This study investigates the relative effects of two types of input modification – linguistic and interactional – on Iranian EFL students' reading comprehension. Eight English reading passages were presented to 248 students in one of the three forms: unmodified (U), linguistically modified (LM), mostly in the direction of elaboration, and interactionally modified (IM). The students were also divided into two proficiency-level groups, i.e. more proficient (MP) and less proficient (LP) groups. Students' comprehension of the passages was measured through a 50-item multiple-choice test which was the same for all the six groups. The data were analyzed by a 2-by-3 analysis of variance (ANOVA). The results show that interactional modifications improve students' reading comprehension scores better than linguistic modifications at both proficiency levels. This suggests that linguistic modifications – even if they are made in the direction of elaboration as suggested by recent studies (Oh, 2001; Urano, 2002; Yano et al., 1994) – do not facilitate reading comprehension as effectively as interactional modifications do. Therefore, it is recommended that instead of making texts comprehensible through commonly-practiced techniques of simplification or elaboration, teachers employ authentic texts, but make them comprehensible through creating interactional modifications.