Experiments were conducted in accordance with the Animal Care and Use Committee guidelines (INSERM, France). Eyes were kept closed by applying clear tape to a thin layer of glue. EO was verified by eye. Waking state was verified by tonic EMG activity. Visual stimulation details are provided in the text and Supplemental Experimental Procedures. Simultaneous VC and sSC recordings used pulled glass microelectrodes (1–2 MΩ) coupled to a direct-current amplifier (Axon Instruments) and multisite

linear array silicon Michigan Probes (Neuronexus Tech) coupled to a custom built AC amplifier (1000×, bandpass 1 Hz–5 kHz). V1 recordings were localized at 3.0–3.2 mm lateral to midline, and 0.0–0.5 rostral to λ, and sSC recordings to 0.5–0.8 lateral, CHIR99021 1 mm rostral. Cortical layer identification was accomplished via multiple criteria. Anatomical sections show layer 4 to be located 400–500 μm below the pial surface. Layer 4 was identified by the peak visual

response (Colonnese et al., 2010). The collicular projecting layer (lower 5a) was defined as 200 μm below L4 and containing large, spontaneously active units (Le Bon-Jego and Yuste, 2007). Multiunit firing was identified by high-pass filtering above 300 Hz and simple threshold discrimination (more than 4.3 times SD of baseline noise). Good discrimination was verified for each channel. We would like to thank R. Desimone for helpful comments, C. Yee and A. Birdsey-Benton for technical assistance, W. Lee GDC0449 for transferring mutant mice, and C. Tunca for biochemistry advice. This work was supported by NIH grant EY006039 to M.C.-P. M.T.C. was supported by a grant to Rustem Khazipov (INSERM, France) from the Agence Nationale de Recherche, France. “
“The intertwined problems of how agents learn about the environment and decide how to act are of central importance in the behavioral, cognitive, and neural sciences. One fundamental question is whether decisions rely on an internal model of the environment, replete with statistical information about the likely causes of outcomes or sensations, or whether they rely on simpler mechanisms, such as learning

the value of one action over another Ketanserin (Daw et al., 2005, Gläscher et al., 2010 and Sutton and Barto, 1998). All decisions are perturbed by multiple sources of uncertainty, but decision making is most demanding when the environment can change rapidly and without warning. An agent that explicitly encodes higher-order statistical information about the changing stimulation history, such as the transitional probabilities among hidden or observable states (Green et al., 2010), their variability (Preuschoff et al., 2008), and rates of change (Behrens et al., 2007), can tailor decision policy to account for this uncertainty, for example by discounting past rewards more steeply when the world changes faster (Rushworth and Behrens, 2008), or by selecting a sure prospect over an equal-valued but risky one (Christopoulos et al.