Knowledge Networks has tested a new approach to market research in the United States by using a random-digidial sample for panel recruitment and equipping panel households with a device connecting their television sets to the Internet. This approach was used to field five experimental magazine reading questionnaires to this panel. Overall, recent reading levels from the test closely reflected those of MRI. Estimates of subscription levels also came close to actual subscription levels. While there were differences between the experiment's recent reading data and MRIs that were related to the position, size, and clarity of the logos in the questionnaire, and differences by publication frequency, these can be addressed through adjustments to the questionnaire. There were also differences between the experiment's and MRI's estimates with respect to genre, but these differences suggest one of the possible advantages of self-administration of magazine audience surveys.