Native Language Experience Shapes Pre-attentive Foreign Tone Processing And Guides Rapid Memory Trace Build-up: An ERP Study Part 5

2.5 | Statistical analysis

2.5.1 | Behavioral data

To test for possible effects of L1–L2 similarity in the behavioral results, we analyzed the behavioral responses to tone mismatch trials focusing on differences between words with contour tones and words with level tones. 

In recent years, researchers have discovered a strong relationship between tone of voice and memory. In particular, contour tones (or "jump tones") tend to make it easier for us to remember a word or phrase. This is because contoured tones allow the listener to better capture key information and emotional content, making it easier to store in long-term memory.

Some studies have shown that people are more likely to remember words with clearly contoured tones, which can make information more clearly processed in the brain than flat tones. For example, when we hear an emphasized syllable in a word or phrase, we notice this information and more easily program it into our memory.

In addition, contour tones can convey emotional information, which is also important for memory. We tend to remember events and information more easily with emotions and emotional experiences. Therefore, contoured tones help us store these emotional memories in our brains.

Overall, the effects of contoured tones do help us remember better. In daily life and work, we can pay attention to the tone of our voice when speaking to others. Especially in various speeches, meetings, advertisements, and other occasions, we should pay more attention to using a contour tone to improve the memory effect. Through training, we can improve our sensitivity to speech and help us better remember important information, improving work efficiency and quality of life. It can be seen that we need to improve memory, and Cistanche deserticola can significantly improve memory because Cistanche deserticola is a traditional Chinese medicinal material that has many unique effects, one of which is to improve memory. The efficacy of minced meat comes from its various active ingredients, including acid, polysaccharides, flavonoids, etc. These ingredients can promote brain health in various ways.

Click know ways to improve brain function

To this effect, we separately submitted mean data for the behavioral variables "Response Accuracy" and "Response Times" to two mixed analyses of variance (ANOVA) with the experimental factor "Tone Type" (contour tones vs level tones), the temporal factor "Day" (day 1 vs day 2) and the between-subject factors "Learner Group" (tonal L1s vs nontonal L1s) and "Target Tone Group" (high/fall vs low/rise). 

Response Times were normalized through log transformations, and for accuracy, d′ scores were computed. As two participants were excluded from the study, the Target Tone subgroups were of different sizes (11 or 12). 

We used mean imputation wherever same-sized target groups were necessary for the statistical analysis. All behavioral analyses were carried out in SPSS 26 (International Business Machines [IBM] Corp., Armonk, NY, United States).

2.5.2 | ERPs

For the ERP data, we selected two-time windows (50–70 ms and 400–600 ms) where we expected to observe influences of L1–L2 similarity on tonal or tone features on word processing, based on previous literature (Gosselke Berthelsen et al., 2018, 2020). 

Using these pre-defined time windows, we conducted cluster-based permutation tests for the factor "Tone Type." We submitted mean ERP amplitudes of block 1 (on day 1) from both participant groups, together and separately, for the selected time windows and conditions (i.e., words with contour tones compared to words with level tones) to a permutation analysis using the nonparametric cluster-based permutation approach implemented in Fieldtrip toolbox for Matlab (Maris & Oostenveld, 2007). 

We ran 1000 random permutations of the data with the Monte–Carlo method to account for large data sets and considered clusters of three or more electrodes with a p-value of <.05 significant. We additionally tested for interactions with "Learning" (target word vs control word) in the permutation analysis to see whether target words differed from control words. 

The interaction was particularly important for the word recognition effect at 50 ms, where tonal learners have earlier been seen to automatically dissociate target words from control words(Gosselke Berthelsen et al., 2020). If significant clusters emerged in any analyses, we carried out mixed ANOVAs to test for possible interactions with temporal and between-subject factors. 

Thus, we computed one mean ERP amplitude across the analyzed time windows and all cluster electrodes. This was done for each participant in each block and day. The thus obtained mean amplitudes were then submitted to a mixed ANOVA with the experimental within-subject factors "Tone Type" and "Learning," the temporal factors "Day" and "Block," as well as the between-subject factors "Learner Group" (if applicable) and "Tone Target Group." 

The ANOVAs were carried out in SPSS 26 (IBM). Greenhouse-Geisser correction was used where necessary. For multiple pairwise comparisons, False Discovery Rate (FDR) corrections (Benjamini & Hochberg, 1995) were applied.

