spectrometer according to the manufacturer’s instructions. Operation is 
facilitated by the use of an autosampler (P.E. Auto 200) and an ASR-33 
teletypewriter. To ensure against instrument drift, a calibration standard is 
included with each 15 samples. To check for unknown matrix effects, a known 
spike is added to an aliquot of one of the samples and an equal quantity of 5% 
nitric acid. From this, a spike recovery is calculated for each group of 15. For 
each group, a sample of the reference material described in this paper is 
included. 
Preparation of Reference Material 
Clams (Arctica islandica) were collected by commercial dredge from Block 
Island Sound, and frozen prior to use. At the time of preparation, they were 
cleaned, thawed and shucked as if for analysis, except that extracellular fluid 
was drained and discarded. A total of 637 clam meats was pureed and 
homogenized in a stainless steel, 40 quart mixer, of the sort found in many 
commercial kitchens (Hobart VCM-40). Samples of approximately 60 g wet 
weight were removed, placed in 120 ml acid-stripped polyethylene bottles, 
sequentially numbered, and frozen for future analysis. This procedure yielded 
476 samples. 
Characterization of Reference Material 
From the 476 samples, a total of 65 were selected for investigation of the 
homogeneity of the material. These consisted of every tenth sample, and two 
blocks of 10 consecutive samples from each end of the sequence. These 
samples were analyzed for 14 trace elements by the above procedure, and the 
results examined for homogeneity. 
RESULTS AND DISCUSSION 
The concentrations of 14 metals in the reference material on a wet weight 
basis are listed in Table 1-1. Only data on a wet weight basis will be discussed 
because of some anomalous wet to dry weight ratios indicating some samples 
were not uniformly dried. On the basis of several criteria, the 14 metals may be 
divided into two groups, A and B. Group A consists of the 10 metals Cd, Cr, 
Cu, Fe, Mg, Mn, Ni, Pb, V, and Zn, for which the relative standard deviations 
are less than 7%, as seen in Table 1-2. Three of these metals are graphically 
represented in Figure 1-1. Note the similarity of the graphs. This similarity may 
be quantified for these metals by determining the 45 pair-wise correlation 
coefficients. Almost all of the coefficients indicate positive correlation at the 
95% confidence level, with many of them being much more highly significant. 
3 
