Dialect Recognition via Lexical Processing: Is It a Viable Litmus Test?
Round 1
Reviewer 1 Report (Previous Reviewer 1)
Comments and Suggestions for AuthorsOverall, the authors have adequately responded to my comments.
However, I would urge the authors to have a critical look at all the claims they make regarding prior research, as there seems to be some mistakes. For instance:
“Declerck and Kirk (2023) investigated bidialectal speakers of Dundonian Scots and Scottish Standard English using voluntary language switch paradigm, and they found symmetrical cost of switching between Dundonian Scots and the Scottish standard English. However, they did not find a significant effect of the switching cost. Therefore, they concluded that the bilingual and bidialectal control mechanisms are different.”
This study did observe switch costs (this is why there were also symmetrical switch costs), but no mixing benefit was observed, unlike bilingual voluntary switching studies. The latter observation led to the conclusion that there might be a difference between bilingual and bidialectal language control.
Author Response
Comment 1: However, I would urge the authors to have a critical look at all the claims they make regarding prior research, as there seems to be some mistakes.
Response: Once again, we are very grateful to the anonymous reviewer. We took the remark seriously and cross-checked works we have cited in the manuscript. We also incorporated the correct interpretation of the findings reported in Declerck and Kirk (2023). We hope that we have now improved the quality of the manuscript.
Reviewer 2 Report (New Reviewer)
Comments and Suggestions for AuthorsThis manuscript seems to be a revised submission, since there are many passages in green font color. I was not part of earlier review rounds. The study reported is very timely and relevant for research on the role of dialects in bilingualism.
On the general topic of 'bidialectalism', the authors mention Lundquist & Vangsnes (2018), but Vangsnes is also part of a research group that classifies 'bidialectal' speakers of Norwegian as 'bilectal' speakers (most recently e.g. Kubota et al. 2023), following the terminology introduced by Rowe & Grohmann (2013). The latter paper has given rise to a lot of research on bilectalism, such as first language acquisition and development or different types of language pathology, incl. the cited Antoniou et al. (2016).
The authors (also in many of their follow-up publications) attribute the term to speakers in diglossia, another aspect discussed in this manuscript. In addition, they zoom in on the term ‘lect’ which also plays a role in the present manuscript (“The Eastern dialect is their native lect” in l. 595).
Subsequently, in the second paragraph, at least the Antoniou et al. study should be cited, arguably the first of its kind: "However, against the general perception in psycholinguistics, studies indicate that bidialectal speakers are more like bilinguals than monolinguals in terms of language representation and processing (Lundquist& Vangsnes, 2018; Poarch et al.,2019; Vorwerg et al., 2019)." (Incidentally, add a space before the ampersand after Lundquist.)
Presentation:
The paper should be double-checked very carefully for expression, typography, and spelling.
l. 3: correct "These studies argues", something to check across the paper (e.g. “Finkbeiner et al. (2006) argues” in l. 176)
l. 101: Is 'alas' really necessary in "However, alas, only"?
l. 844: remove first comma in “Green, & Abutalebi, 2013” and check throughout
l. 848: “use both dialects on the daily bases” ïƒ “use both dialects on a daily basis”?
l. 862: Lunndquist ïƒ Lundquist
l. 895: magnitude ïƒ magnitude
l. 941: “the Universal Bilingualism” ïƒ “Universal Bilingualism”
l. 946: “We rest our case” — Is this really an appropriate way to end the final paragraph?
References:
Kubota, M. et al. (2023). Bilectal Exposure Modulates Neural Signatures to Conflicting Grammatical Properties: Norway as a Natural Laboratory. Language Learning, DOI: 10.1111/lang.12608.
Rowe, C. & Grohmann, K.K. (2013). Discrete Bilectalism: Towards Co-Overt Prestige and Diglossic Shift in Cyprus. International Journal of the Sociology of Language 224, 119–142.
Comments on the Quality of English Languagesee above
Author Response
Comment 1: On the general topic of 'bidialectalism', the authors mention Lundquist & Vangsnes (2018), but Vangsnes is also part of a research group that classifies 'bidialectal' speakers of Norwegian as 'bilectal' speakers (most recently e.g. Kubota et al. 2023), following the terminology introduced by Rowe & Grohmann (2013). The latter paper has given rise to a lot of research on bilectalism, such as first language acquisition and development or different types of language pathology, incl. the cited Antoniou et al. (2016). The authors (also in many of their follow-up publications) attribute the term to speakers in diglossia, another aspect discussed in this manuscript. In addition, they zoom in on the term ‘lect’ which also plays a role in the present manuscript (“The Eastern dialect is their native lect” in l. 595).
Subsequently, in the second paragraph, at least the Antoniou et al. study should be cited, arguably the first of its kind: "However, against the general perception in psycholinguistics, studies indicate that bidialectal speakers are more like bilinguals than monolinguals in terms of language representation and processing (Lundquist& Vangsnes, 2018; Poarch et al.,2019; Vorwerg et al., 2019)."
