Introduction Visual and affective predictions in the orbitofrontal cortex The orbitofrontal cortex (OFC) is known for its role in the generation of outcome predictions. In visual recognition, outcome predictions pertain to the identity of the object being perceived. It has been proposed that the perceptual resemblance of the low spatial frequency (LSF) information with known objects is the trigger of predictions in OFC (Bar 2003). Ambiguity in the visual image would be resolved by identity predictions sent from OFC to the visual cortex. Two anatomo-functional networks in OFC have lead to the proposal that in addition to identity predictions guiding visual recognition, OFC could generate affective predictions meant to guide bodily reaction through the amygdala, brainstem and hypothalamus (Barrett & Bar 2009). In this study, we examine the respective contributions of various identity and affective dimensions in visual information to OFC activity. Specifically, we ask these questions: • Do low level variables like luminance or spatial frequency content activate OFC? Are visual predictions necessary to activate OFC with visual information? • Is it the strength of identity predictions or their number which triggers OFC activity? • Do affective valence / arousal trigger affective predictions in OFC? • What are the targets of these predictions? Behavioral norming fMRI In a first study, we created a set of normed visual stimuli. These stimuli were rated along several dimensions pertaining to meaning and core affective reaction. Subjects (N=40) were presented with the objects, one at a time and answered several questions about it: • What was it? • How confident are you that it was what you said? • How did it make you feel? (on a pleasantness and arousal grid) • What else could it be? (while the object was presented again) In the fMRI experiment, subjects performed a 2-back working memory task with the objects rated during the behavioral norming and with meaningless gabor patches of varying spatial frequencies. No No No No Yes! blocks of 16 images 10s rest x 11 x 3 runs Rapid ER design Stimuli presented for 250ms. One every 2s (TR). 4-5 target 2-back repeats per block 1-2 non target 1-back repeats 6 1 st presentations of object images 4 1 st presentations of Gabor patches Behavior: average d’=2.53 Events of interests: correct rejections. “you will perform better if you try to identify each image. Recognizing the image will make it easier to remember. ” 7 min 42 s x 3 17 healthy subjects Siemens 3T trio scanner 32ch head coil T2* EPI sequence optimized to increase OFC visibitity (Deichman et al. 2003): Slice orientation 30° off AC-PC line Zshimming 33 interleaved slices, 2000ms TR, TE 28ms, 3 mm thickness, 64x64 matrix, 200mm FOV (vox size 3.125x3.125x2.5mm) Controlling for low level information Ratings are correlated Correlation coefficient More confidently identified objects are more consensual. Better confidence corresponds to fewer possible alternative interpretations. More confidently identified objects are more pleasant. Except certain outliers identified with ease as unpleasant. Pleasantness Arousal Consensus Confidence Low Middle High Number of Categories (Frequency of the most used label) (Self rated) (Average number of different basic categories used for an object) (Self rated) (Self rated) More consensual objects are less likely interpreted as belonging to many possible categories. Fewer inter-subject predictions corresponds to fewer intra-subject predictions. High confidence and consensus correspond to stronger but fewer possible alternative interpretations of the current image, hence leading potentially to fewer but stronger predictions. Images luminance T>2.58 p<.01 Gabor spatial frequencies T>2.58 p<.01 None of the low level features of images activates OFC. Effects are found only in early visual areas. Even at very low threshold. Is OFC activated to form visual predictions? Contrasting LSF images vs. Gabor patches Evoked Response (% signal change) Evoked Response (% signal change) Evoked Response (% signal change) T>3.53 p<.05 (FDR) T>3.62 p<.001 Left Medial Right Main effect of image type Evoked Response (% signal change) Evoked Response (% signal change) Evoked Response (% signal change) Effect of ratings: Consensus Confidence % signal change Consensus % signal change Images for which participants try to guess the identity of the object activate OFC. But this activation is not modulated by any of our ratings (only consensus shown here). Maximilien Chaumon 1 , Kestas Kveraga 1 , Lisa Feldman Barrett 1,2 & Moshe Bar 1 1 - Department of Neurology, Athinoula Martinos center for biomedical imaging, Massachusetts General Hospital and Harvard Medical School 2 - Department of Psychology, Boston College Material & Methods Results Conclusions OFC is sensitive to the meaning of LSF information in images. In conditions where the stimulus obviously is not a real object and no identity predictions are formed (gabor patches), no activity is observed in OFC. On the other hand, when subjects try to guess the identity of the presented objects, OFC gets activated. This is consistent with the idea that OFC generates identity predictions to guide visual recognition. Affect did not activate OFC in this task. Possibly because core affect was not relevant to the ongoing task. We need to find a relevant dimension. Connectivity with other brain regions supposed to be involved in visual and affective top-down effects are under investigation. Activation in medial OFC correlates with consensus and confidence ratings. Stronger but fewer predictions activate OFC. Whole brain analysis, parametric modulation by ratings T>3.53 p<.05 (FDR) T>3.62 p<.001 T>3.53 p<.05 (FDR) References Bar, M. (2003). A cortical mechanism for triggering top-down facilitation in visual object recognition. Journal of Cognitive Neuroscience, 15(4), 600–609. Barrett, L., & Bar, M. (2009). See it with feeling: affective predictions during object perception. Philosophical Transactions of the Royal Society B:, 364(1521), 1325-1334. Deichmann, R., Gottfried, J. A., Hutton, C., & Turner, R. (2003). Optimized EPI for fMRI studies of the orbitofrontal cortex. NeuroImage, 19(2), 430-441. Acknowledgements This study is funded by a NIH Pioneer award to LFB.