2.5.3 | Correlations

To test whether ERP effects were affected by individual learning behavior, we carried out two two-tailed Pearson correlations with the variables "Amplitude Change for Lexicality Effect" and "Amplitude Change for Anterior Negativity," on the one hand, and "Response Time Change" and "Response Accuracy Change," on the other. 

The change investigated here was defined as the difference in behavioral or ERP responses between the first and the second half of day 1, where the bulk of learning took place (cf. Gosselke Berthelsen et al.,  2020). Correlation analyses were carried out in SPSS (IBM). FDR corrections were applied to the correlations' p-values.

3 | RESULTS

3.1 | Behavioral results

For Response Accuracy, there was no significant main effect of the Tone Type and no significant interactions with the factors of Tone Type, Learner Group, or Tone Target Group. 

A main effect of the temporal factor Day, F(1,44) = 44.18, p < .001, 휂2 p = 0.501, showed evidence of learning regardless of which sets of tones the learners acquired or which tone types were tested. Thus, Response Accuracy increased significantly from day 1 (M = 60.5%, SD = 25.4, range = 5.6 – 97.1) to day 2 (M = 68.8%, SD = 29.6, range = 2.9 – 100). 

For the descriptive statistics, we use percentages for the sake of simplicity and comparability across results and with other studies, whereas the statistical analysis was carried out on d′ data. For a graph illustrating the accuracy results and the change over time, please refer to Figure 5.

For Response Times, a significant interaction of Tone Type and Tone Target Group, F(1,44) = 4.63, p = .037, 휂2 p = 0.10, broke down into a main effect of Tone Type in the low/rise group, F(1,22) = 6.69, p = .015, 휂2 p = 0.23. 

Mismatches with the pictorial referent based on low tones (M = 1515 ms, SD = 986, range = 219 – 4279) were significantly faster detected than errors based on rising tones (M = 1722 ms, SD = 1093, range = 303 – 4241). There was no main effect of the Tone Type and no significant interactions involving the factors of Tone Type, Learner Group, or Tone Target Group. 

The main effect of the temporal factor Day, F(1,44) = 35.53, p < .001, 휂2 p = 0.45, showed evidence of learning regardless of which tone types the learners were taught or tested on. To this end, there was a significant improvement in Response Times from day 1 (M = 2056 ms, SD = 329, range = 488 – 4279) to day 2 (M = 1427 ms, SD = 153, range = 199 – 4241). The analysis was carried out on log-transformed data; the actual raw Response Times are also reported for data description.

3.2 | ERP results

3.2.1 | 50–70 ms

For the early time window, an interaction between Tone Type and Learning, i.e., comparing level and contour tones in target and control words, produced a significant central electrode cluster (FC2, FC4, C1, Cz, C2, C4, CP1, Cpz, CP2), p = .026, d = 0.87, in the tonal L1 group. 

See ERPs and topographies for the interaction in Figure 6. No comparable cluster was identified in the nontonal L1 group or for all participants, collectively. 

The permutation analysis did not produce any significant clusters for differences between level and contour tones without interaction with Learning (neither for all participants collectively nor for the participant groups separately). For a showing of the subtraction amplitude distribution within and between groups, please see Figure 7.

Further investigating the significant cluster of the tonal L1 group in a mixed ANOVA, a Tone Type * Learning * Target Tone Group interaction suggested that the observed effect for contour tones (i.e., controls were more negative than targets) was significant in the high/fall group only. 

Secondly, an interaction with time indicated that only amplitudes of control words changed over time, turning less negative. For detailed results, see Table 2.

3.2.2 | 400–600 ms

For the second time window, permutation analysis produced two significant clusters for the comparison of words with contour tones and words with level tones for the first 20 minutes of the first session in all participants. 

There was a significant frontocentral cluster(AF3, AF4, AF8, F5, F3, F1, Fz, F2, F4, FC5, FC3, FC1, FC2, FC4, FC6, C3, C1, C2, C4), p < .001, d = 0.38, as well as a significant posterior cluster (FT7, FT8, TP7, CP6, TP8, P7, P5, P8, PO7, POz, PO8, Oz), p < .001, d = 0.33. See Figure 8. 

A mixed ANOVA of the mean ERP amplitudes of the frontocentral cluster for all participants in 20-minute blocks yielded several main effects and interactions for Tone Type, Learning, and temporal factors Block and Day, see Table 3 for details. 

