Data acquisition and preprocessing

Recordings were amplified and bandpass filtered (Plexon Instruments PBX2/16sp-G50, × 1,000 amplification, 0.3–8 kHz bandpass) and digitized at 31.25 kHz. Noise was reduced by common average referencing²⁵; the average voltage signal across all electrodes in the brain was calculated and subtracted from the signal in each electrode. Poly2 electrodes have two vertical rows of recording pads with 50μm inter-pad spacing. These pads were divided into groups of 4 spatially adjacent pads, which were treated as the 4 channels of one tetrode. Adjacent tetrodes used neighboring, non-overlapping sets of pads, and were 50 μm apart. Poly3 electrodes were also divided into 4-channel tetrode groups. Because Poly3 electrodes have three vertical rows of recording pads, they contain 6 pads within each 50μm depth range, and thus there are 15 possible tetrode configurations for each 50μm depth range (i.e., 6 choose 4 configurations). We selected just one tetrode for each 50μm depth range by choosing the 4 pads in each 50μm depth range that exhibited the best signal-to-noise ratio. Signal-to-noise ratio was calculated independently for each electrode pad as the kurtosis (i.e. “peakedness”) of the voltage signal across time: kc=E[(v(t)c−E[v(t)c])4]. Adjacent tetrodes on Poly3 electrodes used non-overlapping sets of pads. Spike detection and sorting were performed on each tetrode separately.