Ontogenetic , spatial and temporal variations in the feeding ecology of Deuterodon langei Travassos , 1957 ( Teleostei : Characidae ) in a Neotropical stream from the Atlantic rainforest , southern Brazil

Information related to the diet of one species always contributes to the knowledge of its bionomy and the functioning of the ecosystem in which the species lives. Therefore, understanding the feeding ecology of one fish population and its ontogenetic, spatial and temporal aspects help to understand the structure of fish assemblages and river communities. Knowledge of this structure is essential for habitat management and biodiversity conservation. The feeding ecology of Deuterodon langei Travassos, 1957 was studied through analyses of diet composition, sharing of resources, feeding strategy and contribution of food items to the width of its niche. The analysis included an assessment of ontogenetic, spatial and seasonal variations. The species was considered omnivorous with great plasticity caused by seasonal variation in food availability throughout the river basin, but mainly through ontogeny. The diet of smaller individuals revealed a predominance of insects and other arthropods, while the diet of larger specimens showed allochthonous plant items as the main components. The low intestinal quotient (IQ) values for the smaller individuals were accounted for their mainly insectivorous diet. The greater relative length of the intestine can account for the greater plasticity of the adult diet, enabling them to use diet items of larger size and more difficult digestion, such as plant items. Therefore, shifts in feeding ecology during growth are confirmed in this omnivorous tropical river fish, strengthening the concept that, when comparing food habits among species, fish size must be taken into account. This study also provides needed information regarding the feeding ecology of fishes intrinsically associated with the Atlantic rainforest, one of the most threatened and biodiverse ecosystems of the planet.


Introduction
The Atlantic Forest is one of the richest and most threatened ecosystems of the planet, and is considered one of the main biodiversity hotspots for conservation priorities (Myers et al., 2000).Several rivers flow through this ecosystem, which contain many endemic fish species (Menezes, 1996).The "Serra do Mar" is a 3000-km mountain ridge covered by the Atlantic Forest which extends along the Brazilian coast.Its network of littoral streams has been poorly investigated (Mazzoni & Lobón-Cerviá, 2000).These streams are among the first environments to suffer from anthropic impacts, potentially threatening their fish fauna (Menezes, 1996;Faria & Marques, 1999;Vitule et al., 2006).
Deuterodon langei is considered a resident of deep or shallow areas, lotic or lentic waters, and various types of substrates (Aranha et al.,1998;Fogaça et al.,2003;Barreto & Aranha, 2005).Studies on D. langei diet had recently been carried out in littoral streams of Paraná State.The fish has been shown to be omnivorous/herbivorous (Aranha et al.,1998;Barreto & Aranha, 2006) and insectivorous, with a predominance of aquatic insects in its diet (Fogaça et al.,2003).A trend toward omnivory with decrease in animal items and increase in plant items has been shown to occur during ontogeny (Vitule & Aranha, 2002).
All the previous work on D. langei had a local focus, using fish sampled from a single site, which is a major problem in understanding its biology in the whole draining basin, considering that some important differences can be found along the same river (Vannote et al.,1980;Allan, 1995;Mazzoni & Lobón-Cerviá, 2000).Therefore, biological characteristics such as feeding ecology interfere directly with fish population dynamics and constitute basic information.The aim of the present observational field study was to analyze the ontogenetic, spatial and temporal variations in the feeding ecology of D. langei in a Neotropical stream of the Atlantic Forest.Understanding the role that this important fish species plays in structuring fish assemblages and all the river community through their feeding ecology is essential for habitat management and biodiversity conservation (Lowe-McConnell, 1999), especially in such important but threatened and fragile Neotropical ecosystems.

Material and Methods
Study area.The study was conducted in rio Ribeirão basin (25°36'S; 48°37'W), east side of the Serra do Mar, sub-basin of Paranaguá bay, Paranaguá city, which is a portion of Atlantic rainforest in Paraná State, Brazil.The spring is located at Serra da Prata, 766 m above sea level.It is a short basin (< 40 km) that drains directly into the Paranaguá bay (Braga et al., 2008;Vitule et al., 2008).
