The authors recognized that little attention had been given to the impact of language context for monolinguals. As a result, they did research to present data detailing the effect of ambient linguistic diversity on monolinguals’ ability to acquire a different language. The authors aimed at using recent research to challenge the traditional assumptions that language processing is uniformly homogeneous and that differences in the performance of native language always result from cognitive resource constraints. They believe that language processing may reflect fluency differences and may not be as stable as it is assumed to be since it changes for bilinguals who are proficient. The authors tend to think that both past and new experiences influence changes in the native language while learning a new one. Before the study, they predicted that monolinguals in contexts with linguistic diversity have a much better language experience apart from English than monolinguals in homogeneous unilingual contexts and hence are less monolingual like.
The authors made a comparison of monolinguals in two different contexts. The first is Central Pennsylvania, which has a relatively homogeneous unilingual setting where the predominant language is English, while the second is Southern California with linguistic diversity where many languages are spoken. They sampled 34 monolingual participants, where 21 of them were females. Out of the 34, 18 were from the Pennsylvania community and the rest from the University of California-Riverside. All participants were native speakers of the English language within the ages of 18-35 and had to be free from epilepsy, color blind, speech disorders, concussion, and have normal vision. A questionnaire on language history was given to the subjects requiring them to report any languages they can speak more than just the greetings and the time proportion they were exposed to each of those languages they reported.
The key finding from the collected data was that California monolinguals were sensitive to non-native phonological contrast in the process of learning the language while Pennsylvania monolinguals did not. This finding suggests critical differences in the learning process in the two locations. Results from learning task behavior showed that the two groups had the ability to learn mappings of the finish words they had studied, although none of the groups could generalize the harmony pattern of vowels to novel words. However, results of the brain activity pattern revealed that the two groups had notable effects of ERP for words studied with Pennsylvania having a broader distribution effect while monolinguals in California were more restricted to the posterior sites. Both behavioral learning task and brain activity results reveal that the two locations were able to differentiate violations in behavior and studied words, with monolinguals from Pennsylvania performing higher.
The results from behavioral changes might be beneficial in assuming that monolinguals were in no way sensitive to differentiating between the vowel and novel harmony violations. This was the case with monolinguals in Pennsylvania, where there were no significant waveform differences, as shown by the ERP results. The differences in ERP results between the studied monolingual groups reveals that linguistic diversity and factors co-involved in it may positively impact on the learning of new languages. The regional dialects might be affected by living among linguistic diversity as well as increasing the interaction numbers with accented speakers. These effects may prepare people for a particular new language learning aspects. In addition, the role played by immersion in learning language can also be explained from the reported results and the period that the monolinguals have lived where they are.
Microstructural plasticity in the bilingual brain
The inspiration to carry out this experiment was the lack of a quantitative evaluation of the vivo microstructural properties from past studies. Most of them did neuroimaging, which only dealt with qualitative analysis as they were exclusively derived from uncalibrated TI-weighted images sensitive to tissue microstructure and organization multiple features. Therefore, to achieve the quantitative evaluation of vivo microstructural properties, the authors utilized the qMRI technique to facilitate computation for macromolecular tissue volume (MTV) of the brain as well as quantitative TI analysis as it contributes linearly to myelin and iron concentrations. The authors also needed to identify executive and bilingual processing relationships. They, therefore, predicted a valid approximation of the myelin volume provided by MTV because the proteins and membranes account for the most substantial part of brain macromolecules.
The experiments involved fifty bilingual participants of native speakers from China who learned the English language as a second language. Among them, twenty-five were bilinguals who had learned the English language between 0-6 years while the other twenty-five were late bilinguals and learned English after nine years. All of them were college students, left-handed and with physically healthy, normal, and neurologically typical children without drug abuse. The participants were given a questionnaire with proficiency and qualitative language experience to complete to allow for their language experience assessment. To evaluate ability of the subjects, the researchers used the reading and listening sub-sections of the IELTS. Administration of cognitive tests was done individually, while the nonverbal intelligence of participants was measured with the Standard Progressive Matrices of the Raven using the standard version of China. Other tests were component search, subsets of the WAIS and rapid automatized naming of numbers, numeric working memory test, phoneme counting, deletion, and Stroop tasks. A 3 T Discovery MR750 system was used to perform the MRI experiments using a head coil of 8 channels. For the TI and MTY quantitative values, spoiled gradient echo (SPGE) images were used to by utilizing various angles. Analysis of the fMRI was performed using SPM12 software package in MATLAB. The mrQ software package was used to process images of SEIR and SPGE, which generated quantitative TI map and macromolecular tissue volume (MTV) for each of the participants. Statistical analysis use of IBM SPSS and ANOVA and evaluated the TI and MTV group difference per ROL.
Results revealed three regions activated strongly in the functional task for the whole of the analysis of the brain using fMRI. Notable variations in the microstructure regarding the AoA were observed in the left middle fusiform and left the anterior region. MTV in the left anterior showed the trend of negative correlation with AoA, while the positive trend was demonstrated with the TI. Researchers found no notable qMRI measure difference between late and early bilinguals in that particular region. This study succeeded in proving that the identification of microstructural plasticity in young bilinguals was possible using the qMRI technique. It confirmed that proficiency and AoA are able to play unique roles that contribute to the bilingual brain microstructure. Further, for the first time, the research demonstrates that acquiring the second language early is attributed to enhanced development of the microstructure within the brain of the bilingual and which can provide significant evidence for the high executive functions in young adult bilinguals as compared to the monolinguals.
