*3.1. Behavioral Performance*

Overall accuracy across all trials was high (Mean percent correct = 89.89%, Standard Error = 1.34%). Mean Hit proportion was 77.5%; mean False Alarm proportion was 6.07% (d' = 2.31; c = 0.40). Mean response time was 685 ms (SE = 35.64 ms). Age was positively correlated with accuracy (*r(*13) = 0.61, *p* < 0.05) but not with response times.

#### *3.2. ERP Data*

Figure 5 shows overall average vincentized ERP waveforms (thick lines) and standard errors (thin lines) recorded over twelve electrode locations (F3, Fz, F4, Cz, T7, T8, P7, Pz, P8, O1, Oz, O2) from the time interval of 200 ms prior to onset of the image to 1000 ms after presentation of the image for both target (blue lines) and distractor (red lines) conditions.

**Figure 5.** Average vincentized event-related potential (ERP) waveforms for targets (thick blue line) and distractors (thick red line) recorded over twelve electrode locations (F3, Fz, F4, Cz, T7, T8, P7, Pz, P8, O1, Oz, O2) for the epoch ranging from 200 ms prior to onset of each stimulus image to 1000 ms after the stimulus image. Standard errors are represented by thin lines. The microvolt value corresponding to the significance threshold is shown as a green line in scale against the *y*-axis in μV of observed ERP amplitude range at the bottom left legend.

For all electrodes, within most bin intervals, larger overall ERP peak amplitudes were observed for targets than for distractors. Significant di fferences in grand peak amplitudes were found (F (1,13068) = 160.124, *p* < 0.01; MSE = 0.468; η2 = 0.13), with larger grand amplitudes for targets than distractors. In addition, there was also a three-way interaction between interval bins, electrodes and condition (F (12,13068) = 154.628, *p* < 0.01; MSE = 20.367; η2 = 0.13).

Given the three-way interaction, to compare di fferences between target and distractor conditions across electrode locations and time intervals we then performed Focused ANOVA *t*-contrasts (see Section 2.1.5). The t-value corresponding to the critical p-value for determining significance threshold after using a Bonferroni correction was tcrit (12) = 5.69 ((MSE = 0.132); *p* = 0.0001 (two-tailed); η2 = 0.71). Accordingly, the minimum significant amplitude di fference between standard and distractor peak amplitudes was computed to be 0.80 μV. The significance threshold is shown in scale against the *y*-axis μV legend in Figure 5.

We also performed traditional peak analysis, which is presented in the Supplementary Materials. There were no substantial discrepancies between the two analyses.

Table 2 reports the di fferences of average peak ERP amplitudes in target and distractor conditions (peak amplitude target − peak amplitude distractor) within interval bins of 100 ms for all electrodes. This analysis focused only on bins capturing processes before motor responses, that is, up to the approximate time of the occurrence of manual response for most children (bins including data up to 700 ms) to exclude e ffects that might be attributed to motor responses (i.e., when participants responded to targets).


**Table 2.** Di fferences of average peak ERP amplitudes and corresponding latencies in target and distractor conditions (absolute value of (peak amplitude target – peak amplitude distractor)) within interval bins of 100 ms of epoch range (0 to 700 ms) for all electrodes.

Note: Superscripts indicate significant differences in peak amplitude between target and distractor condition: "T"indicates larger amplitude for target; "D" indicates larger amplitude for distractor.

Inspection of Table 2 confirms that the e ffects in most of the significant pairwise comparisons (46 out of 84) involved higher amplitudes for targets as compared to distractors. In contrast, larger amplitudes for distractors over targets occurred only for much fewer comparisons (5 out of 84). The di fference between the proportions of significant target enhancement (55%) vs. distractor (6%) is substantial (χ2(1) = 45.05; *p* < 0.0001).

#### *3.3. ERP Activity Paths*

The results of the ERP activity paths analysis are shown in Figure 6. Neural paths of maximum mean difference in neural activity for targets are shown in blue and neural paths of maximum mean difference in neural activity for distractors are shown in red. The paths were constructed from the results of Table 2.

**Figure 6.** Neural paths of maximum ERP activity over seven interval bins of 100 ms time intervals. Top Panel: Target. Bottom Panel: Distractor.

With reference to the onset of target, ERP paths based on maximum mean differences in peak amplitudes first occurred at the parietal central and right electrodes and moved anteriorly to the left temporal, central and right frontal electrodes. Activation subsequently occurred at the frontal and mid parietal electrodes and finally moved to central parietal and occipital electrodes. For the distractor, ERP pathways based on maximum differences in peak amplitudes were only detected at the left parietal and right temporal electrodes. Neural activation finally moved to the mid-frontal electrode.

#### *3.4. Comparison of ERP Activity and Localization: Preschool Data vs. Adult Simulation*

Figures 7 and 8 report the comparison between the localization of dipoles for distractor and target trials. The figures also show the ERP topographical maps of the actual children's data contrasted with the adult simulation data.

**Figure 7.** *Cont*.

**Figure 7.** Comparison between dipole source analysis and topographical mappings of actual preschool children ERP data and simulated ERP data by using an ACT–R modeling architecture in the target condition. Timings are set by the ACT–R model module production schedule simulation (given in Figure 3). Coefficient of determinations show fit results for actual and simulated topographic maps comparisons. "R" represents right, and "L" represents left (Please note that lateral side of brain is showed opposite to perspective of the observer in MRI scans). Z-range represents range of normalized Talairach coordinates and is a measure of margin of error expressed as z score; comparisons between the z-ranges in actual and simulated data showed no significant differences.

**Figure 8.** Comparison between dipole source analysis and topographical mappings of actual preschool children ERP data and simulated ERP data by using an ACT–R modeling architecture in the distractor condition. Timings are set by the ACT–R model module production schedule simulation (given in Figure 3). Coefficient of determinations show fit results for actual and simulated topographic maps comparisons. "R" represents right, and "L" represents left (Please note that lateral side of brain is showed opposite to perspective of the observer in MRI scans). *Z*-range represents range of normalized Talairach coordinates and is a measure of margin of error expressed as *z* score; comparisons between the z-ranges in actual and simulated data showed no significant differences.

The *z*-scores of the variation in matched anatomical localizations did not show significant differences between the children's and the adult simulated data. Similarly, there was a strong fit between the topographic maps of the actual observed children's data and the simulated adult data. For both target and distractor condition, in the regressed data fitting model between actual and simulated data, the coefficients of determination ranged similarly from *R*<sup>2</sup> = 0.70 (F (1,11) = 23.65; *p* = 0.0005) to *R*<sup>2</sup> = 0.95 (F (1,11) = 148.30; *p* < 0.0001), indicating very strong correspondence.

Overall, the results show that the ACT–R model has a very good fit with the actual ERP data, however, two discrepancies are of note. The MRI Talairach coordinates did not match in one instance out of six comparisons concerning the target data, although the spatial coordinate variation was moderate and within satisfactory margins. The match was more imprecise for the distractor data where in both comparisons the Talairach coordinates referred to very proximal but still distinctly different anatomical structures. In addition, differences between ERP time latencies in the actual data and the one derived by simulation might have worsened the fit statistics, since the time of ERP occurrence predicted by ACT–R might actually have not led to sample the most optimal actual data to be fed to the simulation algorithm.
