SPEAKING FROM IGNORANCE 229 eliminate,the SFI effect in the no-prime condition.In the truth of a third iter When I could not see clearly.and i thought other prime d.I did not feel lik was wrong to repeat their answ claimed by values-pragmatics theory. dafncrhtcicier-idtenificaioeaskbtnonceabou Method sible Please give th The mber of im their g the answe ted an First.pat the vas of the Allowing ents of ignorance ince pa who cho his option ng th ntifiable appea it a .10 sed a persis as em to d when they onstrued as their insisting on giving a truthful statement of their ne of two inst ction n both condition they we Results Participants in both truth-prime and no-prime condition nswers (i.e. he win would be selected by random drawing fro M =6.3000f7. 597)They also trusted th ers of the po ins A and B It wi cy仁8，M=17.85 out of18orP A for truth prime no prime) hen introduced to the identification M=13.60 for no prime). e.g. t is the third lett word below the word er All po both truth-prime and and d pattems were agree only. agree-disagree. agree-ignorance.and gre re-ignoran tim truth-pri vas diffen nt (e.ggave a letter and then c ed ignorance.or the ngs thath have nd the conditic under which i urs.When do people cheat esults.For the truth-prime condition,disagreeing answers oc askinaway that wouldn oeteGeitpaicipea the time After the letter-identification task.partici ants answered ques yould diminish any chances of winnine money as much as dis. ions about th grecing answers e counted disagreeing and igno ced the situati and t they thou ions for no nswers a vn in Table 4.An exper imental test of SFI effects was con ted by employing a negativ and no-prime conditions. We expected that this additional reason for agreeing with correct answers would depress, or possibly even eliminate, the SFI effect in the no-prime condition. In the truth￾prime condition, it would present a demanding test of whether concern for truthfulness is a motivating force in the SFI effect, as claimed by values-pragmatics theory. Method Participants. Forty undergraduates (22 females) at a private liberal arts college participated, receiving either course credit or $5. Procedure. The procedure generally followed the same gen￾eral procedure used in Experiment 2, but with a number of impor￾tant changes. First, participants were only seated at Positions 2, 4, and 6 (see Figure 1), with the real participant at 6, and were not moved to other positions during the experiment. Second, six of the visual stimuli were repositioned so that they were visible and could be identified from Position 6, as established by pretesting with other subjects. Targets that were identifiable appeared on Trials 1, 2, 5, 10, 11, and 15. These trials encouraged participants in Position 6 to look carefully at the screen and to check it against what others said was the correct answer. They also served as a control condition for participants’ willingness to repeat correct answers when they could be verified. There were three changes from prior procedure in the instruc￾tions that were given. First, participants were randomly assigned to one of two instruction conditions. In both conditions, they were informed that they could win money by answering questions correctly. Each group was told that the group with the most correct answers (i.e., 3 points if all answered correctly, 2 points if two answered correctly, etc.) would win $30 and that if there was a tie, the winning group would be selected by random drawing from those tied. This monetary incentive was offered in order to make answering correctly even more obviously the socially normative thing to do. It was also intended to create a challenging test bed for the hypothesis that participants are motivated to speak truthfully and with pragmatic warrant. Second, participants were then introduced to the identification task used earlier, except in this experiment, we asked them to identify a particular letter in a particular word, rather than the word (e.g., “What is the third letter in the word below the word en￾gine?”). This change was expected to make the task seem even more precise and demanding, testing the robustness of procedures yielding an SFI effect. Third, participants were randomly assigned to one of two in￾struction conditions. In the truth-prime condition, three sentences were included when introducing the study to participants: “One of the things that we have recently become interested in is cheating and the conditions under which it occurs. When do people cheat? What is about a situation that leads people to cheat?” These sentences were intended to frame the remaining instructions and task in a way that would increase the probability that participants would be especially sensitive to being honest and truthful. After the letter-identification task, participants answered ques￾tions about the number of items they and others were believed to have answered correctly, a one-item index of whether they expe￾rienced the situation as a dilemma and what they thought the experimenter expected them to do (as in Experiment 2). Also included was a three-item Truthfulness index: Two items were the same as those used in Experiment 2 (see Table 2), with the addition of a third item, “When I could not see clearly, and I thought others could, I did not feel like it was wrong to repeat their answer,” which was reverse scored. There were several other items partic￾ipants answered after the letter-identification task but none about which we predicted differences relative to priming condition. Scoring protocol. Unlike prior experiments, there were three possible answers that were accepted. Participants could name a letter that agreed or disagreed with others’ answer or they could claim ignorance. If they said, “I can’t see,” or “I don’t know,” the experimenter reminded them of the instructions: “Please give the letter that is the best answer that you can give.” However, if they persisted in stating their ignorance, the answer was accepted and recorded. On succeeding items, if ignorance was claimed, it was recorded without a reminder of the instructions to name a letter. Allowing statements of ignorance seemed appropriate since par￾ticipants who chose this option were clearly undermining their chances to win money just as much as if they made an incorrect guess. Furthermore, it addressed a persistent question we had been asked about our earlier studies: “Don’t people just say they can’t see?” Finally, given that the instructions explicitly asked them not to give expressions of ignorance as an answer, for them to do so constituted another form of nonconformity, one that could be construed as their insisting on giving a truthful statement of their ignorance. Results Participants in both truth-prime and no-prime conditions strongly agreed that they experienced “a dilemma between saying what I knew (from what others had said) and guessing incorrectly” (Mtruth 6.30 out of 7, Mno prime 5.97). They also trusted the accuracy of the answers of the people sitting at Positions A and B, estimating that they had answered virtually all the questions cor￾rectly (e.g., M 17.85 out of 18 for Position A for truth prime, M 17.00 for no prime). Participants’ estimates of their own number of correct answers were lower (M 10.30 for truth prime, M 13.60 for no prime). Answers given by participants at Position C were scored as agreeing, disagreeing, or claiming ignorance, as described earlier. All possible response patterns occurred in both truth-prime and no-prime conditions except one, disagree only. The most common patterns were agree only, agree-disagree, agree-ignorance, and agree-disagree-ignorance. Three times, participants gave an an￾swer in one category, immediately adding a second answer that was different (e.g., gave a letter and then claimed ignorance, or the reverse). These were scored according to the first answer given, but scoring them differently makes no substantive difference in the results. For the truth-prime condition, disagreeing answers oc￾curred 13.3% and ignorance answers 36.6% of the time; for the no-prime conditions, disagreeing answers occurred 6.7% and ig￾norance answers 12.9% of the time. Given that statements of ignorance were not correct answers and would diminish any chances of winning money as much as dis￾agreeing answers, we counted disagreeing and ignorance answers together as nonconforming or nonagreeing answers. The distribu￾tions for nonconforming answers are shown in Table 4. An exper￾imental test of SFI effects was conducted by employing a negative This document is copyrighted by the American Psychological Association or one of its allied publishers. This article is intended solely for the personal use of the individual user and is not to be disseminated broadly. SPEAKING FROM IGNORANCE 229