Reviews of 2,554 educational software programs by thirty reviewing sources were examined to compare the tendency of different review sources to give generally positive, negative, or neutral ratings. The reliability of language arts and reading software reviews was studied. Thirty-eight percent of the software programs were reviewed by two or more sources. In order to assess the extent to which certain review sources tended to rate the same software positively, negatively, or neutrally compared to other review sources, a probability analysis was undertaken. Results indicated that the review sources failed to agree on the ratings they assigned to software; that certain sources tended to rate all programs positively despite neutral or negative ratings from other sources; and that there was quite a difference between sources that evaluate software using uniform instruments and those that relied on reviews authorized by individuals. In conclusion, it is suggested that educators and parents seek out several reviews of programs they are considering purchasing, one of which should be a systematic review source such as MicroSIFT or EPIE Institute. Preview of the software is also recommended. (GDC)