Luthy et al.: Identification of larval sallfish, white marlin, and blue marlin in the western North Atlantic Ocean 



597 



a single character may be used to separate fish into 

 groups, early work has lacked a means to confirm the 

 identity of the groups. Molecular techniques provided a 

 solution to this problem. A limitation of the molecular 

 identification technique that we used was that only 

 those larvae preserved in ethanol could be identified. 

 Formalin fixation does not always preclude the use 

 of PCR-based methods, but work is usually limited to 

 small fragments; 570 bp is considered large for success- 

 ful amplification (Shedlock et al., 1997). In the present 

 study, DNA quality was too low in the formalin-fixed is- 

 tiophorid larvae for PCR to amplify the 1.2-kb MN32-2. 

 Consequently, only ethanol-preserved larvae could be 

 used for key development and testing. Because of likely 

 differences in length shrinkage between larvae pre- 

 served only in ethanol and those fixed in formalin, it is 

 possible that the regressions presented in the present 

 study are not valid for the latter. 



No longbill spearfish were among the molecularly iden- 

 tified larvae; thus this species could not be included in 

 the key. Very little is known about the longbill spearfish, 

 but it is reported that larvae are found offshore (Uey- 

 anagi et al., 1970), and that even adults are quite rare 

 in United States and Bahamian waters (Robins, 1975). 

 The longbill spearfish spawning season appears to range 

 from late November to early May and peaks in Febru- 

 ary (Robins, 1975; de Sylva and Breder, 1997). Although 

 there is some overlap in the spawning season of longbill 

 spearfish with the spawning seasons of other Atlantic 

 istiophorids, because of the rarity and predominantly 

 offshore occurrence of the longbill spearfish, its absence 

 from the key may not pose major problems for the iden- 

 tification of istiophorid larvae from our study area. 



The larval istiophorids used to create and test the 

 identification key were all captured either in the Straits 

 of Florida or in Bahamian waters and were all smaller 

 than 22 mm SL. Caution must be used when apply- 

 ing the key to larvae from other parts of the world 

 or to larger sizes. Ueyanagi (1963) assumed that spe- 

 cies pairs from different oceans (white marlin and 

 striped marlin [Tetrapturus audax], longbill spearfish 

 and shortbill spearfish [Tetrapturus angustirostris], 

 Atlantic and Pacific blue marlin, Atlantic and Pacific 

 sailfish]) would be identifiable by the same characters. 

 Although these pairs exhibit the same RFLP patterns 

 at the MN32-2 locus (McDowell and Graves, 2002), we 

 have not tested the key with Pacific larvae and cannot 

 be certain that their measurements would fall within 

 the same regression limits or that they would have 

 the same lower jaw pigment patterns. Even within the 

 Atlantic Ocean, spawning seasons vary with location 

 (e.g., Bartlett and Haedrich [1968] collected larval blue 

 marlin off the coast of Brazil in February and March). 

 Month of capture was crucial in our analyses for dis- 

 criminating between small marlins when spawning 

 season overlap is minimal; therefore our key may need 

 adjustment to reflect local spawning seasons when ap- 

 plied to other locations. 



As in Indo-Pacific istiophorid larvae (Ueyanagi, 1964, 

 1974b), snout length, eye orbit diameter, and lower jaw 



pigmentation are important characters for identifying 

 larval istiophorids of the western Atlantic. However, 

 white marlin differ markedly from their Indo-Pacif- 

 ic counterpart, striped marlin. White marlin larvae, 

 long-held as members of the "long-snout group" of istio- 

 phorids, actually more closely resemble the short-snout- 

 ed blue marlin until 17 mm standard length (Fig. 6). 

 After they reach this size, snout length is intermediate 

 between that of blue marlin and sailfish. This result 

 cautions against the assumption that even large larvae 

 with short snouts are blue marlin. Snout length may be 

 useful as a character in phylogeny studies. 



The identification methods presented in the present 

 study reduce subjectivity in the evaluation of charac- 

 ters. This study also brings to light the caveats of using 

 lower jaw pigment patterns as a means of identification 

 and limits which pigment patterns qualify as diagnos- 

 tic. Although there is a family of lower jaw pigment 

 patterns that appears to mark sailfish only, if this char- 

 acter were the only means of identifying sailfish, nearly 

 40% of our sailfish (as confirmed by RFLP analysis) 

 would have been misidentified or escaped classification. 

 Likewise, the preflexion blue marlin pigment pattern 

 will not lead to misidentifications, but too many preflex- 

 ion blue marlin lack the pattern to justify its use as a 

 stand-alone identification character. Lower jaw pigment 

 patterns have also been suggested as potentially useful 

 characters for separation of subspecific populations of 

 both sailfish (Ueyanagi, 1974a, 1974b) and striped mar- 

 lin in the Indo-Pacific (Nishikawa, 1991). The hypoth- 

 esis of pigment-delineated sailfish populations was not 

 borne out (Leis et al., 1987), and the high variability of 

 lower jaw pigments among larvae of each species from 

 our study area casts further doubt on the notion of us- 

 ing pigments alone to distinguish populations. 



Our identification key does not enable separation of 

 species for certain classes of istiophorid larvae. For 

 example, larvae that are caught in June, are less than 

 10 mm SL, and possess none of the diagnostic lower 

 jaw pigment patterns are especially problematic. In 

 these "dead end" cases, discriminant analysis (CVA) is 

 useful. Although a few larvae were misidentified with 

 the CVA, these larvae were plotted near the interface 

 of two species groupings; this position alerts the user to 

 the fact that misidentification is a possibility. One dis- 

 advantage of using CVA (or any discriminant analysis) 

 for identification is that all of the variables must have 

 a value, meaning that a larva with broken preopercular 

 spines, for example, cannot be entered into the analysis. 

 When the species possibilities are narrowed down to 

 blue marlin and either sailfish or white marlin, it may 

 be feasible to identify larvae by vertebral formula. Rich- 

 ards (1974) suggests that this is difficult with larvae 

 less than 20 mm SL, but it is the method that Prince 

 et al. (1991) used to identify blue marlin that were 5-10 

 mm SL. Molecular identification is always an option for 

 resolving dead ends. 



The identification of larval istiophorids has never 

 been an easy task. Molecular identification is reliable, 

 but can be relatively more labor intensive and expensive 



