In fMRI, it is often assumed that fMRI data are normally distributed in real and imaginary parts which lead to Ricean distributed in magnitude, which is approximately normal when the signal-to-noise ratio is large. Here, two fMRI data sets were acquired to examine this assumption. Both data sets image a plastic sphere containing an agarose gel. The first data set utilizes a typical fMRI pulse sequence to acquire data except the RF which produces the signal is omitted. The resulting data should be purely normally distributed noise in real and imaginary parts. Subsequently, a second data set is acquired with the same pulse sequence except the RF was not omitted. The resulting data should be normally distributed in real and imaginary parts which lead to Ricean distributed in magnitude. The noise properties of the data are explored.