Higher cognitive reserve is associated with better neural efficiency in the cognitive performance of young adults. An event-related potential study

To examine the effects of cognitive reserve (CR) and working memory (WM) load on the cognitive performance of young adults, we performed two event-related potential (ERP) experiments. The first experiment aims to show how high CR influences young adult performance as a function of two levels of working memory load (high vs. low) during a Sternberg task. For both positive and negative probes, participants with high and low CR showed larger P300 amplitudes to low WM loads than to high WM loads. Both CR groups showed a longer P300 latency to high WM loads than to low WM loads, but this difference was greater for the low CR group than for the high CR group. The high CR group displayed larger P300 amplitudes for every experimental condition compared to the low CR group. The second experiment analyzed grammatical gender agreement in sentence processing when CR and WM load were manipulated. Sentences varied according to the gender agreement of the noun and adjective, where the gender of the adjective either agreed or disagreed with that of the noun (agreement), and with regard to the number of words between the noun and the adjective in the sentence (WM load). Participants with high CR showed greater modulation of left anterior negativity (LAN) and P600a effects as WM increased than that observed in participants with low CR. The findings together suggest that higher levels of cognitive reserve improve neural efficiency, which may result in better working memory performance and sentence processing.

7 138 P600a, integration of arguments and P600b, mapping of meaning; [43,44]). This 139 atypical processing did not affect the accuracy of their responses, although the 140 response times were longer for older adults than for young people. Older adults 141 also showed a greater modulation of the amplitude in relation to the WM load than 142 young people. Small amplitudes of the P600a and P600b components were 143 observed in the high load condition of the MT, a fact that was not observed in the 144 younger participants. These findings may be consistent with the idea of neural 145 efficiency [9,45] or the idea that older adults may show a less efficient neural 146 response accompanied by compensation.

147
The objective of the present ERP study was to analyze the effect of CR on 148 the performance of young adults. In the same sense as in the studies with older 149 adults, we intend to see the CR effect by dividing the sample of participants into 150 high and low CR groups. We hypothesize that young people with high CR will show 151 better behavioral performance and that their brain electrical response pattern will 152 reflect earlier processing and will be less vulnerable to the complexity of the task 153 compared to the participants with low CR. We think that this beneficial effect of CR 154 would be observed in young people facing two levels of WM load while they are 155 performing tasks involving fluid intelligence (Sternberg WM task) and crystalized 156 intelligence (Reading sentence task).
205   285 The bandwidth of the amplifiers was set to 0.1-100 Hz, and the signal was digitized 286 at a 500 Hz sampling rate. Impedances were kept below 5 kΩ. Two electrodes 287 placed on the external canthus and superciliary arch of the left eye were used to 288 record the electrooculogram (EOG).

289
The EEG recordings were rereferenced offline using the average of the    Fig 3b, the distribution of the differences between WM loads seemed to be 424 wider for the low CR group than for the high CR group when negative probes were 425 used. Furthermore, these topographic differences in the P300 amplitude seemed to 426 have a wider distribution in the negative than in the positive probe conditions for 427 the low CR group than for the high CR group.

428
In another series of analyses, four independent samples t-tests were   492 previous studies [6,9]. The low CR group displayed significantly longer P300 peak 493 latencies to high WM than to low WM loads, and this difference was more marked 494 in this group than in the high CR group for the negative probes.

495
A higher cognitive demand was also observed globally across experimental 496 conditions, even though there were no robust behavioral differences between the 497 groups. The participants with low CR showed a significantly smaller P300 498 amplitude than those participants with high CR for negative probes at high and low 499 WM loads and for positive probes at low WM loads. These findings are supported 500 by previous P300 ERP studies [33,34]  516 the processing of gender agreement would lead to a higher processing cost when 517 there is a higher WM load than when there is not. The high CR group would show 518 better behavioral results than the low CR group when processing sentences with a 519 low WM load. We also think that young people with low CR will base their response 520 on a strategy where the analysis of semantic information is preferred to resolve the 521 gender agreement, because this would result in a smaller amplitude of the LAN 522 effect than that of subjects with high CR. Participants with low CR will also show a 523 larger amplitude of the P600a effect because they will be forced to compensate for 524 the poor processing of the previous stage. In the final stage of the agreement 525 processing, during the P600b window, where the generalized mapping of the 526 sentence is made and the semantic and syntactic information converges, there will 527 probably be no differences between the two groups. The task consisted of 160 experimental sentences with seven words each.
535 Adjective-noun grammatical gender agreement (agree and disagree) and WM load 536 (high and low) were manipulated. Statistical analysis of the behavioral data. 585 The analysis of the behavioral data was performed similarly to that in 586 Experiment 1. Two three-way ANOVAs were performed on the transformed 587 percentage of correct answers and the response times. WM load (high and low) 588 and Gender agreement (agree and disagree) were used as within-subjects factors, 589 and CR level (high and low) was used as a between-subjects factor. HSD post hoc 590 tests were performed for multiple comparisons.

591
A multivariate nonparametric permutation test with 10,000 permutations was 592 used to analyze the ERP amplitude data. The FDR method and Benjamini-593 Hochberg correction were used to correct the global p-alpha for multiple 594 comparisons between experimental conditions.

595
We proceeded first with comparing the agree versus disagree conditions for