Certain aspects of brain function operate with a well-defined timescale. For instance, the timescale associated with a single action potential is reliably around a millisecond; a spike never takes 10 ms or 100 μs. Another reliable well-defined timescale is the day-long period of circadian rhythms. What is the typical timescale of cortical and thalamic neural firing rate dynamics? In this issue of The Journal of Physiology, Munn et al. (2020) show that this might be an ill-posed question. Indeed, they found that many neurons exhibit fluctuations in spike rates composed of many timescales simultaneously and that these diverse timescales are related in a quite peculiar way -like a fractal. As popularized by the visually stunning fractals of Mandelbrot, if one zooms in on a small part of a fractal, the structure looks the same as the structure at larger scales. A cartoon of how such self-similar structure could manifest in temporal fluctuations is illustrated in 9X zoom 1X zoom 81X zoom time spike rate Figure 1. Fractal structure of temporal fluctuations This cartoon illustrates, conceptually, how fluctuations of spike rate could be structured as a fractal.In the real spike rates analysed by Munn et al., the temporal structure is not so perfectly self-similar, rather the statistics (e.g. Fano factor) are the similar across different temporal scales. Fig. 1. Munn et al. showed that the temporal structure of fluctuations in spike rate can also be statistically self-similar as one zooms into shorter temporal scales. Thus, for these neurons, firing rate fluctuations do not have a single typical timescale. Rather, a wide range of timescales are equally important.Interestingly, not all the neurons Munn et al. analysed were equally fractal-like. They analysed spontaneous spiking activity of many single neurons, simultaneously recorded from lateral geniculate nucleus (LGN) and medial temporal visual cortex (MT) of marmosets. The neurons that were most fractal-like were those with the largest population coupling. In other words, the 'loner' cells that fired more independently of the population exhibited less fractal-like fluctuations. The cells that 'follow the crowd' , fluctuating together with the summed population activity, tended to have a large range of temporal scales with fractal structure. Moreover, Munn et al. examined several different cell types and found that konicellular neurons were the most fractal-like in LGN and that MT cells were often more fractal-like than LGN cells.