“Problems “in letters” are susceptible of more, and more interesting, tests than “problems in numbers”