Article Figures & Data
Figures
- Figure 1.
Multielectrode parallel recordings in rat barrel cortex. a, Photograph of a 10 × 10 array of microelectrodes implanted in rat somatosensory cortex in experiment 3. Interelectrode distance is 400 μm, and electrode length is 1.5 mm. Scale bar refers to the approximate 700 μm length of electrode shaft outside the brain; thus, depth of insertion was 700-800 μm. b, Placement of the microelectrode array in the barrel subfield in the same experiment. White-filled circles show the 37 electrodes at which neuronal data were recorded during whisker stimulation. The position of the array relative to the barrel columns was based on identification of the principal whisker input at each electrode of interest (see Materials and Methods). Each barrel column is designated by its row (A-E) and its arc (1-6) (Woolsey and Van der Loos, 1970). Lightning bolts designate electrodes 19 and 45, where the signals shown in c-e were recorded. c, One hundred overlaid action potential waveforms of the neuronal cluster at electrode 19. Further spike sorting was not feasible for this electrode. d, One hundred overlaid action potential waveforms of a putative unit selected from the neuronal cluster at electrode 45. e, Autocorrelogram for the putative single unit shown in d. The plot was constructed with 1 msec time bins from 2000 sec of data including spontaneous and stimulus-induced activity. The absence of spikes within 2 msec of the central bin, not observed for multi-neuron clusters, reflects the absolute refractory period of the isolated single unit.
- Figure 7.
Comparison of two candidate single-neuron codes for vibration frequency. a, Single-unit autocorrelograms for vibrations ranging from 19 to 341 Hz. No phase-locked activity is evident at any vibration frequency. Insets show 50 overlaid spike waveforms from each record, ensuring that the same single unit was measured for each stimulus. b, Single-unit firing rates for the same stimulus set. Note frequency-dependent increase in firing rate. Error bars are the SEM of spike counts across trials. For both a and b responses were measured over the interval 0-500 msec after stimulus onset.
- Figure 2.
Properties of the whisker stimulus set. a, Combinations of seven amplitudes and seven frequencies were used to create 49 different stimuli. Each point represents the frequency-amplitude combination of one sinusoidal stimulus. The stimuli numbered 1-7 are represented in c. b, The frequency-amplitude combinations of a plotted according to loge(Af), where Af is proportional to the mean vibration speed. c, 20 msec segments of the movement of the ladder stimulator, corresponding to stimuli designated 1-7 in a and b. d, Sinusoidal movement of the whisker shaft. The continuous trace shows whisker movement when stimulated by the highest frequency/highest amplitude vibration (denoted as stimulus 7 in a-c). The overlying dots are sampled from a computer-generated sinusoidal waveform of the same frequency and amplitude. Note the match of the observed whisker movement to the sinusoid. Measurements in c and d were made by the optical sensor.
- Figure 3.
Responses to a sinusoidal stimulus recorded from a neural cluster. Raster plot and PSTHs from a neural cluster in barrel column B2 recorded in experiment 1. Stimulus was 211 Hz, 87 μm; stimulus 5 in Figure 2a-c. Dots in the top part show spike times across 200 trials, and the corresponding PSTH (bin size, 4 msec) is aligned below the raster plot. Stimulus presentation was from 0 to 500 msec, as indicated by the thick line below the PSTH. The inset shows the earliest portion of the PSTH with 1 msec bins. The three-peaked structure at response onset did not depend on stimulus parameters.
- Figure 4.
Frequency and amplitude dependence of firing rates for a neural cluster. Data are from barrel column D2 in experiment 1. Spikes were counted over the 0-500 msec time window (vibration onset, 0 msec) and averaged over 200 trials for each stimulus. Error bars are the SEM spike counts across trials. a, Firing rates are shown as a function of stimulus frequency. Each line connects the points generated by increasing vibration frequency with constant vibration amplitude. Amplitudes increase (8, 12, 21, 33, 54, 87, and 140 μm) from the lowest line to the highest. b, Firing rates as a function of stimulus amplitude. Each line connects the points generated by increasing vibration amplitude with constant vibration frequency. Frequencies increase (19, 30, 50, 81, 131, 211, and 341 Hz) from the lowest line to the highest. Note the similar dependence of firing rate on the change in either frequency or the amplitude.
- Figure 5.
Neuronal encoding of the vibration feature Af. a, Color plot of spike counts (0-500 msec after stimulus onset) in response to the 7 × 7 stimulus grid for the same neuronal cluster shown in Figure 4. By comparing neuronal responses with the plot of Af (Fig. 2b), it is clear that firing rates were nearly identical within an iso-Af group regardless of the frequency-amplitude combination that gave rise to the stimulus. Firing rates increased as mean vibration speed grew from one iso-Af group to its neighbor. The 4 × 4 set of responses in the frame are analyzed further in Figure 6. The average firing rate in the 500 msec interval preceding stimulus onset is denoted as spontaneous activity, below the color scale. b, Spike counts 100-500 msec after stimulus onset. Even after excluding the early part of the response, the neuronal cluster robustly encoded stimulus Af.
- Figure 6.
Temporal structure in responses of a multi-neuron cluster. a, PSTHs corresponding to the 4 × 4 grid in Figure 5. Note the similarity in profiles for the top left and bottom right PSTHs (asterisks), which were generated by stimuli of equal values of the product Af but frequencies of 81 and 341 Hz. b, Autocorrelation histograms (1 msec bin size) for the same 4 × 4 grid. The autocorrelograms were modulated as a function of Af (the number of events simply increased), but for the stimuli of a diagonal iso-Af band there was no temporal structure that could characterize frequency or amplitude.
- Figure 8.
Neuronal encoding of mean stimulus speed on multiple time scales. a, Af-coding at a 500 msec time scale. Mean population firing rate for all electrodes positioned in the barrel cortex for experiments 1-5, measured by counting the spikes pooled from all electrodes during the entire stimulus presentation window (0-500 msec). The average firing rate in the 500 msec interval preceding stimulus onset is denoted as spontaneous activity, below the color scale. Neuronal activity was recorded at 16, 24, 37, 35, and 18 electrodes in experiments 1-5, respectively. The strong correlation between Af and response magnitude was highly reproducible in different rats. b, Af-coding by peak spike density counted over a sliding time window of 5 msec (see Results for analysis method). Each plot is matched with the 500 msec spike count for the same experiment plotted on the left. For each of the five experiments, the encoding of the stimulus feature Af by maximum, instantaneous spike density was nearly indistinguishable from the encoding of this feature by a long-time window spike count.
