Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2011 Jan;271(1-2):37-53.
doi: 10.1016/j.heares.2010.06.006. Epub 2010 Jul 11.

How Do Neurons Work Together? Lessons From Auditory Cortex

Free PMC article

How Do Neurons Work Together? Lessons From Auditory Cortex

Kenneth D Harris et al. Hear Res. .
Free PMC article


Recordings of single neurons have yielded great insights into the way acoustic stimuli are represented in auditory cortex. However, any one neuron functions as part of a population whose combined activity underlies cortical information processing. Here we review some results obtained by recording simultaneously from auditory cortical populations and individual morphologically identified neurons, in urethane-anesthetized and unanesthetized passively listening rats. Auditory cortical populations produced structured activity patterns both in response to acoustic stimuli, and spontaneously without sensory input. Population spike time patterns were broadly conserved across multiple sensory stimuli and spontaneous events, exhibiting a generally conserved sequential organization lasting approximately 100 ms. Both spontaneous and evoked events exhibited sparse, spatially localized activity in layer 2/3 pyramidal cells, and densely distributed activity in larger layer 5 pyramidal cells and putative interneurons. Laminar propagation differed however, with spontaneous activity spreading upward from deep layers and slowly across columns, but sensory responses initiating in presumptive thalamorecipient layers, spreading rapidly across columns. In both unanesthetized and urethanized rats, global activity fluctuated between "desynchronized" state characterized by low amplitude, high-frequency local field potentials and a "synchronized" state of larger, lower-frequency waves. Computational studies suggested that responses could be predicted by a simple dynamical system model fitted to the spontaneous activity immediately preceding stimulus presentation. Fitting this model to the data yielded a nonlinear self-exciting system model in synchronized states and an approximately linear system in desynchronized states. We comment on the significance of these results for auditory cortical processing of acoustic and non-acoustic information.


