NOTE Chaloupka and Zug: A polyphasic growth function for Lepidochelys kempii 
853 
function is inconclusive because the data span was 
incomplete, missing not only the first growth cycle 
(if it occurred at all) but also the onset of adult matu- 
ration. By the end of the study the remaining 8 turtles 
were still growing and below estimates of adult size 
(weight) recorded for wild stocks. Moreover, growth 
in captivity might well bear little similarity to the 
growth dynamics of wild stock Kemp’s ridley sea 
turtles, which seem to grow much slower at a given 
size (Caillouet et al., 1986). 
The nonparametric smooth shown here in Figure 
1A fitted to a more complete age and size range for 
wild stock Kemp’s ridley sea turtles implied that 
growth comprised two consecutive phases and that 
an explicit polyphasic model (Table 1; Fig. 2) might 
be a better parametric description of growth than 
monophasic models proposed previously for this spe- 
cies. Nonetheless, two major cautions are warranted 
prior to drawing further conclusions from Figure 2 
about Kemp’s ridley growth dynamics. These cau- 
tions relate to the implications for growth inferences 
due to 1 ) data sparsity in the early growth years for 
this data set and 2) the size composition anomaly 
between stranding subsamples for this data set. 
The data were sparse in the lower region of the 
first inferred growth spurt (see Fig. 1). The growth- 
layer loss protocols used in deriving the skeleto- 
chronological age estimates provided differing cov- 
erage of this growth region (see Zug et al., 1997). 
The age estimates used in our study were based on 
the correction-factor protocol that was considered 
more reliable than ranking protocol estimates (Zug 
et al., 1997) despite providing no coverage of the first 
growth year except for hatchling size. The model that 
was fitted (Fig. 2A) interpolated between known 
hatchling age and the end of the first year on the 
basis of the explicit form implied by the specified 
parametric function (Eq. 1). Although the conclusion 
of the first growth phase completed by ca. 25-30 cm 
SCL (see Fig. 1A) is firm despite sparse data during 
early growth, a specific growth spurt >1 year of age 
is tenuous. Given the lack of data during the first 
year, maximum growth might just as feasibly occur 
immediately following hatching, resulting in a mono- 
tonic decreasing age-specific growth-rate function for 
the first cycle and not the nonmonotonic function seen 
in Figure 2B. On the other hand, a growth spurt 
might occur a little later after hatching, resulting in 
a nonmonotonic age-specific growth-rate function for 
the first cycle similar to that proposed in Figure 2B 
but with the spurt occurring at say 3 or 6 months 
rather than at 12 months of age. Because of a spar- 
sity of data during the early growth years, all these 
growth scenarios for the first growth cycle are fea- 
sible; therefore data for the first 12 months of life 
following hatchling dispersal from the nesting beach 
are essential to resolve this important issue. 
Nonetheless, other sources of information corrobo- 
rate the growth profile proposed for the first phase 
in Figure 2B. First, the polyphasic function described 
by Equation 1 fitted the growth data well, including 
an estimated mean adult size (upper asymptote= 
2(aj+a 2 )=62.6 cm SCL from Table 1) consistent with 
empirical estimates of mean nesting female size of 
64-65 cm SCL (Marquez, 1994). Moreover, the 
polyphasic function also predicted a mean hatchling 
size of 4.3 cm, which is consistent with the empirical 
estimate of mean hatchling size of 4.4 cm (Marquez, 
1994). No other growth model has come close to pre- 
dicting both the mean upper and lower size asymp- 
totes of the postnatal development phase for the 
Kemp’s ridley sea turtle. It is worth noting here that 
it is a common misconception in growth studies (par- 
ticularly sea turtle growth studies) involving more 
than one animal that the upper asymptote of a para- 
metric growth function estimates maximum adult 
size rather than the correct interpretation of mean 
adult size (see Ricker, 1979). 
Second, a growth spurt >12 months of age and a 
growth phase completed by ca. 27 cm SCL (30-36 
months of age) is coincident with developmental 
changes to the blood oxygen system of the Kemp’s 
ridley sea turtle prior to acquisition of an adult blood 
system by 28 months of age (see Davis, 1991) — at 
least this was the case for captive-reared Kemp’s rid- 
ley sea turtles. Davis ( 1991) also found that the oxy- 
gen capacity of the blood had increased substantially 
during the first 12 months of growth. The size range 
and timing of the first growth cycle is also consis- 
tent with apparent dietary and habitat shifts around 
20 cm SCL (ca. 18 months of age: Fig. 2A; Eq. 1 ) from 
a presumed epipelagic habit to a coastal benthic habit 
(see Shaver 1991; Burke et al., 1994; Musick and 
Limpus, 1997). 
The second major caution relates to a lack of infor- 
mative cofactors (sex, geographic subsample) being 
included in the model because of insufficient records 
or small subsamples. For instance, sex was recorded 
for only 37% of the strandings, whereas the Atlantic 
coast subsample (cf. Gulf coast) accounted for 79% of 
the strandings data (see Fig. 1, B and C). Moreover, 
the Atlantic subsample comprised a significantly dif- 
ferent size composition compared with that of the 
Gulf of Mexico (see Fig. 3). The apparent size compo- 
sition anomaly might be due to 1) differential and 
inadequate spatial sampling of strandings and 2) 
developmental migration of Kemp’s ridley sea turtles 
>40 cm SCL from the Atlantic coast to the Gulf of 
Mexico (see Collard and Ogren, 1990; Morreale et 
al., 1992; Epperly et al., 1995; Musick and Limpus, 
