The authors were very responsive to the concerns raised by myself and the other reviewers. Based on a reviewer comment, the paper now includes a criterion that fathers must have read to their child in the past two weeks. I am a bit concerned that this might exclude too many fathers, although the authors do say that they have previous data suggesting that most fathers read to their children. If indeed this criterion proves overly strict, the authors might justifiably deviate from the pre-registration, assuming that this decision is made indepentent of results. Overall, I look forward to seeing the results of the study once it is complete.

Very small wording suggestions:

p. 20, tranlsation of item "Parents may learn babies to talk by talking with them" could be double checked - might be more apt to say "Parents may teach babies to talk by talking to them"

p. 20, "reverted" --> "reversed"?

I appreciate the authors’ responses to prior feedback. Below I outline some remaining concerns/suggestions:

1.     I still find the paragraph (beginning line 286) a bit oddly structured. There seem to be 3 themes – characteristics of Norwegian IDS, cross-cultural/linguistic variation in IDS, and whether vowel space hyperarticulation occurs in IDS. These topics are intersecting but separable and the paragraph in its current form still seems to jump from one to the other.

2.     “it is not fully known whether fathers modulate their IDS when speaking with a child” (line 468). I still feel like this overplays the extent to which we do not know about this. At this point there is robust evidence that fathers do engage in IDS – although there is plenty still to debate and discover with respect to how much relative to mothers and/or differences in quality/kind.

3.     I am not sure “apply[ing] a conservative approach while interpreting the results” (line 504) solves the problem of the possibility that there is not enough reliability in comparing across individuals to be able to pull out the desired results (particularly with respect to the effect of infant experience with father caregiving). I think this is a really interesting and valuable study, but I still worry that there is a strong possibility that this result will be null and given what we are learning about the lack of test-retest reliability in measures of infant preference, that won’t be interpretable (i.e. while the group level preference for IDS may be robustly measurable, individual differences may not be). The authors state that they will report the ICC, but more information on how they will interpret this measure would be helpful.

4.     Small comment: “Parents may _learn_ babies to talk…”  (line 629) I assume “teach” is meant?

5.     More information is needed about how the stimuli for the infant preference study will be created. Who will be recorded? Will the exact ManyBabies procedure (i.e. pulling objects out of a bag, etc.) be followed? Etc.

6.     “each utterance contains 8 words” (line 722) – this is not consistent with the ManyBabies protocol. Please explain?

7.     “The recordings in IDS and ADS” (line 743) I assume this is referring to the father’s recordings, not the experimental stimuli, but perhaps best to explicitly state that for clarity.

8.     I do not see where it is described how the vowel triangle measure will be calculated from the individual formant information. Since it is mentioned that F3 will be measured, will this be included in the analysis in some form, or just F1/F2? Typically vowel space is calculated as the area between 3 point vowels across F1 and F2, but this is not explicitly stated anywhere and there are other details that need to be articulated. [this may have been an issue in the original submission as well, apologies for not bringing this up earlier if so]

9.     In the exploratory analysis, Paternal attitudes and Paternal reading practices were introduced to the model as main effects. Shouldn’t these be interaction effects with register, as with paternity leave duration? I.e. the prediction would be that they would increase the difference between IDS and ADS in the various acoustic measures, not that they would increase overall f0 etc.?

​​​

​​​

Overall, this is a very strong Stage 1 registered report of a study investigating paternal and male infant-directed speech.  As the authors point out, there has been insufficient work on this topic, and the unique parental leave policies in Norway provide an interesting test of several important theoretical questions. The introduction outlines the relevant literature and clearly articulates study hypotheses.  The methods are strong, and have been used in previous high-quality work.  The inclusion of both acoustic measures of fathers’ speech and infant behavioral measures is another strength. In reading the introduction several times I worried about birth order effects, but was reassured when the methods specified that all children would be first-born. It might be useful to mention this design detail earlier in the paper. The authors have provided power analyses, and the study is generally adequately powered.  The planned analyses are appropriate to answer the research question, and the document has specified inclusion/exclusion criteria, plans if data are non-normal etc.

My main concern is with the individual differences analysis, examining whether the duration of paternity leave (presumably a proxy for infant exposure to paternal IDS) predicts infants’ preference for IDS. The authors mention that there “is no known effect size of the interaction between the IDS/AADS differences and the duration of paternity leave”, and this is true. However, one issue that has not been considered here is measurement reliability of the infant task – how stable are individual differences as measured by the IDS-ADS preference task?

In our recent paper (Byers-Heinlein, Bergmann, & Savalei, 2022), we calculate the internal consistency of the preference as reported in the ManyBabies Consortium (2020) paper to be .14, and it is important for the power calculations to account for attenuation of correlation due to unreliability. The observed correlation can be calculated from reliabilities based Spearman’s (1904) formula robserved = rtrue sqrt(rxx * ryy) . Assuming the best case scenario that infants’ exposure to paternal IDS is measured with perfect reliability by duration of paternity leave (itself a questionable assumption), if the true relationship between leave duration and preference is perfect (rtrue  =1), then robserved = .36; if  rtrue  = .7,then  robserved = .26; if If  rtrue  = .3,then  robserved = .11.  The authors will definitely want to run their own calculations and apply them to their linear mixed regression approach, however, I suspect that at any reasonable sample size of infants (besides something on the order of ManyBabies) it will not be possible to detect individual differences as measured by this looking time task, given its low reliability.  It is certainly an interesting research question and a plausible hypothesis, but without a more reliable infant task, it may not be possible to answer.

Byers-Heinlein, K., Bergmann, C., & Savalei, V. (2022). Six solutions for more reliable infant research. Infant and Child Development, e2296. http://doi.org/10.1002/icd.2296