Gender in Unilingual and Mixed Speech of Spanish Heritage Speakers in The Netherlands

Round 1

Reviewer 1 Report

Review of Gender in Unilingual and Mixed Speech of Spanish Heritage Speakers in the Netherlands

The study looks at gender in code-switched Dutch/Spanish DPs. The researcher(s) had three objectives: (i) to study if the heritage speakers of Spanish in the Netherlands produced monolingual-like in their Spanish and Dutch wrt gender; (ii) what strategy these speakers used when assigning gender in code-switched DPs; and (iii) what extra-linguistic factors have an effect on gender assignment. In order to test this, the researcher(s) used a Director-Matcher task. The researcher(s) found that in Spanish the participants were 92.2% accurate. Accuracy in gender in Dutch was even lower. The author(s) found that that the participants were using a mix of the analogical technique and the default technique to assign gender in mixed DPs. More interestingly, they found that the participants were also alternating between pre- and post-nominal adjectives, with varying gender. Finally, they found that there were some extralinguistic factors that had an effect on the assignment of gender in the mixed DPs. For example, age of acquisition did not have an effect for Spanish gender accuracy, but it did for Dutch gender accuracy. They also found that “other” input (e.g., TV, media, books,) did have an effect (e.g., if the participant had more “other” exposure in Spanish, then the gender accuracy in Spanish increased and the analogical gender was used in the mixed DPs).

The main strengths of this paper are that it looks at a language pair (Dutch/Spanish) that has not been frequently explored, thus giving us insight into how these two languages behave wrt grammatical gender. It is especially interesting since Dutch was common/neuter distinction, as well as pre-nominal adjectives (vs. Spanish with mainly post-nominal adjectives). The paper is also nicely organized and easy to follow.

Main comments:

• The paper could benefit from a more in-depth discussion on the participant’s use of the pre- and post-nominal adjectives. First, an explanation of what were the nouns that were used in the pre-nominal constructions. The author(s) show in Figure 11 that the participants used the masculine gender more frequently than the feminine gender. The author(s) concluded that the participants preferred the masculine when the adjective was pre-nominal. However, it is not clear how the author(s) arrived at this conclusion. There needs to be more of a discussion on the nouns that were used in the post-, pre-nominal and even the DP+Relative clause construction. That is, what nouns were used in these constructions? Did the inserted Dutch nouns have common or neuter gender? If they had common gender, then this could explain why the Spanish masculine was used. For example, let us consider “hamer”, which has common gender (and it is masculine in Spanish). If the Spanish masculine gender was used with “hamer” in all constructions (i.e., pre-, post-nominal, and relative clause), then that tells us that the participants are using either the masculine as default, or the analogical strategy. It does not really tell us much else. The same would be for a neuter noun like “huis” (the participants would use the feminine gender rather than the default). The interesting data would be if the participants are using the feminine in the n+adj structure and then used the masculine in the adj+n structure for the same noun. If this is the case, then perhaps there is evidence that having the adjective in pre-nominal position has an effect on gender assignment, thus it can be considered another gender assignment strategy. If the author(s) want to posit that the placement of the adjective has an effect on gender assignment, their analysis needs to be clearer, because I am not convinced that this is the case.
• While the majority of the methodology was well done (such as doing the monolingual trials before the code-switched trial), I would like the authors to address the lack of a proficiency test in both languages or the reason that the author(s) did not include one. The participants had varying profiles (with age ranges being very wide, and some leaving abroad for several years) which could have an effect on the results. I see that there are self-reported proficiency scores but only in Spanish and not in Dutch. The proficiency tests could have also been able to explain the use of the pre- or post-nominal adjective. That is, if the participants were more proficient in Dutch, then it makes sense that they would prefer to use the pre-nominal adjective. On the other hand, if the participants were more proficient in Spanish, then that explains the use of the post-nominal adjective.