Response: We are very grateful to the anonymous reviewer, and we recognize their legitimate concern pertaining to the terminologies we used in the reviewed manuscript. We are aware of the recent works by Kubota et al. (2023), but we could not incorporate into the reviewed manuscript because the work was published online just recently. We have now incorporated both Rowe & Grohmann (2013) and Kubota et al. (2023) in the revised manuscript. Besides, we cited Antoniou et al. (2016) in the proper place.
Regarding the use of terminologies—bidialectalism, bilectalism, lects and diglossia, we certainly agree with the anonymous reviewer; we believe that consistency is pivotal. Hence, in the revised manuscript, we used ‘bidialectalism’ consistently since, in the case of Oromo, the Western dialect is learned in schools, and the target participants are more proficient in the native Eastern dialect than in the school Western dialect—a view also reflected in Rowe & Grohmann (2013). The two dialects are spoken in clearly defined contexts (school vs. outside school), and they are spoken as native dialects in geographically distant locations—the Eastern dialect in the eastern part of the country and the Western dialect in the western part of the country. Hence, in the revised manuscript, we consistently used ‘dialect’ instead of ‘lect’ to describe the target varieties. Given that we are dealing with a sociolinguistic situation in which one of the dialects is used in school and the other in the outside school context, we believe that describing such contexts as ‘diglossia’ is not problematic. As the anonymous reviewer may know, ‘bidialectalism’ is refers to the speakers’ competency whereas as ‘diglossia’ is a sociolinguistic phenomenon. Hence, the two do not necessarily controvert each other.
Finally, we are aware of the apparent difference between Lundquist & Vangsnes (2018) and Kubota et al. (2023. However, from the representation standpoint, we think that both studies make a similar claim; they imply that dialect speakers have a separate grammar for each of the dialects they speak. We do not really think that ‘bidialectal’ and ‘bilectal’ speakers are different in terms of their lexical representation. We assume that both have dialect-specific lexical representation.
Presentation:
Comment 1: The paper should be double-checked very carefully for expression, typography, and spelling.
Response: We tried all our best to fix the problems. We hope we have improved the quality of the manuscript.
Comment 2: correct "These studies argues", something to check across the paper (e.g. “Finkbeiner et al. (2006) argues” in l. 176)
Response: We edited the manuscript and fixed the errors.
Comment 3: Is 'alas' really necessary in "However, alas, only"?
Response: We agree that ‘alas’is not necessary. Hence, we revised the whole sentence.
Comment 4: Remove first comma in “Green, & Abutalebi, 2013” and check throughout.
Response: The comma is removed, and we checked throughout the manuscript.
Comment 5: “use both dialects on the daily bases” à “use both dialects on a daily basis”?
Response: The correct expression is used in the revised manuscript.
Comment 6: Lunndquist à Lundquist
Response: Corrected.
Comment 7: magnitud à magnitude
Response: Corrected.
Comment 8: “the Universal Bilingualism” à “Universal Bilingualism”
Response: Corrected.
Comment 9: “We rest our case” — Is this really an appropriate way to end the final paragraph?
Response: We agree that it is irrelevant. Hence it is excluded.
Reviewer 3 Report (New Reviewer)
Comments and Suggestions for AuthorsSee attached file
Comments for author File: 
Comments.pdf
In general, the language is comprehensible but sometimes too colloquial. In addition, sometimes verbs are in an incorrect form and some typos need to be corrected. I would advise proofreading by an L1-speaker of English - however, the green parts need less improvement.
Author Response
General concept comments
Comment 1: The manuscript is clear and well-structured. Although the manuscript is scientifically sound, the conclusions drawn are not fully justified and seem to be too general (e.g., lines 749-751: “The observed facilitation effect confirms the language-specific lexical representation of bilingual speakers […]”; lines 780-784: “[…] our study nullifies the longstanding contradictions […]”; lines: 936-937: “Our findings show that there is no fundamental difference between the lexical representations of bilingual and dialectal speakers”).
Response: The conclusions are drawn from the assumptions held in the previous studies. In previous studies, for example Melinger (2018, 2021), Dylman and Barry (2018) as well as in earlier studies on bilingual lexical selection—Costa and Caramazza (1999) and Costa et al. (1999), the translation equivalent facilitation effect was used as a tool to determine whether bilingual or bidialectal speakers have independent or co-dependent lexical representation. We did almost the same thing. We realize that there are various perspectives regarding the facilitation effect of translation equivalents, regardless of the limitations, our conclusions are bult on the assumptions of the translation facilitation effect. If we recognize that the translation facilitation effect is an indicator of two-system lexical representation, our conclusions are conceivable. It could be that we did not establish a strong link between our conclusions and the previous works. We tried all the best to establish the link in the revised manuscript.
Comment 2: Too little information is provided about the sociolinguistic situation and the participants to critically evaluate the results. This is highlighted by the manuscript as well as it states on several occasions (e.g., lines 103-105) that ecological and linguistic biases need to be accounted for. This becomes also evident in lines 288-309. Thus, it would be necessary to know more about the participants’ language experience.
Response: The anonymous reviewer is concerned about the target sociolinguistic context and the participants we investigated. We share their concern; we also feel that we have not provided every required information. However, it should also be noted that this limitation is inherently associated with the target language context; here, we are dealing with understudied languages that have not appeared in psycholinguistics literature. We believe that this reality needs to be considered.
