Skip to main content
. 2016 Oct 21;6:35723. doi: 10.1038/srep35723

Figure 1.

Figure 1

(a) Traditional CFS involves a dynamic Mondrian composed of randomly positioned shapes varying in size and luminance presented usually to the dominant eye and a smaller target to the other. Targets may be natural images or simple stimuli. (b) Dynamic Mondrians are commonly updated at 10 Hz by presenting new patterns every 100 ms. Because the grey levels of the Mondrian shapes vary randomly, some undergo large luminance changes between patterns (brown squares) whereas others change little or not at all (blue and yellow squares, respectively). Statistically, over a sequence of frames, the latter is much more likely, and this lengthens the period of the modulation and thus lowers the frequency. (c) Even if strong alternations did occur between the extreme luminance values, a 10 Hz Mondrian update rate would produce a 5 Hz square-wave modulation (grey dashed line, left) with a peak at 5 Hz and lesser peaks at the odd harmonics (grey dashed line, right). The actual waveform, however, is inevi complex with low frequency components due the presence of multiple grey levels (here, n = 5) and non-uniform changes over time (black solid line, left). Consequently, the temporal spectrum is broader with a concentration of energy at frequencies much lower than the intended update rate (black solid line, right). (d) To demonstrate the low-frequency bias, we tracked the pixel timelines of 70 grayscale Mondrian patterns updated at 2, 5, 10 and 20 Hz (randomly sampling from 5 grey levels, 200 pixels each refresh rate), then Fourier transformed the data. The resultant amplitude spectra for all refresh rates have a very strong low-pass profile. For the typical 10 Hz Mondrian, only 1.3% of total stimulus energy occurs at 10 Hz and the peak frequency occurs at 1 Hz, which has more than 20 times the energy (31%) of the 10 Hz component. Raising the Mondrian update rate does little to boost high-frequency content and the strongly lowpass profile remains. Indeed, as the functions decline with frequency, they could be well described as temporal “pink noise”.