Figure 4 A. Growth models for individuals: 1. parabolic asymptotic (Von 



Bertalanffy), 2. sigmoid asymptotic (Gompertz), 3. linear, non-asymptotic, 

 T = Time. 



B. Ford-Walford Plot. Numbers refer to models as in A. 



C. Size-specific orowth vs. size (see text). Numbers refers to models as in 

 A. 



P. Ford-Walford Plot illustrating determinate growth (D) and relatively 

 indeterminate growth (I), as envelopes around clouds of data points. 



COMPARING GROWTH RATES 



Testino the hypothesis that growth rates differ across habitats involves some 

 means of tagging, mapping, or otherwise identifying individuals covering the full 

 range of size classes. It is perfectly possible to have similar asymptotic or 

 maximum sizes in two habitats yet to have significantly different juvenile growth 



rates, for example. Alternatively, statistically indistinguishable growth rates at 

 all sizes could still lead to asymptotic sizes that differ significantly (Fig. 3 A- 

 D). One of the most widely used means of comparing growth increments is the Ford- 

 Walford plot (Ricker 1975; Ebert 1980, 1982; Sebens 1983) which graphs initial size 

 on the abcissa and size after some time interval on the ordinate (Figs. 3C,D;4B,D). 

 This presentation does not show any single individual's growth through all size 

 classes, but instead presents a 'snapshot' of growth in the entire population over a 

 single interval. Variations of the Ford-Walford plot can also be used to fit growth 

 models (von Bertalanffy, Gompertz, Richards functions; Ricker 1975, Yamaguchi 1975, 

 Ebert 1980) (Fig. 4 A, B). A regression line through the points on a Ford-Walford 

 plot (either or both axes can be transformed to produce a linear plot) can be tested 

 for significant differences in slope or intercept, thus testing the hypothesis that 

 differences in growth rates exist between habitats. If a growth -model is to be fit 

 to the data, Ebert (1980, 1982) suggests that the Richards function be used because 

 it has the ability to incorporate deviations from linearity which are especially 

 common in the small size classes. . . 



Growth increments can also be plotted as the specific growth rate (aS*a t^'S -1 ) 

 on the ordinate vs. size (S) at the middle of the time interval At on the abscissa 



13 



