Figure 1.
Stimulus-triggered average LFPs for different tone frequencies at and around the BF (27 kHz).
Peak amplitude is measured from baseline and peak latency from tone pip onset.
Figure 2.
Comparison of frequency-tuning curves (FTCs) for LFP and MSU activity.
The top half shows the LFP-based FTCs in color code; blue represents negative values, red positive ones. The 25% of negative peak amplitude contour lines are drawn in (white). The bottom panels show the MSU spike peak firing rates in 5 ms bins indicated in color. The 25% of negative peak amplitude contour lines are drawn in (white). It is noted that the CF of the spike activity corresponds well with that for the LFPs. Spike thresholds are in about half the recordings a few dB more sensitive than the selected LFP level, and in the other half up to 10 dB less sensitive. Channel 25 did not record spike activity.
Table 1.
Comparison of averaged FTC parameters based on LFP and MSU recordings.
Figure 3.
Comparison of LFP and MSU FTC parameters.
Top figure compares the CFs for MSU and LFPs, the middle part compares the MSU and LFP thresholds, and the bottom part the FTC bandwidths at 20 dB above threshold.
Figure 4.
Example of spike-based and LFP-based STRFs for one 16-electrode array recording.
Each panel is scaled on its own extremes; red colors indicate high spike firing levels and blue colors low firing levels. Maximum values for the spike responses (number of spikes/bin/stimulus) are indicated above each panel. In some cases this value is 0 whereas there is still some response visible. We use 0 when the value is <0.01 spikes/bin/stimulus. Contour lines for LFP amplitude are white for negative values and red for positive values. The frequency axis (log scale) runs from 300 Hz to 10 kHz and covers 5 octaves. The spatial orientation of the 16 electrodes in this Figure is such that the top two rows are physically on the left of the bottom two rows. Within a row the electrodes are separated by 250 µm, and between rows by 500 µm. Although there is very little difference in the LFPs recorded on the various electrodes in the array, the spike STRFs are varying across the array. For instance the neighboring electrodes 19 and 20 show very different spike-STRFs and nearly identical LFP-STRFs. Here the LFP contours, especially the 25% of peak amplitude, cover about 5 octaves, whereas the neural activity covers 1-2 octaves. Spike firing occurs frequently at the short latency edge of the 50% contour (e.g., top row), but can also be within the 50% contour band (third row, last two columns), or extend beyond the negative amplitude contours (second row, last two columns). Channel 25 did not produce spikes but a clear LFP-based STRF starting with a positive phase and indicative of local hyperpolarization.
Figure 5.
Example of STRFs determined on the basis of spikes and LFPs.
STRFs determined on the basis of spikes are shown color coded and LFPs-based ones are indicated by white contour lines for negative levels at 75%, 50% and 25% of maximum, and red ones indicating positive levels. Same plotting conventions as for Figure 4. The frequency range in this example is from 1.2 kHz-40 kHz (5 octaves). As can be seen there are again only minor changes in the LFP profiles across the 8×2 electrode array, whereas the spike-based STRFs are more variable. Note that for a limited frequency range the LFP may start positive, thereby preventing or delaying the spike generation (channels 25 and 26, second row; channels 29 and 32, bottom row).
Figure 6.
STRFs for LFP (left) and for MSU activity (right) averaged over the 16-electrode array.
The two panels represent an average of the data shown in Figure 5. For the left panel blue indicates negative LFP amplitudes, red colors indicate positive amplitude values and yellow corresponds to zero crossings. Contour lines indicate 25, 50 and 75% of LFP negative maximum (white) and positive maximum (red). Note peak spike activity in the 50–75% negative LFP-amplitude contour lines.
Figure 7.
STRFs for LFP (left) and for MSU activity (right) averaged over electrodes in the two arrays that showed clear tuned spike activity.
For the top two panels only 4 electrodes, indicated above the panel, produced clear STRFs for spikes. These were averaged. For the top panels array, the LFP-based STRF starts with a positive part followed by a negative part, whereas for the bottom panels (7 electrodes) the common negative-positive sequence is found. The initial positive LFP delays the spike firings for the top array, which occur on the negative going LFP phase. For the bottom array, spike activity occurs for LFP amplitudes that are at least 50% of negative maximum. Although both electrode arrays were in AI and at approximately the same depth, the differences in latencies are pronounced. Contour lines again indicate 25, 50 and 75% of LFP negative maximum (white) and positive maximum (red).
