The hierarchical organization of the lateral prefrontal cortex
- University of California, United States
Figures
Figure 1 with 1 supplement
Experimental design.
(A) The design orthogonally manipulated factors of Stimulus Domain (verbal, spatial) and two forms of cognitive control: Contextual Control (low – Control, Delay; high – Restart, Dual) and Temporal …
Figure 1—figure supplement 1
Spatial task.
Example blocks of the spatial task wherein participants decided whether the stimulus followed the preceding stimulus in a star trace (Figure 1C). Details are otherwise identical to Figure 1.
Figure 2
Behavioral results.
Reaction times (top) and error-rates (bottom) as a function of condition in the sub-task (left) and return trials (right). Comparisons reflect Bonferroni corrected tests against the Control …
Figure 3 with 5 supplements
Univariate whole-brain results.
Temporal demands of cognitive control produced a gradient of activation along the rostral/caudal axis of the LPFC. Temporal Control activated the rostral LPFC (red), Contextual Control the mid LPFC …
-
Figure 3—source data 1
Statistical parametric maps of the univariate whole-brain results.
Maps contain t-statistics and have been thresholded to correct for multiple comparisons as detailed in the Materials and methods.
- https://doi.org/10.7554/eLife.12112.007
Figure 3—figure supplement 1
Univariate whole-brain results of Stimulus Domain.
Demands on Stimulus Domain produced differences along the dorsal/ventral axis of the LPFC. Effects of verbal processing (verbal > spatial; purple) were observed in the ventral LPFC, while effects of …
Figure 3—figure supplement 2
Effects of delay and dual.
Despite being associated with the fastest reaction times and lowest error-rates, the Delay condition demonstrated activation in the rostral-most areas of the LPFC. These results indicate that the …
Figure 3—figure supplement 3
Activation magnitudes within regions-of-interest.
Bars represent contrasts of each cognitive control condition against the Control condition, collapsing across Stimulus Domain. Data were split to provide unbiased estimates of activation (see …
Figure 3—figure supplement 4
Gradient visualization.
Contrasts of cognitive control overlaid to visualize the gradient and extent of LPFC coverage.
Figure 3—figure supplement 5
Univariate whole-brain robustness.
Data were split into two data sets through an alternating runs procedure. Whole-brain univariate contrasts were repeated for each data set and are depicted above. Details are otherwise identical to F…
Figure 4 with 1 supplement
Stimulus domain-sensitivity along the rostral/caudal axis.
Regions defined by peak effects of Temporal Control (red), Contextual Control (green), and Feature Control (blue). Points reflect the contrast estimates for spatial – verbal such that positive …
Figure 4—figure supplement 1
Stimulus domain-sensitivity along the rostral/caudal axis robustness.
Data were split into two data sets through an alternating runs procedure. Region-of-interest analyses were repeated for each data set and are depicted above. Details are otherwise identical to Figure…
Figure 5 with 1 supplement
Temporal activation-behavior relationships.
Correlations between activation and reaction time (RT). Current RT corresponds to sub-task trials while future RT corresponds to return trials. RT measures have been normalized within-subject across …
-
Figure 5—source data 1
Statistical parametric maps of the activation-behavior correlations.
Maps contain r-values and have been thresholded to correct for multiple comparisons as detailed in the Materials and methods.
- https://doi.org/10.7554/eLife.12112.016
Figure 5—figure supplement 1
Temporal activation-behavior relationships robustness.
Data were split into two data sets through an alternating runs procedure. Correlations were computed for each dataset in the same way as depicted in Figure 5 of the main text. ~puncorrected<0.05; *pc…
Figure 6 with 5 supplements
LPFC dynamic causal model.
Interactions within the LPFC were modeled using dynamic causal modeling. Bayesian model selection indicated that the depicted model was the best model of the dynamics among the models tested. Arrows …
Figure 6—figure supplement 1
Depiction of dynamic causal modeling.
Tiles represent transient stimuli while blocks represent sustained modulations. For simplicity, all modulations are depicted as bidirectional. Top: an example of a Spatial Control block. Spatial …
Figure 6—figure supplement 2
LPFC dynamic causal model robustness.
Data were split into two data sets through an alternating runs procedure. Black numbers indicate input and fixed connectivity parameter estimates Colored numbers indicated modulations by Spatial …
Figure 6—figure supplement 3
Path coefficients in a rostral-to-caudal model.
Inference on parameters was performed on an alternative model of LPFC organization wherein content demands produced feedback modulations from mid areas to caudal areas. Notably, this is a strict …
Figure 6—figure supplement 4
Connectivity as a function of cost.
Examples of connectivity profiles at different costs are depicted. The chosen connectivity profile was at cost 18, falling in the middle of the sampling range. rMFG was disconnected from the rest of …
Figure 6—figure supplement 5
Model comparison.
Modulatory parameter structure for the winning (bottom left) and closest competitor models. The depicted models are the only models with exceedance probabilities (Xp) greater than 0.05, among the 99 …
Figure 7 with 2 supplements
Hierarchical structural dependencies.
Based on fixed connectivity of the model depicted in Figure 6A, hierarchical strength was calculated as the difference between inward and outward projections along the rostral/caudal axis. …
Figure 7—figure supplement 1
Hierarchical structural dependency robustness.
Data were split into two data sets through an alternating runs procedure. Details are otherwise identical to Figure 7 in the main text.
Figure 7—figure supplement 2
Average and total connectivity strength.
(A) Average connectivity strength reflects the mean of rostral/caudal fixed connectivity for each region. (B) Total connectivity strength reflects the sum of rostral/caudal fixed connectivity. Both …
Figure 8 with 3 supplements
LPFC dynamics and higher-level cognitive ability.
Neural metrics were based on modeled estimates of effective connectivity and their modulations (Figure 6). Metrics reflecting top-down LPFC modulations by cognitive control demands (top-down …
Figure 8—figure supplement 1
Model and trait correlations and covariances.
(A) Correlations between top-down parameters. Individual differences in the magnitude of top-down modulations tended to be correlated. The negative correlations reflect anti-correlation between …
Figure 8—figure supplement 2
Dynamic causal modeling derived individual difference measures.
Top-down modulations resulting from Temporal Control (red) and Contextual Control (green) were combined using principle components analysis (PCA) to represent top-down strength. Bottom-up …
Figure 8—figure supplement 3
LPFC dynamics and higher-level cognitive ability robustness.
Data were split into two data sets through an alternating runs procedure. Details are otherwise identical to Figure 8 in the main text.
