272 REPORTS ON THE STATE OF SCIENCE, ETC. 



in which a star shaped piece of metal had to be disengaged from a pair of ' horse- 

 shoes,' was also introduced in the boys' groups. 



(5) School Examination Records were obtained for the boys and girls. In the latter 

 case it was possible to divide these into (a) ability at English, (6) ability at other 

 classroom subjects, and (c) ability at handwork. 

 Subjects : 



(a) An adult group numbering forty-seven subjects drawn from members of the 

 staff of the National Institute of Industrial Psychology and of the City of London 

 College, and from senior students at the College. Thirty of these, divided into four 

 groups, were ' practising ' subjects, taking first all the tests, then practising one of 

 the routine tests (both the ' assembling ' and the ' stripping ' variety) which differed 

 with the group, and finally being re-tested on all of the routine assembling tests. 

 The practice was carried out daily under test conditions for a period of a fortnight, 

 omitting Saturdays and Sundays. The remaining seventeen acted as ' controls,' 

 taking initial and final test only. 



(6) A schoolboy group consisting of the top two classes of a Tottenham elementary 

 school, numbering seventy boys. Thirty-eight of these were ' practisers,' divided into 

 two groups of eighteen and twenty respectively. Each group practised two of the 

 assembling tests (both ' assembling ' and ' stripping ' in each) daily for five con- 

 secutive school days ; they were given all the tests before practice and were re-tested 

 on the unpractised routine tests after the practice. The remaining thirty-two acted 

 as controls. 



(c) Schoolgirl groups drawn from the top two classes of a London elementary 

 school and from a backward class of girls of similar age in the same school. These 

 took the initial tests only. They total fifty-nine ' normals ' and twenty-two 

 ' backwards.' 



E. Chief Conclusions. 



The following conclusions emerge clearly from a preliminary examination of the 

 results of the investigation. 



I. Relating to accuracy of measurement in Routine Assembling. 



(a) A single ' trial,' i.e. the assembling of a single object, such as one ' container' 

 or one ' porcelain,' affords some indication of ' ability,' as shown by its correlation 

 with other single trials at the same operation. Its ' reliability,' however, although 

 tending to be ' significant ' (about thrice its probable error) is so low (about -3) as 

 to render it entirely untnistworthy as a measure. When the scores made at several 

 trials are added together a much more reliable measure is obtained. Thus on adding 

 together the five trials made by adults at the same routine test (' porcelains,' 

 ' containers,' ' wiring '), the reliability rises to over .70. Similarly, the 'reliability ' 

 of ten trials at the screwing test is, for screwing in, .63 and for unscrewing, .66 ; and 

 that of the ' wedges ' rises from .18 for a single trial to .52 for the sum of five trials. 



(6) If the measure must be confined to a single trial, it is much more reliable to 

 choose the best, or next best, or third best, &c., trial, than to choose the first, or 

 second, or third, &c., trial. The best, next best, <L-c., trials are almost as reliable as the 

 sum of all five trials, and there is little to choose between them on this score. 



(c) It follows from the foregoing observation that the disturbing influence of 

 random errors on the reliability of a short test of the kind here referred to is greater 

 than that exerted by systematic factors, such as practice or fatigue incurred during 

 the sitting- — and this in spite of the fact that such factors were discernible in the 

 curves. 



(d) The reliability of these routine tests depends upon the number of repetitions 

 included within the measure rather than on the length of time required for each repetition. 

 Thus the ' reliability ' of a single trial at the ' porcelain ' test, occupying several 

 minutes, is lower than that of the sum of ten trials at the ' screw ' test, occupying a 

 few seconds. When, however, we take as our measure the sum of ten trials at 

 porcelain assembling the reliability rises to a somewhat higher figure than that of the 

 screws (.86 as against .63) — and s imil arly for the other more complex assembling 

 tests. 



(e) If we include the same number of repetitions in our measures of ' ability,' the 

 routine assembling tests possess much the same degree of ' reliability ' when employed 

 with adults as with children. 



(/) The more prolonged period of practice has no clearly marked influence on 



