(reviewed in Kaufmann 1981) (Fin. 4 C). This method has the advantaae of allowina 

 growth to he measured over different time periods (At). Such plots can also be 

 subjected to reoressions and these lines can then be used to fit various growth 

 models. This technique has the disadvantaoe that size (S) is incorporated into both 

 parameters of the regression, making statistical interpretation difficult by violat- 

 ing the assumption of independence (Ricker 1975). The Ford-Walford plot avoids this 

 problem, and can incorporate a small degree of variability in At values by assuming 

 short-term linear orowth and normalizino all data to a constant time interval. 



The Ford-Walford plot allows direct examination of growth and 'degrowth', growth 

 rate as a function of size, and size asymptotes or maxima. The degree of 'indeter- 

 minateness' is also evident as the spread around the regression line (Fig. 4D), as is 

 the fraction of individuals below the zero growth line (showing shrinkage or 

 degrowth). Coral growth, measured as density bands in the skeleton, could be 

 compared in this manner, as could whole colony size increase. Growth can also be 

 compared statistically without resorting to either regression lines (assuming 

 linearity) or any particular growth model. If the data are broken up into size 

 classes (Fig. 3E, F) such that there is a reasonably large number of points in each 

 size class, then the mean growth increments in each size class can be compared 

 across habitats (e.g., by analysis of variance, Sebens 1983). This method allows 

 comparison across several habitats (via a multiple comparisons test) and determines 

 where in the size distribution the significant differences actually occur. This is 

 the only method that would pick up differences between habitats which have equal 

 maximum individual sizes and differ only in early growth rate. 



Growth by fission and fragmentation cannot be incorporated directly into a Ford- 

 Walford plot. A good descriptor of such growth is the rate of biomass change (e.g., 

 exponential function) within a clone, colony, or group of colony fragments. However, 

 some clones or colonies may have reached limits imposed by available space. There- 

 fore, choosing experimental subjects not apparently space limited, or removing such 

 limits experimentally (e.g., clearinq space around the subject, transplanting) may be 

 necessary to compare potential growth rates (Sebens 1983). 



SUGGESTIONS AND LIMITATIONS 



There is no shortcut method that allows an investigator to read causation from 

 static samples of size distributions across populations. Mortality schedules can be 

 arrived at only by following individuals or cohorts over time. It is theoretically 

 possible to calculate an average overall mortality from a size distribution (e.g. 

 Ebert 1982), but this approach assumes constant, mortality and recruitment rates. It 

 is the latter assumption that causes real problems because many sessile inverte- 

 brates have successful recruitment only rarely and sometimes many years apart. In 

 colonial organisms, partial mortality (e.g. corals, Hughes and Jackson 1980) pre- 

 sents another problem for interhabitat comparisons. This process is best treated as 

 'degrowth' or 'shrinkage' and can be compared across habitats directly. Similarly, 

 binary fission does not constitute 'mortality' of the original individual, even 

 though that individual no longer exists as such. Fission rates can also be compared 

 directly either numerically or by using biomass (Sebens 1983). 



Growth rate differences, resulting from habitat or microhabitat variability, can 

 produce some of the observed size gradients within populations. There are several 

 methods for comparing growth and for fitting growth models; however, it is not 

 necessary to choose a growth model if the goal of a study is only to determine 

 whether or not growth rates differ across habitats. Ford-Walford plots of growth 

 increments can be used for this directly, and can also illustrate growth rate 

 variability and the extent of degrowth or shrinkage in the population. An even 

 simpler qrowth comparison between habitats would be to compare only maximum size 

 (mean size of largest N individuals) and nrowth at some earlier staqe, for example 

 at 1/2 the maximum size in the habitat with the smallest maximum size. This would 



14 



