Is Early Bilingual Experience Associated with Greater Fluid Intelligence in Adults?
Round 1
Reviewer 1 Report
The authors present a manuscript in which they compare adult early bilinguals (defined as acquiring an L2 ≤ 6 years of age) vs late bilinguals (i.e. acquired L2 > 6 years) vs monolinguals.
The presentation is clear. I enjoyed reading the manuscript. The authors have done a nice job of explaining the gap in knowledge that this study addresses. Below I list some comments and queries for the authors to address as they revise their manuscript.
For the bilingual groups, what were the language pairs that were spoken? Would be interesting to know that and whether it differed across the early vs late bilingual groups. Typological similarity has been identified as a variable of interest that requires exploration, and it would be interesting to know if there is anything going on here. See: Antoniou, M., & Wright, S. (2017). Uncovering the mechanisms responsible for why language learning
may promote healthy cognitive ageing. Frontiers in Psychology, 8, 2217.
Parent education is important, but this is an incomplete picture of SES. Parent occupation is also needed. I like that the authors describe it as a “proxy for SES” in the Discussion - I think that wording should be used throughout. I do not imagine that the results would turn out very differently if the authors had obtained parents’ occupation data, but strictly speaking, what they have here is not a pure measure of SES; it is parent education. They are correlated, but not the same.
“868 8-12-year-old children” is a bit clumsy. Would be good to rephrase this 8-to-12 somehow to make the reader’s task easier.
The Results section is incredibly brief. Four of the eleven lines of text are about removing outliers (which is fine), but that means that the entire statistical analysis is reported in only seven lines and a figure. When I arrived at the Discussion, I did a double-take because I thought I had skipped a page. Is there really nothing else worth reporting? Seems a bit weak to me, especially considering…
The planned contrasts confirm that the bilinguals > monolinguals. However, the early vs late comparison wouldn’t hold up once alpha is corrected for multiple comparisons (given that the planned contrasts are not orthogonal), but I am not opposed to the authors describing the early > late pattern in the data as long as it isn’t overstated/overinterpreted.
The way that early vs late bilingualism is defined strikes me as rather odd. Early bilinguals acquired their L2 at age 6 or under, which is fine. But, defining someone who acquired their L2 at age 7 as “late” seems wrong to me. It would be helpful to see some stats on the age of L2 acquisition (means, SDs, min, max) so that we know what we’re dealing with here. Relatedly, this could be something to explore statistically. For instance, one possible reason for the weaker results between early vs late bilinguals could be because the authors have drawn an arbitrary line between who they deem to be early vs late. If there are many “late” bilinguals aged under 10 for instance, it could obscure the genuine early vs late effect. I’d be interested to see if intelligence correlates with age of L2 acquisition within the bilingual groups, or within the pooled bilingual sample as a whole.
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
This study tested how adults with different linguistic backgrounds (monolinguals, early bilinguals, late bilinguals) performed on Raven’s Matrices as a measure of fluid intelligence. As predicted, bilinguals outperformed monolinguals, and early bilinguals outperformed later bilinguals. Based on these results, the authors conclude that early bilingual experience shapes attentional processes, leading to lifelong influences on cognition. Additional information about participants’ language experience could provide even more insight into how these effects may develop, and it would also be useful to better understand exactly how language experience might affect fluid intelligence differently than other processes. But overall, this paper addresses a tightly focused topic, and the results appear to be robust and informative in debates about how bilingual experience changes cognitive processes.
Specific comments
- It would be helpful to provide specific evidence that fluid intelligence is likely to “underpin” skills such as cognitive control and attentional monitoring (p.2, ll. 81-81). In general, more clarity on how fluid intelligence is distinct or related to other constructs could help to motivate the theoretical questions, instead of simply introducing a familiar task.
- Relatedly, in the discussion, relations between fluid intelligence and attention are emphasized, making it hard to understand whether the RAPM task is seen as tapping into attention (as a part of fluid intelligence) or if the claim is ultimately that the problem-solving element is fundamentally altered by early bilingual experience. The current study obviously did not include a large battery of cognitive tasks to tease apart different subcomponents of the larger task, but more clarity about the processes that would/would not be expected to be affected by early linguistic environments could strengthen the theoretical contribution.
