Based on complementary metal-oxide semiconductor (CMOS) technology a neurosensor chip with passive palladium electrodes was developed. The CMOS technology allows a high reproducibility of the sensors as well as miniaturization and the on-chip integration of electronics. Networks of primary neurones were taken from murine foetal spinal cord (day 14) and frontal cortex (day 15) tissues and cultured on the silicon surface in a chamber volume of 200 microl with 7 mm diameter. Measurements were performed between days 15 and 59 in vitro. Signals were recorded from both types of cultures. To test the capability of the system to detect pharmacologically induced activity changes two established neuromodulators were applied. The GABA(A)-receptor blocker bicuculline was applied to both tissue cultures, the glycine-receptor blocker strychnine to spinal cord cultures. Four network frequency parameters were analysed: spike rate (SR), burst rate (BR), frequency in bursts (FiB) and peak frequency in bursts (PFiB). Significant changes of spike rate and burst rate were measured with spinal cord cultures after bicuculline application. Significant changes of frequency in bursts and peak frequency in bursts were observed with frontal cortex cultures after bicuculline application. Significant changes of spike rate and frequency in bursts were recorded with spinal cord cultures after strychnine application. These results were compared with results achieved in the same laboratory by using glass-microelectrode arrays (MEAs). This comparison showed for spinal cord similar native spike and burst rate, but higher mean frequency and peak frequency in bursts, whereas frontal cortex activity had higher spike and burst rate and peak frequency in bursts. Application of bicuculline or strychnine to spinal cord networks showed stronger effects on MEAs, whereas with frontal cortex networks the modulation of activity was similar after application of bicuculline.