Maximum short segments of quantity in 120 discrimination tests at the word level were compared between two- and three-choice alternatives and with production test results utilising the same syllable structures. The materials in the perception test were eight kinds of bisyllabic synthetic nonsense words with five different prosodic patterns. These structures were used in the production test. 28 Finnish and Japanese subjects participated in the two perception tests and three speakers of each language in the production test. The results showed that (1) the short vowels/consonants in the two perception tests were longer than in the production tests; (2) the word-structural differences had more effect than the prosodic conditional differences in both Finnish and Japanese in both tests; (3) the short segmental durations were shorter in Finnish in the three choices than in the two choices but the Japanese counterparts were the reverse; (4)there was a correlation in some cases between perception and production.