The problems involved in testing speech-reading ability are studied here from two aspects; first, from that of logical validity of some test procedures, second, from that of applicability of analytical methods to thefollowing two tests. One of these two tests(test A)uses a list of 20 nonsense monosyllables, the other(test B) a list of mutually unrelated 30 words as speech material. The purpose of test B consisting of three sub-tests is to evaluate the effect of word-association upon speech-reading. Subtest 1: For exemple, a group of subjects(55 lower secondary deaf school pupils)receives a speech word "ito(thread)", after having been shown a written stimulus ward "hari(needle)". The speech intelligibility of the word "thread" is counted, and respresented by the symbol "b(thread-needle)". Similarly "b(water-ice)" etc, are derived. Subtest 2: The control group(n=54)receives only speech words such as "thread", "water" etc. The speech intelligibility of each words are represented as "c(thread)", "c(water)" etc. Subtest 3: Each subject is made to give an association-response to the word "needle". The symbol "a(thread-needle)" represents the percent of "theread" responses to "needle". Similarly, "a(water-ice)" etc.are derived. results: (see Table 5, 6.)(a, b, c, in two Tables are above-stated symbols.) Judging from these Tables, the correlation between "a" and "b-c" is doubtful. It is frequently observed that the interacting process of the informations obtained by two different processes, i.e.by the anticipation of succeeding words, and by the visiual perception of oral movement, is not a linear function of these two processes.