Four paradoxes appear in research on learning to read: (1) the ability to name letters is a good predictor of reading readiness, yet letter-naming training does not help children learn how to read; (2) visual discrimination is often better in poor readers than in good readers; (3) learning to read two languages is easier than learning to read one; and (4) it is easier to learn to read in two alphabets than it is in one. These findings which contradict common sense may be easily explained by examining the thought processes at work during the reading process. Too often reading research looks at external aspects of reading such as eye movement, perception, and letter-naming and neglects the central processes of concept formation and reasoning. Examining the first paradox, for example, early experience with letter-naming often indicates an environment in which the parents read a lot and talk about reading, which gives support to the school's effort to teach reading. Learning letter-names for a child from a less stimulating environment, however, is often the rote learning of meaningless symbols. Hence letter-naming can indicate reading readiness but is not a useful method of teaching reading. The other three paradoxes can similarly be explained by looking at the cognitive processes involved. The implication then is that the learning and thought processes of the child must be the starting point for any teaching activity. References are included. (AL)