A CRITIQUE OF CYTOCHEMICAL METHODS 241 



enough data are accumulated for statistical analysis it is found that these 

 values group themselves into normal unimodal distribution curves (Fig. 

 6-14). The cause of this variability is not fully known in any instance. 

 The instrumental and cytological variability can account for but a small 

 part; in many structures (e.g., interphase nuclei) a major portion appears 

 to be due to the distributional error. Until these errors of tec^hnique and 

 unavoidable errors due to cytological structure can be precisely estimated 

 (Sect. 4-3) the magnitude of the biological variation in any one popu- 

 lation of similar cells cannot be estimated; and therefore man}^ of the 

 possible more subtle correlations of nucleoprotein composition with 

 physiological, developmental, pathological, or experimental phenomena 

 cannot be investigated by microspectrophotometry. In the meantime, 

 it has been found possible to make many fruitful studies by utilizing the 

 fact that with proper cytological and photometric techniques the mean 

 values of any population are reproducible to about 10 per cent. This 

 holds when comparing different slides of one material (Table 6-5), when 

 examining the same animal at different times (Table 6-6) , or when study- 

 ing different animals of the same species (Table 6-6). This variability 

 of the means presumably is a measure of the over-all error of the technique 

 of preparing tissues for microscopic examination and of the photometric 

 procedure. In this type of relative quantitative analysis, a major 

 change of composition which may accompany cyclical, experimental, 

 or pathological events is detected as a change of mean value, outside 

 the normal range of variability. This reproducibility of the means 

 has been shown repeatedly for each of the techniques indicated in Fig. 

 6-13; indeed, it is customary to reexamine it for each new material 

 studied.^ 



An important aspect of these relative quantitative studies is that in 

 manj'^ cases it has been possible to compare them with results by inde- 

 pendent methods, such as biochemical analysis of samples of known num- 

 bers of isolated cell components. For example, the correlation of 

 amount of DNA with the chromosome complement of nuclei, first indi- 

 cated by biochemical analysis, has been confirmed and greatly extended 

 by cytological studies made with the Feulgen reaction, whereas the pro- 

 tein composition of the nucleolus, suggested by results with the Millon 

 reaction, has been confirmed by both X-ray absorption data and analysis 

 of masses of isolated nucleoli. Many of these essential validations of the 

 cytochemical analyses are summarized in PoUister (1952a). 



* The approach of the Caspersson school to microspectrophotometry has ])een 

 almost entirely qualitative and semiquantitative. This is emphasized particularly 

 by the fact that in the total of nearly a hundred publications (see list in Caspersson, 

 1950) there is not one demonstration of the statistical reprodui'il)iIity of the ultra- 

 violet absorption terhiii(|U(', nor is the urgency and priority of such a demonstration 

 ever recognized. 



