Measuring public opinion on abortion is an ongoing concern for political scientists, mainly because the public does not always exhibit fixed attitudes on such topics. Most citizens express a centrist viewpoint between the pro choice and pro life extremes. These include a small group whose answers to abortion questions are so inconsistent that they give public officials an inaccurate measure of public opinion on this important issue. Inconsistent responses may result from context effects, such as the order in which the questions are asked or the way they are asked. Usually, researchers ask a battery of questions in which respondents say whether they approve of abortion generally and under a variety of circumstances, citing the reasons for which a woman might seek an abortion. This project includes an independent national survey using questions adopted from the General Social Survey. The sample is divided into four experimental groups with different question orders. Based on these findings, the recommended question order would be the one with the general question last and the remaining specific questions in a somewhat random pattern alternating between the so-called “hard” and “easy” individual abortion situations. One of the more surprising findings is that people didn’t recognize themselves as subtracting the specific situations from the general question when it was asked first; hardly any said that was what they were doing when they gave inconsistent answers. Otherwise, about an equal number of respondents admitted answering the questions off the top of their heads as those who showed ambivalence by claiming they were deeply committed to their inconsistent responses. The study found most people who inconsistent on abortion are moderates leaning towards pro choice. Also, politically conservative regular church-goers can be just as inconsistent on abortion as the non-religious, non-political, low-educated non-church goers, especially if they are basically pro choice. Without a full understanding of who is generating inconsistent answers on abortion, some researchers may be tempted to eliminate these respondents from their sample. This research should allow them to understand these respondents better and develop better question wording and question orders to reduce their numbers.

This study, conducted in the spring of 2005, was designed to assess the contribution of computer experience on primacy effect and satisficing in online survey formats. Although regression analysis found that survey format can predict primacy effect, computer experience did not add to the explanation of the variance in primacy effect. Similarly, survey format was found to be a predictor for satisficing, however, computer experience did not add to the explanation of the variance in satisficing. In addition, there was not a significant correlation between computer experience and the number of words used to answer open-ended questions. Lack of variability of computer experience among respondents was a limitation in this study. Six survey formats (answer choices listed visible vertical, visible horizontal, pull-down menu, or check-all-that-apply with two surveys for each formats; answer choices listed in forward order and answer choices listed in reverse) were used to test for primacy effect. Initial linear regressions revealed that 6 of the 85 questions resulted in a positive and significant beta indicating primacy effect. Further regression analysis compared horizontally listed answer choices and pull-down menu survey formats individually against the vertical and visible answer choice survey format to determine if survey format explained a significant proportion of the variance in primacy effect. The interaction between survey format and answer order did not produce a significant positive correlation with primacy effect, therefore further investigation of the contribution of computer experience on primacy effect in those two survey formats was not necessary. Linear regression showed that the check-all-that-apply answer format does significantly explain the variance in primacy effect, however further analysis showed that computer experience did not significantly explain additional variance in primacy effect. Regression analysis showed that satisficing was more evident in the matrix survey format when compared with the visible vertical survey format, however computer experience did not significantly explain the variance in satisficing in these two survey formats. Finally, regression analysis failed to show that computer experience had an effect on the number of words used to answer open-ended questions (an indication of satisficing). / Ph. D.