samples were prepared in triplicate. The 

 weight of the precipitate was calculated for 

 volumes less than 1.0 ml. on the basis of the 

 mean weight per milliliter of the precipitate 

 as determined from weighing the heavier 

 (i.e., > 1ml.) pads with an analytical balance. 



The diameter of the filter assembly bore 

 was measured, and the weight of precipitate 

 per square centimeter of precipitate cal- 

 culated. The activity (counts per minute) of 

 the precipitate per milligram was also com- 

 puted, and the results plotted on semiloga- 

 rithmic paper, with the counts per minute per 

 nnilligram as the ordinate (see fig. 5). 



An eye -fitted curve was drawn through the 

 points, and the counts per minute at zero 

 weight defined by the point at which the extrap- 

 olated smooth curve intersected the ordinate 

 (i.e., zero weight). Duplicate and triplicate 

 determinations always differ, especially at 

 low weights, but disagreement between zero- 

 weight activity with different samples of the 

 same batch of C-"-* solution has not exceeded 

 + 12 percent or been less than + 8 percent of 

 the nnean value for a particular batch. 



0.2 0.4 06 0.8 1.0 1.2 1.4 1.6 



MG BoCOj PER CM^ 



Figure 5. — Self-absorption curves of BaC O3 at two 

 dilutions of C 



Evaluation of Incubation Techniques 



The purpose of trailing bottle, deck, and 

 laboratory incubation was to obtain unbiased 

 estimates of in situ production without using 

 large amounts of vessel time. The final 

 reference in these comparisons is, then, the 

 in situ measurement itself. The unproven 

 assumption has been made that the in situ 

 observations yield rates which are similar to 

 rates in nature; whether or not this assumption 

 is correct remains to be shown. 



Data for a comparison of these techniques 

 are not abundant, and the inherent incompat- 

 ability of several of the techniques makes the 

 following analysis somewhat indirect. The 

 analysis is in the following order: (1) a 

 connparison between in situ and laboratory 

 incubator rates as applied to the isothermal 

 layer and upper 100 m.; (2) a comparison 

 between surface sample rates obtained by 

 trailing bottle and deck incubation; and (3) a 

 comparison between rates for surface in situ 

 and deck incubated samples. 



The data used to assess the agreement be- 

 tween in situ and laboratory incubator rates 

 have been published elsewhere (Holmes et al., 

 1957; Holmes et al., 1958; Holmes and Black- 

 burn, 1960, Blackburn et al., 1962). These 

 data are considered from two points of view: 

 the agreement between measurements in the 

 isothermal layer; and the agreement between 

 water column (i.e., 0-100 m.) production 

 values. 



Graphs were prepared of observed in situ 

 and laboratory incubator production rates as 

 a function of depth, and the points were 

 joined with a smooth curve. The production 

 values for the water column were obtained by 

 integration with a polar planimeter from a 

 depth of 100 m. to the surface. Isothermal 

 layer rates were integrated in the same man- 

 ner, on the same graphs; the depth of the 

 thermocline was determined subjectively by 

 visual inspection of the station bathyther- 

 mogram. The results of these integrations are 

 given in table 1 3. 



The primary production rates, in situ and 

 incubator, for both the isothermal layer and 

 water column are positively correlated at a 

 highly significant level (see table 13 for 

 Spearman rank-difference coefficients, r^ ). 

 Since the incubator rates are expressed on an 

 hourly basis, they are not directly convertible 

 to in situ rates. Laboratory incubator rates 

 may be converted to approximations of in situ 

 rates by multiplying laboratory rates by the 

 average of the individual laboratory incubator 

 in situ quotients, or by the quotient of the 

 median ratio. The success of this manipula- 

 tion may be judged by the Mann- Whitney U test 

 (Siegel, 1956); in both series the smaller of 

 the U values calculated, 170 and 177 respec- 

 tively, have associated probabilities of greater 



25 



