This article describes an initial study of the effect of focused attention on phonological speech errors. In 3 experiments, participants recited 4-word tongue twisters and focused attention on 1 (or none) of the words. The attended word was singled out differently in each experiment; participants were under instructions to avoid errors on the attended word, to stress it, or to say it silently. The experiments showed that all methods of attending to a word decreased errors on that word, while increasing errors on the surrounding words. However, this error increase did not result from a relative increase in phonemic migrations originating from the attended word. This pattern is inconsistent with conceptualizing attention either as a higher activation of the attended word or greater inhibition of the unattended words throughout the production of the sequence. Instead, it is consistent with a model that presumes that attention exerts its effect at the time of production of the attended word, without lingering effects on the past or the future. (Contains 2 figures and 2 footnotes.)