Three experiments are reported in which a naming task is used to investigate the effects of frequency blocking on lexical access time. The first two experiments examined the long-term effects of a training list of either high-frequency or low-frequency words on a subsequent test list. No consistent effects were observed. The third experiment tested for short-term effects of frequency blocking by comparing performance on the same set of words in pure-frequency and mixed-frequency lists. No advantage was found for pure lists, nor was there any modification of the graded frequency effect found in mixed lists. These results are compatible with a frequency-ordered search model of lexical access in which all words are listed in a single lexicon regardless of frequency, and in which the search always begins at the highest-frequency entry.