Response spectrum matching is commonly used to generate ground motions with response spectra matching a scenario target spectrum. There is some debate in the literature about whether spectrum-matched motions lead to biased structural analysis results. Furthermore, there are no objective, quantitative criteria available for deciding whether a ground motion has been manipulated excessively by spectrum matching, and whether large modification may also lead to bias. This study investigates both of these issues by presenting the results of structural analysis using two reinforced concrete moment frame models and two earthquake scenarios, with suites of unmatched and matched ground motions. Through comparison with a robust benchmark, it is shown that no significant bias is introduced by spectrum matching. The period range and target damping values for matching are also investigated, and matching up to three times the fundamental period is shown to be beneficial in reducing dispersion in the results. Finally, these analyses were also used to investigate whether large changes in the ground motion lead to biased analysis results. Several potential measures of change are investigated, including those based on peak absolute ground motion, cumulative squared ground motion (absolute or normalized), and input energy into single-degree-of-freedom systems. Although no systematic, statistically significant correlation is found for the analysis results in terms of any of these measures of change, tentative criteria are proposed, which may be used by analysts to aid in the decision of whether to accept or reject a spectrum-matched motion. Copyright © 2012 John Wiley & Sons, Ltd.