Knowledge Networks (KN) uses RDD sampling to recruit a national representative panel of 40,000+ members. Panelists participate in 3-4 surveys per month via Internet or WebTV (provided to non-Internet households). KN also conducts cross-sectional RDD surveys for some clients. One such survey offers an opportunity to join the KN panel at the end of a 20-minute interview about Yellow Pages use. This paper compares this post-interview recruiting to KN's usual method of RDD surveys done exclusively for recruiting. Several stages exist in the process of getting "survey-ready" paneled members, with attrition at each stage. This paper examines non-response at each stage for these two recruitment methods. Weighting adjustments are discussed for ameliorating non-response bias across several demographic dimensions to create a merged, nationally representative panel from both recruitment sources.