulative occurrence of the species in the collections for the entire sampling period.
The Shannon-Wiener index (H') was used to calculate species diversity, based on the formula H'= (pi)(log2pi), where S was the number of species collected and pi was the proportion of the total sample belonging to the species i. Evenness (E) was calculated from the equation E=D/DMAX, where D was the observed index of species diversity and DMAX was the maximum possible diversity index, given S species and N individuals. The calculations were processed with the aid of Divers® software, based on the formulas described above and proposed by Krebs (1989).
Correspondence analysis was calculated for sieved fishes to achieve summaries of the patterns in species-sites relationships, as suggested by Jackson (1997). The software used for that was NTSYS-PC®, version 1.50. Similarity (I) was calculated for netted fishes from Sørensen index (Legendre and Legendre, 1983), the formula for which is I=(2a)/2a+b+c, where a is the number of species common to two sites, b is the number of species exclusive to one of the sites being compared, and c is the number of species exclusive to the other site. Fishes collected in Rio Paraná by Nupélia/Finep (1989) were included for comparison in constancy and similarity analysis, both based on qualitative data, only for netted fishes. The Shannon-Wiener index and the correspondence analysis were calculated on CPE data.