- Given the complexity of bilingual experience, more information about the participants would be helpful, e.g., including information from the background questionnaire in a table. Specifically, were the proficiency differences between the bilingual groups? Was any information collected about their language use, or how often they switch languages? Were the participants in the late group primarily exposed to L2 in classroom or immersion settings? Given that the introduction proposes that experience switching between languages may explain the effects of bilingualism on attention, it would be interesting to know whether the key difference between early and late bilinguals is the age at which they were exposed to the language, or if their current experience or proficiency also contribute. It’s possible that not all of this information was available, but given the heterogeneity of bilingual experience, the possibility that factors other than the age of exposure contribute differences should be considered and discussed.
- The results were presented very clearly and directly addressed the primary research question. But, exploratory analyses could be informative. For instance, did the late bilingual group differ from the monolinguals? Was RAPM performance related to any other factors, such as proficiency, age of acquisition, or a specific type of bilingual experience? Knowing whether the late bilinguals differed from the monolinguals in particular significantly changes the interpretation about how, and when, language experience has the potential to shape more general cognitive processes.
- It is valuable to see results connected to a variety of topics in the literature. However, given that this was a purely behavioral study and the introduction was focused primarily on cognitive theory, the focus of the discussion on neural data was somewhat unexpected. It might be useful to also connect these results to behavioral and developmental findings (e.g., Byers-Heinlein, Morin-Lessard, & Lew-Williams, 2017 describe how challenges in early bilingual experience might yield cognitive benefits; Kovacs & Mehler, 2009 describe early differences between monolingual and bilingual infants’ flexibility, etc.).
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
This is a well-designed and interesting study comparing early bilingual, late bilingual, and monolingual groups using Raven’s Advanced Progressive Matrices – Set 1. Overall, it is a well-designed study and carefully conducted. I appreciate the fact that the authors shared their data in the OSF. I believe that the paper will be promising and have important implication to the literature.
I only have several comments to improve the manuscript further:
- Since participants are young adults, I feel that their current education may provide more information about their SES than their parental education. Is this information available. Or at least, the authors should provide more information on how participants were recruited in all three groups. Are they coming from the same pool? What is the inclusion and exclusion criteria. These information is important and should be elaborated in the participant section
- A lot of linguistic characteristics of the participants are missing. What is their first and second language? Is it heterogeneous? The authors should describe the linguistic characteristic in details. Currently, it is very lacking. If these information were not collected in the study, the authors should acknowledge this limitation and should at least expand and qualitatively elaborate the socio-cultural context of the bilingual and monolingual community recruited in this study.
- The Bayesian analysis is informative and important. It should be moved to the results section for clarity.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 1 Report
Having now read the revised version of the manuscript, I thank the authors for addressing each of my comments.
I understand that it has been challenging to keep data collection going during the COVID pandemic and compromises had to be made by all labs around the world.
My opinion about the paper remains largely the same as it did after my first reading: it is an interesting study with some caveats:
- The way SES is measured (parent education but not occupation) isn’t perfect
- The way early vs. late bilingualism is defined is slightly odd, and (I think) hurts the authors in their attempt to describe the underlying effects
- I like the expanded Results section
- I don’t buy the orthogonal contrasts argument because the variance in the data for the early and late bilinguals is being used in both planned contrasts which will increase the family-wise error rate, but ultimately, it doesn’t matter. I actually think that the effect is probably real, but obscured because of imperfect experimental design decisions (rather than not correcting alpha)
So, where does that leave us then? I find this to be a potentially good, but imperfect study. The results are fairly subtle, and I would argue are obscured by some design choices. I am satistifed with the tempering of conclusions, and do not see any outlandish claims.
The authors have addressed my comments as best they can. I don’t have anything further to add.
Author Response
We thank the reviewer for their comments. We agree that the study is not perfect. A larger study with more detailed measures is definitely needed. With more detailed measures, we could also do away with the need to compare two groups, and all the theoretical and practical problems that come with such basic comparisons (where you draw the line between early and late, etc.).
For the planned contrasts, we did run and interpret the analyses according to Andy Field’s Discovering Statistics Using SPSS – Third Edition. It’s what is taught in many of our universities in the UK. I discussed the relevant chapter with colleagues from different universities, but none found issue with what Andy Field wrote. If I obtain a large grant to run a similar study, I will budget for a statistician to check over our analyses and interpretations.
Reviewer 3 Report
The authors have sufficiently addressed all my concerns.
Author Response
We are pleased we have sufficiently addressed the reviewer's concerns, and thank the reviewer again for their review and comments.
