854 
Fishery Bulletin 95(4), 1 997 
• extreme outlier 
Atlantic Gulf 
coast coast 
Figure 3 
Boxplots of the size distribution of the Kemp’s 
ridley sea turtle from the Atlantic Coast and 
Gulf of Mexico subsamples. The boxes show 
the interquartile range (25th to 75th percen- 
tiles), bars show 10th and 90th percentiles, 
and the notches show comparison-wise 95% 
confidence intervals. The notches on the two 
boxplots do not overlap each other, indicat- 
ing a significant difference in median size 
between the two subsamples. 
1997). If the Atlantic and Gulf coast subsamples in 
the Zug et al., (1997) data set represent two discrete 
populations with population-specific growth behav- 
iors, then the growth model here (Table 1; Fig. 2) is 
applicable to the Atlantic group only although a simi- 
lar polyphasic model is apparent for both subsamples 
despite a sparsity of data for the >50 cm SCL group 
of the Atlantic subsample (Fig. IB) and <40 cm SCL 
group of the Gulf subsample (Fig. 1C). The inclusion 
of the Gulf subsample serves to provide sufficient 
data to derive the upper asymptote for estimating 
mean adult size (see Fig. 1C). The Atlantic subsample 
comprised only immature Kemp’s ridley sea turtles 
consistent with recorded size distributions for popu- 
lations resident in various habitats along the US At- 
lantic coast (see Burke et al., 1994, Epperly et al., 
1995; Schmid, 1995). 
If the Zug et al. (1997) data set is representative 
of a single panmictic interbreeding stock displaying 
some form of staged developmental migration, then 
the model presented here is a reasonable approxi- 
mation of the growth dynamics of the endangered 
Kemp’s ridley sea turtle. There is compelling sup- 
port for this view given current knowledge of Kemp’s 
ridley sea turtle movement patterns (Musick and 
Limpus, 1997). Further support comes from the dis- 
covery of a Kemp’s ridley sea turtle (70 cm CCL, 67 
cm SCL) nesting at Rancho Nuevo in 1996 (116 eggs 
laid) that had been tagged as a juvenile (51 cm CCL, 
49 cm SCL) seven years earlier in Chesapeake Bay 
(Musick 1 ). Growth for this nesting ridley was con- 
sistent with the polyphasic growth function (Fig. 2A) 
although clearly a single record is not sufficient to 
provide conclusive evidence. Nonetheless, if the At- 
lantic and Gulf subsamples represent a single pan- 
mictic interbreeding stock, then a juvenile growth 
spurt at 46 cm SCL (ca. 8 years old, Fig. 2B) would 
indicate an ontogenetic shift associated with devel- 
opmental migration from juvenile foraging habitats 
in the South Atlantic Bight (Musick and Limpus, 
1997) and from within the Gulf of Mexico (Collard 
and Ogren, 1990) to foraging grounds in habitats along 
the Gulf coast prior to the onset of sexual maturity. 
It is also conceivable, given the dispersal scenarios 
proposed by Collard and Ogren (1990), that the 
Kemp’s ridley sea turtle is a single panmictic inter- 
breeding stock that comprises two distinct post- 
hatchling developmental groups. One group remains 
within the Gulf of Mexico displaying relatively rapid 
growth owing to the warmer water (see Caillouet et 
al., 1995b). The second group represents the 
posthatchlings swept from the Gulf of Mexico that 
settle as juveniles (ca. 20 cm SCL) in the inshore 
developmental habitats of the mid-Atlantic (Morreale 
et al., 1992; Burke et al., 1994) and South Atlantic 
Bights (Epperly et al., 1995; Schmid, 1995). In this 
case the polyphasic growth model presented here 
(Table 1; Fig. 2) would be applicable to describing 
the mean stochastic growth dynamics of the cohorts 
swept each year from the Gulf of Mexico and under- 
going growth in the Atlantic Bights prior to migrat- 
ing back to the Gulf of Mexico. A separate growth model 
would need to be derived for the Gulf of Mexico devel- 
opmental group although polyphasic growth behavior 
is also apparent for that subsample in our study (see 
Fig. 1C). 
Clearly, a better understanding of the dispersal and 
developmental dynamics of the Kemp’s ridley sea 
turtle based on a mark-recapture program with a 
high recapture likelihood is needed to resolve these 
complex issues. Although several local tagging pro- 
grams have been undertaken (e.g. Caillouet et al., 
1995a; Schmid, 1995; Burke et al., 1994, and refer- 
ences therein) a more comprehensive spatial and 
sampling-intensive program spanning the distribu- 
tional range of this species is needed. 
1 Musick, J. 1997. Virginia Institute of Marine Science, Col- 
lege of William and Mary, Gloucester Point, VA. Personal 
commun. 
