2000 Special Issue Tracking functions of cortical networks on a millisecond timescale V. Jousma ¨ki * Brain Research Unit, Low Temperature Laboratory, Helsinki University of Technology, P.O. Box 2200, FIN-02015 HUT, Espoo, Finland Received 7 May 2000; accepted 11 July 2000 Abstract The human cerebral cortex, consisting of six layers and billions of neurons and synapses, processes sensory input from numerous sensory receptors. Noninvasive magnetoencephalographic (MEG) recordings provide a view through the skull to electrophysiological signals of the cortex on a millisecond timescale. For example, magnetic somatosensory evoked fields (SEFs) to a given peripheral somatosensory stimuli, reflect sequential activation of an extensive cortical network. Several cortical areas contributing to the SEFs can be evaluated in time and space by using source modeling. This brief review focuses on MEG studies of the human somatosensory networks with a special emphasis on tactile stimulation.  2000 Elsevier Science Ltd. All rights reserved. Keywords: Primary somatosensory cortex; Secondary somatosensory cortex; Magnetoencephalography; Mechanoreceptors; Vibration; Airpuffs; Tapping; Human 1. Introduction 1.1. Magnetoencephalography In magnetoencephalography, weak magnetic fields due to the neuronal currents of the brain are measured outside of the head by means of a neuromagnetometer (Fig. 1). Neuro- magnetic measurements are typically carried out in a magnetically shielded room to attenuate ambient noise from the environment. The theory and instrumentation of the MEG method have been reviewed in detail by Ha ¨ma ¨la ¨inen, Hari, Ilmoniemi, Knuutila, and Lounasmaa (1993). The MEG method has developed gradually since the first measurements with induction coils in the late 1960s and the invention of magnetometers based on the superconductive quantum interference device, SQUID (Josephson, 1962), in 1970s. Multichannel sensor arrays were soon available providing simultaneous mapping of the magnetic field above a given brain region. The first whole-scalp neuro- magnetometers were introduced in 1992 (Ahonen et al., 1993; Vrba et al., 1993). Modern whole-scalp MEG instru- mentation, with hundreds of channels and an user-friendly control interface, provides a noninvasive tool to investigate human brain functions with a millisecond time resolution. MEG signals are mainly generated by the postsynaptic currents of the pyramidal neurons of the cortex. Both anatomical and electrophysiological factors contribute to the generation of neuromagnetic fields. The magnetic signal of a single postsynaptic potential (PSP) is far too small to be detected outside the head with current instruments (Vvedenski, Hari, Ilmoniemi, & Reinikainen, 1985) However, one PSP lasts about 10 ms providing a temporal window in which magnetic fields may summate. When thousands of adjacent pyramidal neurons are activated, a net current normal to the cortex and corresponding magnetic field are generated. Since the head resembles a sphere and the currents are normal to the cortical surface, some cortical areas are magnetically silent due to radial currents. However, fissural cortex, giving rise to magnetic signals, covers roughly 2/3 of the cortex and thus most of the cerebral cortex can be reached by MEG recordings. The layered structure of the head distorts the electric potential distribution on the scalp, i.e. electroencephalo- gram (EEG), whereas the magnetic field distribution is not affected. Thus, MEG is more suitable to study multiple spatially and temporally overlapping activations of the brain. The usefulness of MEG depends largely on source model- ing, i.e. the ability to estimate the current distribution from the measured magnetic field distribution. The ambiguity of this neuromagnetic inverse problem can be restricted by introducing constraints based on physiology and anatomy of the brain. For example, the sources can be constrained to Neural Networks 13 (2000) 883–889 PERGAMON Neural Networks 0893-6080/00/$ - see front matter  2000 Elsevier Science Ltd. All rights reserved. PII: S0893-6080(00)00061-7 www.elsevier.com/locate/neunet * Tel.: +358-9-451-5693; fax: +358-9-451-2969. E-mail address: veikko@neuro.hut.fi (V. Jousma ¨ki).