Working memory training involves learning new skills
The field of intensive training for intellectual capacity expansion has been advancing progressively towards the understanding of its effects on not only structure but also the functioning of the neural networks. However, detailed the cognitive changes that occur have not been accounted for fully, which made the authors of this article to explore a new framework and describe what they could be including how to constrain and enable novel situations transfer. To do this, they explored the Working memory, which is among the cognitive training areas. The research characterizes the task elements that stimulate transfer within the working memory. They first predicted that this transfer takes place significantly when one acquires cognitive skills that are new and complex from training and which are easy to apply to activities no trained.
For study 1, the researchers carried out a meta-analysis of randomized controlled trials (RCTs) to experiment if after the transfer that comes after working memory comes up when certain task features are shared by transfer tasks as well as training activities. Various studies were carried out to achieve the objectives of the study. First, for visuospatial material, stimulus input modality, and verbal recall modality, the extent for transfer mediation by usual components of both the trained and untrained tasks was assessed. The primary research method and source of evidence for their argument was the literature search. Collected searches were collated, and duplicates taken away to review the data’s abstract meaning. Single working memory tasks were paired with every untrained one during the training program, after which the two tasks were coded as per the response modality and paradigm, complex and backward, stimulus type, stimulus domain, and Type of the stimulus. For particular tasks, it necessitated the coding of several features within one category. The meta-analytic procedure, on the other hand, involved the recording of data for each task transfer. Version 3.3 of the Meta-analysis comprehensive program was used for conducting the data analysis. The analysis plan included conducting analysis for the summed comparisons in all categories, unmatched features, and matched conditions for every element. Moderator analyses tested any significant effect on the effect size multitude. Crucial results of the moderator analyses were the R^2 and p-value. For study 2, the boundary conditions to task transfer within the working memory were comprehensively analyzed to transfer within the working memory. Participants used for this study had low working memory scores.
Across all the one hundred and thirteen task pairs used for trained and untrained working memory, the mean effect size was found to be 0.42, SD=0.54. The analysis mainly evaluated the statistical importance of the effects sizes as per the matched and unmatched features with the condition of feature matching acting as the transfer moderator. The results produced revealed a large effect size for matched pairs of tasks that is, d=0.994 and for unmatched pairs, it was smaller, however significant (d=0.357). For letters, the effect sizes were significant, although small, for matched stimuli in trained and untrained tasks. Based on words and non-words, the effect size was moderate and quite substantial for matched pairs, although, for unmatched pairs, it was non-significant. Objects also showed comparable and reasonable magnitudes for both matched and unmatched stimuli.
Through meta-analysis, the features associated with working memory transfers were evaluated in RCTs of adaptive working memory training using current conditions of control training. The transfer strength for 113 pairs after the 24 studies ranged between small to moderate for both trained and untrained tasks. The high transfer was observed in cases where tasks used similar paradigm, i.e., either complex span, backward span, or serial recall paradigms. The research findings broadly support cognitive routine network predictions. This basis showed that quite large transfer following working memory training occurred in cases when trained and untrained activities imposed similar and unfamiliar demands of tasks not supported by the already existing working subsystems of the working memory. This research helps in new cognitive routines construction and transfer only for those tasks that are exposed to have the same routines.
Balancing Type 1 error and power in linear mixed models
The authors of this article carried out a simulation to show that the additional cost in the potential overfitting of Linear mixed models (LMM). The cost comes about in that protection against Type I errors means a considerable increase in the Type II error rate. This means a loss of the power of statistics to detect the importance of fixed effects. Additionally, the researchers aimed at showing that selecting a parsimonious linear mixed model is a quite promising option to the maximal model in balancing the rate and power of Type I error. The experiment involved demonstrating the problem of maximal models approach introducing costs i.e., statistical power through simulating, estimation of the Grand Mean, and single fixed effects.
The fixed effects used were and a residual error, where c represents the condition, s subject, and I item. The generating process with y as the dependent variable and subscripts mimicked an experiment where twenty items were presented to fifty subjects under conditions. That meant collecting times for response with 25 milliseconds as experimental effects and 2000millisecond grand mean. They were mainly concerned with the fixed effects significance and the way the complex effects of the random structure of the linear mixed model affects the estimate concerning the Type I error and power. The generating process involved subjects where each had a random intercept and slope for the conditions. Each of the items was also given a slope and an intercept. Both standard deviations for subject and items were varied on an interval of 1-120, which found a correlation of the item specific random effects.
Moreover, the model residuals were independent and also identical distribution. A sample was drawn from the process of generation, and the researchers fitted five linear mixed models to the data producing a difference in the random structure effects part. Next was the estimation of the models under the null hypothesis alternative of zero fixed effect which had a value of 25. The maximal model was the first one and included estimating the correlation parameters fixed in the generating process. It was found that the model exactly matched the generating process apart from cases where random slope variances were set to zero.
The determination of error rates simulation involved two runs. The first run involved estimation of the model Type I error. Each iteration simulation consisted of drawing a sample from the generating process with set to zero. After getting the estimates of the type one error and power, a comparison was made for the maximal model specification and parsimonious model performance. In the second run, power was determined by drawing samples from the generating process with set to 25. Simulation scenarios were two, the worst-case scenario and the small random slopes scenario. The worst-case scenario showed evident deficits of the maximal model, while in the second scenario, the maximal model showed the worst performance than the parsimonious model even when with the generating process matching the maximal model. Yields from the simulations were in agreement with theory in showing that the maximal linear model guarded against increasing the Type I error by ignoring the significant variance component. The research is vital in that it showed that determining a parsimonious model by selecting a standard model can be a logical choice in finding the model ground between power and Type I error.