We described the sociolinguistic context and the participants’ background on page 9-10 of the revised manuscript—rather than repeatedly describing them in the method section. We did so to provide unified information about the target context and participants. Given the concern of the anonymous reviewer, we have incorporated additional explanation into the revised manuscript.
Having said these, we feel that we have provided necessary evidence which justifies that the Ethiopian language ecology is indeed different. For example, the special linguistic features such as non-concatenative morphology and converbs make the context linguistically different. In the introduction of the manuscript, we also highlighted that the target bidialectal speakers are bidialectal bilingual speakers, unlike the monolingual bidialectals studied in most previous studies. Besides, we pointed out that the Oromo dialect context represents a special type of diglossia since there is little difference between the school and local dialects in terms of social status. Crucially, we indicated that the Oromo dialect context allows equal use of both dialects unlike some other dialect contexts that allow frequent use of only one of the dialects. These explanations are included in the revised manuscript.
Comment 3: Experiments using the picture-word interference paradigm and those using a language switching task should be kept apart more clearly as these tasks might yield very different results and/or use different cognitive mechanisms.
Response: We agree with the anonymous reviewer, and we tried to differentiate the two experiments in the revised manuscript.
Comment 4: The data would be easier to interpret if the reaction times would (in addition) be presented in a table instead of figures 2 and 3. Furthermore, the identity effect should be calculated as well.
Response: During the early submission of the manuscript, we presented the results both in tables and figures. Later, we realized that combining the two results in unnecessary repetition. Besides, we believe that tables with full of RTs discourage especially readers that are unfamiliar to the subject area. Hence, in the revised manuscript, we preserved the graphs, but we also provided the RTs in the appendix (see appendix B.1 and B.2). The identity effect and the direction of the inter-lexical facilitation are also presented in the appendices.
Comment 5: The bilingual/bidialectal nature of both experiments (i.e., switching languages/dialects between blocks) might have had a very strong influence on the activation of the nontarget variety.
Response: We agree with the anonymous reviewer, but we do not really know how this affects the contents of our argument. We found facilitation effect of translation equivalents even when the pictures are named in the nonnative language/dialect. This implies that it is the overall response times of the nonnative language/dialect that is primarily affected by the parallel activation of the target and nontarget language/dialect.
Comment 6: I do not understand why it’s mentioned several times throughout the manuscript that interference effects are beyond the scope (e.g., lines 736-737) as the experiments are conducted to investigate co-activation during picture naming for which interference effects would be an indication.
Response: Thank you for bringing this point into our attention. We were referring to the language switch cost—we now realize that our description needs some clarifications. We thought and still think that our experiment is not suitable to make a valid claim about the language switch cost. We think so because of two reasons. First, in our experiment, we did not have item-level language/dialect switch. In our case, the switch was at the block level. Second, there was a practice session of 15-20 minutes between the first and the second blocks. This time gap could potentially lead to the activation decay. Because of these, we decided not to take the language switch cost seriously—two other reviewers also share our concern.
Specific comments
Comment 7: Line 2: If the title is kept, I would advise to take up the question again more directly or (at least) make a clearer reference to Melinger (2018).
Response: We understand why the anonymous reviewer is concerned. As we indicated in the introduction of the manuscript, we want to investigate both the theoretical and practical aspects. Hence, we want to keep a balance between the two. Based on this, we have decided to preserve the middle ground. We want to benefit from Melinger’ approaches, but without neglecting other alternatives.
Comment 8: Lines 3-5: To me it does not seem that the main argument of the manuscript is on distinguishing dialects from languages but rather on lexical selection in two groups of participants who speak (according to current definitions) two different languages. belonging to different Afro-Asiatic branches or two closely related dialects of the same Cushitic language.
Response: We believe that our study serves both—see our response to comment 7. We view lexical selection and dialect recognition as two sides of the same coin. The lexical selection mechanism is a tool to determine whether the target varieties are languages or dialects. Once the type of the lexical selection mechanism is determined, the differentiation between language and dialect is automatic. In other words, lexical selection is a means whereas dialect recognition is the outcome. Hence, we believe that lexical selection and dialect recognition have a process - outcome relationships.
Comment 9: Lines 68-72 seem to be overstating the aim of the processing-based proposal by Melinger (2018, 2021).
Response: We share the concern. Neutral expressions are used in the revised manuscript.
Comment 10: Lines 83-86: No, they did not use a picture-word interference paradigm.
Response: That is right. Vorwerg et al. (2019) used language switch experiments. Hence, we revised the statements.
Comment 11: The outline in lines 129-135 does not seem necessary.
Response: We agree. We removed the outline.
Comment 12: In 1.1., the manuscript should briefly also talk about timing issues regarding the semantic interference effect as well as about other moderating factors (such as whether the distractors are part of the response set or not and whether the relationship between target and distractor is a PART-WHOLE relationship or not).
Response: They were parts of our initial manuscript. We later decided not to dive into the vast literature of semantic interference effect–mainly because we are interested in the facilitation effect. Given the concern of the anonymous reviewer, we included a brief summary in the introduction part of the revised manuscript.
