Skip to main page content
Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2020 Mar 6;10(1):4216.
doi: 10.1038/s41598-020-61171-3.

The Lateral Prefrontal Cortex of Primates Encodes Stimulus Colors and Their Behavioral Relevance During a Match-To-Sample Task

Affiliations
Free PMC article

The Lateral Prefrontal Cortex of Primates Encodes Stimulus Colors and Their Behavioral Relevance During a Match-To-Sample Task

Philipp Schwedhelm et al. Sci Rep. .
Free PMC article

Abstract

The lateral prefrontal cortex of primates (lPFC) plays a central role in complex cognitive behavior, in decision-making as well as in guiding top-down attention. However, how and where in lPFC such behaviorally relevant signals are computed is poorly understood. We analyzed neural recordings from chronic microelectrode arrays implanted in lPFC region 8Av/45 of two rhesus macaques. The animals performed a feature match-to-sample task requiring them to match both motion and color information in a test stimulus. This task allowed to separate the encoding of stimulus motion and color from their current behavioral relevance on a trial-by-trial basis. We found that upcoming motor behavior can be robustly predicted from lPFC activity. In addition, we show that 8Av/45 encodes the color of a visual stimulus, regardless of its behavioral relevance. Most notably, whether a color matches the searched-for color can be decoded independent of a trial's motor outcome and while subjects detect unique feature conjunctions of color and motion. Thus, macaque area 8Av/45 computes, among other task-relevant information, the behavioral relevance of visual color features. Such a signal is most critical for both the selection of responses as well as the deployment of top-down modulatory signals, like feature-based attention.

Conflict of interest statement

The authors declare no competing interests.

Figures

Figure 1
Figure 1
A delayed match-to-feature task for monkeys. (A) The animals started an individual trial by fixating the central fixation point and releasing a manual push-button. Next, a sample was presented for 1 second. The sample cued the monkeys as to the trial type (either conjunction-matching, motion-matching or color-matching). Simultaneously, the sample stimulus contained the relevant visual feature(s) (color and/or motion) that had to be remembered and then matched to the test stimulus. Stochastic motion or a grey color signaled that the respective feature was irrelevant for the current trial. After a variable delay (800–1600 ms) we presented a test stimulus for 250 ms. The test always moved coherently in a cardinal direction and was always colored in one out of four isoluminant colors. The animals responded by depressing the response button when the test stimulus matched the feature(s) of the sample and received a liquid reward upon correct responses. If the test did not match the sample, the animals were rewarded for not responding within 600 ms of test stimulus onset. Eye fixation had to be maintained throughout sample, delay and test epochs. (B) We recorded local field potentials from the animals left lPFC by means of chronically implanted 96-channel microelectrode arrays. For both animals we analyzed the simultaneously recorded data from all available channels. Drawings by Klaus Lamberty, Deutsches Primatenzentrum GmbH. (C) We arbitrarily grouped colors and motion directions into pairs of features. This ensured that during test presentation, the likelihood of a target presentation was 50%.
Figure 2
Figure 2
Both color and motion features were relevant for the monkey’s behavior. (A) Sensitivity indices d’ were calculated separately for each task-relevant stimulus feature. In conjunction trials, we contrasted the responses to target stimuli matching both sample features (color and motion) with responses to non-target stimuli matching either none, or only a single feature. We compared those values with the sensitivity in single task trials, in which only one visual feature was behaviorally relevant. Error bars represent standard errors across n = 16 and n = 25 sessions, for monkey EDG and SUN, respectively. Stars indicate significant differences. (B) Median reaction times plotted separately for stimuli matching both searched-for features (hits in conjunction trials), for stimuli matching either color or motion, or stimuli matching none of the sample features (false-alarms). Those data are plotted as a function of matching features and thus contain hit trials and false-alarms, as indicated by the bar labels. Error bars indicate first to third quartile ranges.
Figure 3
Figure 3
Test color can be decoded from 8Av/45 activity. (A) We trained support vector machines to separate four different stimulus colors of the test stimulus based on time-locked 8Av/45 data. We estimated the performances of the classifiers with a 20-fold cross-validation procedure and contrasted the results with distributions of 500 runs of the same data, but with randomly shuffled trial labels. Here, we plot performance as the moving average in 28 ms sliding-windows with corresponding 95% confidence intervals. Chance performances are plotted as 99.9% confidence intervals of the shuffle distributions as overlapping, shaded areas. Above the x-axis, dots in corresponding colors indicate for which time-bins the decoding probability significantly differs from chance, evaluated at a α = 1e-4. The grey, shaded area illustrates the test stimulus duration. Orange, right-hand axes correspond to cumulative reaction-time distributions for each trial group and their corresponding target trials. Here, 100% equals to all target trials being terminated by the monkey. The orange, dashed line illustrates the fastest possible reaction time (300 ms, as defined by trial inclusion criteria). (B) Like A, but for the four cardinal motion directions.
Figure 4
Figure 4
Trial-by-trial behavior can be predicted from 8Av/45 activity. We trained support vector machines to separate correct trials with button presses (hit trials) from trials without button presses (correct rejections) based on time-locked LFPs recorded from 8Av/45 during test stimulus presentations. When the monkeys’ behavior was based only on stimulus color or a conjunction of color and motion (conjunction task), this decoding became significant between 148–164 ms after test stimulus onset for both monkeys. For decisions based only on stimulus motion, we observed lower decoding performances and later onsets (216–220 ms). Orange, right-hand axes correspond to cumulative reaction-time distributions for each trial group and their corresponding target trials. Dashed, orange lines at 300 ms indicate the start of response windows. No button presses occurred before this time. Other analysis details and plot layout like in Fig. 3.
Figure 5
Figure 5
The behavioral relevance of stimulus colors can be read out from 8Av/45 activity. For conjunction matching trials, we contrasted presentations of correctly ignored test stimuli that mismatched the sample in color and motion with test stimuli that were also correctly ignored but matched either the searched-for color or motion, but not both. We trained support vector classifiers to separate these two types of non-matching test stimuli and plotted the decoding performance over time and against corresponding noise distributions. Significant time-bins are indicated by color-corresponding dots above the x-axis. Other analysis details and plot layout like in Fig. 3.

Similar articles

See all similar articles

References

    1. Zaksas D, Pasternak T. Directional Signals in the Prefrontal Cortex and in Area MT during a Working Memory for Visual Motion Task. Journal of Neuroscience. 2006;26:11726–11742. doi: 10.1523/JNEUROSCI.3420-06.2006. - DOI - PMC - PubMed
    1. Hussar CR, Pasternak T. Flexibility of Sensory Representations in Prefrontal Cortex Depends on Cell Type. Neuron. 2009;64:730–743. doi: 10.1016/j.neuron.2009.11.018. - DOI - PMC - PubMed
    1. Bullock KR, Pieper F, Sachs AJ, Martinez-Trujillo JC. Visual and presaccadic activity in area 8Ar of the macaque monkey lateral prefrontal cortex. Journal of Neurophysiology. 2017;118:15–28. doi: 10.1152/jn.00278.2016. - DOI - PMC - PubMed
    1. Everling S, Tinsley CJ, Gaffan D, Duncan J. Filtering of neural signals by focused attention in the monkey prefrontal cortex. Nature Neuroscience. 2002;5:671–676. doi: 10.1038/nn874. - DOI - PubMed
    1. Bichot NP, Rossi AF, Desimone R. Parallel and serial neural mechanisms for visual search in macaque area V4. Science. 2005;308:529–534. doi: 10.1126/science.1109676. - DOI - PubMed
Feedback