Several concepts have been developed to explore the function of attentional systems, such as an attentional "beam" proposed by Posner, Snyder, & Davidson (1980), which refers to an attentional system that enhances the efficiency of the detection of events in a directed region, and "perceptual glue" proposed by Treisman and Gelade (1980), which refers to the ability of an attentional system to bind separable features into unified objects. For comparing the attentional systems' functions of these two models, Briand and Klein (1987), using central and peripheral cues to direct subjects' attention to the likely locations of stimuli, instructed subjects to decide whether an R was presented in the stimulus display, which contained a pair of letters either from the set RPB (feature set) or from the set RPQ (conjunction set). The response latency showed a greater peripheral cue effects for RPQ set than for RPB set. It was suggested, therefore, that the attentional system oriented with peripheral cues performed the role of perceptual glue in Treisman' s feature integration theory. But from the perspective of topological hypothesis (Chen,1982) that topological discrimination occurs earlier than that of other geometric properties, R is topologically equivalent to P and Q (all of them possess one hole), while R is topologically different from B that possesses two holes. So it led us to reason that selective attention may playa different role-probably a weakerrole-in topological discrimination from that in discrimination of local geometric properties. Our present experiments demonstrate that while Treisman's feature integration theory fails to explain our data, the topological hypothesis provides an account good for both our and Briand and Klein's results.