The study was conducted in the lower Djoue River, downstream of the hydroelectric dam. The experimental station is located at the confluence of Djoue River with Mfilou River (Figure 1).

The sampling station is located at the confluence of the Djoue River with the Mfilou River. It is located between 315 m and 292 m of altitude; 4.29˚S and

4.30˚S latitude, 15.23˚E and 15.22˚E longitude. This station is dominated by Echinochloa pyramidalis.

Specimens of C. polli were caught monthly (September 2012 to August 2013). The standardized sampling technique was the cash net. All captured fishes were fixed with 10% formalin and stored in 5% formalin. The standard length and length of the intestine were measured to the millimeter using a Vernier brand caliper and weights (not eviscerated and eviscerated) using an Oaus brand balance to the nearest milligram. Stomach contents are collected, examined and sorted under a Motic brand binocular loupe before being weighted. The identification of preys was made according to keys proposed by various authors [11] [12] [13] . Ingested preys were identified using a binocular microscope to the species level where possible; the excessively weathered remains were classified as the indeterminate group. For each prey, specimens were counted and measured. The fishes were grouped by size class according to the Sturge rule [14] . The number of size classes (NC) and the range of classes (IC) are given by the following relation:

N C = 1 + ( 3.3 log 10 N ) and I C = L S maxi − L S mini N C

where N is the number of all specimens examined.

LS is the maximum and minimum standard length.

The intestinal coefficient (Ir) is the ratio of the length of the intestine to the standard length predicts the diet of a species [15] [16] [17] [18] .

It is given by the relation I r = L i n t L S , where Lint is bowel length and LS is standard length in millimeters.

The following indices were used for the qualitative and quantitative analysis of the diet.

This is the ratio expressed as a percentage between the number of empty stomachs (Ev) and the total number of stomachs examined (N). Its relation is:

V = E v / N × 100 [19]

The corrected percentage of occurrence proposed by [20] [21] is given by the relation:

F c = F i ∑ F i × 100 and F i N i ∑ N i

where

N_i is number of stomachs containing the element i;

F_i is Frequency of a prey i.

The weight percentage whose relation is:

P = P i ∑ P i ,

where

Pi is the total weight of a prey.

The preponderance index of [22] , gives the relative importance of each category of prey. It combines the corrected percentage of occurrence and the weight percentage:

I p = F c × P ∑ ( F c × P ) × 100 .

This index varies between 0 and 100.

where

Ip <10: accessory preys;

10 ≤ Ip < 25: secondary preys;

25 ≤ Ip < 50: important preys;

Ip ≥ 50: main preys.

The Schoener similarity index is given by the relation:

S x y = 1 − 0.5 ( ∑ | P x i − P y i | ) [17]

where P_xi and P_yi are the relative proportions of prey (i) for specimens of classes x and y.

Statistical analyzes of the results by size classes and seasons were made from two statistical tools: Excel and Statistica-7.1. Statistical tests (student test and ANOVA) were used to check the links between parameters. A Cluster analysis based on Euclidean distance is performed from the Prey Occurrence—Classes size matrix to identify similarities between diets of specimens at different sizes.

The very short oesophagus continues with a well developed stomach. Beyond the pyloric region the intestine wraps around itself, describing convolutions, and then becomes rectilinear to the anus. Intestinal coefficient and vacuity coefficient of the 662 stomachs of C. polli examined, the intestinal coefficients calculated ranged between 5.63 and 8.02; with an average of 6.83 ± 0.5. Specimens of C. polli have an intestines length six times longer than the body. This value is predictive of an herbivorous diet [23] . Of 662 stomachs analyzed, 145 stomachs were empty, representing an overall vacuity coefficient of 22%. This coefficient is higher in the dry season (24.61%) than in the rainy season (18.12%).

Twelve types of preys were identified in the 517 full stomachs examined: fish remains (scales, bones and fins), insects (imagos and fats), insect larvaes, nematodes, annelids, trematodes, algaes, plant debris, mud, sand grains and other preys that include all unidentified preys. Two preys were frequently seen in the stomachs: algaes and plants debris with occurrence percentages of 28% (Figure 2).

