Should instruments, such as Likert-type scales, contain both positively worded and negatively worded items within the same scale (i.e. mixed format)? Recent evidence suggests that the use of scales with a mixed format can adversely affect the psychometric properties of scales. In particular, the mean item response to the positively worded items has been found to be statistically and practically significantly different than the corresponding mean item response to the negatively worded items with the same scale. Because positively and negatively worded items may evoke different responses for some individuals, it is likely that reading ability plays an important role in the quality of responses to scales with a mixed format. Indeed, Marsh (1984) found a relationship between inappropriate responses to negatively worded items and reading ability among pre-adolescents. Likewise, Williams and Swanson (2001) found a similar relationship for adults. Yet, no other studies have investigated this relationship. Thus, the present study extends this line of research by examining this association among 153 graduate students. A canonical correlation analysis revealed a moderate relationship between degree of differential responses between positively and negatively worded items on four dimensions of a measure of library anxiety and components of reading ability (i.e. reading comprehension and reading vocabulary).