Abstract: Visual mental image generation had been defined as the process by which long-term memory knowledge of the visual appearance of objects or scenes is used to create a short-term percept-like image. It was a complex and multi-componential cognitive process, and a most basic process in visual mental image process. That researched systematically on it would not only favor us deeply understanding the law of human visual mental image process, but examining and evaluating visual spatial cognition functional level most effectively in practice. The localiza-tion of mental image generation had been a controversial topic, whether hemispheric specialization really ex-isted in visual mental image generation process? What roles that both hemispheres played in image generation process? All these questions needed to be explored again.
Using two tasks dissociated paradigm developed by Kosslyn, three behavior experiments were conducted to reveal the Hemispheric Specialization during visual image generation process. Seventy two right – handed un-dergraduates (aged between 19 and 21) took part in the experiments. In three experiments, a classical image generation task was adopted to investigate the cognitive processing mechanism of simple letters image genera-tion. Subjects were cued to form images within a grid or within a set of four corner brackets; a single X mark was enclosed within each stimulus, and the subjects were to determine whether the X mark would have fallen on an imaged pattern. Reaction time (measured from the end of the question to the onset of the participants’ re-sponse) and error rates were recorded and analyzed. The aim of these experiments was to clarify the different roles of two hemispheres in generating the mental images.
Some evidence was obtained by having subjects memorize letters in grids(which are easily encoded using descriptions of the positions of segments) or within a space delineated by four brackets(which require memo-rizing the precise positions of the segments). In Experiment 1, we found evidence that the right hemisphere was better at generating images in brackets, whereas there was no such hemispheric difference for the grids. In Ex-periment 2, we found no evidence that size per se affected image generation differently in the two hemispheres, indeed, this experiment only replicated earlier results, with a right hemisphere advantage for brackets stimuli. In experiment 3, the subjects required less time when the stimuli were presented initially to the left hemisphere, which was as expected if categorical spatial relations were used to arrange segments into an image. Through these experiments, Subjects were relatively more accurate when cued in the left visual field with bracket stimuli, but tended to be relatively more accurate when cued in the right visual field with grids stimuli. These results were predicted by the theory that images are built up by arranging parts, and that two different processes can be used to arrange them. One process used stored descriptions to arrange parts, and was more effective in the left cerebral hemisphere; The other process used stored memories of metric positions to arrange parts, and was more effective in the right cerebral hemisphere.
The present results suggested that visual mental images could be generated by either the left or right cere-bral hemisphere, but in different ways. The left hemisphere more effectively generated images by arranging parts according to descriptions (using categorical spatial relations), whereas the right hemisphere more effec-tively generated images by positioning parts in precise locations in space(using coordinate spatial relations).

Key words: image generation, hemispheric specialization, spatial relations, categorical /coordinate