(National Bureau of Standards/NOAA standards). Values of analytical blanks were 

 subtracted and final concentrations were corrected for recovery of internal standards. 

 Analyte identifications were confirmed using a Hewlett Packard Mass Selective Detector 

 (MSD), model 5979, operating in the scan mode. The MSD was interfaced with a Hewlett 

 Packard gas chromatograph, model 5880, equipped with a DB-1 fused-silica capillary 

 column. Instrumental conditions included an ionization voltage of 2800 eV and scan 

 conditions of m/z 45-450 at one scan per second. Selected ion searches were used to obtain ion 

 chromatograms for compounds with known retention indexes that were suspected to be present 

 in the samples. If necessary, the mass spectrum and retention time of an identified peak was 

 retrieved and compared with an authentic standard or to a mass spectrum library to aid 

 identification 



The PCBs were identified on the basis of International Union of Pure and Applied 

 Chemists (IUPAC) congener numbers (Ballschmitter and Zell, 1980; Mullin et al, 1984). Since 

 there are many more PCB congeners in contaminated marine environments than we could 

 quantify in this study, 19 major congeners were chosen for analysis. These congeners represent 

 the different degrees of chlorination encountered in aroclor mixtures, are indicative of 

 specific aroclors, or are known to be biologically active. Congeners 18, 87, and 180 were used 

 to estimate the concentrations of Aroclor 1242, 1254, and 1260, respectively (Capel et al, 

 1985). For example, congener 18 represents 9.38 percent of the total congeners in Aroclor 1242. 

 In order to estimate the concentration of Aroclor 1242, the measured concentration of congener 

 18 is multiplied by (100/9.38). Congener 87 represents 3.32 percent of the congeners in Aroclor 

 1254. Congener 180 represents 6.5 percent of the congeners in Aroclor 1260. Congener numbers 

 66, 87, 118, 128, 153, 180, 195, and 206 are included since they are possible inducers of MFO 

 activity in animals (Clark, 1986). Further, congener numbers 66, 118, 128, 180, 195, and 206 

 have chlorine atoms in positions 4 and 4' and are thought to be preferentially degraded in 

 marine sediments (Brown et al, 1987). Congener numbers 87, 101, and 187 lack chlorines in 

 the 4 and 4' positions and are included in order to compare 4,4' congeners and non-4,4' 

 congeners. For example, in an unaltered Aroclor 1254, congener 118 is expected to be about 3.5 

 times more abundant than congener 87 (11.5%/3.3% = 3.5). For Aroclor 1260, the expected 

 proportion of congener 180 to 187 is 2.6. In the past, use of these analytical methods results 

 in 70 percent recoveries being within 50 to 90 percent. Analysis of split samples have 

 produced values within 10 percent of the mean for 80 percent of the chlorinated compounds 

 analyzed (Spies et al, 1988). 



Data Analysis Methods 



Sediment Toxicity Tests. Three attributes of the candidate toxicity tests were considered to 

 be of primary importance and were evaluated: (1) sensitivity of each end-point to test 

 sediments relative to respective controls, (2) within-sample analytical precision (i.e., low 

 analytical variability among replicates), and (3) total range in biological response to the test 

 samples relative to analytical precision (referred to hereafter as "discriminatory power"). 

 Of less importance, the tests should demonstrate some concordance in response with a range in 

 chemical contaminant concentrations. However, the evaluation of concordance between 

 toxicity results and chemical data assumes that the etiological agents in the environmental 

 sediments are among or co-vary with the chemical analytes that were quantified. This 

 assumption may or may not be correct, since the most sensitive toxicity tests may identify 

 some samples as "toxic" that otherwise would not be suspected as such based upon 

 quantification of a limited number of chemical analytes. Other unquantified chemicals may 

 occur in complex media such as sediments that are equally or more toxic than those that are 

 quantified. Finally, if all tests with similar end-points (e.g., acute toxicity) are responding 

 to related mechanisms of toxicity, toxicity data among the bioassays should demonstrate 

 concordance. Outlier toxicity tests may be responding to only "nuisance" variables or may be 

 insensitive . 



All sediment toxicity test data were tested for normality by either the Kolmogorov- 

 Smirnov test (Zar, 1984) or an approximation of the Shapiro and Wilke test. Since all end- 

 points were shown to be non-normally distributed (0.01 < p < 0.025), and could not be 

 transformed to a normal distribution, non-parametric tests were used for further data 



28 