Figure 8.
Examples of marginal frequency distributions for STRFs based on 2–40 Hz, 4–8 kHz, 8–16 kHz and 16–40 kHz LFPs, and for spikes for 4 different recordings.
In most cases the various filtered LFP data are much broader tuned than the spike based data. A peak in the spike-based frequency distribution always correspond to a peak in the LFP-based distributions. The LFP-based frequency distributions nearly always contain more peaks that the spike-based ones.
Table 2.
Pearson product-moment correlation coefficients for marginal frequency distributions.
Figure 9.
Distributions of the Pearson product-moment squared correlation coefficients.
The correlation coefficients are shown (with sign preserved) between the spike-based frequency marginals and the 2–40 Hz, 8–16 Hz, and 16–40 Hz LFP-based frequency marginals.
Figure 10.
Scattergrams of the Pearson squared product-moment correlation coefficients.
The correlation coefficients are shown (with sign preserved) between spike-based and 2–40 Hz LFP-based marginals and spike-based and 16–40 Hz LFP-based marginals (top), and spike-based and 8–16 Hz LFP-based marginals (bottom). The Pearson product-moment correlation coefficients of spikes with 2–40 Hz and 8–16 Hz LFPs are very similar (bottom) whereas those with 16–40 Hz filtered LFPs are clearly different; the Pearson product-moment correlation coefficients with 16–40 Hz LFP frequency-marginals are nearly always larger than those with the 2–40 Hz LFP based frequency-marginals.
Figure 11.
Distance dependence of spike-spike correlograms.
In the left column, cross-correlation coefficient functions are show for all pairs from array 1 (top), array 2 (middle) and between electrodes located in different arrays (bottom). The second column shows the coherence-corrected correlograms. Note the extensive overlap of these correlograms. Note that the peak values in the correlogram are smaller between arrays than within arrays. The fourth column upper two rows show the dependence of the corrected peak values as a function of distance for the within-array correlations; one notices only a moderate effect. The mean values are indicated with a thin line. The third column upper two rows show the change in peak cross-correlation coefficient as a function of electrode distance within an array. In the fourth column the corrected peak cross-correlation coefficients for spike-spike pairs are shown.
Figure 12.
Distance dependence of 2–40 Hz LFP-pair correlograms.
In the left column, cross-correlation coefficient functions are show for all pairs from array 1 (top), array 2 (middle) and between electrodes located in different arrays (bottom). The second column shows the coherence-corrected correlograms. Note the extensive overlap of these correlograms. Note that the peak values in the correlogram are similar between arrays than within arrays. The fourth column upper two rows show the dependence of the corrected peak values as a function of distance for the within-array correlations; one notices only a moderate effect. The mean values are indicated with a thin line. The third column upper two rows show the change in peak cross-correlation coefficient as a function of electrode distance within an array. In the fourth column the corrected peak cross-correlation coefficients for spike-spike pairs are shown.
Figure 13.
Distance dependence of 16–40 Hz LFP-pair correlograms.
In the left column, cross-correlation coefficient functions are show for all pairs from array 1 (top), array 2 (middle) and between electrodes located in different arrays (bottom). The second column shows the coherence-corrected correlograms. Note the extensive overlap of these correlograms. Note that the peak values in the correlogram are similar between arrays than within arrays. The fourth column upper two rows show the dependence of the corrected peak values as a function of distance for the within-array correlations; one notices only a moderate effect. The mean values are indicated with a thin line. The third column upper two rows show the change in peak cross-correlation coefficient as a function of electrode distance within an array. In the fourth column the corrected peak cross-correlation coefficients for spike-spike pairs are shown.
Figure 14.
Scattergram of peak correlation coefficients for LFP pairs and spike pairs.
Natural logarithm are used for coherence-corrected values. Regression line are drawn separately for within array electrode pairs and for between array electrode pairs.
Figure 15.
Correlation coefficient distance dependence.
Scatterplot of the distance dependence of the natural log of the coherence-corrected peak cross-correlation coefficients for 2–16 Hz LFP-pairs (green symbols), 16–40 Hz LFP-pairs (Blue symbols), and spike-pairs (red symbols) recorded on the same electrodes. The slopes of the regression lines are steeper for the spike data compared to the LFP data.