We discuss a robust yet at first sight surprising fact: individuals who have problems understanding sentences with object A-bar movement cannot use overt Case marking of the object to interpret these sentences and to associate the DPs with thematic roles. We tested the effect of overt Case marking of the object in typically developing Hebrew-speaking children by comparing their comprehension of which object questions with and without the object Case marker et, and found that there was no difference in comprehension between the two. A similar pattern was found in an adolescent with syntactic SLI. We then tested the comprehension of object topicalized structures in the order OVS, where the only element identifying these sentences as objectfirst sentences and distinguishing them from simple SVO sentences was the object marker. We tested this in three populations with object A-bar movement problems: individuals with agrammatism, adolescents with syntactic SLI, and orally-trained children with hearing impairment, as well as in analysis of previous data on typically-developing children acquiring Hebrew. All populations failed to understand the sentence, but did not consistently reverse the thematic roles of the two noun phrases. This suggests that they were sensitive to the presence of the Case marker but could not use it for interpretation. We argue that these findings immediately follow from the way intervention and locality are computed, under the featural Relativized Minimality approach. Case is not among the features triggering movement, therefore a Case difference is not taken into account in trying to build a movement chain across an intervener. As a result, the object chain cannot be built across the intervening subject in the relevant cases, and overt Case marking of the object cannot help rescue the structure. Thematic role assignment in complex movement configurations requires the building of movement chains; if chain formation fails, strategies based on overt morphological cues do not help. These results argue for a featureselective approach to locality and for encapsulated syntactic computation of movement.