Njoyment and conducted a one-factor within XR9576 chemical information subject ANOVA for each item. Table 1 compares the statistical results. For the “degree of easiness to make a request,” we found a significant difference among the conditions (F(2,52) = 6.538, p = .003, partial 2 = .201). Multiple comparisons with the Bonferroni method revealed significant differences: social > simple (p = .007) and social > caregiver (p = .01). No significance was found qhw.v5i4.5120 between caregiver and simple (p = 1.0). For the “degree of RR6 web comfort,” we found a significant difference among conditions (F(2,52) = 10.419, p <.001, partial 2 = .286). Multiple comparisons with the Bonferroni method revealed significant differences: caregiver > simple (p = .001) and social > simple (p = .001). No significance was found between caregiver and social (p = 1.0). For the “degree of enjoyment,” we found a significant difference among the conditions (F (2,52) = 12.026, p <.001, partial 2 = .316). Multiple comparisons with the Bonferroni method revealed significant differences: caregiver > simple (p = .029) and social > simple (p <.001). A significant trend was found between social and caregiver (p = .096). For the "total evaluation," we found a significant difference among the conditions (F(2,52) = 13.326, p <.001, partial 2 = .339). Multiple comparisons with the Bonferroni method revealed significant differences: caregiver > simple (p = .004) and social > simple (p <.001). No significance was found between caregiver and social (p = 0.33). Regarding prediction 1, the questionnaire results of the wheelchair robot without social behaviors were significantly lower than the caregivers in "the degree of comfort," "the degree pnas.1408988111 of enjoyment,” and “the total evaluations’, opposite to our hypothesis. The rest of items did not show significant differences between conditions. Thus, these results did not support prediction 1.PLOS ONE | DOI:10.1371/journal.pone.0128031 May 20,9 /Effectiveness of Social Behaviors for Autonomous Wheelchair RobotRegarding prediction 2, the questionnaire results indicate that the wheelchair robot with social behaviors showed a significant difference in terms of “the degree of easiness to make a request.” Other measurements are not significant and generally resemble those of the caregivers. We thought that these differences would increase the perceived ease of use, and the “intention to use” would also become higher than the caregiver condition. Thus, these results support prediction 2-a.InterviewsWe analyzed the interview results to investigate the reasons behind the different impressions among the conditions. Two coders analyzed and classified the transcribed interview results. We gathered 118 sentences of their impressions (S2 Table includes the coded data). Cohen’s kappa coefficient [31] from the classifications of the two coders was 0.726 and yielded the following: 1. I could more easily request moving support from the robot than the humans. 2. I could more easily request moving support from the humans than the robot. 3. The robot was safe for moving support. 4. The robot was not safe for moving support. 5. I liked/wanted speaking behaviors from the robot. 6. I did not like/want speaking behaviors from the robot. 7. The robot’s locomotion capability was adequate. 8. The robot’s locomotion capability was inadequate. 9. The locomotion capability of the humans was adequate. 10. The locomotion capability of the humans was inadequate. 11. Others Table 2 shows the number of each category i.Njoyment and conducted a one-factor within subject ANOVA for each item. Table 1 compares the statistical results. For the “degree of easiness to make a request,” we found a significant difference among the conditions (F(2,52) = 6.538, p = .003, partial 2 = .201). Multiple comparisons with the Bonferroni method revealed significant differences: social > simple (p = .007) and social > caregiver (p = .01). No significance was found qhw.v5i4.5120 between caregiver and simple (p = 1.0). For the “degree of comfort,” we found a significant difference among conditions (F(2,52) = 10.419, p <.001, partial 2 = .286). Multiple comparisons with the Bonferroni method revealed significant differences: caregiver > simple (p = .001) and social > simple (p = .001). No significance was found between caregiver and social (p = 1.0). For the “degree of enjoyment,” we found a significant difference among the conditions (F (2,52) = 12.026, p <.001, partial 2 = .316). Multiple comparisons with the Bonferroni method revealed significant differences: caregiver > simple (p = .029) and social > simple (p <.001). A significant trend was found between social and caregiver (p = .096). For the "total evaluation," we found a significant difference among the conditions (F(2,52) = 13.326, p <.001, partial 2 = .339). Multiple comparisons with the Bonferroni method revealed significant differences: caregiver > simple (p = .004) and social > simple (p <.001). No significance was found between caregiver and social (p = 0.33). Regarding prediction 1, the questionnaire results of the wheelchair robot without social behaviors were significantly lower than the caregivers in "the degree of comfort," "the degree pnas.1408988111 of enjoyment,” and “the total evaluations’, opposite to our hypothesis. The rest of items did not show significant differences between conditions. Thus, these results did not support prediction 1.PLOS ONE | DOI:10.1371/journal.pone.0128031 May 20,9 /Effectiveness of Social Behaviors for Autonomous Wheelchair RobotRegarding prediction 2, the questionnaire results indicate that the wheelchair robot with social behaviors showed a significant difference in terms of “the degree of easiness to make a request.” Other measurements are not significant and generally resemble those of the caregivers. We thought that these differences would increase the perceived ease of use, and the “intention to use” would also become higher than the caregiver condition. Thus, these results support prediction 2-a.InterviewsWe analyzed the interview results to investigate the reasons behind the different impressions among the conditions. Two coders analyzed and classified the transcribed interview results. We gathered 118 sentences of their impressions (S2 Table includes the coded data). Cohen’s kappa coefficient [31] from the classifications of the two coders was 0.726 and yielded the following: 1. I could more easily request moving support from the robot than the humans. 2. I could more easily request moving support from the humans than the robot. 3. The robot was safe for moving support. 4. The robot was not safe for moving support. 5. I liked/wanted speaking behaviors from the robot. 6. I did not like/want speaking behaviors from the robot. 7. The robot’s locomotion capability was adequate. 8. The robot’s locomotion capability was inadequate. 9. The locomotion capability of the humans was adequate. 10. The locomotion capability of the humans was inadequate. 11. Others Table 2 shows the number of each category i.