This region has a tropical, sub-equatorial, super humid climate without dry season and frost.In general, annual rainfall rates are above 1000 mm, and average annual temperatures range between 17º C and 21º C (IAPAR, 1978;Maack, 1981).Rainfalls are distributed differently during the year; January and February were the rainiest months, and July was the driest month during the investigation period (Fig. 1).
Like other coastal clear water streams in the Atlantic rainforest, rio Ribeirão is dependent on the high rainfall of the area; it displays characteristics of rain waters, neutral pH and low level of nutrients.In the rainy season (summer), the river is subject to flash floods, significantly increasing water volume and flow which occurs suddenly, causing great alterations in physical and biological parameters due to the water drag (Por, 1992).
In general, sample sections were 2 -5 m wide, and the depths were between 10 cm deep in the shallower margin, to 1.80 m deep in central deeper portions.The bottom was mainly composed of sand, with some presence of clay, leaves, gravel, rocks, and tree trunks and branches.The vegetation of the river bank is composed partially of submerged shrubs, trees and grass which were usually covered with an algal layer.Aquatic macrophytes were rare in the area studied.Most of the sampling sites were very shaded, and there were some beach formations.The main characteristics of each of the three sampling sites are presented in Table 1.
Fish sampling and data analyses.Fish were sampled monthly from February of 2002 to February 2003.The samples were obtained using sieves, a small trawl net of 1.30 x 1.40 m with 2mm mesh, and five gillnets of 1.50 x 5 m, three with 15-mm mesh and two with 20-mm mesh between consecutive knots.To gain a better knowledge of feeding tactics and strategies utilized by the species, some sporadic in situ observations were made using free diving techniques.
Samples were always taken during daytime, since personal underwater observations and information from the literature (Fogaça et al., 2003;Sabino & Castro, 1990;Barreto & Aranha, 2005) indicated that D. langei were active during daytime.Still, in the field, specimens were immersed in 10% formalin with the objective of stopping the digestive process, and after 48 h they were transferred to 70% alcohol for better conservation and to facilitate handling during dissection.In the laboratory, Lt (total length to 0.01 cm) was measured, and fish were eviscerated.Voucher specimens were deposited in the Museu de História Natural Capão da Imbuia, Curitiba, Brazil: MHNCI 11698.
After evisceration, the intestines were removed and measured (total length to 0.01 cm) to determine the intestinal quotient (IQ = intestine length / total body length) sensu Knöppel (1970) and Zavala-Camin (1996).The IQ values were statistically evaluated by the Kruskal-Wallis, nonparametric test to determine possible ontogenetic variations in putative response in feeding behavior.Stomach fullness was estimated and four categories were created: full (> 70% of the stomach filled up), almost full (50 to 70% of the stomach filled up), almost empty (< 50% of the stomach filled up) and empty.Only full stomachs were used in diet analysis.Stomach contents were observed under a stereomicroscope (here called  macroscopic analyses) and light microscope (here called microscopic analyses).Items were identified to the lowest possible taxonomic level, with the aid of specialized bibliography.When necessary, specialists were consulted.
Based on previous studies (Vitule & Aranha, 2002) of ontogenetic analyses, the fish were grouped into three total length (Lt) categories: juveniles (Lt < 3 cm); semi-adults(3 < Lt < 8 cm) and adults(Lt > 8 cm).For temporal analyses, the samples were seasonally grouped, with December, January and February representing summer, March, April and May representing autumn, June, July and August representing winter, and September, October and November representing spring.
For analysis purposes, food resources were grouped into taxonomic and/or ecological categories: microscopic filamentous algae, protozoa, diatoms, unicellular chlorophytes, cyanophytes, allochthonous plants (leaf fragments, flowers, stems), autochthonous plants (macroscopic algae and bryophytes), fruits and seeds, non identifiable arthropod parts, autochthonous insects (semi-aquatic and aquatic insects), allochthonous insects (terrestrial insects), other allochthonous arthropods (arachnids, amphipods, and terrestrial arthropods), other autochthonous arthropods (crustaceans and others), detritus (non identifiable organic matter), remaining items (non representative, low frequency items and/or items that do not fit in any of the above categories such as fish scales and eggs).
