hide
Free keywords:
-
Abstract:
While short-term memory is an essential requirement for behavior, knowledge about its neural code remains limited. In particular, how the brain organizes maintenance of multiple items at the same time has received little attention in this context. We trained two macaque monkeys to perform well in a memory task with load 1 and then, without further training had them memorize one to four visual sample stimuli presented sequentially over a period of 900 ms. Behavioral performance was above chance for all load conditions from the first session on. After a 3 second delay, the monkeys had to decide whether a newly presented test stimulus was part of the memorized set or not. During all subsequent sessions in which the monkey performed with load > 1, we recorded multi-unit activity and LFPs in prefrontal cortex using up to 16 micro-tetrodes. We performed spike sorting with a new algorithm (bayes optimal template matching) and analyzed rate modulations across task time: preliminary results show the firing rate of most sorted single units is significantly modulated during stimulus presentations and delay, compared to the level of baseline activity. Separating units from the compound tetrode multi-unit signal provides additional information: many multi-units exhibit no significant (Friedman test, p<0.05) rate modulation for different memory loads, while we find multiple load-selective single units after sorting the respective spikes. Most selective sorted units modulate their firing rate during the first second of the delay period. For the majority of these units, the rate during this period is increased compared to the baseline activity. Later in the delay period, the selective units exhibit no preference towards an increase in firing rate. These results suggest an active memory encoding phase shortly after the stimulus presentation. We are now starting to assess single trial classification performance and include analyses of LFP oscillations and correlations between units.
