Limits...
Distinct Neural Substrates for Maintaining Locations and Spatial Relations in Working Memory

View Article: PubMed Central - PubMed

ABSTRACT

Previous work has demonstrated a distinction between maintenance of two types of spatial information in working memory (WM): spatial locations and spatial relations. While a body of work has investigated the neural mechanisms of sensory-based information like spatial locations, little is known about how spatial relations are maintained in WM. In two experiments, we used fMRI to investigate the involvement of early visual cortex in the maintenance of spatial relations in WM. In both experiments, we found less quadrant-specific BOLD activity in visual cortex when a single spatial relation, compared to a single spatial location, was held in WM. Also across both experiments, we found a consistent set of brain regions that were differentially activated during maintenance of locations vs. relations. Maintaining a location, compared to a relation, was associated with greater activity in typical spatial WM regions like posterior parietal cortex and prefrontal regions. Whereas maintaining a relation, compared to a location, was associated with greater activity in the parahippocampal gyrus and precuneus/retrosplenial cortex. Further, in Experiment 2 we manipulated WM load and included trials where participants had to maintain three spatial locations or relations. Under this high load condition, the regions sensitive to locations vs. relations were somewhat different than under low load. We also identified regions that were sensitive to load specifically for location or relation maintenance, as well as overlapping regions sensitive to load more generally. These results suggest that the neural substrates underlying WM maintenance of spatial locations and relations are distinct from one another and that the neural representations of these distinct types of spatial information change with load.

No MeSH data available.


Related in: MedlinePlus

Mean difference Beta weights (Corresponding Quadrant – Opposite Quadrant) from visual cortex ROIs (V1–V3) for Location and Relation trials under low (left) and high (right) load conditions. For low load, the significant trial type × quadrant interaction suggests less quadrant-specific BOLD activity for Relation compared to Location trials, which replicates Experiment 1. Error bars represent within-subject standard error of the mean. ∗p < 0.001.
© Copyright Policy
Related In: Results  -  Collection

License
getmorefigures.php?uid=PMC5121279&req=5

Figure 7: Mean difference Beta weights (Corresponding Quadrant – Opposite Quadrant) from visual cortex ROIs (V1–V3) for Location and Relation trials under low (left) and high (right) load conditions. For low load, the significant trial type × quadrant interaction suggests less quadrant-specific BOLD activity for Relation compared to Location trials, which replicates Experiment 1. Error bars represent within-subject standard error of the mean. ∗p < 0.001.

Mentions: To explore this 3-way interaction we tested two separate 2 (trial type: Location vs. Relation) × 2(quadrant: corresponding vs. opposite) repeated-measures ANOVAs, one for each load. For low load, a main effect of quadrant emerged, F(1,29) = 78.33, p < 0.001, with more activity in the corresponding compared to opposite quadrant. The main effect of trial type did not reach significance, F(1,29) = 2.56, p = 0.12. Importantly, the trial type × quadrant interaction was significant, F(1,29) = 50.12, p < 0.001, with the difference in the BOLD response between corresponding and opposite quadrants being larger for Location trials than Relation trials. Follow-up t-tests confirmed that BOLD activity in low load Location and Relation trials were significantly different in the corresponding quadrant, t(29) = 3.96, p < 0.001, but not for the opposite quadrant, t(29) = 1.31, p = 0.20. As Figure 7 illustrates, this low load result is a direct replication of Experiment 1’s ROI results. For high load, the main effect of quadrant approached significance, F(1,29) = 2.86, p = 0.10, with more activity in the corresponding compared to opposite quadrant. However, neither the main effect of trial type, F(1,29) = 0.07, p = 0.80, nor the trial type × quadrant interaction, F(1,29) = 0.16, p = 0.69, approached significance (Figure 7).


Distinct Neural Substrates for Maintaining Locations and Spatial Relations in Working Memory
Mean difference Beta weights (Corresponding Quadrant – Opposite Quadrant) from visual cortex ROIs (V1–V3) for Location and Relation trials under low (left) and high (right) load conditions. For low load, the significant trial type × quadrant interaction suggests less quadrant-specific BOLD activity for Relation compared to Location trials, which replicates Experiment 1. Error bars represent within-subject standard error of the mean. ∗p < 0.001.
© Copyright Policy
Related In: Results  -  Collection

License
Show All Figures
getmorefigures.php?uid=PMC5121279&req=5

Figure 7: Mean difference Beta weights (Corresponding Quadrant – Opposite Quadrant) from visual cortex ROIs (V1–V3) for Location and Relation trials under low (left) and high (right) load conditions. For low load, the significant trial type × quadrant interaction suggests less quadrant-specific BOLD activity for Relation compared to Location trials, which replicates Experiment 1. Error bars represent within-subject standard error of the mean. ∗p < 0.001.
Mentions: To explore this 3-way interaction we tested two separate 2 (trial type: Location vs. Relation) × 2(quadrant: corresponding vs. opposite) repeated-measures ANOVAs, one for each load. For low load, a main effect of quadrant emerged, F(1,29) = 78.33, p < 0.001, with more activity in the corresponding compared to opposite quadrant. The main effect of trial type did not reach significance, F(1,29) = 2.56, p = 0.12. Importantly, the trial type × quadrant interaction was significant, F(1,29) = 50.12, p < 0.001, with the difference in the BOLD response between corresponding and opposite quadrants being larger for Location trials than Relation trials. Follow-up t-tests confirmed that BOLD activity in low load Location and Relation trials were significantly different in the corresponding quadrant, t(29) = 3.96, p < 0.001, but not for the opposite quadrant, t(29) = 1.31, p = 0.20. As Figure 7 illustrates, this low load result is a direct replication of Experiment 1’s ROI results. For high load, the main effect of quadrant approached significance, F(1,29) = 2.86, p = 0.10, with more activity in the corresponding compared to opposite quadrant. However, neither the main effect of trial type, F(1,29) = 0.07, p = 0.80, nor the trial type × quadrant interaction, F(1,29) = 0.16, p = 0.69, approached significance (Figure 7).

View Article: PubMed Central - PubMed

ABSTRACT

Previous work has demonstrated a distinction between maintenance of two types of spatial information in working memory (WM): spatial locations and spatial relations. While a body of work has investigated the neural mechanisms of sensory-based information like spatial locations, little is known about how spatial relations are maintained in WM. In two experiments, we used fMRI to investigate the involvement of early visual cortex in the maintenance of spatial relations in WM. In both experiments, we found less quadrant-specific BOLD activity in visual cortex when a single spatial relation, compared to a single spatial location, was held in WM. Also across both experiments, we found a consistent set of brain regions that were differentially activated during maintenance of locations vs. relations. Maintaining a location, compared to a relation, was associated with greater activity in typical spatial WM regions like posterior parietal cortex and prefrontal regions. Whereas maintaining a relation, compared to a location, was associated with greater activity in the parahippocampal gyrus and precuneus/retrosplenial cortex. Further, in Experiment 2 we manipulated WM load and included trials where participants had to maintain three spatial locations or relations. Under this high load condition, the regions sensitive to locations vs. relations were somewhat different than under low load. We also identified regions that were sensitive to load specifically for location or relation maintenance, as well as overlapping regions sensitive to load more generally. These results suggest that the neural substrates underlying WM maintenance of spatial locations and relations are distinct from one another and that the neural representations of these distinct types of spatial information change with load.

No MeSH data available.


Related in: MedlinePlus