For the diet analysis, the methods of Frequency of Occurrence (FO) and Percentage Composition (PC) (Hynes, 1950) were employed.The method of Points (Hynes, 1950;Hyslop, 1980) was adapted in a such way that the relative amount of the macroscopic item was estimated by the area it covered on a grid paper.The abundance of each item corresponded to a score proportional to its area in relation to the area covered by other items.The total area of food items was considered the total volume (100%).Calculations of the volumetric proportion of each food item were made according to the formula: where Pix is the proportion by volume of item i in the gut of individual x and Nj is the number of individuals of the species j.
The feeding index (IA i ) was used to ponder the results obtained by the methods of frequency of occurrence and the method of points (relative amounts or volume) of each item (Kawakami & Vazzoler, 1980): where F i is the frequency of occurrence of item i (%) and V i is the relative volume of item i (% of total).
The macroscopic (observed under stereomicroscope, 25 x) and microscopic (observed under light microscope, 400 x) analyses were performed separately.For microscopic analyses, a relative abundance series of degrees was created, which consisted in the counting of 36 fields of 2 mm 2 on a 24x24 mm coverslip, using a light microscope at 400 x.Each item was scored according to the number of fields where it was present: items that were present in more than 26 fields were considered very abundant (score = 4); items that appeared on 18 to 25 fields were considered abundant (score = 3); items that were present in 8 to 17 fields were considered little abundant (score = 2); items that were present in 1 to 7 fields were considered rare (score = 1); and finally, items with no presence were called absent (score = 0).
The analysis of macroscopic and microscopic data performed using different methods (Frequency of Occurrence, Percentage Composition and Feeding Index) produced rather similar results.Thus, only the feeding index IA i was used in subsequent statistical analyses.
The dietary data from index IA i were used to construct a similarity matrix, using the Morisita-Horn similarity coefficient, which was further processed in an agglomerative cluster analysis through the unweighted pair group method, UPGMA -Euclidean distance, using the NTsys v5.0 package.The differences in size, sites and seasonal groups were defined by Non-Metric Multidimensional Scaling (MDS) and one-way ANOSIM, using the PRIMER-E v5.0 package (Clarke & Gorley, 2001;Clarke & Warwick, 2001).
The statistical comparison of diets was made using the c 2 test (a = 0.05), "Comparison with more than two proportions" (Zar, 1999).The H 0 hypothesis corresponded to the same item proportions for distinct samples (sample sites, length categories, and seasons).For samples where H 0 was rejected, the test of "Comparison between two proportions" (Zar, 1999) was applied to detect significant variations in diet categories (P<0.05).Only significant variations were presented in the results.

Results
Ontogenetic and spatial foraging patterns in macroscopic analyses.The stomach contents of 260 fish were analyzed, where 235 were considered full, 19 almost full, 6 almost empty and no empty stomachs were found.The comparison of Intestinal Quotient (IQ) values demonstrated significant differences (P<0.0001) between the three length categories; Juveniles (J) showed an average of 0.72 for the IQ value (n=43); Semi-adults (S) showed an IQ of 0.89 (n=124); and Adults(A) had an average IQ of 1.06 (n=93).
Fifty-three macroscopic diet items were , and they were grouped into 10 ecological and/or taxonomic categories, as described in Table 2.The macroscopic diet of the juvenile size class demonstrated a predominance of autochthonous insects, allochthonous insects, and non identifiable arthro-pod parts.The semi-adults showed a predominance of detritus and allochthonous and autochthonous insects.For adults, there was a predominance of allochthonous plant parts and detritus (Table 2).
The diet analysis of the three length categories showed significant differences (P<0.05)(χ 2 : J x S = 48.87,J x A = 114.63,S x A = 51.87).The main differences when comparing diet between the three length categories were a greater proportion of autochthonous insects and non identifiable arthropod parts in the juvenile category, greater role of detritus and autochthonous plants in the semi-adults, and greater contribution of allochthonous plants, seeds and fruits in the adult category.