Comment 13: Lines 163-181, these models do not seem to be considered any further in the manuscript.
Response: We agree. Meanwhile we wanted to show the readers that there are alternative thoughts. We still believe that it is safe to keep the models for not to convey a wrong message—a message that only models of our interest are relevant.
Comment 14: - Are there any copyright issues with Figure 1 in lines 225-231?
Response: We are not certain about this. To be safe, we will contact either Melinger or the publisher.
Comment 15: Unify the spelling of Scottish S-/standard English (e.g., lines 245, 247) and s- /Standard Scottish English (e.g., lines 251, 253).
Response: The spelling is unified (see the revised manuscript).
Comment 16: I do not understand the claim in lines 389-391: “[…] since there is no legally endorsed standard Oromo dialect, the stereotyping associated with the dialect status is either minimal or nonexistent”.
Response: The sentence is modified in the revised manuscript.
Comment 17: - I do not understand the claim in lines 394-395: “Since, the native speakers of Eastern Oromo dialect have exposure to the Western Oromo dialect, they are bidialectal speakers”. Mere exposure to a language/dialect does not lead to knowledge/proficiency.
Response: We are referring to an extensive exposure to the Western dialect—via school and media. May be our explanation was not informative. Hence, we revised the explanation. We also incorporated additional information.
Comment 18: Lines 442-443: Were the distractors written using the Latin or the Ge’ez script? According to line 460, “[t]he distractor words were always written in […] capital letters”. Therefore, I assume the Latin script was used. Isn’t this unusual for Standard Amharic?
Response: Thank you very much for pointing out this. For the bidialectal experiment (Experiment I), Latin script was used to write every distractor word. However, for the bilingual experiment (Experiment II), Amharic distractor words were written in Ge’ez script, and the Oromo distractor warders were written in the Latin script. In other words, the combination of the Latin and Ge’ez scripts were used in the first experiment. This information is included in the revised manuscript. Apologies for not explaining this in the reviewed manuscript.
Comment 19: Line 486: How big were the pictures? How far did participants sit from the screen?
Response: We sized the pictures into 600 by 600 pixels. The font size of each written distractor was 28, Time News Roman. We did not instruct the participants to sit at a particular distance from the computer. However, we told them to sit comfortably and to avoid movements once the experiment begins. This information is included in the revised manuscript.
Comment 20: Lines 511-512, 677-678: Please include percentages for the errors and outliers.
Response: We have included the percentage of the errors in the revised manuscript.
Comment 21: Lines 548-560, 712-722: What kind of errors were made?
Response: Trials in which the participant produced the wrong word, stuttered, and hesitated or failed to respond, including any trial with RT faster than 200ms and slower than 2500ms, were classified as errors. This information is included in the manuscript.
Comment 22: Lines 567-569: Shorter RTs than in the translation equivalent condition when naming the pictures in Amharic or shorter RTs than in the same condition when naming the pictures in Oromo? Is this irrespective of block order?
Response: We mean shorter reaction times than in the translation equivalent condition when naming the picture in Amharic, irrespective of the block order.
Comment 23: Lines 586-589: It could also be that participants were in a more bilingual mode which led to stronger activation of both languages.
Response: We agree with the anonymous reviewer. Since both languages are active during the second block, it can be that the participants are in a more bilingual mode. We included this alternative explanation in the revised manuscript.
Comment 24: Lines 597-604: What about their competence in the Western Oromo dialect?
Response: We did not directedly measure the participants’ competence in the Western Oromo dialect. However, from studies recently conducted by Feleke (2024) and Feleke and Lohndal (2024), we know that the bidialectal speakers are proficient in both Oromo dialects—though the Eastern dialect is the dominant one. Most children acquire the Western dialect at about the age of ten. This information is included in the revised manuscript.
Comment 25: - Lines 615-623: Pleas re-write this part as it is very difficult to follow.
Response: We revised the subsections. We hope that we have reduced the inconveniences.
Comment 26: - Lines 624-687: These subsections can be shortened substantially by referring to the first experiment.
Response: It is important to remember that there is a slight difference between Experiment I and Experiment II. In the second experiment, the written distractors are the labels of the pictures themselves, which is not the case in the first experiment (Also see Appendix A.1 and A.2). Hence, we feel that it is better to maintain the details in both experiments.
Comment 27: - Line 695: Do you mean the within-dialect control condition?
Response: No, in this case we are comparing the dialect equivalent condition to the cross-dialect control condition. In other words, we found a significant difference only between the dialect equivalent condition and the cross-dialect control condition. We did not find a significant difference between the dialect equivalent and within-dialect control conditions.
Comment 28: - Lines 761-762: Do you mean that there is “no conclusive explanation”?
Response: Yes, that was what we wanted to say.
Comment 29: - Lines 811-819: More evidence is needed to follow this train of thought.
