This is an investigation into the predictability of FLAT (Foreign Language Aptitude Test). It is also an attempt to make some improvement. (1) FLAT was planned out in 1955 by one our colleagues and tested on the students (who were then in the 7th grade of a junior high school in Tokyo.) (2) We kept records of the students each year. (3) The school marks record of each year showed a significant correlation with each other. This may be a proof that the teachers' rating is reliable. [table] Our self-made English word test showed a correlation coefficient. 7 of with the school records. To find out the new items which should be added to FLAT, the relation was examined between the tests included in two kinds of intelligence tests and those of our English achievement tests. The results showed that : (4) ELAT did not show as high a predictability as the schoo mark records. In order to examine the predictability of each part of FLAT, the following investigation was made. (5) The correlation was examined between the tests included in FLAT and those of our self-made English achievement test. (6) The following items showed rather high correlations with each other. (7) The sub-tests of two intelligence tests which test verbal factors had more relation with English ability than non-verbal items. (8) Some sub-tests using figures and other quantitative items had a high relation with English ability, but no logical reason could be seen for this. FLAT To find the words which have the similar meaning to the key words (in Japanese) To arrange the words which were given at random into correct order (in Japanese)ENGLISH TEST TOTAL word test translation into Japanese translation into English special word test (words both in isolation and in sentences)