The Keyword Method (KWM) of vocabulary learning is a mnemonic method designed to help students learn foreign vocabulary. It has been shown to be effective over several other memorization strategies. The present experiment was designed to explore the cognition underlying the effectiveness of the KWM. Specifically, both the degree of cognitive engagement with the method and visual encoding were examined as possible sources. Subjects were tested in a 2 (imagery level)×2 (processing strategy) mixed design. Each subject was asked to memorize 30 Latin vocabulary words, divided evenly among high- and low-imagery value words. Subjects were either provided with both keywords and interactions (the Given condition) or instructions to generate their own keywords and interactions (the Self-Generated condition). Retention was tested in both immediate and delayed post-tests. Results revealed a strong effect of imagery level in both post-tests. No significant main effect of processing strategy or interactions between imagery and processing were observed in either post-test. Results indicate that the KWM is effective because it provides a meaningful visual image upon which to base memory for a new word’s meaning. They suggest that there is some flexibility in how the KWM is used.