The diet analysis of the three length categories showed differences (P<0.05)along the sample sites (χ 2 : JP1 x JP2 = 45.83,JP1 x JP3 =47.66,JP2 x JP3= 38.72, SP1 x SP2 = 21.29,SP2 x SP3 = 42.34,AP1 x AP2 = 56.20,AP1 x AP3 = 86.76,AP2 x AP3 = 51.87) and revealed the following pattern.For the juvenile category, there was a greater inclusion of allochthonous and autochthonous insects and detritus at sample point 1, allochthonous and autochthonous insects at sample point 2, and allochthonous and autochthonous insects, and non identifiable arthropod parts at sample point 3.For the semi-adults, there was a predominance of detritus and allochthonous insects at sample point 1, detritus and autochthonous insects at sample point 2, and allochthonous insects and detritus at sample point 3.For the adults, there was a great contribution of allochthonous insects, allochthonous plants, seeds and fruits at sample point 1, allochthonous plant parts and detritus at sample point 2, and allochthonous insects, non identifiable arthropod parts and allochthonous plants at sample point 3 (Table 3).
An analysis of similarity revealed that the major shifts in diet occurred due to ontogeny rather than seasonality.The three primordial groups in cluster group A with a predominance of autochthonous insects, allochthonous insects and non identifiable arthropod parts; B with a predominance of detritus, allochthonous and autochthonous insects; and C with a predominance of allochthonous plants and detritus were formed due to ontogenetic developmental state, thus  langei grouped into three total length (Lt) categories: Juveniles (J = Lt < 3 cm), Semi-adults (S = 3 < Lt < 8 cm) and Adults (A = Lt > 8 cm), and three sites sampled along the catchment.Sampling was carried out at: site 1 (P1), an upstream first order section of the basin near its spring; site 2 (P2), a second order middle section of the basin; and site 3 (P3), a downstream third order section of the basin.The group codes mean respectively: firmed a predominance of autochthonous insects, allochthonous insects and non identifiable arthropod parts in juveniles; a predominance of detritus, allochthonous and autochthonous insects in semi-adults; and a predominance of allochthonous plants and detritus in adults, with diet depending on the ontogenetic developmental state .
Ontogenetic and spatial foraging patterns in microscopic analyses.Microscopic analyses of the diet revealed the presence of 20 items grouped in 11 categories, as described in Table 5.All length categories showed a predominance of diatoms (Bacylariophyta) and cyanophyte algae.The analysis of the diet of the length categories along the sample sites showed that in the juvenile category there was a predominance of diatoms, filamentous algae and unicellular chlorophytes at sample point 1; diatoms and cyanophyte algae at sample point 2 and 3. Semi-adults had greater amounts of diatoms and cyanophyte algae at the three sample sites.Adults showed a greater proportion of diatoms and cyanophyte algae at sample sites 1 and 3, and sample point 2 showed a homogeneous distribution of filamentous algae, diatoms, cyanophyte algae and unicellular chlorophyte algae (Table 5).
Seasonal analyses of the microscopic diet demonstrated that in the juvenile category there was a greater role of diatoms in the summer of 2002, autumn of 2002 and summer of 2003; cyanophyte algae predominated in the winter of 2002; and autochthonous micro-arthropods dominated in the spring of 2002.For the classes of semi-adults and adults, diatoms predominated during the whole study period, but an increase in the contribution by cyanophyte algae in the winter of 2002 for both categories (semi-adults and adults) was noticeable (Table 6).
