148 
Fishery Bulletin 99(1 ) 
parts as reported and observed by Zhang and Runham 
(1992). Otoliths of both species are difficult to interpret 
and attention should be paid to the need for trained read- 
ers capable of properly interpreting microincrements in ti- 
lapia otoliths. 
Validation of microincrement deposition in tilapias 
Tetracycline remains a universal marker for otoliths 
(Beamish and McFarlane, 1987; Brothers, 1990; Geffen, 
1992). In our study only one S. melanotheron did not 
reveal any tetracycline deposit on its otolith. Because the 
fish is necessarily handled, injecting tetracycline induces 
a strong calciotraumatic effect on the otolith (Meunier and 
Boivin, 1978; Pannella, 1980; Campana and Neilson, 1985; 
Panfili and Ximenes, 1992). As a result, the mark is easily 
recognizable in the otolith and corresponds to a check in 
the otolith structure. This check probably reflects a ces- 
sation in the growth of the otolith that could last several 
days. In our study, underestimation of age was constant 
(equal to 5 days) for adult O. niloticus which indicates that 
the growth of the otolith was stopped with the tetracy- 
cline injection and resumed five days later. When using 
microincrement counts, the effect of marking should be 
considered, but it is difficult to estimate the time elapsed 
between injection and resumed growth of the otolith. 
We estimated the age of juvenile and adult O. niloticus 
and S. melanotheron accurately using the same microin- 
crement examination and interpretation on thin otolith 
transverse preparations. Tanaka et al. (1981) and Zhang 
and Runham ( 1992) had similar results on transverse sec- 
tions of juvenile O. niloticus otoliths. The former observed 
otoliths with scanning electron microscopy and the latter 
observed stained otoliths. Karakiri and Hammer (1989) 
also reported daily increment deposition on sagittal sec- 
tions of Oreochromis aureus otoliths observed with scan- 
ning electron microscopy. Our technique of preparation is 
the only one that has been validated for two different spe- 
cies and for several developmental stages. 
Precision in age estimation with otolith microincre- 
ments was calculated for both species and showed that the 
error ranges between 4.4 d in juveniles and 8.1 d in adults 
for S. melanotheron , and between 0.9 d in juveniles and 5.1 
d in adults for O. niloticus. 
Validation of back-calculation and influence of 
individual growth rates 
We validated the back-calculation model with a body pro- 
portional hypothesis (BPH) developed by Whitney and 
Carlander (1956, in Francis, 1990) and commonly rec- 
ommended and used in the literature (Francis, 1990; 
Smedstad and Holm, 1996; Horppila and Nyberg, 1999). 
Back-calculation models rely on the assumption that oto- 
lith size and fish size are related and that a relation 
between them can be established. It is assumed that 1) 
the frequency of formation of each structure is constant 
and 2) the width of each increment is proportional to 
the growth of the fish (Campana and Jones, 1992). Cau- 
tion should be taken when calculating the relationship 
because, as stated by Francis (1990) and Campana and 
Jones (1992), if the aim is to backcalculate a mean fish 
length from any otolith dimension, the resulting regres- 
sion must have fish length as a dependent variable and 
otolith dimensions as independent variables. It is there- 
fore very important to set up the most suitable relation- 
ship relating fish length and otolith length. Some works 
have shown that many factors influence this relation- 
ship. Wright et al. ( 1990) reported that the relation of fish 
length to otolith length was linear for smolts belonging 
to the high mode (fast growth) and curvilinear for smolts 
belonging to the low mode (slow growth). The relation- 
ship is also affected by food supply (Rice et al., 1985) or 
seasonal changes (Thomas, 1983). Furthermore, Reznick 
et al. (1989) showed that slow growing guppies ( Poecilia 
reticulata) have larger otoliths than fast growing gup- 
pies of similar lengths, even though both groups of fish 
shared the same genetic background, had the same feed- 
ing schedule, and were reared under the same conditions. 
In this context, the use of the model of Whitney and Car- 
lander is particularly suitable because it assumes that if 
a fish is 10% smaller than the mean length of the popula- 
tion for a given otolith size, this deviation will be constant 
throughout the life of the fish. We chose the curvilinear 
model for the relationship between fish length and oto- 
lith length because it fitted the existing data better. Brad- 
ford and Geen ( 1987) also found no significant difference 
between the curvilinear model and the linear model and 
therefore used the former because it adjusted total data 
better. Smedstad and Holm ( 1996) compared several back- 
calculation formulae for cod otoliths and concluded that 
the nonlinear one was better. These relationships seem to 
depend on the axis of the otolith chosen for back-calcula- 
tion. Back-calculation results could have been less vari- 
able if the diameter of the whole otolith had been used 
in the relationship with fish length instead of the radius 
on the transverse otolith section (Fig. 4; Table 6). Unfortu- 
nately, because it was impossible to interpret the microin- 
crements along the anteroposterior axis (diameter), that 
axis was discarded for back-calculation. A prerequisite of 
back-calculation is the assumption that the frequency of 
formation of microincrements is constant along the axis 
of analysis (Campana and Jones, 1992) and that assump- 
tion could not be made in the anteroposterior axis otoliths 
of our study. 
Our study shows that back-calculated fish lengths are 
greater than measured fish lengths at marking among fish 
that have grown between one and two months. These re- 
sults agree with those obtained by Rijnsdorp and Visser 1 
on plaice ( Pleuronectes platessa ) grown for 19 months, al- 
though the back-calculation model used by these authors 
was the Dahl-Lea model (1920, in Francis, 1990). Two 
main points can be related to the observed overestima- 
tion. First, it is related to fish growth: the larger the fish 
growth, the larger the overestimation, although previous 
authors found an inverse relationship. This overestima- 
1 Rijnsdorp, A. D., and T. A. M. Visser. 1987. Tetracycline label- 
ling of otoliths in plaice. ICES, C.M. 1987/G:33, 12 p. 