Other comments:

• The paper needs a little more clarification on the connection between the gender of Dutch and Spanish. That is, I am assuming that common gender in Dutch is the equivalent to the masculine gender in Spanish (and neuter is equivalent to feminine). For this reason, I am also assuming that the common gender is the default in Dutch. This, however, needs to be clearer in the text. Perhaps this could be included in the section where grammatical gender is discussed.
• In Table 3 (page 9), in the first column, the author(s) wrote “Default gender (masc.)” and “Default gender (fem.)”. This is a bit confusing, for in Spanish masculine is the default and feminine is the marked gender. Furthermore, in several places in the text, the author(s) writes “common default strategy” (line 393) and “neuter default strategy” (line 394). Again, this is confusing as there should be only one default strategy. This is also seen in the first column of table 4. Another example is in line 754, where the author(s) wrote “the use of the feminine default was not attested”. This is confusing because the feminine is not the default. The issue should be addressed to make clear what is the default gender strategy in Dutch and in Spanish.
• In table 5 (page 10), “sombrero” and “gorro” were both accepted. I am not sure how, or if, it affects the results, but “gorra” (fem) is also a possible translation of “hat”.
• Line 43-43: There might be a typo in the sentence, “…there are no studies [regarding?] gender assignment…”
• Line 54-55: The sentence that begins with “Aalberse et al (2009)…” is difficult to parse.
• Line 96: Typo – “…also indicative of e gender…”
• Line 144: Typo – 3d -> 3^rd
• Line 228: The citation Balam el at (2021) is not in the references
• Line 230: The word “mesa” should be italicized
• Line 731: Missing closing parenthesis
• Line 886: There is an extra period at the end of the parenthesis

Author Response

Main comments:

• The paper could benefit from a more in-depth discussion on the participant’s use of the pre- and post-nominal adjectives. First, an explanation of what were the nouns that were used in the pre-nominal constructions. The author(s) show in Figure 11 that the participants used the masculine gender more frequently than the feminine gender. The author(s) concluded that the participants preferred the masculine when the adjective was pre-nominal. However, it is not clear how the author(s) arrived at this conclusion. There needs to be more of a discussion on the nouns that were used in the post-, pre-nominal and even the DP+Relative clause construction. That is, what nouns were used in these constructions? Did the inserted Dutch nouns have common or neuter gender? If they had common gender, then this could explain why the Spanish masculine was used. For example, let us consider “hamer”, which has common gender (and it is masculine in Spanish). If the Spanish masculine gender was used with “hamer” in all constructions (i.e., pre-, post-nominal, and relative clause), then that tells us that the participants are using either the masculine as default, or the analogical strategy. It does not really tell us much else. The same would be for a neuter noun like “huis” (the participants would use the feminine gender rather than the default). The interesting data would be if the participants are using the feminine in the n+adj structure and then used the masculine in the adj+n structure for the same noun. If this is the case, then perhaps there is evidence that having the adjective in pre-nominal position has an effect on gender assignment, thus it can be considered another gender assignment strategy. If the author(s) want to posit that the placement of the adjective has an effect on gender assignment, their analysis needs to be clearer, because I am not convinced that this is the case.

We thank the reviewer for raising this concern. First, we would like to point out that the gender assignment strategies discussed in this paper apply to individual participants, not to the group as a whole. We checked “if the participants are using the feminine in the n+adj structure and then used the masculine in the adj+n structure for the same noun”. Each participant only used each noun once. What we try to make clear with Figure 11 is that in pre-nominal constructions, the masculine gender is almost always assigned, even when the Spanish translation equivalent of the noun in question is not necessarily masculine (translation equivalent mismatch n=48). On the other hand, very few cases are assigned the feminine gender (n=6). These data suggest that the masculine gender is preferred in adj+n constructions, regardless of the gender of the translation equivalent.

The assumption that “If they (the Dutch nouns inserted into Spanish) had common gender, then this could explain why the Spanish masculine was used” is not correct, given that common gender does not equal masculine gender; it is a gender category containing words that used to have either masculine or feminine gender in Dutch, but in present day Dutch, this distinction is largely lost. So, common nouns would not be more likely to be assigned masculine gender in Spanish than neuter ones (or vice versa for masculine insertions being assigned common gender into Spanish). In any case, we did check whether there was an effect of the Dutch noun on the gender that was assigned and there was barely a difference: both common and neuter nouns were assigned masculine gender 83% of the time. Similarly, we checked whether the gender of the Spanish noun influenced the gender assigned in Dutch, and this did not appear to be the case: 63% of all feminine nouns vs. 70% of all masculine nouns was assigned common gender in Dutch. To be sure, we also included gender of the noun in the statistical models and it did not significantly improve the model in each case. This information is added in footnotes 9 and 11.

