290 REPORTS ON THE STATE OF SCIENCE, ETC. 



3. Conformity to Previous Measurements. 



Whatever other measurements may be made, the value of the work for com- 

 parative purposes will often be much increased by the inclusion of measurements 

 which are comparable, as strictly as possible, with those taken in the same or related 

 species by previous workers. 



It is desirable that some quantitative measurements should always be presented 

 with general biological data. Length measurements are the most usual. However, 

 a frequent ' failure to record ' is seen in microscopical figures to which no record 

 of magnification is appended. For example, in the article on Botifera in the Cam- 

 bridge Natural History and in the Encyclopaedia Britannica no magnification is 

 given in any of the figures, nor are any measurements given in the text, so that the 

 reader (inter alia) will not be told, nor able to find out for himself, the interesting 

 biological fact that the Rotifera have the lowest average size and the smallest size- 

 range of any considerable Metazoan group. 



The most satisfactory way of giving magnifications is to reproduce with the figure 

 some unit of length magnified to the same scale. This obviates the error which 

 frequently creeps in when figures from one source are reproduced in another publi- 

 cation on a different scale, but without altering the statement as to magnification 

 in the legend. 



In addition measurements of weight or volume should be made whenever possible 

 as a matter of routine, since they provide the best standard of quantitative comparison 

 between differently shaped organisms or organs. 



4. Specification of Precise Conventions. 



It is essential to specify the conventions, including any points of reference adopted, 

 by which each measurement is defined. This can often best be done by the aid of a 

 diagram. When satisfactory standard terms, conventions or points of reference 

 already exist they should be adopted whenever possible. The aim should be to 

 ensure that a second observer, working over the identical material, and guided only 

 bv such specifications, should normally obtain significantly similar results. 



The state of preservation of the material may often affect the measurements, 

 especially in the case of soft parts. Accordingly the method of preservation and the 

 degree of contraction or relaxation of the parts should be noted. 



Observations of colour should when possible be referred to one of the standard 

 scales in general use, e.g. ' Nomenclature of colours for naturalists,' R. Ridgeway, 

 U.S. National Museum, 1912; ' Code des couleurs a Fusage des naturalistes, artistes, 

 cornmercants et industriels,' P. Klinksieck, Paris, 1908. 



5. Tests of Significance. 



The critical stages of the statistical examinations of a body of data are reached 

 in the application of what are known as tests of significance, (See Section D, 3-8.) 

 These are essentially tests whether the difference between two (or the variance among 

 several) groups is or is not greater than can with reasonable probability be ascribed 

 to the variability found within each group. The ultimate value of the conclusions 

 to be drawn from any data depends upon the precision and validity with which such 

 tests can be carried out ; consequently it is advisable that investigators, whether or 

 not they undertake the work of statistical analysis, should have a general acquaintance 

 with the nature of such tests, and, where the case does not seem clear, should seek 

 the advice or co-operation of a statistician. 



B. Presentation of Data. 



1. An incomplete specification of a sample is never to be preferred to a complete 

 specification, e.g. greatest, least and mean length is an incomplete specification 

 (see below). 



2. Single-variate Data. 



For a single measurement a complete specification of a sample may be given by 

 recording the number of cases observed to fall in successive intervals of magnitude. 



