600 



Fishery Bulletin 93(3), 1995 



swordfish have no scales and otoliths are not ame- 

 nable to traditional ageing techniques owing to their 

 small size (Ovchinnikov, 1970; Becket, 1974). Suc- 

 cessful otolith readings have been performed only 

 through the application of scanning electron micros- 

 copy (Radtke and Hurley, 1983; Wilson and Dean, 

 1983). Moreover, the use of spines has the advan- 

 tage that the material can be obtained relatively eas- 

 ily without reducing the economic value of the fish. 

 The main problems associated with the spine method 

 are the existence of multiple bands and the missing 

 first annulus. An experienced reader can overcome 

 the problem of multiple bands by determining 

 whether they continue around the entire circumfer- 

 ence of the spine and by recording their distance from 

 the preceding and following annuli. The problem of 

 the missing first annulus in larger fish can be re- 

 solved by identifying its position on sections from 

 younger specimens where the annulus is visible. Dif- 

 ficulties in defining the location of the first annulus 

 in older animals have also been reported for Atlantic 

 swordfish (Berkeley and Houde, 1983) and Pacific 

 blue marlin, Makaira nigricans (Hill et al., 1989). 



Ehrhardt (1992) suggested that the standard von 

 Bertalanffy growth function does not adequately rep- 

 resent the growth of swordfish and proposed the use 

 of the generalized growth function of Chapman 

 ( 1961). The present estimates of r 2 indicate that both 

 models describe swordfish growth equally well over 

 the age ranges considered and give almost identical 

 results for the first eight years of growth (Fig. 4). 

 Although the extrapolation of regression curves be- 

 yond the data is not advisable, it is interesting to 

 note that the generalized model gives much more 

 realistic results for ages less than one because it 



forces the growth curve to pass close to the origin of 

 both axes. On the other hand, estimates of asymp- 

 totic length from this model are questionable because 

 it is unlikely that females have lower asymptotic 

 lengths than males, given that all the large animals 

 in the sample were female. Therefore, the general- 

 ized model appears to overestimate asymptotic length 

 (especially that of males). 



Although results of both back-calculation methods 

 generally agreed with the values predicted by the 

 growth equations, the estimates of the nonlinear 

 method are in slightly closer agreement (Table 3). 

 The use of a nonlinear function for the LJFL spine- 

 radius relation is preferable because it is more real- 

 istic for small animals, which have near zero spine- 

 radius values and very small LJFL values. Ehrhardt 

 ( 1992) also suggested the use of a nonlinear function 

 to describe the LJFL spine-radius relation. 



Growth studies carried out in the Atlantic together 

 with previous work (Tsimenides and Tserpes, 1989) 

 have shown that females grow faster than males 

 (Table 4). Therefore, the calculation of a common 

 growth equation is not valid and may be useful for 

 management purposes only under certain circum- 

 stances. However, it should be noted that in previ- 

 ous work (Tsimenides and Tserpes, 1989), lengths at 

 age were overestimated because the position of the 

 first annulus was misidentified. 



In the case of swordfish it seems that the use of a 

 relatively wide range of size data and mean values 

 for fitting the growth curves has resulted in rela- 

 tively unbiased estimates of swordfish growth pa- 

 rameters in the eastern Mediterranean. Since all 

 Mediterranean swordfish are proposed to belong to 

 the same stock (Magoulas et al., 1993), our findings 