• While the majority of the methodology was well done (such as doing the monolingual trials before the code-switched trial), I would like the authors to address the lack of a proficiency test in both languages or the reason that the author(s) did not include one. The participants had varying profiles (with age ranges being very wide, and some leaving abroad for several years) which could have an effect on the results. I see that there are self-reported proficiency scores but only in Spanish and not in Dutch. The proficiency tests could have also been able to explain the use of the pre- or post-nominal adjective. That is, if the participants were more proficient in Dutch, then it makes sense that they would prefer to use the pre-nominal adjective. On the other hand, if the participants were more proficient in Spanish, then that explains the use of the post-nominal adjective.

We agree with the reviewer that a proficiency test could have been used to explain the outcomes of our research, especially when it comes to the differences in use of pre- or post-nominal adjectives by the participants. However, taking the duration of the procedure into account, we thought it would be best to limit the time that especially younger participants have to spend concentrating on the tasks. Moreover, previous research with heritage speakers of Spanish in the Netherlands (van Osch, 2019) has shown self-reports to correlate significantly with other measures of proficiency such as the DELE (Diploma Español de Lengua Extranjera) and lexical decision tasks. We have added a comment about this to the paper in footnote 5.

As the reviewer rightfully notices, we asked our participants to report their proficiency in Spanish but not in Dutch, as we assumed Dutch to be their dominant language. This may have been a wrong assumption, which is why this is something that we could have done differently.

Other comments:

• The paper needs a little more clarification on the connection between the gender of Dutch and Spanish. That is, I am assuming that common gender in Dutch is the equivalent to the masculine gender in Spanish (and neuter is equivalent to feminine). For this reason, I am also assuming that the common gender is the default in Dutch. This, however, needs to be clearer in the text. Perhaps this could be included in the section where grammatical gender is discussed.

We thank the reviewer for raising this concern. It was not our intention to suggest that common and masculine are the default genders, and that common gender in Dutch is equivalent to the masculine gender in Spanish (and neuter equivalent to feminine). Spanish and Dutch have two different gender systems, that only have the number of gender categories in common (2). Although common and masculine gender are often overgeneralized by bilinguals and used in a default strategy when code-switching, this is not necessarily the case, as for example studies on Spanish-Basque described in section 3.2. suggest a feminine default strategy. Moreover, we analysed gender assignment strategies for each individual participant separately, and the data show that while some participants may have a common default strategy, others may adopt a neuter default strategy.

• In Table 3 (page 9), in the first column, the author(s) wrote “Default gender (masc.)” and “Default gender (fem.)”. This is a bit confusing, for in Spanish masculine is the default and feminine is the marked gender. Furthermore, in several places in the text, the author(s) writes “common default strategy” (line 393) and “neuter default strategy” (line 394). Again, this is confusing as there should be only one default strategy. This is also seen in the first column of table 4. Another example is in line 754, where the author(s) wrote “the use of the feminine default was not attested”. This is confusing because the feminine is not the default. The issue should be addressed to make clear what is the default gender strategy in Dutch and in Spanish.

Thanks to the reviewer for this observation. We understand why this would be confusing. Our intention was to point out all the possible assignment strategies. We explain in our study that there is a tendency to overgeneralize the masculine gender in Spanish and common gender in Dutch. We also point out that this is not always the case, as studies on Spanish-Basque (section 3.2.), for example, suggest the use of a feminine default strategy. This means that we cannot say with certainty that there is a default gender used by every community or individual. As we studied code-switched structures with a new language pair (Spanish-Dutch), where no pattern of a default strategy is reported yet, we included all possible strategies and intended to explain these in section 5.1. Table 3 & 4. As we do understand the confusion, we decided to add that these strategies are hypothetical (Line 429 & 438).

• In table 5 (page 10), “sombrero” and “gorro” were both accepted. I am not sure how, or if, it affects the results, but “gorra” (fem) is also a possible translation of “hat”.

We thank the reviewer for noticing this. We did find the use of “gorra” in the data; three times in the unilingual Spanish mode, but not in the code-switched modes. We included these instances in the analysis, and treated them as feminine instead of masculine.

• Line 43-43: There might be a typo in the sentence, “…there are no studies [regarding?] gender assignment…”

We thank the reviewer for pointing out this mistake. We have added “regarding” to the sentence (line 46). 

• Line 54-55: The sentence that begins with “Aalberse et al (2009)…” is difficult to parse.

