Model-based fMRI reveals dissimilarity processes underlying base rate neglect
- Texas Tech University, United States
- Texas A&M University, United States
- University of Sydney, Australia
Figures
Figure 1
Abstract task design and an example trial.
In the headings, I = imperfect predictor, PC = common perfect predictor, PR = rare perfect predictor. The second row refers to the visual category used for each stimulus feature: F = face, S = scene, O = object. Each following row corresponds to a learning trial, with a ‘1’ indicating the presence of the feature and ‘0’ indicating its absence.
Figure 2
Learning curves.
Points depict proportions correct for common and rare disease predictions over the 12 blocks of the training phase (mean ± SEM). *p < 0.05.
Figure 3 with 1 supplement
Multivoxel pattern similarity to common and rare stimulus features for ambiguous trials in which participants made common (left) and rare (right) responses (mean ± SEM).
Purple squares correspond to the objects/scenes associated with the common category, while green squares correspond to the objects/scenes associated with the rare category in a given trial. Gray squares depict mean pattern similarity to the non-present face dimension for both response types. *p < 0.05. No error bars are included for the gray bars because face dimensions were not included in the overall mixed effects model.
Figure 3—figure supplement 1
Multivoxel pattern similarity to common and rare objects, scenes, and faces for ambiguous trials in which participants made common (left) and rare (right) responses (mean ± SEM).
Purple corresponds to objects/scenes associated with the common category, while green corresponds to the objects/scenes associated with the rare category on a given trial. Triangles indicate objects, and circles indicate scenes. Gray squares depict mean pattern similarity to the non-present face dimension for both response types. *p < 0.05. No error bars are included for the gray bars because face dimensions were not included in the overall mixed effects model.
Figure 4
Multivoxel pattern similarity to each feature type during the learning phase (mean ± SEM).
The left panel is for trials predictive of a common disease, and the right for trials predictive of a rare disease. Red points represent perfect predictors, purple points represent imperfect predictors (faces), and blue points represent non-present features. **p < 0.01, ***p < 0.001.
Figure 5
Associations between multivoxel pattern similarity to stimulus dimensions during the learning phase and individual differences in base rate neglect.
For each graph, the y-axis depicts the proportion of rare responses made by each subject on ambiguous test trials, while the x-axis depicts subjects’ mean BOLD pattern similarity to a respective stimulus dimension over the course of learning. **p < 0.01.
Figure 6
Results from the model-based univariate analysis.
(A) Depicts activation that tracks similarity-based contributions to choice (summed similarity to the winning category). Red depicts activation positively correlated with similarity-based contributions and blue depicts negatively correlated activation. (B) Depicts brain regions that are positively correlated with dissimilarity-based contributions to choice (summed dissimilarity to the non-winning category).
Tables
Table 1
Observed and dissGCM-predicted response probabilities for the test phase.
The feature combinations presented at test are listed in the leftmost column: F = face, S = scene, O = object. In the headings, D1–D4 correspond to the four possible category responses (diseases). Bold, italicized values indicate results for the key ambiguous stimuli in which a scene was paired with an object.
| Behavior | dissGCM | |||||||
|---|---|---|---|---|---|---|---|---|
| Test item | D1 | D2 | D3 | D4 | D1 | D2 | D3 | D4 |
| F1 + S1 | .972 | .018 | .008 | .003 | .971 | .014 | .008 | .008 |
| F1 + O1 | .031 | .962 | .008 | 0. | .063 | .901 | .018 | .018 |
| F2 + O2 | .005 | 0. | .987 | .008 | .008 | .008 | .972 | .013 |
| F2 + S2 | .023 | .008 | .069 | .901 | .018 | .018 | .059 | .905 |
| F1 | .667 | .295 | .023 | .015 | .667 | .283 | .025 | .025 |
| F2 | .107 | .038 | .550 | .305 | .027 | .027 | .640 | .307 |
| S1 | .848 | .061 | .008 | .083 | .955 | .015 | .015 | .015 |
| O1 | .008 | .908 | .069 | .015 | .040 | .880 | .040 | .040 |
| O2 | .008 | .069 | .908 | .015 | .012 | .012 | .965 | .012 |
| S2 | .023 | .053 | .008 | .916 | .032 | .032 | .032 | .904 |
| S1 + O1 | .414 | .487 | .073 | .027 | .419 | .496 | .042 | .042 |
| O2 + S2 | .035 | .047 | .453 | .465 | .038 | .038 | .460 | .464 |
| F1 + O2 | .131 | .238 | .631 | 0. | .156 | .174 | .658 | .013 |
| F1 + S2 | .264 | .062 | .008 | .667 | .178 | .237 | .023 | .563 |
| F2 + S1 | .608 | .031 | .138 | .223 | .657 | .013 | .155 | .175 |
| F2 + O1 | .008 | .674 | .302 | .016 | .024 | .567 | .170 | .239 |
| O1 + O2 | .008 | .514 | .475 | .004 | .040 | .444 | .477 | .040 |
| S1 + S2 | .397 | .065 | .011 | .527 | .400 | .040 | .040 | .520 |
Table 2
Activated clusters and peaks for the fMRI results in Figure 6.
