The purpose of the present study was to disentangle the neurocognitive subprocesses underlying different spatial reasoning problems with event-related potentials (ERP) technique. Fourteen healthy Chinese young students (6 males and 8 females, mean age 25.2±2.2 years) without history of neurological disorders or chronic disease participated in this study. There were four kinds of three-term spatial reasoning task and one kind of memory task (baseline, BS). The term orders are AB-CB (A refers to the end term of the first premise, and C refers to the end term of the second premise) in all tasks. All stimuli were generated with E-prime 1.0 software and displayed in black with a white background on a computer screen. EEG was recorded from the scalp through 32 non-polarizable Ag/AgCl sintered electrodes in pre-configured caps, with a 500 Hz sampling rate using the Neuroscan EEG system. The ERP waveforms were time-locked to the onset of premise 2. Mean amplitudes in each time window were analyzed using three-way repeated-measures analyses of variance. The behavioral results indicated that memory task (BS) were easier than reasoning tasks (AO and AT-V) except for no valid multi-model problems (AT-NV). But both multi-model reasoning tasks were easier than one-model reasoning task. ERP data showed that reasoning tasks elicited more positive amplitudes than baseline task from 200ms to 900ms.The content-based spatial reasoning elicited similar waveform to that of no-content spatial reasoning. One-model problem elicited different waveform with valid two-model problem. The latter elicited a more positive waveform than the former in the time window of 300-600ms. Valid two-model problems elicited different waveform with two model and no-valid problems in the time window of 200-600ms. In conclusion, the behavioral data show that the place of indeterminate premises and the difficulty of spatial problems affected the process of resolving uncertain spatial reasoning. ERP data suggest that spatial reasoning and memory might reflect different requirement of attention resources. Reasoning need integrate information of two premises, while memory only maintain premise in working memory. When confronted with two-model problem, people process the stimulus and make a primary decision after 200ms, then construction of a single mental model with an annotation. Moreover, visual image do not affect spatial reasoning significantly.