We agree with this observation. We have changed the sentence from “Aalberse et al. (2019) mention several dimensions along which definitions of heritage languages and speakers have been noted to differ in the literature...” to “Aalberse et al. (2019) notice several dimensions along which definitions of heritage languages and speakers differ in the literature...” (line 57-58).

• Line 96: Typo – “…also indicative of e gender…”

Thanks to the reviewer for this observation. We have corrected the typo.

• Line 144: Typo – 3d -> 3rd

Thanks to the reviewer for noticing. We have made the change.

• Line 228: The citation Balam el at (2021) is not in the references

We appreciate the attention paid to checking the references. Balam et al (2021) can be found in between Balam (2016) and Bellamy et al (2018) in the reference list (Line 1188-1190).

• Line 230: The word “mesa” should be italicized

This remark is correct, we have made the change.

• Line 731: Missing closing parenthesis

We thank the reviewer for noticing this, and we added the closing parenthesis.

• Line 886: There is an extra period at the end of the parenthesis

The reviewer is right, we removed the extra period.

Reviewer 2 Report

SUMMARY

This paper investigates gender production in Spanish-Dutch bilinguals in the Netherlands with Spanish as a heritage language. Data was gathered through a director-matcher task in four modes: two unilingual and two CS modes. Results are largely consistent with what has been reported for Spanish heritage speakers elsewhere and speakers of other heritage languages in the Netherlands. In the CS modes, evidence for both default and analogical gender agreement strategies is found, together with the newly attested strategy of uninflected adjectives in a non-default/ungrammatical position. The influence of extralinguistic factors on gender production was also investigated.

The paper is an extension of the literature on heritage speakers and gender assignment strategies in CS. While the paper could benefit from some clarifications, especially with regards to the statistical modelling, I feel the foundations of the paper merit publication.

MAJOR COMMENTS

1. In the discussion of Dutch agreement, the adjectival agreement is presented more straightforwardly than it actually is (not all adjectives exhibit agreement for example, see Broekhuis (2013), section 5.1), though I understand this is perhaps the information that is needed for the paper.

My main issue is that adjectival agreement and determiner agreement are presented as if they are equally salient/stable. The sensitivity to definiteness in the adjectival agreement for example is eroding, especially for frequent items. Many native speakers of Dutch (such as myself) would accept both “het grote huis” and “het groot huis” (though the distinction between common and neuter gender is maintained: *de groot tafel/*een groot tafel).

2. The statistical modeling is not adequately reported. Numerous times the “best fitting model” is mentioned, but it is not said what effects this model exactly contained and nor how fit was determined (r2? AIC(c)? BIC?). It is also unclear if the same model was used to investigate the linguistic and extralinguistic variables. And was only one model used for all modes, or a different model for each mode. 

3. Why was in the analysis of the CS mode the gender of the noun itself (i.e the Dutch noun in section 6.3.1 and the Spanish noun in section 6.3.3) not included as a dependent variable? Treffers-Daller’s work suggests that this may be relevant. Neuter Dutch nouns may also be more susceptible to masculine default strategies (see Klassen 2016)

MINOR COMMENTS (see also attached PDF)

line 115-116 The section also discusses heritage speakers in the Netherlands with other heritage languages

In the section on Spanish gender, words like mapa and puente are described as gender being “less predictable”. However, mapa is a case where gender is the opposite as what would be predicted, whereas the gender of puente is just not predictable

line 128: GMT is not explained and not sufficiently well-known/transparent to leave it unexplained such as GJT and WRT

line 157 Dutch predicative adjectives do not show agreement (Broekhuis 2013:326) — reading the sentence again, I realise that agreement of Spanish agreement is meant: clarify this on line 153 

line 178: Spanish should be English?

throughout section 3.1 it is not always made explicit which gender agreement system the discussed study reports on. While the reader can often guess, making your reader guess is never a good idea ;) — I recommend making it explicit (lines 153 / 190-191 / 195-196 / )

line 334 which one was the HL/L2?

line 340: the description of the “different analogical strategy” is confusing - isn’t it just “regular agreement”: masculine nouns get masculine agreement on determiners
This is explained slightly better in the following lines, but could still be improved 

line 353: I would not say there is no conclusion to be drawn about the astrategy for gender assignment here; rather that (it seems like) the default strategy is used when a translational equivalent is not available

