| Objective Working memory(WM) is one of important cognitive functions. When WM load reaches the capacity limit, behavioral errors increase. Why WM function declines when WM load reaches the capacity limit, is still one of important questions that neuroscience research have focused on. From the view of functional connectivity among neural signals, the aim of this study is to investigate the change of the functional connectivity among multi-channel electroencephalographs(EEGs) when WM load reaches the capacity limit. EEGs were recorded from healthy subjects, while they performed a WM task with different loads. This may provide support for studying the brain network mechanism underlying the capacity limit in WM.Methods 1. Experimental data: 32-channel EEGs and behavioral data were recorded from healthy subjects(age ranging from 19 years to 29 years), while they performed a visual WM task with WM load 1-6. The data came from 16 subjects, 208 trials for each WM load. 2. Analysis of behavioral data: Reaction time(RT) and accuracy were calculated across all the subjects for load 1-6, respectively. Individual WM capacity limit was estimated according to behavioral results. 3. EEGs preprocessing: Baseline drift, 50 Hz noise, and the artifacts caused by vertical and horizontal electrooculography were excluded from the original EEGs. 4. Time-frequency and spatial distributions of power: Short-time Fourier transform was applied to analyze the critical channels and principal frequency band related to WM. EEG components in the principal frequency band were extracted. The topographies of power in the principal frequency band were calculated. 5. Directed transform function(DTF) of EEG θ components: Based on multi-variable Granger causal connectivity analysis, DTF of EEG θ components during WM was calculated for load 1-6 and for 16 subjects. 6. Functional connectivity strengths of the causal network: Given the causal network identified by the DTF matrix, the functional connectivity strengths of the whole brain network(DTFglobal), within and among brain regions(frontal, central and parietal regions) were calculated. Besides, the topographies of functional connectivity strengths were analyzed. 7. Statistical analysis: Statistical differences among six loads(208 trials from 16 subjects for each load) were evaluated by using one-way ANOVA analysis, followed by Student Newman Keuls(S-N-K) test.Results 1. Behavioral results for load 1-6 Mean RT from load 1 to 6 was 407.25 ± 14.33 ms, 491.83 ± 16.35 ms, 568.96 ± 16.10 ms, 635.76 ± 15.97 ms, 694.30 ± 20.88 ms, 764.88 ± 29.61 ms, respectively(p<0.001).Mean accuracy from load 1 to 6 was 98.06 ± 0.55%, 96.46 ± 0.93%, 96.18 ± 0.55%, 96.01 ± 0.96%, 87.43 ± 1.39%, 84.69 ± 1.19%, respectively(p<0.001). The averaged WM capacity limit over 16 subjects was 4.13 ± 0.15.2. Time-frequency and spatial distributions of EEGs for load 1-6 Theta(4-8 Hz) power was strong and sustained through the delay period, and focused in the frontal midline region. The theta power in the frontal midline region increased from load 1 to 4, and leveled off thereafter.3. Functional connectivity among EEGs in the θ band for load 1-6 3.1 Causal matricesFunctional connectivities in fronto-parietal and fronto-frontal regions strengthened with increasing load from 1 to 4, and weakened as load increased beyond 4.3.2 Functional connectivity strength Functional connectivity strength of the whole brain network(DTFglobal) from load 1 to 6 was 0.0136 ± 0.0007, 0.0143 ± 0.0005, 0.0148 ± 0.0004, 0.0154 ± 0.0004, 0.0144 ± 0.0006, 0.0135 ± 0.0005, respectively(p<0.001). Increasing load from 1 to 4 resulted in an increase in DTFglobal. By contrast, increasing load from 4 to 6 resulted in an decrease.Functional connectivity strength within the frontal region(DTFFF) from load 1 to 6 was 0.0182 ± 0.0021, 0.0197 ± 0.0018, 0.0213 ± 0.0018, 0.0276 ± 0.0029, 0.0236 ± 0.0022, 0.0236 ± 0.0023, respectively(p<0.001). The value of DTFFF increased from load 1 to 4, peaked at load 4, then decreased after load 4. However, functional connectivity strengths within central and parietal regions(DTFCC and DTFPP) presented no significant change with load increasing from 1 to 6(p>0.05).Functional connectivity strengths from the frontal to the central and parietal regions(DTFCF and DTFPF) were calculated. The value of DTFCF from load 1 to 6 was 0.0147 ± 0.0017, 0.0148 ± 0.0019, 0.0155 ± 0.0017, 0.0195 ± 0.0025, 0.0174 ± 0.0019, 0.0169 ± 0.0018(p<0.001). The value of DTFPF from load 1 to 6 was 0.0112 ± 0.0018, 0.0128 ± 0.0019, 0.0136 ± 0.0020, 0.0186 ± 0.0024, 0.0156 ± 0.0020, 0.0153 ± 0.0018(p<0.001). DTFCF and DTFPF shared the common tendency: increased from load 1 to 4, peaked at load 4, and then decreased.3.3 Spatial distributions of functional connectivity strengths The connectivity strength was strongest over the frontal midline region for each load. Meanwhile, a load-dependent change was observed in functional connectivity strengths of the frontal midline region.Conclusion 1. The WM capacity limit was 4 calculated from the behavioral results. The decline of the accuracy was not statistically significant until the load exceeded 4. 2. The principal frequency band related to WM was theta band. Theta power was focused in the frontal region, which indicated that the frontal region is important in WM. 3. The fronto-frontal and fronto-parietal connectivities, which were calculated among EEG theta components, were activated during WM. The estimations of functional connectivity strength showed the capacity-constrained responses: increased with increasing load, peaked at the capacity limit and declined as the load increased beyond the capacity. Therefore, when WM load reached the capacity limit, functional connectivity among neural signals declined, which resulted in increasing of behavioral errors. |
PDF Full Text Request