A paper self-administered questionnaire (SAQ) can be a reliable and cost-effective approach to data collection. For the last few decades, the Total Design Method developed by Dillman has been the gold standard for effectively and efficiently collecting survey data. However, relatively low cost and ongoing improvements in Internet accessibility seem to be driving researchers to include self-administered online surveys conducted over the Internet (Web) as a data collection option. In many ways, SAQ and Web surveys are very similar. Each selfadministered mode allows respondents to answer privately and at their own pace, which Bishop (1988) suggests provides better quality data than interviewer-administered surveys. Further, a respondent's ability to comprehend both question text and response options is more similar in the SAQ and Web survey modes, than the interviewer administered mode. But these two self-administered modes are inherently different. In order to determine the effect of self-administered mode on data quality, this paper will make a qualitative comparison of data collected in the SAQ and Web modes in the 2003 Survey of Doctorate Recipients. The SDR is a panel survey administered biennially to a national sample of 40,000 doctorate-holders in the fields of science and engineering; the Web survey option was introduced in the 2003 round. More specifically, we will compare the data quality of 18,040 complete SAQs and 6,179 compete Web surveys. Additionally, we will compare the data quality between rounds for 3,580 sample members who completed a SAQ in the 2001 SDR and a Web survey in the 2003 SDR. Quantitative data evaluations will compare the amount of data editing required, overall item non-response, the amount of contacting information provided, and length of the key analytic verbatim data item.