https://doi.org/10.7554/eLife.36395.011| Contrast | Regions | Peak t-value | Peak MNI coordinates (x,y,z) | Number of voxels | Cluster P |
|---|---|---|---|---|---|
| Similarity > 0 | |||||
| Rostral and ventral medial prefrontal cortex | 7.76 | 0, 54, –2 | 2268 | p<0.001 | |
| Middle temporal gyrus | 5.74 | 64, –6, −14 | 1050 | p=0.016 | |
| Precentral gyrus | 7.65 | 42, –16, 62 | 1031 | p<0.001 | |
| Precentral gyrus | 4.81 | 0, –30, 58 | 463 | p=0.008 | |
| Middle temporal gyrus | 4.98 | −62, –2, −16 | 322 | p=0.016 | |
| Parietal operculum cortex | 5.74 | −34, –30, 18 | 282 | p=0.027 | |
| Hippocampus | 5.25 | −22, –18, −16 | 271 | p=0.019 | |
| Lateral occipital cortex (inferior) | 4.68 | 50, –72, 12 | 157 | p=0.039 | |
| Similarity < 0 | |||||
| Superior parietal and lateral occipital cortex (superior) | 8.43 | −44, –46, 58 | 5123 | p<0.001 | |
| Middle frontal gyrus | 7.01 | −52, 12, 36 | 2491 | p<0.001 | |
| Dorsal medial PFC | 9.00 | −2, 18, 46 | 917 | p<0.001 | |
| Cerebellum | 6.11 | 28, –64, −28 | 769 | p=0.002 | |
| Middle frontal gyrus | 6.23 | 32, 2, 62 | 730 | p=0.002 | |
| Middle frontal gyrus | 5.89 | 44, 36, 30 | 391 | p=0.009 | |
| Inferior temporal gyrus | 6.09 | −54, –52, −12 | 375 | p=0.009 | |
| Inferior temporal gyrus | 6.53 | 60, –52, −10 | 271 | p=0.016 | |
| Cerebellum | 6.08 | −28, –60, −32 | 186 | p=0.031 | |
| Thalamus | 4.88 | −10, –18, 10 | 175 | p=0.038 | |
| Dissimilarity > 0 | |||||
| Occipital cortex | 7.55 | 12, –78, 12 | 1372 | p<0.001 | |
| Fusiform and lateral occipital cortex (inferior) | 6.82 | 42, –60, −18 | 865 | p=0.002 | |
| Fusiform and lateral occipital cortex (inferior) | 5.97 | −36, –52, −18 | 575 | p=0.003 | |
| Middle frontal gyrus | 5.00 | 54, 16, 34 | 255 | p=0.020 | |
| Frontal pole (rostrolateral PFC) | 5.93 | −42, 52, –6 | 130 | p=0.047 |
Additional files
-
Supplementary file 1
Number of trials where common and rare responses were made for each participant over the 24 ambiguous scene-object test trials.
- https://doi.org/10.7554/eLife.36395.012
-
Transparent reporting form
- https://doi.org/10.7554/eLife.36395.013
Download links
A two-part list of links to download the article, or parts of the article, in various formats.
Downloads (link to download the article as PDF)
Open citations (links to open the citations from this article in various online reference manager services)
Cite this article (links to download the citations from this article in formats compatible with various reference manager tools)
Model-based fMRI reveals dissimilarity processes underlying base rate neglect
eLife 7:e36395.
https://doi.org/10.7554/eLife.36395
