relationship has been demonstrated for T. nilotica (Hughes and
Behrends, 1983) and T. mossambica (Uchida and King, 1962).
 Snow et al. (1983) obtained a fry production rate of
 17.8/female/day for T. area by using a fry transfer method in
 small tanks without hapas. A lower sex ratio (three females to one
 male) and stocking density (1.6/m2) were utilized. Hughes and
 Behrends (1983) obtained production of seed of 20.4/females/
 day for T. nilotica in hapas. These production rates were substan-
 tially higher than the rates obtained in the present study.
 At least two factors contributed to low fry production. First,
water quality deteriorated as the result of high standing crops of
brood fish and high feeding rates. For example, the final stand-
ing crop in Treatment 3 was equivalent to 18,600 kg/ha for the
hapa and 6,600 kg/ha for the entire tank. The initial feeding rate
in Treatment 3 was equivalent to 129 kg/ha/day for the hapa and
46 kg/ha/day for the entire tank. As dense algal blooms
developed, algae clogged the mesh, thereby reducing water circu-
lation and exchange between the hapa and the limited tank
volume. Batches of dead eggs were occasionally found during fry
collection. Poor water quality reduced fry production.
 A second reason for low fry production was that the brood fish
were fed at a maintenance level and did not receive the nutrition
required for maximum fry production. The daily feeding rate
never exceeded 1% of the initial body weight, and as the fish
grew, they received less than 1% of their actual weight. The
feeding rate in Treatment 3 had to be lowered to 0.5% of the in-
itial body weight to limit deterioration of water quality. Hughes
and Behrends (1983) and Snow et al. (1983) fed their brood fish
at rates of 3% and 2-3% of body weight, respectively. Uchida
and King (1962) found that adequate nutrition was very impor-
tant in obtaining good fry production.
 Hapas and small tanks are appropriate fry production units for
small operations. To meet production goals, data generated from
this and other studies should be utilized to determine production
schedules and unit requirements. If fry are needed continuously,
daily averages may be adequate for annual projections. However,
fry are usually required in large quantities on a given date to coin-
cide with nursery and growout cycles. In this case, mean produc-
tion figures are less realistic as the result of high variability. The
required number of production units, as determined by using
mean fry production rates, should probably be doubled to ensure
adequate fry production.

Fingerling Production
 The objective of this experiment was to raise 5-g fingerlings, a
size that could tolerate restocking at lower rates for additional
growth to a larger fingerling size. After 63 days of intensive
feeding, the largest group of fingerlings reached a mean weight
of 2.3 g. These fingerlings had been stocked at the lowest density
-26 fish/m2 (Table 5). The initial mean weight of fry for all
stocking densities was 0.019 g.
 A longer growing period would be necessary to obtain 5-g
fingerlings under present experimental conditions. Under dif-
ferent conditions a higher growth rate has been obtained. Snow
et al. (1983) produced 4.4-g fingerlings (T. aurea) in static tanks
in a 40-45 day growing period at a stocking density of 62/m2.
Weight gain averaged 55.5 kg/ha/day. The highest gain in the
present study was 16.4 kg/ha/day at a stocking rate of 104/m2
(Table 5). In both experiments Purina Trout Chow (crumbles)
was fed at a daily rate of 15% of body weight for the first three
weeks. The feeding rate was then reduced to 2% in the earlier
study and 5% in the present study. The difference in growth be-
tween the experiments was the result of fertilization. Snow et al.
(1983) fertilized the fry rearing tanks with a commercial inorganic
fertilizer (20-20-5 for N-P-K) at a rate of 44.8 kg/ha/week until
Secchi disc visibility was less than 30 cm. Natural food production
that resulted from fertilization was an important component in


the fry diet and promoted rapid early growth. Algal blooms
eventually developed in the present study as the feeding rate
increased, but the advantage of abundant amounts of natural
food for initial growth was lost. If fertilization is not employed,
recent findings suggest that the initial feeding rate should be in
the range of 40 to 50% (Kubaryk, personal communication).
Feed that is not consumed will serve as organic fertilizer and pro-
mote the growth of natural foods.
 Survival in this study was generally poor, ranging from 32.7 to
 87.8% (Table 5). Density did not affect survival since survival of
 less than 50% occurred at the lowest and highest densities. The
 fry were the same size and age when they were stocked so that
 cannibalism was probably minimal. It seems that predation by
 dragonfly larvae was the major cause of poor and variable sur-
 vival. Of the 18 tanks utilized in this study, 15 tanks had been
 full of water for several weeks before fry were stocked and three
 tanks were drained, cleaned and refilled the day before stocking.
 These three tanks were stocked at different rates, but survival was
 93.9, 97.3 and 97.4%, rates that were uniformly higher than
 those in any of the other 15 tanks. Large populations of dragonfly
 larvae had developed in the previously filled tanks but none
 developed in tanks that were cleaned prior to stocking.
 Variable and poor survival greatly affected the results. There
was a large difference between the stocking density and the actual
density of fish recovered at harvest (Table 5). The density at
harvest probably existed throughout most of the experiment since
dragonfly larvae prey on fry right after stocking when they are
small in size. The feeding rate was calculated on the assumption
of 100% survival and therefore the actual feeding rate was much
higher than desired (5%) in the treatments with low survival.
Table 5 shows the actual feeding rate for the fish surviving at
harvest as well as the feed conversion ratio and the final feeding
rate in terms of kg/ha/day. Fish at stocking densities of 52 and
104/m2 had high survival rates, actual feeding rates of less than
6% and feed conversion ratios of less than 0.9. Fish at these den-
sities converted feed well and grew well through week 9 (Figure
1). However, fish at the higher density grew at a much lower rate,
possibly because there was less natural food available per fish.
Fish at a stocking density of 78/m2 were fed at an actual rate of
8.0% and exhibited a feed conversion of 1.57. Apparently the
quantity of feed was adequate as indicated by the relatively high
conversion ratio. The harvest density (48/m2) was nearly the same
as the harvest density (45/m2) of the fish stocked at a rate of
52/m2, but the growth rate was slightly lower (Figure 1). The
final feeding rate was 41.8 kg/ha/day, which may have caused
some deterioration in water quality although no major differences
in water quality were detected among any of the stocking den-
sities (Table 6). Fish stocked at densities of 130 and 155/m2 were
fed at actual rates of 15.3 and 10.4%, respectively, and exhibited
poor feed conversion (4.39 and 2.29). A sufficient amount of
food was available and yet the final mean weights were low (1.4
and 1.3 g) as the result of water quality deterioration. The final
feeding rates were 102.7 and 93.6 kg/ha/day, respectively.
Tucker and Boyd (1979) found that a maximum feeding rate of
56 kg/ha/day reduced growth and survival of channel catfish in
ponds as the result of water quality deterioration. The absence of
bottom sediments in tanks and poor water circulation may cause
water quality to deteriorate at lower feeding rates than in ponds.
Fish stocked at 26/m2 were fed at an actual rate of 13.6% and ex-
hibited a poor feed conversion (2.61). However, their growth rate
was the highest because the final feeding rate was low (25.3
kg/ha/day) and much natural food was available.
 Although poor water quality slowed the growth of fish fed at
high rates near the end of the experiment (Figure 1), these fish
gained a temporary advantage earlier in the experiment when
feeding rates (as kg/ha/day) were lower and the fish were at a size
where they were able to consume feed at a higher daily percen-


PROCEEDINGS of the CARIBBEAN FOOD CROPS SOCIETY-VOL. XX


238