Response: We partly agree with the anonymous reviewer since we are not certain whether Oromo or Amharic is the dominant language of the target bilingual group—we just assume that the first language/Oromo is the dominant one. However, as to the Oromo bidialectal speakers, Feleke (2024) shows that the Eastern dialect is the dominant one. The influence of dominance on the inter-lexical facilitation effect is explicitly discussed in Kroll and Stewart (1994). Hence, our argument is not unfounded, but we agree that it is not supported by a substantial amount of evidence.
Comment 30: Lines 864-895: I do not understand why this model was not part of the introduction.
Response: Dlyman and Barry (2018) illustrated the involvement of Inhibitory Control (IC) in the lexical selection process. Our modified version—the Dynamic Lexical Selection Model adds two extra components; (1) recognizes the role of language/dialect context, and (2) acknowledges the involvement of IC in the inter-lexical facilitation. The link between dialect ecology and IC is central to the dynamic model we proposed. We argued that due to the strong link between dialect contexts and IC, every bidialectal speaker is different when it comes to lexical selection and representation. We did not include this argument in the introduction since we wanted to keep a balance between the theoretical and practical aspects of our study.
Comment 31: Appendix: Please translate the distractors.
Response: The English translations of the distractor words are included in the revised manuscript (see Appendix A.1 and A.2).
Round 2
Reviewer 3 Report (New Reviewer)
Comments and Suggestions for AuthorsThe revision improved the quality of the manuscript substantially. However, there are still minor changes I would like to suggest (see attached file).
Comments for author File: Comments.pdf
See attached file.
Author Response
Once again, we are very grateful to the anonymous reviewer. Thank you very much for the feedback. They are constructive and extremely useful.
Specific comments (minor):
Comment 1: In several cases, spaces are missing for references in brackets (e.g., line 19 ‘Macnamara1967’; line 139 ‘Savoy1998’; line 368 ‘1972,1977’; line 788 ‘et al.1999’).
Response: We have fixed all the problems associated with the space.
Comment 2: Similarly, some references include a comma between author(s) and year (e.g., line 523 ‘Forster and Forster, 2003’; lines 763-764 ‘Feleke, 2004; Feleke and Lohndal, 2024’). However, mostly they do not.
Response: We have fixed the problems associated with references. We have written the references without comma, based on the instructions provided in the author´s guideline.
Comment 3: Maybe this was unclear in my first review. When I said “the conclusions drawn are not fully justified and seem to be too general”, I mostly referred to the language used. This is still an issue (e.g., line 73 ‘proves’; lines 92, 267, 961 ‘showed’; line 359 ‘succeed in discovering’; line 367 ‘will come to end for once and forever’; lines 799-802 ‘The observed facilitation effect of dialect equivalent means that [...] Kirk (2023)’; line 803 ‘have’; lines 805, 820, 967 ‘show’; line 975 ‘we have demonstrated’). All the obtained results could just be due to chance. An excerpt taken from Drager (2018: 8, Experimental Research Methods in Sociolinguistics) that I use in teaching how to report statistics: “we work with likelihoods instead of absolute certainties. Therefore, when we frame our results, we say things like ‘The results provide evidence that ...’ instead of stronger claims like ‘The results prove...’ The more independent studies that provide evidence in support of some hypothesis, the more confident we become that the hypothesis is true.”
Response: Thank you very much for the elaboration. We now clearly see the concern of the anonymous reviewer. In the revised manuscript, we tried to avoid words of certainty and consistently used more suitable expressions.
Comment 4: The umlaut (Öcher) in line 81 is distorted in my version.
Response: Corrected.
Comment 5: Line 96: ‘first and foremost’
Response: Corrected.
Comment 6: Lines 149-151: What semantic interference effect is there, when distractor words are not semantically related to the target?
Response: The picture is ‘FISH’ and the distractor words is ‘whale’. We thought and still think that they are semantically related. Are not they?
Comment 7: Line 158: La Heij (2005: 289, 290, 297, ...) refers to ‘complex access, simple selection’ (not ‘easy’).
Response: Corrected.
Comment 8: Line 221: speakers
Response: Corrected.
Comment 9: Line 227: space between ‘D1’ and ‘and’
Response: Corrected.
Comment 10: Line 258: either ‘found a similar pattern of switch costs’ or ‘found similar patterns of switch costs’
Response: Corrected. We used ‘a similar pattern’ in the revised manuscript.
Comment 10: Line 281: ‘Kroll and Stewart (1994 who’
Response: Corrected.
Comment 11: Line 289: ‘lexical links’
Response: Corrected.
Comment 12: Line 294: ‘bidialectal speakers’
Response: Corrected.
Comment 13: Lines 310, 865: ‘adopt’
Response: Corrected.
Comment 14: Line 377: ‘Amhara and SNNP’ What does that mean?
Response: We were referring to two administrative regions or states in Ethiopia— ‘the Amhara Regional State’ and ‘the South Nation Nationalities People Regional State’. From the anonymous reviewer’s feedback, we have learned that the abbreviation in the reviewed manuscript was confusing. Hence, we avoided the abbreviation and replaced by the full names of the states.
Comment 15: Line 395: ‘It is also serves ...”
Response: Corrected.
Comment 16: Line 408: “it is a de facto a second language ...”
Response: Corrected.
Comment 17: Lines 440-443: repetition
Response: We have excluded the repeated sentences.
