The present study investigated capacity limitations of visual search. In a series of experiments, participants searched for a singleton target among homogenous distractors, a conjunction target defined by combination of two features, or a feature target among heterogeneous distractors. Using the simultaneous-sequential paradigm, I found that singleton search proceeded in a capacity-unlimited manner. By contrast, the performance of the conjunction search was found to depend on a capacity-limited process. For feature searches, the performance of searching for a specific color was not affected by how the stimuli were presented, while the orientation search performance was enhanced as the number of distractors simultaneously presented with the target increased. These results imply that distinct colors are individually coded, whereas multiple orientations are encoded as an ensemble in a structured way. Taken together, the present study clarifies which type of search process are capacity-limited and reveals how this limit can be overcome.