COGNITIVE NEUROSCIENCE NEUROREPORT 0959-4965 & Lippincott Williams & Wilkins Vol 11 No 10 14 July 2000 2313 Evidence for fusion and segregation induced by 21 Hz multiple-digit stimulation in humans Li Chan Liu, William C. Gaetz, Daniel J. Bosnyak and Larry E. Roberts CA Department of Psychology, McMaster University, Hamilton, Ontario L8S 4K1, Canada CA Corresponding Author Received 18 April 2000; accepted 5 May 2000 Subjects were trained to detect changes in the frequency of 21 Hz tactile stimulation applied to digits 2  3  4 (fusion group) or 2  4 (segregation group) of the right hand. The 21 Hz steady-state response for digit 3 was measured by 64 channel EEG on mapping trials before and after training. Discrimination improved over 3 days, con®rming that subjects attended to the training stimuli. The 21 Hz response was larger on training than on mapping trials, indicating sensitivity of the response to the strength of cortical activation. Under these conditions the 21 Hz response for digit 3 decreased after training in both groups on day 1. On day 3 this effect reversed in a subset of fusion subjects while segregation continued to yield decreases. The ®ndings suggest that somatosensory representations are dynamically modi®ed by the sensory input experienced on a task. NeuroReport 11:2313±2318 & 2000 Lippincott Williams & Wilkins. Key words: Electroencephalography (EEG); Hebbian learning; Somatosensory cortex; Steady-state responses; Use-dependent plasticity INTRODUCTION Recent animal studies have shown that the receptive ®elds (RFs) of neurons in somatosensory cortex can be altered in adult organisms by behavioural training. For example, Recanzone et al. [1,2] found that training owl monkeys on a tactile frequency discrimination task enlarged the neural representation for the trained skin site by a factor of 1.5±3 in cortical area 3b. An enhancement of similar magnitude was observed by Jenkins et al. [3] when monkeys were trained for 50±100 sessions to touch a rotating disk with the correct pressure in order to earn food reward. Remo- delling of sensory representations by behavioural proce- dures appears to be gated by neuromodulatory systems in the thalamus and basal forebrain that condition rapidly to cues signalling task relevant events. Remodelling does not occur unless the animal is required to process the task stimuli by discriminative training (see [4] for review). Fusion and segregation of somatosensory representations for different skin sites has also been demonstrated by animal studies. Wang et al. [5] applied a series of brief, phase-coherent tactile stimuli alternatively to the distal or proximal phalanges of digits 2  3  4 of owl monkeys. The task of the monkey was to detect when pulses were delivered twice in succession to either phalange. Invasive mapping of area 3b carried out after 4±6 weeks of training revealed a high proportion of neurons (44±56%) with multi- ple-digit RFs that were rarely seen in untrained animals. At the same time, a band representing dorsal skin emerged between the zones representing the non-simultaneously excited distal and proximal skin surfaces, suggesting a concurrent segregation effect. Experiments on rat barrel ®elds indicate that lengthy training may not be necessary for the occurrence of fusion effects. Whisker-pairing leads within 24 h to fused representations which appear initially in the super®cial (II, III) and deep (V, VI) layers of the barrel cortex and 10±30 days later in the input layer (IV), signify- ing reorganization at the level of the thalamus [6]. Wang et al.'s report that multiple-digit RFs were induced in 50% of sampled neurons suggests that fusion and segregation may be detected in electrical (EEG) or magnetic (MEG) ®elds which can be recorded non-invasively in human subjects. In the present study we used EEG to study short-term plasticity induced by training for tactile frequency discrimination in humans. Subjects were trained to discriminate small changes in the frequency of a 21 Hz standard stimulus applied for 1 s simultaneously to the ®nger tips of digits 2  3  4 (fusion condition) or digits 2  4 (segregation condition) of the right hand. Based on previous animal research we hypothesised that synapses conveying input from the ®ngertips receiving multiple digit stimulation would be strengthened by simultaneous depolarisation of their postsynaptic targets, through a process of Hebbian learning. If so, stimulating digit 3 after training for fusion was expected to elicit a response from neurons tuned previously to digits 2 and 4, thereby augmenting the 21 Hz response. In contrast, segregation training was expected to decrease the response for digit 3 (segregated from 2 and 4) through anti-Hebbian mechan- isms. Subjects were trained for 3 days to allow for the development of remodelling effects.