Fig. 4. ALM neurons show comprehensive activities of sensory and choice information. (A) Steps for inferring Sensory Mode (SM) and Choice Mode (CM). Step 1: A firing rate matrix was constructed every 50 ms. Note that we concatenated the firing rates across sessions (Methods). Step 2: A projection vector was inferred that discriminates sensory or choice information maximally. In inferring SM, we excluded the error trials. Step 3: Linear discriminant vectors were inferred over the sample period (or delay period) and averaged across time within the period. To infer SM (or CM), we used the firing rates during the sample period (or delay period). Step 4: SM and CM were kept orthonormal to each other for the specificity of the extracted signals. (B) Absolute dot products between SM and CM across time in the tactile delayed-response task. SM and CM were inferred at each time point before orthonormalization between them [steps 1~3 in (A)]. The sample and delay periods are marked by red boxes. (C) SM and CM in the auditory delayed-response task, as described in (B). (D) Sensory and choice signals obtained by projecting population activities onto one-dimensional spaces via SM (top) and CM (bottom) during the tactile delayed-response task. A solid line indicates mean sensory (bottom) or choice (top) signals across repetitions of calculating SM and CM. Shaded regions represent the standard deviation. Sensory signals exhibited similar temporal patterns between HR and ER or between HL and EL (top). Choice signals exhibited similar temporal patterns between HR and EL or between HL and ER (bottom). (E) Sensory and choice signals in the auditory delayed-response task, as described in (D).
© Exp Neurobiol
