Very Young Children Learning German Notice the Incorrect Syllable Stress of Words
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsIn this paper, the authors test German learning infants and toddlers on ability to recognize familiar disyllabic words when they are produced with the correct and incorrect stress. The study is well powered with 69 infants and 28 toddlers (though see the comment below on the attrition). Overall, they find minimal evidence that infants attend to stress: infants do not recognize items based on non-parametric tests; only in the Growth Curve Analysis is there some evidence that infants orient faster when stress is on the first syllable. Toddlers on the other hand, consistently use stress and also orient faster and more to trochaic targets.
The question whether infants use stress to recognize words (and thus represent them) is interesting. The methods are sound. There are however some substantive issues that make the claims in the paper not compelling.
First, the statistical analysis choices, seem unmotivated. What is the rationale behind the by-participant and by-item analyses? Why are the infant and toddler data combined in the mixed effect model, when they are separated for every other analysis? They should be separated here as well. In fact, I was surprised that the results are not reported separately for infants and for toddlers. In fact, I suspect there is no effect of stress in the infant data at all. That effect is being carried by the toddlers. This needs to be made clear (the absence of the interaction frankly is not compelling). In fact, the only evidence that 4- to 14-month-olds may be sensitive to stress comes from the GCA – and the time course of looking for infants looks nothing like the usual time course data (for e.g, in the toddlers) – so what is going on with the data?
Second, it is also unclear is what the input in German is like for infants, and who the stimuli align with how stress is produced in German. How is stress acoustically realized in German? How does it relate to the stimulus items produced here? Similarly, what is the distribution of stress in lexical items in German. And how does it contrast with English, only because the results are being compared to Campbell et al’s findings on English learning infants. Some of the latter is presented in the General Discussion, but would be better at the beginning.
Third, the attrition in the infant experiment is ~40% and in the toddler experiment is about ~33%. This seems very high. Minimally the exclusion rate in this experiment should be compared with previously published research.
Finally, what is the rationale to predict that infants would not look at the target when presented the incorrectly stressed label? It is a less good, but definitely not a match for the other word. And infants and toddlers do look at the target object when listening to mispronunciations. How is this different or similar to that?
Other comments
- Fixed picture pairs – referred to as ‘yoked’
- How many of the items have stressed minimal pairs in German? This seems pertinent to determine if the kids ever hear minimal pairs.
- It is very strange to have English word bank data in Table 1. Why are they presented? How are they relevant to explain the experimental results in German infants?
- Please present German target words – they are not mentioned anywhere in the manuscript (for e.g., in Table 1).
- Please present German instructions instead of English translations like “Look at the”. This should be done in the text as well as in Figure 3. Without the German frames it’s hard to see the phonological contexts in which targets were represented. The English translations could be placed in parentheses for readers.
- Figure 5: has missing dots (each column should have 4; car-plate has 2 and cup-glasses only has 3). If there is a reason for this, it should be mentioned in the figure caption.
Missing or incomplete reference
Marimon, M., Langus, A. & Höhle, B. (2024). Infancy, 29(5), 750–770. https://doi.org/10.1111/infa.12593
Author Response
COMMENTS 1: In this paper, the authors test German learning infants and toddlers on ability to recognize familiar disyllabic words when they are produced with the correct and incorrect stress. The study is well powered with 69 infants and 28 toddlers (though see the comment below on the attrition). Overall, they find minimal evidence that infants attend to stress: infants do not recognize items based on non-parametric tests; only in the Growth Curve Analysis is there some evidence that infants orient faster when stress is on the first syllable. Toddlers on the other hand, consistently use stress and also orient faster and more to trochaic targets.
The question whether infants use stress to recognize words (and thus represent them) is interesting. The methods are sound. There are however some substantive issues that make the claims in the paper not compelling.
RESPONSE 1: We would like to thank the reviewer for the valuable comments on our manuscript, which helped us to improve it further. According to the issues raised, we rewrote and restructured large parts of the introduction and discussion.