Figure 1
Figure 1. Structured spontaneous activity in auditory cortex
Raster plots show activity of simultaneously recorded layer 5 neurons, blue traces show local field potentials (LFP) recorded from the a subset of the channels from which spikes were detected. Bottom black trace indicates multiunit activity (MUA), i.e. the population-averaged firing rate of all recorded cells. Red rectangles indicate “downstates,” i.e. periods of global network silence. Green traces show sync pulses, with positive values indicating times of tone presentation. (A, B) show two periods from a single recording under urethane anesthesia, (C, D) show two periods in a passively listening unanesthetized animal. Note that structured population activity is seen in all cases, but that the nature of this activity (such as the length of downstates) is variable both between and within recordings. Adapted from Luczak et al (2009).
Figure 2
Figure 2. Sequential population activity patterns
Representative raw data plots, showing responses to five tones (3, 7, 12, 20, 30 kHz respectively) and five different natural sounds (insect vocalizations), together with spontaneous activity patterns seen after stimulus offset, taken from a single recording under urethane. The green line indicates the duration of the stimulus; blue traces show local field potentials from one of the recording sites; underneath is a raster plot showing the spike trains of simultaneously recorded neurons. Neurons are sorted vertically by average spontaneous mean spike latency to reveal sequential firing patterns (unlike Figure 1, where they were arranged by recording shank). Neurons are displayed in the same vertical order in all plots. Although population firing events can vary significantly in firing rate and duration, stereotyped sequential patterns of timescale ∼100ms typically accompany evoked and spontaneous population spiking events, as well as occurring during the presentation of extended tones and natural sounds. Adapted from Luczak et al (2009).
Figure 3
Figure 3. Preservation of sequential structure between sensory-evoked and spontaneous events in unanesthetized animals
(A, B) Representative raw data plots from an unanesthetized, head-fixed subject in a passive listening paradigm. (C,D) Top two rasters (black ticks) show spike times for two individual neurons, triggered by tone onsets (C) and upstate onsets (D). Bottom panels show average activity of all simultaneously recorded neurons triggered by tone or upstate onsets: grey bars show pseudocolor representations of each neuron's perievent time histogram (PETH), red dots denote each neuron's mean spike latency in the 100ms after tone onset. Neurons are ordered vertically by the mean latency over all stimuli, to illustrate sequential spread of activity. Neurons are sorted in the same order in C and D, to illustrate the similar sequential order of tone-evoked and spontaneous activity. (E) Conservation of latency measure μcc across tones and spontaneous events. Adapted from Luczak et al (2009).
Figure 4
Figure 4. Combinatorial constraints on population firing rate vectors
(A) Spike counts of two neurons (recorded from separate tetrodes) during the first 100ms of spontaneous upstates (black), responses to a tone (green), and natural sound (red). Data were jittered to show overlapping points. Note that regions occupied by responses to the sensory stimuli differ, but are both contained in the realm outlined by spontaneous patterns. (B) Contour plot showing regions occupied by points from (A). The blue outline is computed from spike counts shuffled between upstates, indicating the region that would be occupied in the absence of spike count correlations. (C) Firing rate vectors of entire population, visualized using multidimensional scaling; each dot represents the activity of 45 neurons, nonlinearly projected into two-dimensional space. (D) Contour plot derived from multidimensional scaling data, with responses to individual stimuli marked separately. Sensory-evoked responses again lie within the realm outlined by spontaneous events. (E) Scatter plot showing the Euclidean distances from each evoked event to its closest neighbor in the spontaneous events (Espont), and in the shuffled spontaneous events (Eshuf). Dashed red line shows equality. (F, G) Histogram showing the difference between distances to shuffled and spontaneous events (Eshuf - Espont). Top and bottom: data from all anesthetized and unanesthetized experiments, respectively. Almost every evoked event was closer to a true spontaneous vector than to a shuffled vector. Adapted from Luczak et al (2009).
Figure 5
Figure 5. Visualization of population firing rate vectors
The panels show projections of the mean firing rate vector trajectory of 282 cells pooled from 4 experiments, evoked by tones of three frequencies, plotted for increasingly longer time periods from left to right. (A1) Trajectories viewed with principal component analysis (PCA), which finds the projection of maximum variance; all three plots are in the same projection. In this projection, onset responses are dominant. (A2) PSTHs of the five cells contributing most to the PCA projection. Arrows above each PSTH indicate factor loadings in the projections above. (B1) Trajectories viewed with multiple discriminant analysis (MDA), to maximize the differences between sustained responses. In this projection, onset, sustained, and offset responses have approximately equal magnitude. Dashed circles: baseline activity, dotted circles: sustained activity. (B2) PSTHs of the five cells contributing most to the MDA projection. Arrows as in A2. Adapted from Bartho et al (2009).
Figure 6
Figure 6. Population vector rotation
(A) Schematics of hypotheses tested in this figure. Hypothesis I: the population vector in the sustained period is a linearly scaled version of that at stimulus onset. Hypothesis II: both the magnitude and direction of the population vector changes. (B1) Pseudocolor plot showing the angle in degrees between the mean population firing rate vectors for all times and tone frequencies, and a reference vector produced by a 14.4 kHz tone during stimulus onset (indicated by the arrow). The reference vector is more similar to onset response vectors for other frequencies, than to sustained responses for the same tone. (B2) Similar analysis for a reference vector computed during the sustained response, here showing greater similarity to sustained responses of other stimuli, than to onset responses to the same stimulus. (B3-4) Same as B1-2, with 27 kHz tone response as reference. Adapted from Bartho et al (2009).
Figure 7
Figure 7. Cell-type dependent sparseness of population activity
