Functional specialization of medial and lateral orbitofrontal cortex in inferential decision-making

Abstract

Inferring prospective outcomes and updating behavior are prerequisites for making flexible decisions in the changing world. These abilities are highly associated with the functions of the orbitofrontal cortex (OFC) in humans and animals. The functional specialization of OFC subregions in decision-making has been established in animals. However, the understanding of how human OFC contributes to decision-making remains limited. Therefore, we studied this issue by examining the information representation and functional interactions of human OFC subregions during inference-based decision-making. We found that the medial OFC (mOFC) and lateral OFC (lOFC) collectively represented the inferred outcomes which, however, were context-general coding in the mOFC and context-specific in the lOFC. Furthermore, the mOFC-motor and lOFC-frontoparietal functional connectivity may indicate the motor execution of mOFC and the cognitive control of lOFC during behavioral updating. In conclusion, our findings support the dissociable functional roles of OFC subregions in decision-making.

Keywords: Behavioral neuroscience; Neuroscience; Psychology.

Graphical abstract

Figure 1

Illustration of the experimental task and the definition of the trial identities (A) Task design. In this task, the participants were required to judge the age of a certain category. First, a text introduction was presented on the screen, indicating the category that needed to be focused on (face, as shown in the Figure 1A). Then, participants needed to continuously judge the age of the instructed category (face). When the age of the face changes (cue for the category switch), participants should switch their attention to another category in the next trial (category switch). Based on the changing categories (inferred outcomes), there were two conditions in this task, outcome-stable (cue and outcome-stable trials) and outcome-switching (switch trials). (B) Stimulus presentation procedure. (C) Identity definition for each trial. According to the task rules, we defined the identity of each trial according to the category and age information from both the previous trial and the current trial. Therefore, each trial contained the information of four pieces (previous category, previous age, current category, and current age) which led to a total of 2⁴ = 16 trial identities in the task.

Figure 2

Behavioral performance during the task-fMRI scanning (A) Response accuracy for the whole task across all participants. (B) Response accuracy for both the face trials and house trials. (C) Response time for both the face trials and house trials. Whisker plots indicate the distribution of the response accuracy or response time, and the whiskers and error bars represent the variability of data points (Mean ±2 standard deviations, M ± 2SD). The dots indicate the raw data of the participants. M, mean; n.s., non-significant; ^∗∗∗, p < 0.001.

Figure 3

Classification of decision-related information in the subregions of the orbitofrontal cortex (OFC) (A) Anatomical location of each of the OFC subregions. According to the Brainnetome atlas, we defined 5 subregions of the OFC. They were located in the medial part (mOFC₁ and mOFC₂) and the lateral part (lOFC₁, lOFC₂, and lOFC₃); their sizes and MNI coordinates are listed in Table S1. (B) Classification accuracy of the previous category and previous age in each of the OFC subregions. (C) Same as (B) except for the inferred outcome (current category) and the current age. The chance level of the classification accuracy was 0.50. Whisker plots indicate the data distribution of the classification accuracy, and the whiskers represent the variability of data points (Mean ±2 standard deviations, M ± 2SD). ^∗, p < 0.05/5 = 0.01 (Bonferroni correction). (D) Partial correlation between classification accuracy and response accuracy. Dots correspond to the raw data from the participants and the shadow indicates a 95% confidence interval (CI). Histograms indicate the classification accuracy for the mOFC₁ (blue) and lOFC₃ (red) as well as the response accuracy (yellow).

Figure 4

Significant clusters from the whole-brain multi-voxel pattern analysis (MVPA) searchlight on decision-related information The displayed clusters were significant at p_FDR < 0.05, with the cluster size determined by threshold-free cluster enhancement (TFCE, 5,000 iterations). No voxel showed significant classification accuracy on the previous age. The anatomical label was determined according to the HCP template. The peak coordinates are listed in Table S4. Abbreviations: m/lOFC, medial/lateral orbitofrontal cortex; FuG, fusiform gyrus; STS, superior temporal sulcus; PhG, parahippocampal gyrus; V1, primary visual cortex; V2, second visual cortex; V3, third visual cortex; V4, fourth visual cortex; Tha, thalamus; SAC, sensory association cortex.

Figure 5

Within- and cross-condition classification on the inferred outcomes for each of the orbitofrontal cortex (OFC) subregions (A) The procedure of the within-cross-condition classification. Within-condition classifiers were trained to classify the activation patterns of the OFC subregions into “faces” or “houses” within the outcome-stable (classifier 1) and the outcome-switching conditions (classifier 2) separately. Then, we estimated the accuracy of these classifiers in predicting the labels of the OFC activation patterns within the same condition (e.g., trained in the outcome-switching and tested in the same condition). The within-condition classification accuracy was obtained by averaging the prediction accuracy of the two classifiers. For the cross-condition classification, we trained the classifier in outcome-switching and tested it in the stable condition or vice versa. The cross-condition classification accuracy was obtained by averaging the prediction accuracy. (B) Accuracy of the within- and cross-condition classification in each OFC subregion. The chance level of the classification accuracy was 0.50. The whiskers represent the variability of data points (Mean ± 2 standard deviations, M ± 2SD). ^∗∗, p < 0.01; ^∗∗∗, p < 0.001. Abbreviations: n.s., not significant.

Figure 6

Psychophysiological interactions (PPI) for the seeds at the mOFC₁ and lOFC₃ and their behavioral relevance (A) Significant brain regions that interacted with the mOFC₁ and lOFC₃ during behavior updating, obtained from GLM2 and GLM3. The significance threshold was set at the voxel level p < 0.025 with Gaussian random field (GRF) correction at the cluster level Z > 3.09. Peak coordinates are reported in MNI space (See also Table S3). Color bars indicate the Z-value for seed regions in the mOFC (cold color) and lOFC₃ (warm color). (B) Pearson’s correlation between the PPI βs obtained from significant clusters in the visual area and the response accuracy. Dots correspond to the data of the participants and the shadow indicates a 95% confidence interval (CI). Histograms indicate the interactions (PPI βs) between V1 and mOFC₁ (blue) as well as lOFC₃ (red). Abbreviations: FPC, frontal polar cortex; dlPFC, dorsal lateral prefrontal cortex; OPC, opercular cortex; SMA, supplementary motor area; PMC, premotor cortex; M1, primary motor cortex; dACC, dorsal anterior cingulate cortex; SMC, somatosensory cortex; SMG, supramarginal gyrus; MTG, middle temporal gyrus; V1, primary visual cortex; V3, third visual cortex; m/lOFC, medial/lateral orbitofrontal cortex; L(R), left (right) hemisphere.