In the last few years, fractal analysis has found widespread application in
the field of neuroscience and some investigators are starting to use multi
fractals as a methodology that may provide information about the distributi
on of fractal dimensions in biological structures. This is so, despite of t
he technical difficulties of multifractal analysis. In this paper, we inves
tigate the theoretical and practical aspects of studying and measuring the
multifractal dimensions of neurons. Patterns were analysed by means of the
standard box-counting method and a generalised sand-box method. Our results
show that odd behaviours of D-q reported in the literature are a consequen
ce of numerical deficiencies of the box-counting method and cannot be assoc
iated to peculiar geometrical characteristics of neurons. Instead the sand-
box method gives a D-q which monotonically decreases with q. Although this
result may indicate that neurons are multifractals, it is argued that size
effects may in fact be the origin of this apparent multifractality. (C) 199
9 Elsevier Science B.V. All rights reserved.