Shepard's tones are a typical example for auditory illusion. They consist in a series of computer generated tones, which prohibit relative pitch discrimination. As a result, when repetitively played in sequence, the illusion of an ever-ascending scale is evoked. In order to investigate this aural phenomenon, fMRI time series were acquired during presentation of a conventional block-designed paradigm as well as during continuous presentation of Shepard's tones. With respect to the different setups of the two experiments, two fundamentally different methods were applied in order to conduct data analysis. Common Statistical Parameter Mapping served to evaluate the time series obtained with the block-designed paradigm. For the continuous experiment, a novel wavelet-based multifractal analysis was used, recently proposed as a classification tool for fMRI time series. This approach applies the wavelet transform to extract multifractal spectra from time-signals. For reasons of quantification, we introduced an ameliorated method for visual inspection of the multifractal properties. The results proved existence of characteristic neural responses to continuously presented Shepard's tones. Interestingly, the same was not restricted to the auditory cortex, but also involved areas of the visual cortex. Related impact on the imaged cognitive areas, primary motor cortex, and primary sensory cortex could not be observed. We further provide evidence that pitch misjudgment does not occur in temporal concurrence with the repetition of the whole scale, but according to whether the main perceived frequency is located in the sensitive range of auditory perception or not. We remark that this is the first time, continuously stimulated brain areas could be detected by means of fMRI.