Nominative and accusative case markers are frequently dropped in colloquial Korean. Prior studies demonstrate a nominative-accusative asymmetry in frequency and acceptability of case-drop sentences, which correlates well with the implications of the actor identification strategy (AIS) of the extended argument dependency model (eADM). This study investigated the reading of case-drop sentences by native Korean speakers using the eye tracking methodology. While the nominative-accusative asymmetry was only partially observed in the reading time data, our results showed an important trade-off effect in the processing of case-drop sentences: a decrease in reading time on the bare noun phrase (NP) was balanced by an increase in reading time on the case-marked NP. A critical exception to this trade-off effect suggested that extra processing costs can be avoided when the AIS works in an optimal condition.