Section 4 does not seem to have the correct title! (though besides the confusing title, this section is particularly clear)

line 407-408 Dutch does have morphophonological gender cues which may interfere with gendr assignment in CS. The (historical) suffix -ing for example derives feminine nouns, leading meeting to be feminine in Dutch, despite most English loans are masculine in Southern Dutch. Though for the nouns used in the stimuli, this is not a problem.  

line 430 dialectal variation in Dutch, Spanish or both? In fact the sentence that spans 429-431 can be deleted entirely, as the next sentences explain it again, but better.

line 571-572 - while there are two significant effects, there is a huge difference in effect size, which deserves a mention (same with lines 595-597)

figures 2-5: I personally find these redundant, but perhaps they are useful for the more visually inclined?

line 618 it si not made explicit through which test the significance was determined (though because the beta-value is reported, I suspect it is the output of the model).

line 618 - again, effect size (and in this case comparatively large standard error) deserves a mention

line 660 - that is average *reported* skill, right?

lines 668-670 - how exactly?

line 691 before the ( — a word seems to be missing

line 692 - do those numbers indicate gender of assigned gender or translational equivalent?

line 721-722 - unclear formulation (similarly to what was decribed for the monolingual modes)

section 6.3.3 doesn’t seem to have the right title

footnote 4 - it is unclear what “the construction” refers to 

line 819 - again, the size of this effect deserves a mention

lines 926 and 931 - there seems to be a contradition here - see attached pdf

References

Broekhuis, H. (2013). Syntax of Dutch: Adjectives and adjective phrases. Amsterdam University Press.

Klassen, R. (2016). Asymmetric grammatical gender systems in the bilingual mental lexicon (Doctoral dissertation, Université d'Ottawa/University of Ottawa).

Comments for author File: Comments.pdf

Author Response

1. In the discussion of Dutch agreement, the adjectival agreement is presented more straightforwardly than it actually is (not all adjectives exhibit agreement for example, see Broekhuis (2013), section 5.1), though I understand this is perhaps the information that is needed for the paper.

We thank the reviewer for this comment. A footnote (2) has been added pointing out that there is more complexity to Dutch adjective inflection, and providing references to more complete descriptions of adjective inflection.

My main issue is that adjectival agreement and determiner agreement are presented as if they are equally salient/stable. The sensitivity to definiteness in the adjectival agreement for example is eroding, especially for frequent items. Many native speakers of Dutch (such as myself) would accept both “het grote huis” and “het groot huis” (though the distinction between common and neuter gender is maintained: *de groot tafel/*een groot tafel).

We thank the reviewer for raising this issue. The authors - the majority of whom are native speakers of Dutch - were not aware of this variety, and have not been able to find sources describing this development. As described in the running text (line 647-649), “Of the 120 incorrect utterances, 51 were errors with just the determiner, 66 with just the adjective (all of which had an indefinite determiner), and 3 with both the determiner and the adjective.” If the suggestion made by the reviewer were applicable to our data, we would have expected a greater number of errors with the adjective with a definite determiner, of which there were none.

2. The statistical modeling is not adequately reported. Numerous times the “best fitting model” is mentioned, but it is not said what effects this model exactly contained and nor how fit was determined (r2? AIC(c)? BIC?). It is also unclear if the same model was used to investigate the linguistic and extralinguistic variables. And was only one model used for all modes, or a different model for each mode.

We thank the reviewer for pointing this out. Models were compared using the anova () function in R, which compares the chi squared value. This information has been added to the paper (line 572-574).  Moreover, different models were used for the different modes. This has been been made explicit in various parts of the paper (e.g. line 595, 626, 652, 661, 698, 762, 858, 878)

3. Why was in the analysis of the CS mode the gender of the noun itself (i.e the Dutch noun in section 6.3.1 and the Spanish noun in section 6.3.3) not included as a dependent variable? Treffers-Daller’s work suggests that this may be relevant. Neuter Dutch nouns may also be more susceptible to masculine default strategies (see Klassen 2016)

We thank the reviewer for this suggestion. We did not take into account the gender of the noun in the original language, given that the two gender systems – Spanish and Dutch – differ in terms of the gender values: while Spanish has masculine and feminine, Dutch has common and neuter. This is different from the situation reported by Klassen (2016), in which both German and Spanish have masculine and feminine gender, and where one thus might expect the gender of the word in the source language to effect gender assignment in the target language. In our case, we did not expect to find this effect. However, to be sure, we checked whether there was an effect of the Dutch noun on the gender that was assigned and there was barely a difference: both common and neuter nouns were assigned masculine gender 83% of the time. Similarly, we checked whether the gender of the Spanish noun influenced the gender assigned in Dutch, and this did not appear to be the case: 63% of all feminine nouns vs. 70% of all masculine nouns was assigned common gender in Dutch. We also included gender of the noun in the statistical models and it did not significantly improve the model in each case. This information is added in footnotes 9 and 11.