The comparison of the microscopic items of the diet between the three length categories demonstrated significant differences (P<0.05) between the diets of juveniles and adults(χ 2 : J x A = 22.16).The cause of this difference was the greater proportion of autochthonous micro-arthropods in the juvenile category.An association between abundance and Fig. 5. MDS on the different sizes, sites and seasonal groups.Total length (Lt) categories compared were Juveniles (J = Lt < 3 cm), Semi-adults(S = 3 < Lt < 8 cm) and Adults (A = Lt > 8 cm) and seasons were: VER=summer (December, January and February), OUT=Autumn (March, April and May), INV=Winter (June, July and August) and Spring=PRI (September, October and November).showing that despite the importance of all other variations, ontogeny was the most important factor (Fig. 3).There was a gradual shift from insectivory to herbivory during development, as shown by MDS analysis (Figs. 4, 5 and 6) and oneway ANOSIM using the ontogeny factor (Table 4).This con-Table 4. One-way ANOSIM testes using ontogenetic factor for independent groups: comparing the total length (Lt) categories Juveniles (J = Lt < 3 cm), Semi-adults (S = 3 < Lt < 8 cm) and Adults (A = Lt > 8 cm); three sites sampled along the catchment: site 1 (P1), an upstream first order section of the basin near its spring; site 2 (P2), a second order middle section of the basin and site 3 (P3), a downstream third order section of the basin; and seasons: VER = Summer (December, January and February), OUT = Autumn (March, April and May), INV = Winter (June, July and August) and Spring = PRI (September, October and November).occurrence of detritus and/or macroscopic algae and microscopic algae was observed.

Discussion
The richness of macroscopic and microscopic items found in this study indicates the opportunistic character of this species.This pattern is usual among fresh-water fish and supports the idea of Lowe-McConnell (1999) that most Neotropical fish have wide variable diets.The fact that almost all stomachs analyzed were full demonstrates an intense feeding activity during the sample period and is evidence of the efficiency of the methods employed.As the digestive process was paralyzed almost immediately, the amount of non identifiable digested material was reduced, thus increasing the reliability of the later analyses (Hynes, 1950;Amundsen et al.,1996).Underwater observations and literature data (e.g., Sabino & Castro, 1990) confirm a diurnal feeding activity for D. langei.It could be observed that individuals retreat to shelters during the night, probably to avoid nocturnal predators such as Hoplias malabaricus (Bloch, 1794), Rhamdia quelen (Quoy & Gaimard, 1824) and Synbranchus marmoratus (Bloch, 1795).
Most of the diet studies using species of the genus Deuterodon have been conducted employing only qualitative methods, although the importance of quantitative or semiquantitative abundance estimates has always been emphasized by many authors (e.g., Hyslop, 1980;Esteves & Aranha, 1999).Considering that quantitative analyses of microscopic items are very difficult (e.g., Aranha, 1993), a semi-quantitative analysis of microscopic items was chosen, allowing greater confidence in the final results.
The Intestinal Quotient (IQ) can be a very useful method to evaluate inter-specific or ontogenetic variations that reflect the feeding behavior (Knöppel, 1970;Zavala-Camim, 1996).The variations found in the D. langei IQ values for the three length categories were marked in this work, differing from what Vitule & Aranha (2002) found for the same species in another study.That difference is probably the result of a smaller sample size and a greater number of length classes employed in the previous study.In D. iguape, the smaller IQ values of juveniles (2.8 to 4.1 cm standard length) have been related to a diet mainly composed of insects (Sabino & Castro, 1990).Similarly, the smaller IQ values found for juveniles of D. langei in this study may be associated with a mainly insectivorous diet in the rio Ribeirão basin.On the other hand, the greater relative length of the intestine can explain the greater plasticity of the adult diet, enabling them to use diet items of greater size and more difficult digestion such as plant items with thick cellulose cell walls (Knöppel, 1970;Sabino & Castro, 1990;Zavala-Camim, 1996;Amundsen et al., 2003).In fact, many tropical stream fishes show an ontogenetic increase in their gut length relative to