Comment 18: Line 475: “the Oromo distractor ...”
Response: Corrected.
Comment 19: Line 476: “the Ge’ez script’
Response: Corrected.
Comment 20: Line 483: longer dash before ‘same condition’
Response: Corrected.
Comment 21: Line 494: ‘congruency facilitation effect’ aÌ€ this terminology was not used before.
Response: It was unnecessary. Hence, we have excluded from the revised manuscript.
Comment 23: I would advise the authors to always report exact p-values; not only when they are non-significant (e.g., lines 578 and 581).
Response: We are not sure that we understand this feedback. For example, we reported p-value of 0.0008 as 0.001. Is not this the standard way of reporting statistical results?
Comment 24: Line 649: ‘Moreover importantly, the length of the letters ...”aÌ€ Please revise.
Response: Corrected.
Comment 25: Lines 684-704: I still think that this subsection can be shortened substantially by referring to the first experiment.
Response: We agree with the anonymous reviewer. We have now shortened the section by referring to Experiment I (see the revised manuscript).
Comment 26: Appendix: Don’t you think that it might have had an influence on the results that some (or in Experiment 2 almost all) control distractors were also target names?
Response: Here we recognize the legitimate concern of the anonymous reviewer. Simultaneously, we believe that this approach avoids the challenge of matching the distractor words with the labels of the target pictures in terms of their frequency. This is substantially difficult particularly when it comes to understudied dialects such as Oromo dialects which do not have well organized electronic corpora. More importantly, the same approach was used in previous studies (see Melinger 2021), presumably for the same issue of frequency.
Comment 27: References: The order of first and last names is not consistent.
Response: All the references are cross-checked, and the errors are now corrected.
This manuscript is a resubmission of an earlier submission. The following is a list of the peer review reports and author responses from that submission.
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThis is an interesting re-evaluation of the work of Melinger with bidialectals. While I really liked the paper overall, I do have some comments:
In the introduction, it would be good to consistently provide the language variety combinations of the bidialectals for each of the studies that you discuss.
Some other bidialectal studies might be good to add to the introduction, to give the reader a more comprehensive view of this literature (e.g., Declerck & Kirk, 2023; Vorweg, Suntharam, & Morand, 2019).
I would not discuss the carry-over effect in the introduction, since this deals with something different than the main focus of the study. If you want to discuss this at all, I would suggest to do so in the general discussion maybe? I am not even sure this should be added at all, unless the authors can make a better connection with this finding and the overall question regarding dialect processing and its relationship to bilingual processing.
Related to this, when discussing the carry-over effect, it might be good to explain what is going on here early on. Also, it might be good to include some recent, empirical studies that have looked into this (for a recent review, see Declerck & Koch, 2023).
I would suggest to shorten section 3 and integrate it into section 1. Now it feels like we are going from one thing to another and then back to the first subject.
It would be good to indicate that you mimic the methodology of Melinger (2018), which is not a bad thing at all!
In the participant section of Experiment 1 there is mention of a questionnaire. If numerical data was acquired from this questionnaire (for instance how good they would rate their language production in a specific language or how well they can comprehend sentences in a specific language), it would be good to add the averages here (with SD).
Regarding the Figures, the data seems to be pressed against the left and right side, making it difficult to see any differences. I would redraw these Figures so that it’s easier to see the pattern.
“The frequency of the Amharic labels, the Oromo labels and the distractor words was obtained from SKELL.” What is SKELL? Is there a reference?
Why were slightly more participants investigated in Experiment 2 than Experiment 1?
Do both dialect varieties have a formal way in which each dialect variety is written?
“Bilingual lexical selection and representation can be influenced by an interference 842 from the nontarget languages. In this regard, we observed a bidirectional between-lan-843 guage switch cost; there exit both L1-to-L2 and L2-to-L1 interferences.”
These are not called switch costs, but have several names (e.g., blocked language order effect (Declerck, 2020) or language after-effect (Wodniecka et al., 2020)).
Comments on the Quality of English LanguageThere are a couple of mistakes, but these should be easy to pick out before resubmitting.
Author Response
The response is attached
Author Response File: Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsSummary:
The manuscript under review delves into the challenge of objectively distinguishing between languages and dialects, building upon a recent psycholinguistic approach proposed by Melinger (2018 & 2021), which contends that bilingual speakers store dialect words as a co-dependent, not independent, representation.
The research objective is to apply and test Melinger's theory in the context of underrepresented language varieties. The authors study bilingual speakers of Oromo and Amharic and bidialectal speakers of Oromo using the picture-word interference paradigm. The fact this work involves speakers of underrepresented varieties was a particular strength of the paper and has the potential to add an important contribution to the literature, which often focusses on a rather limited number of language varieties.
The major findings reveal language-specific lexical representations in both the bilingual and the bidialectal cohorts, challenging previous work that demonstrates a different finding for bidialectals. The results highlight similarities between these two groups in terms of lexical representation and selection. This implies that bidialectal speakers can have flexible lexical representation and selection mechanisms influenced by their previous language experiences, similar to bilinguals.
