12 



I add any modifying phrases necessary to make 

 these subtitles precise. The analysis, then, 

 consists merely in sorting the details of in- 

 formation into as many groups as there are 

 distinct subjects involved. Then for each 

 group, the scope of the information included 

 is determined and a subtitle is written just 

 as one would write a title for an article which 

 contained the information in that group. Each 

 subtitle should cover all the details in its group 

 and yet should not imply the inclusion of 

 material not actually contained; that is, it 

 should be both complete and precise. 



Rule 2. Each subtitle should describe the corre- 

 sponding information so precisely that the chance 

 of any investigator being misled into thinking the 

 article contains the particular information he de- 

 sires when it does not, or vice versa, may be small. 

 " Zeeman effect for metallic spectra " is too broad 

 unless all metals have been studied, for the investi- 

 gator may be interested at the time in only one 

 metal; but " Infra-red arc spectrum of iron to 

 3/(f ' ' evidently satisfies the rule. In general, a sub- 

 title is sufficiently precise if it carries the classi- 

 fication of the information involved three stages 

 or the equivalent, for instance if it gives {a) the 

 elements and substances, (6) the property, and (c) 

 the phase or range studied. 



Rule 3. The subtitles should together form a 

 complete index of tJie new information; that is, 

 they should include every measurement, observa- 

 tion, method, suggestion and theory which is pre- 

 sented as new and of value in itself. They should 

 be complete in themselves and independent of the 

 main title of the article. 



3. Text. — Not until the subtitles are formu- 

 lated, do I proceed to the writing of the text 

 of the abstract. I devote a paragraph to each 

 group of details beginning, as a rule, with 

 those relating to the main subject of the 

 article. In case several phases of a subject 

 are dealt with as in the second sample abstract 

 above, it is often convenient to begin the para- 



