Neuropil signals were removed by first selecting a neuropil ring surrounding each neuron (excluding adjacent rings; Kerlin et al., 2010), estimating the common time course of all such rings in the image (first principal component), and removing this component from each cell’s time course (scaled by the fluorescence of the surrounding ring). For subsequent analyses, only cells that were significantly driven by at least one stimulus
type were included (t tests with Bonferroni correction; for example, p < 0.05/96 in Figure 3). Estimates for orientation preference (Figure 3) were obtained by vector averaging (1/2 × tan−1(ΣRi(θi)sin(2θi)/ΣRi(θi)cos(2θi)), where θi is the orientation of each stimulus, and Ri is the average response to that stimulus; Kerlin et al., 2010). Similarly, direction preference (Figure 2) was defined as tan−1(ΣRj(θj)sin(θj)/ΣRj(θj)cos(θj)),
where θj denotes the direction of AC220 manufacturer each stimulus, and Rj is the average response to that stimulus. Direction selectivity (Figure 2) was defined as (Rpeak − Rnull)/(Rpeak + Rnull), where Rpeak is the peak response (across 16 directions) and Rnull is the average response Z-VAD-FMK ic50 at 180° from peak. We thank Lindsey Glickfeld and Vincent Bonin for assistance with resonance scanning and data analysis, Vladimir Berezovskii for assistance with histology, Demetris Roumis and Christine Mazur for surgical contributions, Jeff Curry for behavioral training, and Sergey Yurgenson, Peter O’Brien, Aleksandr Vagodny, Anthony De Simone, and Matthias Minderer for technical contributions. We thank Chris Deister, Aaron Kerlin, and members of the Reid and Andermann labs for advice, suggestions, and discussion. This work was supported by the NSF CAREER Award DBI-0953902 (M.J.L. and N.G.), Kavli Center for Neuroscience (M.W. and D.A.M.), Swebilius award (R.N.S.S), NIH (D.A.M.), NIH R01 EY018742 and EY010115 (R.C.R.),
the Ludcke Foundation and Pierce Charitable Trust (M.L.A.), and the Smith Family Foundation (M.L.A.). “
“Why is our behavior at times automatic and driven by habit and at other times deliberative and focused on a specific goal? Although most of us seamlessly switch between these modes of behavior, it has been suggested that a relative dominance whatever of either habit-like or goal-directed modes of behavior underpin a range of disorders that span addictions (Everitt and Robbins, 2005) through to Parkinson’s disease (de Wit et al., 2011). This renders understanding the parsing of control between these two modes of decision making a pressing issue. Here we address whether it is possible to causally manipulate their relative dominance. An elegant computational framework that captures the presence of (often competing) habit-like and goal-directed behaviors is provided by a formulation of model-free and model-based control (Daw et al., 2005 and Dayan and Niv, 2008).