FISHERY BULLETIN: VOL. 80, NO. 3 



mean values of the first and second canonical 

 variates for each stage (Anderson 1958) (Fig. 2) 

 shows that stages 1-4 follow a logical sequence 

 from immaturity to ripeness. The first two ca- 

 nonical variates are the orthogonal pair of linear 

 combinations of the variables which best dis- 

 criminates between the groups. Mean values for 

 the final discriminant function variables are 

 listed in Table 3. Typical DML's for each of the 

 sex-stage groups have also been included for ref- 

 erence. 



CANONICAL VARIABLE 1 



Figure 2. — Canonical variates for each stage of sexual matur- 

 ity evaluated at the stage or group means for each sex. Trajec- 

 tories between subsequent stages are indicated by arrows. 



The Classification Process 



To determine the stage of maturation of a 

 squid, the sex must first be determined. If this 

 cannot be done by visual inspection, the squid is 

 considered juvenile (stage 0). Otherwise, DML 

 and MW, plus TL and SPL for males, or NGL, 

 AGL, and E V for females, should then be deter- 

 mined. From these four or five parameters the 

 appropriate indices needed (Table 2) can be cal- 

 culated. As indicated in Table 2, a set of four lin- 

 ear equations is then constructed by combining 

 the measured parameter values with the ap- 

 propriate discriminant function coefficients 

 (weighting factors) and constant listed for each 

 stage. The equation with the largest solution in- 

 dicates the stage into which the squid should be 

 placed. Thus, once the raw data have been mea- 

 sured and recorded, the actual classification can 

 be done at a later date using even a simple hand 

 calculator. If the raw data are stored on a com- 

 puter-readable medium, as was done in this 

 study, the process is particularly efficient. 



To investigate the broader applicability of the 

 classification, additional squid from the 1976 

 Narragansett Bay and Cryos samples were clas- 

 sified. The discriminant analysis was then per- 

 formed on these new data to see if the same four 

 stages were identifiable. Over 98% of theindivid- 



Table 3.— Typical mean values (1 SD) of selected maturity stage vari- 

 ables of Loligo pealei, listed by sex and stage of development. With the 

 exception of mantle length values (DML), which are from a much larger 

 pooled sample from 1976 to 1978. the reproductive character means are 

 those of the 1976 Narragansett Bay and Cryos samples used to compute 

 the final discriminant functions listed in Table 2. Variable coding fol- 

 lows that of Table 1. 



452 



