The attentional boost effect (ABE), also called fast task-irrelevant perceptual learning, refers to a phenomenon that transient increasing in attention to one task enhances performance in a second task. The previous studies mainly investigated the temporal characteristics of ABE. The purpose of the present study is to investigate the spatial distribution of it. Specifically speaking, whether the centriod distance between the items and background images will influence the phenomenon or not. RSVP paradigm is employed and Participants are required to remember a series of scene images (500 ms/ image) while they performed a simple detection task (detect a white square, 100 ms/square) that was irrelevant to the scene images memory task. For the detection task, participants pressed a spacebar whenever the fixation point was a white square (target) rather than a black square (distractors). There are two conditions in this study: in one condition, squares overlap on scene images, and in other condition, squares are presented on the left or right side of the images (2.0°gap). After completing the dual-task encoding phase, participants performed a four-alternative-forced-choice (4AFC) recognition test on the scenes. Twenty undergraduates participated in the overlap condition and eighteen undergraduates participated in the gap condition. The stimuli were presented on a 19'' monitor (1024×768 pixels, 75 Hz) with a Lenovo computer (Pentium (R) Dual-Core CPU E5700, 3.00 GHz) using E-Prime. The main results were as following: (a) in the overlap condition, the likelihood that a scene was correctly recognized depended on the scene’s serial position during encoding (F(2, 62) = 4.36, p = .017), that is, scene identity memory was more accurate for scenes encoded at the target position than at the pre- and post- target positions (p = .040, p = .024), consistent with Swallow & Jiang (2010) 's results. (b) in the gap condition, the likelihood that a scene was correctly recognized depended on the scene’s serial position during encoding (F(2, 62) = 6.68, p = .002), scene identity memory was less accurate for scenes encoded at the target position than at the pre- and post- target positions (p = .041, p = .041), showing contrary to ABE. Two possible explanation for these results are purposed: reduced encoding account and attentional shift account. It was concluded that the centriod distance between items and background scene images has an impact on the ABE. When items (squares in this study) overlap on the background scene images (centriod distance equal zero), the observers have more accurate for scene encoded at the target position than at the pre- and post- target positions, showing ABE. And when items are presented on the left or right side of the images (2.0°gap), the observers have less accurate for scene encode at the target position than at the pre- and post- target positions.