For each trial, we defined a test set, which HIF inhibitor contained the trial, and a training set, which contained all the trials except the trial in the test set. We ranked the cells according to their RT selectivity computed using an ANOVA based on the training set and fit a linear discriminant to the training set. We then decoded the test set on an increasing subset of cells from two to the maximum number available ranked according to the results of the ANOVA on the training set. We repeated this analysis for each trial and computed the probability of correct classification. We report the results of decoding across all conditions based on the same number of cells (eight cells)

in Figure 5 and present the data across increasing subsets of cells in Figure S2. This procedure ensured that the decoding results were not influenced by overfitting. Significant differences between the performance of the decoding for each group were determined using a binomial test. The mean firing rate of cells in the coherent and not coherent groups was different. To test whether the mean firing rate affected the decoding probability, we subtracted the mean firing rate across all trials from each cell and reran the decoding algorithm. Additionally, we performed the same decoding analysis for the significantly coherent units using firing rates

that were decimated check details by removing each individual spike with 50% probability in order to match the mean firing rate of the units that were not coherent with the LFP. We thank Eva Tsui for assistance with animal training, Gerardo Moreno for surgical assistance, Roch Comeau, Stephen Frey and Brian Hynes for customizations to the Brainsight system and Bob Shapley for comments on the manuscript. This work was supported, in part, by CRCNS Program award R01 MH-087882, NSF CAREER Award BCS-0955701, a Fellowship in Brain Circuitry from the Patterson Trust (HLD), NIH Training grant T32 MH-19524 (HLD), NIH Training grant T32 EY-007136 (MAH), a Career Award in the Biomedical Sciences from the Burroughs Wellcome Fund (BP), a Watson Investigator Program Award from NYSTAR

(BP), a TCL McKnight Scholar Award (BP), and a Sloan Research Fellowship (BP). “
“A long-debated and critical question in schizophrenia and other neuropsychiatric illnesses is whether the underlying neural impairments of the disorder are immutably fixed, or whether they can respond in a significant and enduring manner to targeted behavioral interventions. Here, we demonstrate that intensive neuroscience-informed cognitive training can improve brain function in patients who have been ill for decades. Specifically, we show that it can improve a complex and clinically meaningful “reality monitoring” process defined as the ability to distinguish the source of internal experiences (self-generated information) from outside reality (external information) (Bentall et al., 1991, Johnson et al., 1993, Keefe et al.