The effects of superficial and structural information on online problem solving for good versus poor anagram solvers

Abstract

The two experiments reported here tested two predictions concerning the sensitivity of good and poor problem solvers to superficial and structural information during online problem solving: (a) Superficial features have a greater effect on solution difficulty for poor problem solvers, whereas (b) structural features have a greater effect on solution difficulty for good problem solvers. The tests were conducted in the domain of anagram solution by manipulating or measuring several superficial and structural characteristics in this domain. The results supported both predictions. They also indicated that better problem solvers have access to structural information from the earliest stages of processing (within the first 2 s). The authors discuss the implications of their results for the types of solution strategies used by more and less competent anagram solvers.

We would like to thank Elizabeth Lanthier, Marshall Lewis, and Megan McKinley for their help in collecting the data for Experiment 1, Katy Humphreys for collecting the data for Experiment 2a, and Adam Mudge and Angela Dobbins for collecting the data for Experiment 2b. Woo-kyoung Ahn, Miriam Bassok, Roger Dominowski, and Howard Sandler provided helpful comments on earlier versions of the manuscript.

Notes

1 The solutions are: minus, gavel, impel, gaily, axiom, genus, louse, gnome, milky, naive, power, tweak, rivet, smoky, chute, spurn, madly, scour, depot/opted, probe.

2 Because the rated pronounceability of the anagrams varied reliably across the three spelling constraint categories, it is important to consider whether the interactions in the solution time data of anagram competence with spelling constraint category and with pronounceability are really telling different stories as we predicted. We believe this is the case. Looking at the anagram competence by pronounceability interaction, we see that the difference in solution times for good versus poor solvers is larger for the easier to pronounce anagrams (mean difference for poor solvers minus good solvers of 8.10 s) than for the harder to pronounce anagrams (mean difference of 6.53 s). The manipulation check data showed that the anagrams for the unlikely solution words were somewhat easier to pronounce than were the anagrams for the other two types of solution words. Thus, if the two anagram competence interactions in the solution time data are alternative descriptions of the same phenomenon, one would expect the difference in solution times between poor and good solvers to be greater for the unlikely solution words than for the convergence and nonconvergence words. Contrary to this prediction based on the pronounceability ratings, the unlikely items exhibited a smaller expertise difference than did the other items.

3 The cut-off scores defining the anagram competence groups in this experiment were slightly different from those used in Experiment 1. Most of the participants in Experiment 1 were from Indiana University, where the average score on our anagram test is lower than that at Vanderbilt University, the source for all of the Experiment 2 participants. It was not possible to get enough students at Indiana with scores greater than or equal to 12, the cut-off we used in our earlier research (Novick & Sherman, 2003), so we had to reduce the cut-off for good solvers to 11 in Experiment 1. To compensate for this decrease, we also reduced the cut-off for poor solvers from 6 to 5 in Experiment 1. In both experiments, there was approximately a 10-point difference between the mean test scores for the two groups.

4 An analysis of these errors revealed no consistent patterns.

5 Novick and Sherman (2003) published a similar graph showing just the results from the present Experiment 1.