COMMENTS 2: First, the statistical analysis choices, seem unmotivated. What is the rationale behind the by-participant and by-item analyses? Why are the infant and toddler data combined in the mixed effect model, when they are separated for every other analysis? They should be separated here as well. In fact, I was surprised that the results are not reported separately for infants and for toddlers. In fact, I suspect there is no effect of stress in the infant data at all. That effect is being carried by the toddlers. This needs to be made clear (the absence of the interaction frankly is not compelling). In fact, the only evidence that 4- to 14-month-olds may be sensitive to stress comes from the GCA and the time course of looking for infants looks nothing like the usual time course data (for e.g, in the toddlers) so what is going on with the data?
RESPONSE 2: We would like to thank the reviewer for these comments and tried to justify our statistical analyses in more detail and adjust them accordingly. As normality distribution was violated for some subtests (according to visual plots and Shapiro tests, see R-script in osf), we first conducted Wilcoxon-tests to test general word comprehension in infants and toddlers. As is common in psycholinguistic research with adults, we tested whether the effects were robust for participants and items by conducting separate analyses by participants and items. As we also intended to test for age effects, we run mixed effect models. In the revised version, we added single mixed effect models for each group. Indeed, there was no stress effect in the infants (see p. 12., l.497 (footnote) and Table S1 and S2 in the supplementary material). We are now discussing the results even more cautiously (see restructured discussion), and we also reformulated (to strong) conclusions in the abstract p.1 and in the title p.1. Furthermore, we added more explanatory sentences to the results of the GCA in infants, p.12 l. 511-513.
COMMENTS 3: Second, it is also unclear is what the input in German is like for infants, and who the stimuli align with how stress is produced in German. How does it relate to the stimulus items produced here? Similarly, what is the distribution of stress in lexical items in German. And how does it contrast with English, only because the results are being compared to Campbell et al s findings on English learning infants. Some of the latter is presented in the General Discussion, but would be better at the beginning.
RESPONSE 3: We realized that this important information is missing in the original version – in particular in the introduction. In the revised ms., we added a new paragraph on statistical regularities of suprasegmental cues in German as compared to English and other languages (see p.2, l .58-76) and we also discuss these issues in the discussion in more depts (see p. 19, l. 775-780).
COMMENTS 4: Third, the attrition in the infant experiment is ~40% and in the toddler experiment is about ~33%. This seems very high. Minimally the exclusion rate in this experiment should be compared with previously published research.
RESPONSE 4: We agree that an attrition rate of 33 to 40 % is high, though it lies in the wide variation of attrition rates reported across infant studies (if they are specified). High attrition rates depend on several factors like fussiness or distraction of the infants and related inclusion criteria for minimum data points. In a systematic review, Slaughter and Suddendorf (2007) reported an average attrition rate of 13.7% in infant studies using visual paradigms, with a range from 0 to 62 %. The authors also reported that experimental outcomes (i.e. effects and their sizes) did not correlate with infant attrition due to fussiness or to any other factor. Therefore, they concluded that differential exclusion rates do not systematically influence experimental outcomes. On the contrary, if ambiguous data are excluded (in particular very short fixation times of the infants), this can even lead to larger effect sizes (Byers-Heinlein et al., 2022).
We would like to mention that there are also studies with even higher attrition rates compared to our study. For example, Bergelson et al. (2017) tested 108 infants between 12-20 months with eye tracking and only 46 infants went into the final sample (58% attrition rate). Similar, Bánki et al. (2022) assessed 4-6-month-olds and reported attrition rates of 60% when recorded in the lab and 53% when recorded online. Testing 3-to-9-month-olds Frank et al. (2009) reported attrition rates of 48%. There are certainly studies with fewer exclusions, but the example studies also show that the exclusion rate in our study is not exceptionally high.
We added a note on attrition rates on p.6 l. 265-269 of the revised ms.
COMMENTS 5: Finally, what is the rationale to predict that infants would not look at the target when presented the incorrectly stressed label? It is a less good, but definitely not a match for the other word. And infants and toddlers do look at the target object when listening to mispronunciations. How is this different or similar to that?
RESPONSE 5: Thank you for pointing out the relation of our study on misplaced stress to studies on mispronunciation at the segmental level. We added a paragraph relating to mispronunciation studies in the introduction, p.3, l. 141-155. Finally, we also referred to a recent mispronunciation study investigating very young infants (as we did) in the discussion at p. 16., l. 627-629.
