Students (N = 124) viewed 14 timed Web pages that distinguished 14 confusing word pairs. In a 2 x 2 factorial design, the authors gave all of the students matrices containing representational pictures for each pair of words, as well as examples of each word in use. One factor in the design was the absence or presence of rules of usage for each pair, which did or did not prompt rule construction, and the other factor was audio or text presentation of the examples or rules + examples, which could affect cognitive load. Results were consistent with the construction hypothesis and previous research. Students in rule-absent conditions (a) generated more novel examples of the words in use and (b) explained the rules of usage better on written tests than did students in rule-present conditions. (Contains 1 table and 5 figures.)