(A) Examples of five juxtacellularly recorded pyramidal cells (PCs), digitally superimposed. (B) Spectral tuning of the neurons shown in (A). Each plot shows a pseudocolor representation of the cell's mean firing rate in a 50ms period following tone onsets, as a function of tone frequency and intensity. The number above each plot indicates maximum firing rate. L5sPC, L5 slender PC; L5tPC, L5 thick PC. (C) Tuning of four representative cells identified from silicon probe recordings. Left, schematic drawing of electrode, and average spike waveform profiles of a putative deep PC, superficial PC, deep interneuron(IN), and superficial IN. Right, spectral tuning of these cells. (D) Left, schematic drawing of recording by a 32-site linear electrode. Right, raster plot of multi-unit activity (MUA) for each channel, superimposed on local-field potentials (gray traces). (E,F) Sparseness of evoked and spontaneous activity was assessed using a “response probability” measure, for which smaller values indicate sparser firing. Bars above and below dotted line indicate cell-classes identified morphologically by juxtacellular recording (“juxtacells”), and silicon probe-recorded units putatively classified by spike waveform (“extracells”), respectively. Asterisks denote pairwise post-hoc lsd tests, indicating a significant difference (p<0.05) to the class corresponding to that color. Post-hoc comparisons were performed for juxtacells and extracells separately. sP, superficial PCs; dP, deep PCs; sI, superficial INs; dI, deep INs. Error bars indicate SE. (G) Sparseness is correlated between sensory responses and upstates. Each symbol shows the response probability of one cell to tone and click stimuli, with large symbols indicating juxtacells. Adapted from Sakata and Harris (2009).
Figure 8
Figure 8. Difference in propagation of activity across cortical layers and columns
(A) Example laminar profiles of upstates and evoked responses. Rasters indicate MUA of all channels on a 32-site linear probe for individual upstates and evoked responses. Shaded periods indicate tone presentations. Red dots indicate “peak latency,” computed as the median MUA spike time in a 50-ms window after event onset. (B) Laminar profiles of peak latency for tone-evoked responses (best frequency, 60-80 dB SPL) and upstates. The graphs show a pseudocolor histogram of the distribution of peak latency as a function of depth. (C) Two-shank multisite electrodes (2×16 linear probe) were inserted parallel to the layers of auditory cortex. A part of the drawing was replicated from (Paxinos & Watson 1997). (D,E) Examples of spatiotemporal patterns for upstates (D) and click-evoked responses (E). Each plot shows rasters of MUA on all recording sites, with superficial and deep shanks on top and bottom. The sites on each shank are arranged from dorsal (D) to ventral (V). (F) Distribution of propagation speeds for upstates (top) and evoked responses (bottom), estimated as the regression slope of median MUA time across recording sites. Arrows indicate the median, and the x-axis is log-scaled. Propagation speed was faster for evoked responses than for upstates in both layers (ANOVA with post hoc lsd test, p<0.0001). (G) Hypothesized flow of sensory-evoked and spontaneous activity through auditory cortical circuits. Each sheet represents a population of the corresponding layer, with cones and spheres representing PCs and INs, respectively. Colored symbols represent active neurons. Adapted from Sakata and Harris (2009)
Figure 9
Figure 9. Trial-to-trial variability across a range of cortical states
(A) Six examples of population responses to click stimuli, from a rat that exhibited stable dynamic state throughout the recording. Vertical green lines denote stimuli (time 0); LFP (black trace), activity of simultaneously recorded single neurons (rasters) and smoothed multi-unit activity (MUA; red trace) all show a pattern of population activity characteristic of the synchronized state. Right column shows an expanded view of the smoothed MUA in the response period for each trial; gray shaded areas denote “initial” (10-35ms; dark gray shading) and “persistent” (40-135ms; light gray shading) response periods. The stimulus may arrive during a downstate (trials 1, 2), at the beginning of an upstate (trials 3, 4), or well into an upstate (trials 5, 6). While preceding activity does not have a clear effect on peak activity levels in the initial response period, the timing of the stimulus relative to up/down transitions appears to modulate activity in the persistent response period. (B) Same conventions as in (a); all data are selected from a different recording session that showed variable dynamic state. In the synchronized state (trials 1,2), persistent responses are anticorrelated with activity levels in the 200-300ms preceding the stimulus. In intermediate states (trials 3,4), the stimulus induces a large initial response followed by a transient downstate. In the most desynchronized states (trials 5,6), responses exhibit a small but reliable initial response followed by a return to baseline, with no discernible persistent response. Adapted from Curto et al (2009).
Figure 10
Figure 10. Stimulus-evoked responses can be predicted from models fit on prior spontaneous activity
(A) Methodology. 3s of spontaneous activity preceding the stimulus is used to fit the model parameters. The model-fit dynamic state (illustrated by the corresponding phase diagram), together with the activity state at the time of the stimulus (green star), is then used to simulate an evoked response (blue), shown superimposed on the true response (red). As in Figure 9, time 0 corresponds to presentation of click stimulus and shaded regions correspond to initial (dark gray) and persistent (light gray) response periods. (B, C) Estimated dynamic states and simulated responses for each trial displayed in Figure 9. (D) Histograms of prediction error for a single trial from rat 1 (red line), compared to predictions for the response on this trial made from states estimated for all other trials. Estimates from other trials, both within the same recording (left) and from other recordings (right), produced worse predictions. (E) Box plots of model fit percentiles for each trial, within and across recordings. Median percentiles (red) are in each case significantly above chance level (50%). (F) Same as in (e), but performance is compared using only dynamic states from other trials, keeping activity state at stimulus onset fixed. For recording sessions with high cortical state variability (1 and 2), percentiles were still high within and across recordings. For recording sessions with very stable dynamic state (3 and 4), percentiles were not significantly above chance within each recording, but remained high across recordings. Adapted from Curto et al (2009).

Similar articles

See all similar articles

Cited by 24 articles

See all "Cited by" articles