body length; and those that consume relatively more plant items show disproportional ontogenetic growing, with high rates of increase in gut length (Kramer & Bryant, 1995b).In fishes from forest streams in Panama, the ontogenetic allometry in intestine length was related to increasing body size and absorptive capacity of the digestive tract (Kramer & Bryant, 1995a).In general, herbivorous fish ingest food resources with morphological and chemical defenses, encased in largely indigestible fibrous cell walls, which are thereby considered nutrient poor, and thus, ontogenetic increases in gut length are well known in these fishes.Along this line, many fishes appear to begin life as carnivores or omnivores and shift to a more herbivorous diet as they grow (e.g., Sabino & Castro, 1990;Zavala-Camim, 1996).Their gut lengths usually increase accordingly (Kramer & Bryant, 1995a,b;German & Horn, 2006).Besides this morphological relationship between gut length and diet, behavioral aspects can also be considered.Apparently, ontogeny and phylogeny more than diet influence fish gut dimensions, thus favoring genetic adaptation over phenotypic plasticity as the major force acting on the digestive system (Germam & Horn, 2006).The greater necessity of protein for juveniles relates to a shift from insectivore to omnivore during ontogenetic development.As fish grow they can explore new food sources contributing to a more omnivorous diet (Sabino & Castro, 1990;Vitule & Aranha, 2002, Barreto & Aranha, 2006).These results confirm that ontogenetic dietary shifts occur widely among omnivorous tropical stream fishes (Kramer & Bryant, 1995b).Trends in fish diet have also been detected in other tropical and temperate zones (Amundsen et al., 2003;Katunzi et al., 2006;Rezsu & Specziár, 2006;Alcaraz & García-Berthou, 2007).In this sense, shifts in the feeding ecology during growth strengthen the notion that species may not be an adequate unit when comparing food habits among species that occupy the same area during ontogeny (Stoner & Livingston, 1984;Livingston,1988).
The great abundance of allochthonous plant and animal items in the diet of D. langei reflects the importance of the conservation of river-margin vegetation and nearby forest.The presence of fruits and seeds in the diet reinforces the importance of vegetation to the fish diet (Goulding, 1980;Araujo-Lima et al.,1995).Vilella et al. (2002) suggested that related characins of the genus Astyanax can have an active role in dispersion of seeds, and such role cannot be dismissed for D. langei.
Diet variations throughout sample sites demonstrated a predominance of animal items downstream, and an increasing inclusion of other items, mainly detritus and plant items in the more upstream sample sites.These differences were not due to sample size of length classes in each sample point, because the differences persisted even when the analyses were performed separately for each length class.These differences can be related to availability of feeding resources.
Tropical rivers and streams have large spatial heterogeneity that range from microhabitats to landscapes across a continuum of spatial scales.In fact, succession of organisms in rivers and streams can be the cause of spatial and seasonal variation of occurrence and abundance (Allan, 1995;Allan et al., 1997;Winemiller & Jepsen, 1998).Fish populations and community dynamics show many responses to physical habitat disturbance and seasonal changes in habitat quality and quantity (e.g., Harrell, 1978;Winemiller & Jepsen, 1998).
Tropical fishes show the most diverse morphological and physiological adaptations to dry-wet seasons.A notable example is that reproductive seasons coincide with rising water levels, allowing juveniles to feed in the wider rich resources of floodplains (Lowe-McConnell, 1999).There are other examples of fish adaptations to dry-wet seasons in feeding ecology (e.g., Winemiller & Jepsen, 1998;Katunzi et al., 2006).In D. langei, seasonal diet variation revealed a pattern where there was a greater participation of detritus during the dry season (autumn/winter), and this was true for all length classes.This pattern can be related to lower availability of other feeding resources as the environment retracts due to the lack of rain during that period.In tropical streams, omnivorous fish (e.g., Astyanax fasciatus) and flash floods have a strong role in structuring aquatic insects and benthic algal assemblages (Pringle & Hamazaki, 1997).
