In a word association task, the probability of producing a certain response to a cue is considered to be a direct measure of associative strength between words in the mental lexicon. The common single word association procedure is limited, since the number of words connected to a cue might be underestimated when a single response is asked. The continued association task overcomes this limitation by asking a person to generate multiple associative responses. To test whether continued strengths allow a better approximation of our lexicon, an experiment was conducted in which participants judged the associative strength between words. Our results show that in contrast to other semantic tasks, continued strength predicts weak to moderate judgments only. Two explanations based on the sampling of information and differential semantic activation of later responses in continued association are proposed. Theoretical implications for semantic activation and methodological implications for derivation of strength are discussed.