We examined whether typical frequency effects observed in normal reading would also occur in a target search task using non-linguistic Landolt-C stimuli. In an initial learning session, we simulated development of frequency effects by controlling exposures participants received of Landolt-C clusters during learning. In a subsequent scanning session, we manipulated the cluster demarcation form of linear strings of Landolt-C clusters (i.e., spaced vs. unspaced vs. shaded unspaced). Participants were required to scan and search for pre-learnt target clusters that were embedded in longer Landolt-C strings. During learning, frequency effects were successfully simulated such that targets with more exposures received shorter fixation time than those with fewer exposures. Participants were unable to successfully detect the pre-learnt targets when they were embedded in the strings during scanning. No evidence of frequency effects was observed in the scanning session. In contrast, eye-movement control was significantly influenced by cluster demarcation form, with increased difficulty for unspaced strings, less for shaded strings, and least for spaced strings. Furthermore, typical landing position distributions that occur in reading of spaced languages also occurred during scanning of spaced Landolt-C strings but not for the shaded or the unspaced strings. In conclusion, exposure frequency effects were successfully simulated during learning but did not carry over to target search during scanning of Landolt-C strings. Possible reasons why frequency effects did not occur in the scanning session are discussed.