Of all these preys, algaes are the main preys of Ctenochromis polli, with a preponderance index equal to 65% (Table 1).

Percentage of occurence (Fc); percentage of weight (Pc) and preponderance index (Ip).

Plants debris and mud were secondary preys, while all other preys were incidental. In the diet of Ctenochromis polli, there is a clear dominance of algaes. [24] point out, however, that phytoplankton (Diatoms, Cyanophyceae) have been observed frequently in the stomach contents of H. (Paralabidochromis) sp. Ctenochromis polli has a diet similar to that of H. (Paralabidochromis) sp. [24] . The diet of this last species is more diversified than that of Ctenochromis polli while remaining mainly insectivorous. It consists of insect larvae and, to a large extent, phytoplankton

Despite the seasonal fluctuations of preys available in the environment, the food spectrum consisted of more algaes in the dry season and insects in the rainy season. Figure 3 shows the grouping of preys items by season.

In the rainy season, the rising water provides to fishes a wide range of foods as [25] . The drop of water levels during the dry season, however, seems to have an impact on the availability of food. An ANOVA (Main effects) by the method of speciation (Analysis Winzard) gives the results in Table 2.

At the 95% threshold (p = 0.05), the difference was very significant in the seasonal consumption of the following preys: plants debris, sand grains and mud. The Cichlidae therefore has a larger number of preys at its disposal, so it presents an opportunistic diet, adapting its diet according to available preys that vary qualitatively and quantitatively seasonally. In tropical environments, many studies suggest an abundance of food resources in the rainy season and a reduction in the quantity of these food resources in the dry season [26] [27] [28] . The seasonal fluctuations of available preys in the environment would explain the food spectrum of Ctenochromis polli which consists more of algaes in dry season and insects in rainy season. In the rainy season, the upwelling makes available

to the species a wide range of foods as [25] . The drop of water level in dry season, however, seems to have a big impact on food availability.

The standard length of the specimens studied vary between 15.37 and 109.52 mm. Seven (7) interval classes size 13.45 were defined by the Sturge rule. The modal class is that consisting of specimens whose standard lengths are between 45.27 and 55.72 mm (Class 4). Table 3 illustrates the quantitative and qualitative changes in the diet of C. polli by size.

It appears that the qualitative composition of the food changed during the growth of the fish. Small specimens (Class 1, Class 2 and Class 3) have a preference for benthic preys, of plant origin (algaes and plants debris and mud). The consumption of benthic preys is usually accompanied by sand. Class 4 specimens incorporate insects of different stages into their diet. At sizes greater than

69.17 mm (classes 5 - 7), a gradual change in diet was observed with an entomophagous tendency. The tendency to ichthyophagy or scaliphagia also increased with the size of the fish (from 69.17 mm). The same observations were reported by [29] [30] [31] [32] , who argue that during fish growth, the qualitative and quantitative composition of food changes. A grouping of prey items in fraction allows to obtain, four fractions: the vegetable fraction, the animal fraction, the sedimentary fraction and the unidentified preys fraction. It seems likely that the sedimentary fraction was collected at the same time as the benthic feed (algaes, plants debris and sand). The same observation is made by [17] [25] [33] , these authors also indicate that ingested sand may play a role in shredding food exoskeletons. Larger specimens consume almost no sand and have a preference for animal fractions. The increase in insect preference with fish size may also be related to the ability to consume very mobile and large preys by large specimens [34] [35] believe that this change in diet with fish size is an asset in decreasing competition between small and large specimens. After Cluster analysis of the preys-class size occurrence matrix (Figure 4), four class size groups discriminate at aggregation threshold 120.

The group 1 consists of specimens from class 4; the group 2 makes individuals of classes 2, 3 and 5; the group 3 composed of those of classes 6 and 7, then the group 4 formed mainly specimens of class 1. On the basis of this grouping of size classes, it appears that the diet of C. polli varies according to the size of the specimens.