Considering microscopic items, the diet showed little variation between sample sites and seasons and diatoms were always predominant.The most representative variation was detected in the dry season (autumn/winter) when a greater presence of blue-green algae (cyanophytes) occurred in the diet, probably because it was more available in the dry period.In the dry period there are no flash floods or storms that may result in decreases in total biovolume of algae diatom biovolume and algal taxon biodiversity.In small tropical streams, storms with high discharge events in the absence of omnivorous fishes would play an important role in structuring benthic algal assemblages, resulting in extreme fluctuations in algal biomass.In this sense, omnivorous fishes show a key role in maintaining the stability of benthic algal assemblages and their resistance to flash flood or storms (Pringle & Hamazaki, 1997).Many successions of organisms in rivers and streams have seasonal causes (e.g.Allan, 1995;Allan et al.,1997;Winemiller & Jepsen, 1998).On the other hand, risksensitive foraging can be a general phenomenon that influences food web dynamics at spatial scales in tropical streams (Winemiller & Jepsen, 1998).Moreover, the presence of an association between abundance and occurrence of detritus and/or macroscopic algae and microscopic algae was observed, suggesting that the greater the proportion of detritus and/or macroscopic algae, the greater will be the proportion of microscopic algae.In this sense, the abundance of microscopic algae can be related to the presence of detritus and/or macroscopic algae in benthic food resources such as periphyton, which increase with gradual drying.
In general, D. langei is an omnivorous and opportunistic species with great diet flexibility.However, it is noted that small individuals show a trend toward insectivory and that as fish develop the feeding spectrum widens, mainly with the addition of plant items of more difficult digestibility and lower protein content.The differences observed in the Intestinal Quotient during ontogenetic development can be related to the increase in plant items in larger individuals.Despite the importance of other diet variations observed, the variation of diet through ontogenetic development was the most prominent aspect for the population dynamics of this species.Most diet variations found between sample sites and seasons were the result of fluctuations in availability of resources.Seasonal diet variation revealed patterns where there is a greater role of detritus and algae during the dry season (autumn/ winter), and this held true for all length classes.That pattern can be related to lower availability of other feeding resources as the environment retracts due to the lack of rain during that period; however, more studies on the availability of feeding resources are essential to corroborate this hypothesis.
Ontogenetic shifts were the most prominent occurrence in the feeding ecology of the species.These shifts in feeding ecology during growth confirmed in an Atlantic Forest stream fish the concept that species may not be an adequate unit when comparing food habits without considering length.This study also contributed needed information regarding the feeding ecology of endemic tropical fishes intrinsically associated with one of the most threatened ecosystems of the planet.

Fig. 3 .
Fig. 3. UPGMA cluster analysis using Morisita-Horn similarity coefficient on the different sizes, sites and seasonal groups.Total length (Lt) categories compared were Juveniles (J = Lt < 3 cm), Semi-adults (S = 3 < Lt < 8 cm) and Adults (A = Lt > 8 cm).The three sites sampled along the catchment were: site 1 (P1), an upstream first order section of the basin near its spring; site 2 (P2), a second order middle section of the basin; and site 3 (P3), a downstream third order section of the basin.The seasons are: VER=summer (December, January and February), OUT=Autumn (March, April and May), INV=Winter (June, July and August), and Spring=PRI (September, October, November).A (> 70 % similarity) = with a predominance of Autochthonous Insects, Allochthonous Insects and Non Identifiable Arthropod Parts; B (> 80 % similarity) = with a predominance of Detritus and Allochthonous and Autochthonous Insects; and C (> 80 % similarity)= with a predominance of Allochthonous plants and Detritus.

Fig. 4 .
Fig.4.MDS on the different sizes, sites and seasonal independent groups.The total length (Lt) categories compared were Juveniles (J = Lt < 3 cm), Semi-adults (S = 3 < Lt < 8 cm) and Adults (A = Lt > 8 cm) .The three sites sampled along the catchment were: site 1 (P1), an upstream first order section of the basin near its spring; site 2 (P2), a second order middle section of the basin; and site 3 (P3), a downstream third order section of the basin.Seasons compared were: VER=summer (December, January and February), OUT=Autumn (March, April and May), INV=Winter (June, July and August) and Spring=PRI (September, October and November).

Table 1 .
Major characteristics of the three sampled sites along the rio Ribeirão basin, a Neotropical stream of Serra do Mar, Atlantic Forest, in southern Brazil.

Table 2 .
Contribution of macroscopic diet categories in Percentage Composition (CP) and Feeding Index (IA i ), comparing D.