MINOR COMMENTS (see also attached PDF)

line 115-116 The section also discusses heritage speakers in the Netherlands with other heritage languages

We thank the reviewer for pointing this out. The sentence has been changed to include these heritage speakers.

In the section on Spanish gender, words like mapa and puente are described as gender being “less predictable”. However, mapa is a case where gender is the opposite as what would be predicted, whereas the gender of puente is just not predictable

We appreciate this remark. The examples have been more accurately described according to the reviewer’s suggestion: (...) there are nouns where the gender is the opposite of what would be predicted, such as masculine nouns ending in -a (e.g. el mapa “the.M map”), and feminine nouns ending in -o (e.g. la mano “the.F hand”), or unpredictable, such as nouns ending in other vowels or consonants (e.g. la nariz “the.F nose”, el puente “the.M bridge”).

 line 128: GMT is not explained and not sufficiently well-known/transparent to leave it unexplained such as GJT and WRT

We agree with the reviewer that this should have been explained better. We added an explanation to section 3.1. line 135-136 that says: “...in which the participants are asked to determine the gender of the target noun, masculine or feminine.” We hope that this is clearer.

line 157 Dutch predicative adjectives do not show agreement (Broekhuis 2013:326) — reading the sentence again, I realise that agreement of Spanish agreement is meant: clarify this on line 153

We agree with the reviewer that the wording may be unclear. We have specified that we are talking about Spanish gender agreement in line 162, and about Spanish heritage speakers in line 158.

line 178: Spanish should be English?

We thank the reviewer for noticing this, and we have changed the mistake.

throughout section 3.1 it is not always made explicit which gender agreement system the discussed study reports on. While the reader can often guess, making your reader guess is never a good idea ;) — I recommend making it explicit (lines 153 / 190-191 / 195-196 / )

We appreciate the reviewer’s suggestion to provide more clarity, and have added ‘with respect to Dutch gender’ in line 203, in addition to the changes suggested in a previous comment.

line 334 which one was the HL/L2?

Radford’s et al. (2007) study is about French-German bilinguals in Germany. This has been clarified in the text.

line 340: the description of the “different analogical strategy” is confusing - isn’t it just “regular agreement”: masculine nouns get masculine agreement on determiners

This is explained slightly better in the following lines, but could still be improved

The authors agree with the reviewer that the sentence was confusing, and changed the wording to: Treffers-Daller (1993), who studied 34 Dutch/French bilinguals in Brussels by analysing natural speech and elicited production data, found a strong preference for an analogical strategy based on the gender of the noun in the original language.

line 353: I would not say there is no conclusion to be drawn about the astrategy for gender assignment here; rather that (it seems like) the default strategy is used when a translational equivalent is not available

We thank the reviewer for this suggestion. We have removed the phrase, and added the following sentence: In conclusion, these bilinguals resorted to a common default strategy when inserting Moroccan Arabic nouns without a clear translation equivalent.

Section 4 does not seem to have the correct title! (though besides the confusing title, this section is particularly clear)

We thank the reviewer for noticing this mistake and we have corrected it.

line 407-408 Dutch does have morphophonological gender cues which may interfere with gendr assignment in CS. The (historical) suffix -ing for example derives feminine nouns, leading meeting to be feminine in Dutch, despite most English loans are masculine in Southern Dutch. Though for the nouns used in the stimuli, this is not a problem. 

We thank the reviewer for this comment and the interesting example provided. The authors are aware of the discussion about the ending/suffix -ing, which also behaves irregularly with respect to plural formation, and it is for this reason that we avoided it in the stimuli.

line 430 dialectal variation in Dutch, Spanish or both? In fact the sentence that spans 429-431 can be deleted entirely, as the next sentences explain it again, but better.

We agree with the reviewer that the next sentence explains it better, which is why we decided to remove “dialectal variation” from Line 452.

line 571-572 - while there are two significant effects, there is a huge difference in effect size, which deserves a mention (same with lines 595-597)

We thank the reviewer for pointing this out, and we agree that it should be mentioned. We included two footnotes (6&7) on this.

figures 2-5: I personally find these redundant, but perhaps they are useful for the more visually inclined?

