Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
, 100 (3), 1407-19

Dynamic Population Coding of Category Information in Inferior Temporal and Prefrontal Cortex

Affiliations

Dynamic Population Coding of Category Information in Inferior Temporal and Prefrontal Cortex

Ethan M Meyers et al. J Neurophysiol.

Abstract

Most electrophysiology studies analyze the activity of each neuron separately. While such studies have given much insight into properties of the visual system, they have also potentially overlooked important aspects of information coded in changing patterns of activity that are distributed over larger populations of neurons. In this work, we apply a population decoding method to better estimate what information is available in neuronal ensembles and how this information is coded in dynamic patterns of neural activity in data recorded from inferior temporal cortex (ITC) and prefrontal cortex (PFC) as macaque monkeys engaged in a delayed match-to-category task. Analyses of activity patterns in ITC and PFC revealed that both areas contain "abstract" category information (i.e., category information that is not directly correlated with properties of the stimuli); however, in general, PFC has more task-relevant information, and ITC has more detailed visual information. Analyses examining how information coded in these areas show that almost all category information is available in a small fraction of the neurons in the population. Most remarkably, our results also show that category information is coded by a nonstationary pattern of activity that changes over the course of a trial with individual neurons containing information on much shorter time scales than the population as a whole.

Figures

FIG. 1.
FIG. 1.
Organization of the stimuli and behavioral task. A: time course of the delayed match to category experiment. B: an example of 1 of the 9 morph lines of the stimuli from the cat 1 prototype to the dog 1 prototype (the actual stimuli used in the experiment were colored orange) (see Freedman et al. 2002). C: the 6 prototype images used in the experiment. All the stimuli used in the experiment were either the prototype images, or morphs between the cat (C) and dog (D) prototypes.
FIG. 2.
FIG. 2.
Basic decoding results for 4 different types of information. A–D: blue lines indicates results from inferior temporal cortex (ITC) and red lines indicate results from prefrontal cortex (PFC; red, and blue shaded regions indicate one SD over the bootstrap-like trials). The 3 vertical black lines indicate SAMPLE-STIMULUS onset, SAMPLE-STIMULUS offset, and DECISION-STIMULUS onset from left to right respectively. E and F: comparison of SAMPLE-STIMULUS category decoding accuracy (purple), DECISION-STIMULUS category decoding accuracy (green), and whether a trial is a match or nonmatch trial (brown) for ITC (E) and PFC (F).
FIG. 3.
FIG. 3.
Decoding task-relevant “abstract” category information. A: decoding accuracies for ITC (blue) and PFC (red) when training on data from 2 dog and 2 cat prototype images and testing on the remaining dog and cat prototype images. The results are the average over all 9 permutations of training/test splits and the shaded results show the SDs over the 9 permutations (the individual traces are shown in Supplementary Fig. S4A). B and C: comparison of visual plus category stimulus decoding accuracies (purple line) to abstract category information (orange line) for ITC (B) and PFC (C). Note that there is a larger difference between these two types of information in ITC compared with the difference between these information types seen in PFC. This is a strong indication that the high SAMPLE-STIMULUS category decoding accuracies seen in ITC in Fig. 2B are largely due to visual information and not abstract category information during the sample period. During the decision period, for both ITC and PFC, most of information about the category of the SAMPLE-STIMULUS is in a more abstract representation, as there is little difference between abstract category information and “basic” category information during this period.
FIG. 4.
FIG. 4.
Readout using the “best” 2, 4, 8, or 16 neurons, compared with readout using all 256 neurons, for ITC (A) and PFC (B). As can be seen for almost all time periods, the abstract category information available in whole population is available in only ≤16 neurons. The best neurons were determined based on t-test between cats and dogs using the training data. Because the algorithm used to select the best neurons works in a greedy manner and is not necessarily optimal, the information reported in the subsets of neurons is an underestimate of how much information would be present if the optimal k neurons were selected.
FIG. 5.
FIG. 5.
Illustration of redundant information in ITC (A) and PFC (B). The magenta line indicates the readout performance when the top 64 neurons were used, and the green line indicates when the top 64 neurons were excluded and the remaining 192 neurons were used. As can be seen, the top 64 neurons achieve a performance level that is as good as using the whole population of 256 neurons. However, even when these neurons are excluded, readout is above chance, indicating that there is redundant information in these populations.
FIG. 6.
FIG. 6.
Evaluating whether the same code is used at different times for abstract category information. A: in ITC there is some similarity in the neural code for abstract category information in the sample and the decision periods, as can be seen by the green patches near the top right and bottom left of the figure. Also there appears to be two different codes used during the sample period as can be seen by the two blob regions occurring 775–1,275 ms after the start of the trial. B: for PFC, the code for abstract category information seems to be constantly changing with time as indicated by the fact that the only high decoding accuracies are obtained along the diagonal of the plot. C and D: examples of decoding accuracies using 3 fixed training times from the sample, delay and decision periods (colored lines) compared decoding accuracies obtained when training and testing using the same time period (black line) for ITC (C) and PFC (D); (each of these plots corresponds to 1 row from the from A or B and the black line corresponds to the diagonal of this figure and is the same line as shown in Fig. 3A). These figures again illustrate that the highest performance is always obtained when training and testing is done using the same time bin relative to stimulus/trial onset, which suggests that the neural coding of abstract category information is time-locked to stimulus/trial onset.
FIG. 7.
FIG. 7.
Elimination of the best 64 neurons from the time period t1 (specified on the y axis) and then training and testing with all the remaining 192 neurons at time period t2 (as specified by the x axis) for ITC (A) and PFC (B). Eliminating the best neurons from the training set at one time period only has a large affect on decoding accuracy at that same time period and leaves other time period unaffected as can be seen by the fact that there is only lower performance long the diagonal of the figure. This indicates that the neurons in the population that carry the majority of the information change with time. Additionally, one can see a decrease only along the diagonal even during periods where the stimulus is constant (areas between the black vertical bars). This indicates that the neural code is changing at a faster rate than changes in the stimulus.
FIG. 8.
FIG. 8.
Illustration showing that many individual neurons have short periods of selectivity for ITC (A) and PFC (B). The figure plots the 4 neurons for ITC and PFC that had the highest the mutual information between the category of the SAMPLE-STIMULUS and neuron's firing rate (firing rates where calculated using 100-ms bin periods sampled every 10 ms). As can be seen, most neurons show high mutual information (MI) values for only short time periods, which is what is expected for a population code that changes with time. It is also interesting to compare neurons 1 and 4 in ITC (A) because it shows that individual neurons have different peak selectivity times even when the stimulus being shown is constant. Thus the changing of the neural code is not just due to changes in the stimulus.

Similar articles

See all similar articles

Cited by 117 articles

See all "Cited by" articles

Publication types

LinkOut - more resources

Feedback