The aim of functional magnetic resonance imaging (fMRI) studies of perception is to unravel the neural basis of the computational processing of the targeted perception. A standard fMRI study of perception generally uses simple stimuli, such as geometrical graphics in vision and pure tones in audition, and measures brain responses to such artificial stimuli, which are generated carefully. The target brain activity of percep- tual processing has been investigated using a contrast between stimulus conditions, to cancel out other processes. In contrast, the brain’s response to naturalistic stimuli is considered not to be a combination of responses to simple stimuli. The uncontrolled per- ceptual processes that are evoked in parallel hamper the analysis of the data in a simple factorial manner. In this article, I briefly reviewed fMRI studies that used naturalis- tic stimuli (e.g., photos and movies) and introduced nonstandard analytical methods. One was the computational model-based analysis of a hypothesis-driven study, and the other was the intersubject correlation of data-driven research. A model-based fMRI study can directly predict the brain responses to the processing of dynamic perception (e.g., motion perception in the middle temporal area). Intersubject correlation can be used to evaluate the reliability of fMRI signals in response to naturalistic stimuli. In addition, I introduced a decoding technique using pattern recognition, which has been used widely not only in the engineering but also in the neuroscience fields. Although the number of studies using naturalistic stimuli and novel analytical methods has in- creased, the standard fMRI study using simple stimuli and analysis remains the most effective approach to identify the neural bases of perception. I expect that these stan- dard and novel fMRI studies will contribute complementarily to the elucidation of the brain processes for naturalistic stimuli.