The Guided Search theory by Wolfe (1994) argues that information extracted preattentively could guide the shift of attention during visual search. This was supported by several recent studies in which the bias of the distribution of the saccadic endpoints was examined. The current study examined the flexibility of guidance during the search process by manipulating the ratios of the distracters within an array of fixed display size. The participants were asked to decide whether a green X was present or absent among same-shape distracter (red X's) and same-colour (green O's). Results suggest strong influence of distracter ratio on RT and patterns of eye movements. The reaction time, number of fixations and latency to move are longer when the two types of distracters were equally represented on the display than when one type of distracters was relatively rare. Furthermore, the distribution of saccadic frequency was flexibly biased by the distracter ratio. When there were few same-colour items on the display, the saccadic selectivity was biased towards the same-colour distracters. In contrast, when most of the distracters shared colour with the search target, the saccadic selectivity was biased toward the same-shape distracters. Larger saccadic bias was found in the first saccades than in the subsequent saccades. The results are discussed in terms of the top-down and bottom-up activation during visual search.