The manuscript culminates with the proposition of a dynamic lexical selection model that accommodates a variety of dialect ecologies, thus contributing a new perspective to the discussion on language and dialect differentiation. I enjoyed reading this paper and think it would make a valuable contribution to the literature once the following points have been addressed:
Major Comments:
1) What was the justification for the sample sizes used in both studies? Experiment 1 has 24 participants whereas Experiment 2 has 26 and neither study reports the exclusion of any participants. While I recognise that these samples are consistent with Melinger (2018, 2021) it would be useful to provide a justification for these sample sizes in the absence of having conducted an a priori power analysis. Likewise, the exclusion criteria for removing outliers would benefit from a similar justification. I feel these are particularly important to outline given the increasing awareness of “researcher degrees of freedom” (such as optional stopping and outlier manipulation) that have the potential to change study outcomes considerably.
2) The figures are very difficult to read, particularly from an accessibility perspective. I would strongly encourage the authors to re-do these figures to ensure that they are a larger size, of higher quality and that there is greater contrast to differentiate between the conditions. I would recommend the following paper which provides a tutorial for data visualisation using R and is aimed towards those who may not have prior experience using this software:
Nordmann E, McAleer P, Toivo W, Paterson H, DeBruine LM. Data Visualization Using R for Researchers Who Do Not Use R. Advances in Methods and Practices in Psychological Science. 2022;5(2). doi:10.1177/25152459221074654
3) In Line 335 the authors cite Declerck and Kirk (2021). However, this paper was a Stage 1 Registered Report protocol outlining a proposed study, rather than one that demonstrated specific findings (which it appears the authors have implied here). It would be more appropriate to cite the Stage 2 paper instead (with an awareness that the conclusions for this paper indicate that there are potential differences in bilingual and bidialectal language control, unlike the message the authors have conveyed in this section):
Declerck, M., & Kirk, N. W. (2023). No evidence for a mixing benefit—A registered report of voluntary dialect switching. Plos one, 18(5), e0282086.
4) As mentioned, a strength of the paper is that the authors have included speakers of underrepresented and understudied language varieties. I would encourage the authors to emphasise this importance as part of the introduction. I direct them to the following paper discussing the over-reliance on studying speakers of English, as well as the editorial article for a special journal issue titled “Towards a just an equitable psycholinguistics”, where they may find other appropriate papers to emphasise the significance of this current work.
Blasi, D. E., Henrich, J., Adamou, E., Kemmerer, D., & Majid, A. (2022). Over-reliance on English hinders cognitive science. Trends in cognitive sciences, 26(12), 1153-1170.
Kutlu, E., & Hayes-Harb, R. (2023). Towards a just and equitable applied psycholinguistics. Applied Psycholinguistics, 44(3), 293-300.
Minor Comments:
Some minor comments I feel are worth addressing:
1) I understand the requirements for this journal do not place an emphasis on Open Science practices such as pre-registering study designs etc. However, with a general increasing awareness of the issues with replicability and reproducibility of research findings, I would strongly encourage the researchers to make their data available in a repository such as the Open Science Framework, as well as any code or syntax that would allow others to reproduce their analyses. Given that this paper is a potentially important contribution to a relatively small research area - which already has conflicting results - it would be extremely useful for this data to be available for future meta-analyses.
2) The introduction is very comprehensive, providing a nice overview of the literature in this field. A particular strength is studying speakers of language varieties that have received little attention in the literature, providing some very important sociolinguistic context. I would suggest that some restructuring of this section would strengthen the narrative being presented. I would suggest that Section 2, outlining the sociolinguistic context should be presented later in the paper, once the overview of the wider psycholinguistic theory has been discussed.
3) While I would expect these to be amended at a future copy editing stage, it’s worth pointing out that there are several typos throughout the document (e.g. in the Figure 1 caption) and some missing references (e.g. Kirk, 2022, does not appear in the reference list).
Comments on the Quality of English LanguageSee above.
Author Response
The response is attached
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for Authors
The manuscript reports the results of a series of picture-word interference tasks with Oromo-Amharic bilinguals and Oromo bidialectals to explore the lexical representation of cross-language and cross-dialect translation equivalents. The results show clear patterns of facilitation for both cross-language and cross-dialect translation equivalents relative to unrelated distractors. The primary strength of the paper is the examination of language varieties that are severely under-represented in psycholinguistic research. Unfortunately, the manuscript has a number of major problems that outweigh this strength.
Overall Framing:
The big-picture question that this study aimed to address is how to distinguish languages from dialects, but it wasn’t clear to me what’s at stake for psycholinguistic theory in making this distinction. Why does it matter for psycholinguistic theory or method if we can tell languages from dialects? There’s some discussion on pp. 3-4 of concerns about comparisons between bilinguals and monolinguals, especially with respect to the lack of homogeneity of monolingual comparison groups, but it’s not clear to me how distinguishing languages from dialects will help resolve these concerns about homogenous participant groups.
Introduction and background:
The abstract and the first few pages of the introduction set up the paper to be a test of Melinger’s proposal about co-dependent lexical representations in bidialectals. It’s not until p. 3 that we learn that there is an alternative proposal, also supported by psycholinguistic evidence, that parallels the generally-accepted proposal about independent lexical representations in bilinguals. The paper should be framed around these competing proposals from the start, so that the alternative (which ends up being supported by the data in the manuscript itself!) is not introduced almost as an after-thought.
