Figure 1.
Behavioral paradigm and attentional modulation of MUA.
A. A trial started when the monkey touched a lever and fixated the central fixation point. The trial would abort without any reward if the monkey released the lever or broke its fixation at any time. After a short (130 ms) blank screen a static RDP was shown for a 455 ms interval indicating the position of the target. After another blank period of 325 ms two moving RDPs were presented peripherally. At a random time between 680–4250 ms after the onset of the RDPs, one or both of the stimuli made a direction change of 30 degrees and turned back after 130 ms. The monkey had to respond to the direction change in the target within a time window of 150–650 ms while ignoring any direction change in the distracter. The plus sign indicates the fixation point, the filled circle is the RF and the dashed circle marks the target. The circles were not presented in the experiment. B. Normalized MUA aligned to target onset. The dashed line represents the MUA recorded in the unattended condition and the solid line shows the MUA in the attended condition. Error bars show the standard error of mean (SEM).
Figure 2.
Attentional modulation of LFP power at different frequencies.
LFP power at different frequencies is shown in the two attention conditions. Solid lines represent the LFP power in the attended condition and the dashed lines show the LFP power in the unattended condition. Significant differences between the LFP powers in the two conditions are indicated by stars. A. Normalized power of frequencies below 48–200 Hz. C. Logarithm of the normalized power for 1–200 Hz excluding 48–52 Hz. Error bars show SEM.
Figure 3.
Performance of decoding attention condition using LFP power at different frequencies.
A. Decoding performance for trials in which any of 8 possible directions (0 to 2π radians with steps of π/4) were presented. B. Decoding performance for trials in which the preferred direction was shown as the target. C. Performance for trials with the anti-preferred direction as the target. Performances are shown given different sizes of training data. The solid lines indicate the performance using LFP powers between 1–200 Hz (marked as WB) and the dashed lines show the performance using LFP powers between 1–500 Hz (marked as EWB). Error bars show SEM.
Figure 4.
Performance of decoding for trials with the target moving in any of the 8 equally separated directions.
Each plot shows the decoding performance for the frequency band written above it given different sizes of training data. Error bars represent SEM.
Figure 5.
Comparison of decoding across delta, theta and low gamma bands.
A. ROC curves of the low frequency bands (delta & theta) vs. low gamma band using the features extracted by the SVM algorithm. B. The learning curve of the three bands limited to the size of training data less than 30. Delta, theta and low gamma bands are represented using dotted, dashed and solid lines respectively. Error bars show SEM.
Table 1.
Peak value of decoding performance and coefficient of variation for different LFP bands in trials with any of the 8 motion directions.
Figure 6.
Performance of decoding for trials with the target moving in the preferred direction.
Each plot shows the decoding performance for the frequency band written above it given different sizes of training data. Error bars represent SEM.
Figure 7.
Decoding performance for trials with the target moving in the anti-preferred direction.
Each plot shows the decoding performance for the frequency band written above it given different sizes of training data. Error bars represent SEM.
Table 2.
Peak value of decoding performance and coefficient of variation for different LFP bands in trials with preferred vs. anti-preferred targets.