As a result of this collection of information it was possible to compile a number of questions, usually from forty to sixty, each of which called for an answer that showed knowledge of the trade. Experience in the formulation of such questions has shown that a good question meets the following requirements: 1. It must be in the language of the trade. 2. It must be a unit, complete in itself and requiring no further explanation. 3. It must not be a chance question that could be answered by a good guess. The extreme example would be a question calling for the answer “yes” or “no.” 4. It must be as short as possible and must be capable of being answered by a very short answer. 5. It must not be ambiguous; the meaning must be unmistakable. After the large number of questions originally formulated has been sifted down by application of the requirements stated above and others of less importance they are used in a preliminary sampling on a number of artisans engaged in the trade, usually from nine to twelve, whose answers indicate the merits of the different questions and their grades, from easy to difficult. In this sampling, tradesmen from different shops or plants are tried in order to guard against specialized methods or modes of expression confined to a single locality. At least two examiners worked on each set of questions at this stage, in the Army’s work, to get the benefit of more than one point of view for revision. This preliminary sampling affords a means of checking on the following points: 1. Is the test applicable to trade conditions? 2. Does the test represent good trade practice? 3. In what way can parts be profitably modified, supplemented, or eliminated? 4. Does the test represent the whole range of the trade, from the novice to the expert? 5. Is it a representative sampling of the whole range of trade processes? In the light of the answers to these questions, the test is revised during this sampling process and is then ready to be formulated. This formulation consists of limiting the questions to a small enough number to be handled in a short space of time and to a wide enough range to represent every possible degree of trade skill. The questions are tabulated and are then ready to be used in the final sampling process. Final sampling is made by testing twenty men who are known to be typical representatives of each group (novice, apprentice, journeyman, expert). Among the novices tested are some highly intelligent and mature men of good general knowledge but no trade ability. Three testing stations were used in the Army’s work: one in Cleveland, one in Newark, and one in Pittsburgh, in order to get the benefit of wide geographical distribution. Examinations were given to men whose record in the trade was already known and who were tested as nearly as possible in the same manner as men in the camps.