VOLUME 77, NUMBER 19 PHYSICAL REVIEW LETTERS 4NOVEMBER 1996 Stochastic Resonance in a Neuronal Network from Mammalian Brain Bruce J. Gluckman, 1,2 Theoden I. Netoff, 2,3 Emily J. Neel, 2 William L. Ditto, 4 Mark L. Spano, 1 and Steven J. Schiff 2,3 1 Naval Surface Warfare Center, Silver Spring, Maryland 20903 2 Department of Neurosurgery, Children’s National Medical Center and The George Washington University School of Medicine, Washington, D.C. 20010 3 Program in Neuroscience, The George Washington University, Washington, D.C. 20052 4 School of Physics, Georgia Institute of Technology, Atlanta, Georgia 30332 (Received 24 June 1996) Stochastic resonance, a nonlinear phenomenon in which random noise optimizes a system’s response to a signal, has been postulated to provide a role for noise in information processing in the brain. In these experiments, a time varying electric field was used to deliver both signal and noise directly to a network of neurons from mammalian brain. As the magnitude of the stochastic component of the field was increased, resonance was observed in the response of the neuronal network to a weak periodic signal. This is the first demonstration of stochastic resonance in neuronal networks from the brain. [S0031-9007(96)01583-9] PACS numbers: 87.22.Jb, 02.50.Ey, 05.40. + j, 87.50. – a The brain is a noisy processor, and the idea that the brain might make use of such noise to enhance informa- tion processing is not new [1]. In stochastic resonance (SR), the response of a nonlinear system to an other- wise subthreshold signal is optimized with the addition of noise. Since its proposal as a mechanism for amplifying the effects of the Earth’s small periodic orbital variations by random meteorological fluctuations leading to ice age periodicity [2], SR has been observed in a diverse range of physical systems [3]. Despite theoretical work predicting that SR might be found in single neurons [4] and neu- ronal networks [5,6], and experimental evidence sugges- tive of SR from interspike interval histograms (ISIH) [7], there has been no experimental confirmation in the brain. SR has previously been observed in the activity of single mechanoreceptive sensory neurons from crayfish [8], rat skin [9], and from single interneurons from cricket ab- dominal ganglia [10]. Each of these previous demonstra- tions of SR involved the processing of mechanosensory information, when signal and noise were encoded into en- vironmental pressure fluctuations. Adjusting the noise of neurons directly has been dif- ficult. In the crayfish two approaches have been taken for optimizing detection sensitivity to pressure fluctua- tions. Raising the temperature failed to show optimiza- tion as a function of noise level [11], while raising the light level on the caudal photoreceptor has been success- ful [12]. Nevertheless, because of the technical difficulty of delivering signal and noise directly to neurons, the ex- perimental study of SR in mammalian brain has remained an intractable problem. In recent work we demonstrated that an electric field could be used to either suppress or enhance epileptiform activity in mammalian brain slices [13]. The effect of an imposed electric field on neurons has been worked out in detail, and it is well known that the amplitude of an electric field required to modulate the action potential timing of an actively firing neuron is much less than that required to initiate an action potential in a neuron from rest [14]. The physics can be understood by considering a field aligned parallel to the axis between the dendrites, where signals come in from other neurons, and the soma, where these signals are translated into action potentials. The field induces ionic currents both inside and outside the neurons, but the cell membranes act as containers (albeit leaky ones) causing charge to build up and thereby changing the transmembrane potential at the somata. The result on each neuron is a shift in the effective threshold for action potential initiation, and therefore a modulated response to incoming signals. Because the electric field interacts with neurons even at magnitudes insufficient to trigger action potentials, it provides a means to introduce a subthreshold signal into an entire network of neurons to probe for SR. A schematic of the experimental setup is shown in Fig. 1(a). Longitudinally or transversely cut hippocampal slices (400 mm thick) from rat temporal lobe [15] were placed in the center of a field produced by parallel nonpolarizing Ag-AgCl electrode plates submerged in the perfusate. The neural layers of the slice, which are visually identifiable, are oriented with respect to the field. The potential between the plates was set by a computer generated signal applied through an isolation amplifier. The resulting field in the chamber, and within a slice, was measured and calibrated to the potential applied to the plates [Figs. 1(b) and 1(c)]. The field is quite uniform in the central region of the chamber where the slices are placed, and is proportional to the potential applied to the plates over the range of amplitudes and frequencies used in these experiments. Hippocampal slices in a high (8.5 mM) potassium per- fusate, as used in these experiments, demonstrate increased neuronal synchrony and spontaneous ensemble activity [16] in which large populations of the main excitatory 4098 0031-9007 96 77(19) 4098(4)$10.00 © 1996 The American Physical Society