NOTE Toole and Nielsen: Microprobe precision associated with SrCa ratios 



423 



standard deviation of that measurement. For a single microprobe analysis, 

 this is calculated as 



CV = 



i=i n 



n \ 2 



0.5 



\ n / 



Np + 



N, 



(Np-Nb) 



Otolith 



Standard 



0.5 



where n = number of samples taken on the standard, 



Nj = x-ray count (corrected for background count) from ith sample 



on the standard, 



Np = x-ray count for peak wavelength of element in sample, 



Nb = x-ray counts from background wavelengths of element in sample, 



tp = peak wavelength counting time, and 



tb = background wavelength counting time. 



Approximate 95% confidence limits for each element measured in each 

 sample were considered ± 2 * CV, since the Poisson distribution underlying 

 these calculations approximates a normal distribution when sample size (the 

 number of x-rays detected by the spectrometer during an analysis) is high 

 (Williams 1987). X-ray counts in this experiment were on the order of 



102-103 for Sr and lO^-lO^ for 

 Ca. Confidence limits for the 

 Sr/Ca ratio were also calculated 

 as ± 2 * CV, but in this case the 

 standard deviation of the k-ratio 

 was calculated as 



OSr/Ca = 



OSr 



k-ratiosr 



OCa 



k-ratioca 



0.5 



Sr and Ca were analysed using 

 the TAP (Sr L-a) and PET (Ca K- 

 a) crystals. Background counts 

 were taken at ±(0.005* sin 6) 

 (where G is the angle of the spec- 

 trometer crystal when it is detec- 

 ting peak counts) for the same 

 length of time as the peak count. 

 Due to interference with a sec- 

 ond-order Ca K-a peak, only one 

 background count was made for 

 Sr. Strontianite (NMNH R10065) 

 and calcite (USNM 136321) were 

 used as standards. 



30 sec 



10,5,10,7nm 



20 sec 

 5,7, 10 Mn 



100 urn 



m 



Figure I 



Photomicrograph of otolith from 65.7 mm SL juvenile Dover sole Microstomus pacificus, 

 showing location of 12 microprobe transects used for analysis. Each circular area 

 represents one analysis. Note hyaline area near central primordium at inner end of 

 transects and more opaque area towards outer end. The 13th transect was an acciden- 

 tal repetition of the lOfim, 30-sec transect. Bar indicates lOOjim. 



Counting time and precision 



Counting time refers to the 

 length of time a spectrometer is 

 collecting counts of character- 

 istic x-rays for an element during 

 one analysis. Counting times of 

 10, 20, 30, and 40 sec were com- 

 pared for each beam-power den- 

 sity. The most commonly used 

 counting time for both elements 

 in previous studies was 20 sec (R. 

 Radtke, pers. commun. 1990), al- 

 though Kalish (1989) analyzed Sr 

 at 100 sec and Ca at 20 sec. Pre- 

 cision was determined as with 

 beam-power density. 



Transects of twelve analyses 

 each were made for the 12 com- 

 binations of beam power density 

 (4) and counting time (3) (com- 

 bined Af= [12*3*4] =144). These 

 transects passed from an area 

 near the central primordium to 

 an area just inside the discontinu- 

 ity created by accessory primor- 

 dia (Fig. 1). This discontinuity 



