Thresher et al.: Otolith analysis of Nemadactylus macropterus 



835 



of beam conditions, which differ among elements (e.g. 

 Na analysis was more sensitive to effects of pitting 

 than was analysis of Sr) (see Gunn et al., 1992), and 

 the low precision of estimates for elements at low 

 mean concentrations (e.g. CI and S). The effects of 

 these factors can be estimated from standard formu- 

 lae and empirically by comparing 'replicate' analy- 

 ses on the same otolith. In practice, true replication 

 is impossible, owing to the effects of beam damage 

 and small-scale heterogeneity in composition, but it 

 can be approximated by comparing points in two 

 parallel life history scans. Our comparison (Fig. 6) is 

 also a worst-case scenario in that it also includes the 

 effects of different beam-power densities and point 

 spacing in the two scans, which can be expected to 

 have a marked effect on the estimated concentrations 

 of some elements, such as Ca. Nonetheless, for the 

 five elements other than Ca, differences between 'rep- 

 licate points' are still on the order of the theoretical 

 analytical precision (Table 2) and suggest a conser- 

 vative difference criterion between point analyses 

 that ranges from 331 ppm for Sr to 73 ppm for CI 

 (Table 2). 



Ontogenetic comparisons also require that otolith 

 sections be accurately duplicated among individuals. 

 Our test of this accuracy — a comparison of left and 

 right otoliths from single individuals — leads to three 

 conclusions. First, despite our best efforts we could 

 not guarantee duplication of the life history track 

 between otoliths. Pairwise comparisons suggest vari- 

 able compression and expansion of the ontogenetic 

 signal between pairs, which presumably reflects 

 slight differences between otoliths in the beam path 

 relative to the main growth axis. The accuracy of 

 duplication was generally high, but also differed 

 among individuals and declined as distance from the 

 primordium increased. Second, nonetheless, the over- 

 all pattern of peaks and troughs in the pairs of 

 otoliths compared was generally quite similar. As a 

 result, the principal ontogenetic patterns in, for ex- 

 ample, Sr concentrations would be reflected in both 

 otoliths, but examination of one alone could lead to 

 erroneous conclusions about the life history stage at 

 which a particular change in concentration occurred. 

 The variability in life history scans induced by dif- 

 ferential compression renders statistical comparisons 

 of ontogenetic patterns extremely difficult and liable 

 to subjective interpretation. In theory, these difficul- 

 ties could be overcome by calibrating ontogenetic 

 changes in composition against real age — as opposed 

 to distance along the growth axis — but difficulties in 

 resolving the ages of larger individuals are likely to 

 make this approach problematic for most species. 



Third, for at least some elements, mean concen- 

 trations and ontogenetic patterns appear to differ 



between otoliths even within the same individual. 

 In two of the three pairs examined, mean CI and S 

 concentrations differed significantly between otoliths 

 over relatively large sections of the main growth axis 

 and at levels well above measurement error. This 

 asymmetry was so surprising that we repolished and 

 reanalyzed one pair of otoliths (specimen #312) to 

 confirm the results; the second series of data were 

 virtually identical to the first. The implication is that 

 otolith pairs do not encode life history information 

 in the same way. As yet, sample sizes for this com- 

 parison are much too small to assess the generality 

 of mismatches and the scale of the problem, but the 

 available information suggests treating with caution 

 data obtained from single point analyses in otoliths. 

 Another potential source of methodological errors 

 is specimen contamination. J. Calaprice (in press), 

 for example, discounted CI as a stock discriminator 

 in his studies on Atlantic bluefin tuna, Thunnus 

 thynnus, because the element is widely present in 

 the laboratory environment and easily transferred 

 during specimen preparation and handling. Given 

 this, the dependence of our site separation on varia- 

 tion in CI concentrations is of concern. Although con- 

 tamination is a critical issue (particularly at the sub- 

 ppm level), several of our observations are not im- 

 mediately consistent with the contamination hypoth- 

 esis. First, CI does not vary independently; its con- 

 centrations in otoliths covaries among specimens 

 with Na and K. If CI concentrations are principally 

 contaminants, the same contamination must affect 

 Na and K concentrations, which is unlikely. Second, 

 samples collected in the same region but at different 

 times, different places, and with different gear types 

 (e.g. juveniles and adults from the Tasmanian sites 

 collected by hook-and-line and trawling) exhibit simi- 

 lar concentrations of CI, suggesting that observed 

 variability is not a consequence of the way individu- 

 als are caught and handled. And third, the order in 

 which the specimens were prepared and analyzed 

 was randomized to check for systematic error; none 

 was detected. 



Evidence for regional variation in otolith 

 composition 



Several previous studies, using probe microanalysis 

 (e.g. Radtke, 1989; Kalish, 1990) and whole otolith 

 analysis (e.g. Gaudie et al., 1986; Edmonds et al., 

 1991 ), have demonstrated that otoliths vary in com- 

 position ontogenetically and regionally. Our data 

 permit a detailed evaluation of the interaction be- 

 tween these components of variability. However, be- 

 cause we could not collect and analyze otoliths of lar- 

 vae from known spawning areas our results only test 