With regards to Tone Type, level tones were more negative than contour tones.
This difference was stronger in the tonal L1 group than in the nontonal L1 group. 

For Learning, we found that target words elicited larger negativities than control words (see Figure 8). The difference was again greater in the tonal L1 group and was also stronger on day 2 than on day 1. Finally, a general decrease of the negativity was observed over time.

For the posterior cluster, the effects were virtually indistinguishable from those of the frontocentral cluster but reversed in polarity: all of the main effects were near-identical to those above, as were the crucial interaction clusters. 

Only two unique interactions emerged in the posterior cluster. We, therefore, chose to treat the positivity as a dipole effect and, for the sake of brevity, present the observed significant effects and interactions for the posterior positive cluster as Supporting Information S1 instead of in the main text.

3.2.3 | Correlation results

For the early effect, no significant correlations were found between changes in behavioral and neurophysiological data (p > .8). Further analysis revealed no significant correlations for the tonal learner's H/F subgroup for this effect either (p > .2). 

A significant correlation was observed between Amplitude Change for Anterior Negativity and Response Accuracy Change, r = −.353, p = .040, such that the larger the improvement in accuracy on day 1 was, the smaller the difference amplitude for the anterior negativity became. There was no significant correlation with Response Time Change (p > .7).

4 | DISCUSSION

4.1 | Word recognition component: Transfer effects

There was a clear effect of native language experience and familiarity in the pre-attentive lexicality gating component at ~50 ms. The facilitation effect at this latency was only found at the highest degree of L1–L2 similarity. Our tonal learners did not show indications of facilitated word acquisition for all tonal target words or all target words with contour tones but instead only for target words with a falling tone. 

This became apparent in a reduced negativity which we assume reflects a successful, rapid word trace formation for target words with falling tones such that they became processed real-word-like already within the first 20 minutes of acquisition. A trend toward a similar amplitude decrease has previously been seen for real words in Mandarin speakers (Yue et al., 2014). 

After four minutes of word and legal pseudoword repetition, neural activity to real words appeared to become reduced. Note, however, that Yue et al. did not use an analysis that could cancel out frequency effects caused by the comparison of frequent standard and infrequent deviant stimuli (cf. Shtyrov & Lenzen, 2017) and, therefore, did not detect the amplitude change for real word deviants in their statistical analysis. 

Also, their time window was longer than ours. We mention the trend here because Yue et al. (2014) is the only study to look at this early component in the context of tonal words. Notably, for words without tones, the same effect was previously reported by Kimppa et al. (2015). 

By these previous studies, we interpret the decreased effect size for narrowly L1-facilitated target words as evidence that the words were acquired and processed like real words exceptionally quickly. 

Pseudowords, even those with the same pitch pattern, and nonfacilitated target words, on the other hand, could not be acquired equally rapidly and therefore evoked an increased negativity. 

This negativity likely signals an ongoing, incomplete memory trace formation process for untaught and nonfacilitated words. Consistent with the idea of an ongoing word trace formation process, the negativity for both target words and control words decreased slightly throughout the learning sessions (target words: MB1 = −1.02 μV ± SD = 0.8 μV, MB6 = −0.99 ± 0.8 μV; control words: MB1 = −1.31 ± 1.0 μV, MB6 = −0.90 ± 0.8 μV). 

This decrease was significant only for the control words where the amplitude was highest initially. Interestingly, the decrease in amplitude proceeded in a step-wise pattern for both word types such that the amplitude increased again after the breaks between blocks 2 and 3 and blocks 4 and 5 (cf. e.g., target words: MB1 = −1.02 μV ± SD = 0.8 μV, MB2 = −0.99 ± 0.9 μV, MB3 = −1.12±0.9 μV, MB4 = −0.88 ± 0.9 μV, MB5 = − 0.94± 0.8 μV, MB6 = −0.99 ± 0.8 μV). 

Together with the fact that we found no effect of the learning session for the response amplitude at this latency, this suggests that word traces were only formed temporarily for words with nonnative phonology (i.e., tones). 

Thus, only words with native phonology (and a familiar function) had a consistently reduced amplitude (M = 0.78 μV), which suggests that the rapid word trace formation process is dependent on the native neural phonology network and that word traces for L1-like novel words were formed almost instantly and permanently.

For more information:1950477648nn@gmail.com