Study design: Psychometric study. Objectives: The objective of this study was to examine the reliability of the Spinal Cord Independence Measure III (SCIM III) by interview and compare the findings with those of assessment by observation. Setting: This study was conducted at Loewenstein Rehabilitation Hospital, Israel. Methods: Thirty-five spinal cord lesion (SCL) patients who underwent rehabilitation at Loewenstein Rehabilitation Hospital in Israel were assessed during the last week before discharge with SCIM III by observation and by interview. Nineteen of the patients were also assessed by interview by a third rater to examine inter-rater reliability. Total agreement, kappa, Bland–Altman plots and intraclass correlation (ICC) were used for comparison between interviewers and between interviews and observations. Results: Total agreement between the interviewers' scores and between interviews and observations was low to moderate (kappa coefficient 0.11–0.80). Bland–Altman analysis revealed good agreement, with low mean difference for almost all SCIM III subscales and total scores, between pairs of interviewers (bias −4.15, limits of agreement −22.51 to 14.19, for total score) and between interviews and observations (bias 1.62, limits of agreement −20.55 to 23.81, for total score). ICC coefficients for the SCIM III subscales and total scores were high (0.637–0.916). Conclusion: The findings of this study support the reliability and validity of SCIM III by interview, which appears to be useful for research of SCL patient groups. Individual scoring of SCIM III by interview, however, varied prominently between raters. Therefore, SCIM III by interview should be used with caution for clinical purposes, probably by raters whose scoring deviation, in relation to observation scores, is known.