We thank the reviewer for this input. Nevertheless, the authors decided to keep the figures.

line 618 it si not made explicit through which test the significance was determined (though because the beta-value is reported, I suspect it is the output of the model).

Thank you for pointing this out. Indeed, it is the main effect in the output of the model, which has now been clarified in the text.

line 618 - again, effect size (and in this case comparatively large standard error) deserves a mention

We thank the reviewer for pointing this out. We have added a footnote (footnote 8) on the relatively large standard error.

line 660 - that is average *reported* skill, right?

We thank the reviewer for pointing out the inconsistency and added the word ‘reported’.

lines 668-670 - how exactly?

By means of a Pearson correlation analysis. This has been added to the paper.

line 691 before the ( — a word seems to be missing

Thank you, we have changed this sentence.

line 692 - do those numbers indicate gender of assigned gender or translational equivalent?

We thank the reviewer for pointing this out, it is indeed unclear. We are referring to the assigned gender. This has been added to the text.

line 721-722 - unclear formulation (similarly to what was decribed for the monolingual modes)

Thank you for pointing this out. We now described the analysis more precisely in the following way: “An additional analysis was performed on a subset of the data which contained only the mismatch cases, to check whether there were significantly more mismatch cases for masculine gender than for feminine gender. A model was used on these data, with the assigned gender as the dependent variable, and without fixed effects, only random effects. The intercept of this model was significant  (β=14.49, SE=2.82, z=5.15, p<.001), indicating that, within this subset, the masculine gender was assigned more often than the feminine gender, thus confirming the default masculine strategy.”

section 6.3.3 doesn’t seem to have the right title

We thank the reviewer for noticing this, and have changed the title.

footnote 4 - it is unclear what “the construction” refers to

We appreciate the reviewer bringing up this unclarity. The word ‘construction’ has been replaced with ‘adjective’.

line 819 - again, the size of this effect deserves a mention

We thank the reviewer for pointing this out and we added a footnote on this (footnote 12)

lines 926 and 931 - there seems to be a contradiction here - see attached pdf

We thank the reviewer for this comment. The section only deals with unilingual speech, and concludes that although there is a positive correlation between self-reported Spanish

Reviewer 3 Report

The authors appropriately link to a range of papers from bilingual acquisition, and the methods are thorough and appropriate.  Framing of age and acquisitional questions are relevant but not entirely clear.

Specifically, there are two issues and one area where the authors could have (and perhaps should in future papers) explore further.

Use of "differential acquisition" (paragraph starting on line 132)

The use of the term “differential” here is too broad and obscures the predictions somewhat.  Two types of claims are made in this paragraph:  first, that heritage speakers were less accurate than monolinguals at all ages; and second, that with age, the accuracy drops (is this differential acquisition or something else?).  Since age of acquisition is studied as a covariate, the predictions need to be crystal clear.  Are the authors predicting a change over time or not?

Characterization of the age variable

For the most part, the authors were careful about the way they framed age of onset of bilingualism (discussion at the end of the intro is accurate, but see comment above with regards to the section 2).  However, the interpretation of the results with regards to age, such as in the conclusion, is a bit weak.  The authors refer to the later bilinguals as "late arrivals".  This characterization has certain flaws:

1. The term "late arrival" is already used in L2 research for adult or teenage arrivals.
2. No interpretation of the age-related factors from section 2 are explored here.
3. It's unclear why the authors refer to the variable as if it were nominal (early vs. late) when they have coded it as continuous.

(To explore) More fine-grained analysis of bilingual grammar

It exceeds this paper slightly, but fruitful explorations of bilingual grammar have been made (see for example López, 2020; full citation below) that this type of analysis could benefit from.  There's reason to believe that seemingly directional code-switching of this type is actually more like borrowing.

López, L. (2020). Bilingual grammar: Toward an integrated model. Cambridge University Press.

Author Response

Use of "differential acquisition" (paragraph starting on line 132)

The use of the term “differential” here is too broad and obscures the predictions somewhat.  Two types of claims are made in this paragraph:  first, that heritage speakers were less accurate than monolinguals at all ages; and second, that with age, the accuracy drops (is this differential acquisition or something else?). 