More generally, the introduction needs to be more fully fleshed out. The first full paragraph on p. 3 articulates a large number of goals for the current study and each of those goals needs to be motivated by a review of the relevant literature, including not just the theoretical conclusions, but a description of the results and, to the extent that they might be relevant, the language varieties that were studied and the methods that were used. Some of this additional content is provided later in section 3, but it is needed to motivate the current study. Some restructuring of sections 1 and 3 is therefore needed to provide the necessary background before the goals and research questions are articulated.
The section on the sociolinguistic context of Amharic and Oromo is sparsely cited. What is the basis for the various claims about the major dialects, their sociolinguistic status, their acquisition, and their inclusion in various domains (government, education, etc.)?
Also, in this preliminary section, the two groups of participants are not clear. There are three language varieties of interest: (standard?) Amharic, Western Oromo, and Eastern Oromo. Do the bilingual participants speak only the first two? Do the bidialectal participants speak only the last two? Or are the bidialectal participants also bilingual? There is some discussion of Amharic in the Oromo context, but it’s not clear here whether or not the bidialectal participants speak Amharic (or, as it turns out later, any other languages, like English).
Some additional explanation of Figure 1 (on p. 6, note that there are three figures called Figure 1 in the paper) is needed. What do the two levels of boxes with circles represent? What do the arrows indicate? Why is D2, but not D1, connected to a box with circles in Figure 1a? Where are D1 and D2 in Figure 1b? How does the structure in Figure 1a lead to facilitation for dialect equivalents?
I don’t follow the arguments about dialect ecology on pp. 7-8. What are the relevant factors that shape dialect ecology? What are the predictions about how dialect ecology affects lexical representation in bidialectals? And, in the context of section 4, what about the dialect ecology of the participants in this study is different from those in previous studies? I find the issue of dialect ecology, especially in the context of these varieties that are under-represented in psycholinguistics, very interesting and would like to see these ideas elaborated considerably throughout the manuscript.
Analysis and results
More information about the linear mixed effects modeling is needed. Were both random intercepts and slopes included for subjects and items or only random intercepts? How were the fixed effects coded (e.g., using treatment or sum contrasts)?
The pairwise comparisons on pp. 11-12 are unnecessary. The condition x language interaction was not significant, so there is no reason to look at the effects of condition separately by language.
It is stated on p. 12 that the number of errors was large relative to previous work, but neither the number of errors in this study nor the number of errors in previous work are provided. Some numbers to support this claim are needed. More generally, I don’t understand the analysis of the errors. First, given the significant condition x language interaction, which conditions differed in terms of errors in Oromo and which differed in Amharic? What do these differences mean in the context of the competing theories of lexical representation? Second, what is the block effect that was not significant? Isn’t block the same as language (which was significant in this analysis)? The difference between block and language is explained later in the context of Figure 1 (on p. 13) and response times, but it should also be explained for the error analysis.
I don’t understand what is shown in Table 1 or the arguments that follow it on p. 12 about the direction and magnitude of facilitation within and across languages. First, what are the numbers reported here? Are they mean response times? If so, why don’t they match the values shown in Figure 1 (on p. 11)? Second, what does it mean to talk about L1-to-L2 or L2-to-L1 facilitation in the within-language comparisons? Third, given the lack of a condition x language interaction in the response time analysis, the numerical differences here are not supported by the statistical model. That is, there’s really no evidence of an asymmetry in the magnitude of facilitation across language blocks.
Is the analysis reported below Figure 1 (on p. 13) the same as the response time analysis reported at the beginning of this section? If not, why are two models needed to examine the response time data? A single model with all of the relevant factors can be reported instead. Regardless, the entire response time analysis should be presented together, either before or after the error analysis. Currently, the error analysis is presented in the middle of the response time analysis.
Note that many of these comments about the analysis and the results also apply to the second experiment, but I do not repeat them here.
Discussion and conclusions
The conclusions in section 6 about larger switch costs for L2-to-L1 and in section 5.2.5 about larger switch costs for D2-to-D1 than D1-to-D2 are not supported by the analyses. In both analyses involving block, there was a robust block effect, in which responses were slower in the second block than the first block, suggesting a switch cost, but that effect was the same magnitude across languages (in the first experiment) and dialects (in the second experiment) because the block x language interaction was not significant in either analysis. The switch cost was therefore equivalent in both cases, regardless of the order of the blocks.
Some further discussion of Figure 8 is needed. First, what is the target dialect in this example and how do the various connections lead to the target utterance ‘adurree’? Second, how does the model depicted here capture variation in bidialectal lexical representation as a function of language ecology? As I noted above, the preliminary ideas in this paper about dialect ecology are quite interesting and it would be good to flesh them out to enhance the potential impact of the work.
Comments on the Quality of English Language
The manuscript would benefit from careful proofreading to correct grammatical infelicities and odd word choices.
Author Response
The response is attached
Author Response File: Author Response.pdf