We thank the reviewer for this comment. We had initially used the term “differential acquisition” to include any type of development that is different from the typical monolingual developmental path, including the possibility of loss of knowledge. However, we agree that the term can be confusing, which is why we have removed this term everywhere, and have replaced it with the following:

• “Several studies involving child heritage speakers of Spanish in the US also report differences in the development of gender in heritage Spanish as compared to monolingual acquisition.” (line 139-140)
• “we hypothesize that the heritage speakers of Spanish in this study will deviate from what is typically reported for the acquisition of gender in monolingual Spanish” (line 389-391)

Since age of acquisition is studied as a covariate, the predictions need to be crystal clear.  Are the authors predicting a change over time or not?

We thank the reviewer for this suggestion, we have thought about this. However, we think our data do not allow to test any prediction about a change over time. Testing for a change over time would require including ‘age at testing’ as a variable in the statistical models. However, ‘age at testing’, ‘length of residence’ and ‘age of onset’ are three variables that are directly related to each other, as the length of residence is the difference between the age at testing and the age of onset. Effectively, both ‘length of residence’ and ‘age of onset’ correlated statistically with ‘age at testing’. Therefore, we decided to include length of residence and age of onset in the models, rather than age at testing. We have, however, included a comment about the relation between these three variables in the discussion section (line 1027-1033)

Characterization of the age variable

For the most part, the authors were careful about the way they framed age of onset of bilingualism (discussion at the end of the intro is accurate, but see comment above with regards to the section 2).  However, the interpretation of the results with regards to age, such as in the conclusion, is a bit weak.  The authors refer to the later bilinguals as "late arrivals".  This characterization has certain flaws:

1. The term "late arrival" is already used in L2 research for adult or teenage arrivals.
2. No interpretation of the age-related factors from section 2 are explored here.
3. It's unclear why the authors refer to the variable as if it were nominal (early vs. late) when they have coded it as continuous.

We thank the reviewer for this comment. It is true that age of arrival was included as a continuous variable in this study, it was not our intention to suggest it was used as a nominal variable. Therefore, we changed “late arrival” to “a later arrival” in line 1125, and we deleted the following sentences from section 7:

• “In fact, the 5 lowest scoring participants were all born in the Netherlands (Ages at testing: 8, 9, 10, 13, and 15 years old), and of the 5 latest arrivers (between age 8 and 12), 4 obtained scores above 97% (Ages at testing: 15, 15, 20 and 28 years old). However, there were also many participants with scores in that high range who were either born in the Netherlands or arrived before age 4.” (was placed on line 1007)
• - “Most of the early arrivers (i.e. before age 5) scored above 80% accuracy in Dutch, while those who arrived after age 5, tend to reach an accuracy of around 55%”. (was placed on line 1026)

As for an interpretation of the age-related factors: for Spanish, age of arrival was not found to have an effect, and we take this as an argument to view age of arrival as a gradual variable, rather than a way to distinguish between two different categories of bilinguals: child L2 vs. heritage (line 1008-1014 ). For Dutch, there was an effect of age of arrival, but as mentioned above, this variable is directly related to the variables length of residence and age at testing. It is thus difficult to draw strong conclusions from this effect, and this is emphasized now a bit more in the discussion section (line 1027-1033, see also our earlier response to the reviewers remark about a prediction w.r.t. change over time)

(To explore) More fine-grained analysis of bilingual grammar

It exceeds this paper slightly, but fruitful explorations of bilingual grammar have been made (see for example López, 2020; full citation below) that this type of analysis could benefit from.  There's reason to believe that seemingly directional code-switching of this type is actually more like borrowing.

López, L. (2020). Bilingual grammar: Toward an integrated model. Cambridge University Press.

Thanks for this comment. Indeed, López (2020) does not make a distinction between code-switches and borrowings in his integrated model. While some linguists agree that “code-switching and borrowing are two distinct phenomena” (Poplack & Meechan, 1998:132), others contend that
“code-switching and borrowing fall on a continuum” (Myers-Scotton, 1993:176). Yet others argue that there is no distinction between the two processes (López, 2018). Our focus in this paper is on gender assignment strategies, both in unilingual and bilingual grammars (see also chapter 5 in López 2020). The doubt about the classification of other-language materials falls beyond the scope of this paper.

Round 2

Reviewer 1 Report

Author(s) has addressed the concerns that I expressed in my first review.

Author Response

Thanks for the useful feedback.