Other comments
COMMENTS 6: Fixed picture pairs referred to as yoked
RESPONSE 6: Thank you for pointing this out, we have corrected it.
COMMENTS 7: How many of the items have stressed minimal pairs in German? This seems pertinent to determine if the kids ever hear minimal pairs.
RESPONSE 7: None of the target words has minimal stress pairs. We presented the canonical form (initially stressed) and a word pronounced with incorrect stress (stress on the second syllable). p.4. l.181-186.
COMMENTS 8: It is very strange to have English wordbank data in Table 1. Why are they presented? How are they relevant to explain the experimental results in German infants?
RESPONSE 8: We added more explanation to that decision within Table1 on p. 9 and clarify the use of the samples in more detail in the text p. 8 l.339-344.
COMMENTS 9: Please present German target words they are not mentioned anywhere in the manuscript (for e.g., in Table 1).
RESPONSE 9: We added the German words to Table 1 and refer to Table 1 with the complete list of German target words used in the experiment in the beginning of the Stimuli section. p.6 l. 285, see also Table1 on p.9.
COMMENTS 10: Please present German instructions instead of English translations like Look at the . This should be done in the text as well as in Figure 3. Without the German frames it s hard to see the phonological contexts in which targets were represented. The English translations could be placed in parentheses for readers.
RESPONSE 10: We replaced Figure 3, p.10. and inserted the desired information to the text, p. 9 l. 374-376.
COMMENTS 11: Figure 5: has missing dots (each column should have 4; car-plate has 2 and cup-glasses only has 3). If there is a reason for this, it should be mentioned in the figure caption.
RESPONSE 11: Thank you for bringing this to our attention, we have now scattered the points as they previously overlapped, see Figure5 on p.15.
COMMENTS 12: Missing or incomplete reference
Marimon, M., Langus, A. & H hle, B. (2024). Infancy, 29(5), 750 770. https://doi.org/10.1111/infa.12593
RESPONSE 12: Thank you for pointing this out, we have added the missing title.
REFERENCES
Bánki, A., de Eccher, M., Falschlehner, L., Hoehl, S., & Markova, G. (2022). Comparing Online Webcam- and Laboratory-Based Eye-Tracking for the Assessment of Infants’ Audio-Visual Synchrony Perception [Original Research]. Frontiers in Psychology, Volume 12 - 2021. https://doi.org/10.3389/fpsyg.2021.733933
Bergelson, E., & Aslin, R. N. (2017). Semantic Specificity in One-Year-Olds' Word Comprehension. Language learning and development : the official journal of the Society for Language Development, 13(4), 481-501. https://doi.org/10.1080/15475441.2017.1324308
Byers-Heinlein, K., Bergmann, C., & Savalei, V. (2022). Six solutions for more reliable infant research. Infant and Child Development, 31(5), e2296. https://doi.org/https://doi.org/10.1002/icd.2296
Frank, M. C., Vul, E., & Johnson, S. P. (2009). Development of infants' attention to faces during the first year. Cognition: International Journal of Cognitive Science, 110(2), 160-170. https://doi.org/10.1016/j.cognition.2008.11.010
Slaughter, V., & Suddendorf, T. (2007). Participant loss due to “fussiness” in infant visual paradigms: A review of the last 20 years. Infant Behavior and Development, 30(3), 505-514. https://doi.org/https://doi.org/10.1016/j.infbeh.2006.12.006
Reviewer 2 Report
Comments and Suggestions for AuthorsThis study explores the role of stress patterns in early word recognition in infants and toddlers. An eye-tracking investigation revealed that infants aged 4-5 months demonstrate some sensitivity to stress violations, although their word recognition abilities remain fragile. In contrast, toddlers exhibited robust word comprehension accompanied by a distinct and consistent stress effect. These findings indicate that German-speaking children utilize stress cues to build word form representations from an early age, contributing to our understanding of early word processing.
The manuscript is well written, and the study demonstrates careful design and thorough consideration. The eye-tracking method is implemented effectively, and the statistical analysis is appropriate and clearly presented. I have no fundamental criticisms of this interesting work, only a few follow-up questions and comments.
- Participants: The study references prior work by Campbell et al., yet the current study employs age groups that differ from the former (younger infants and older toddlers). What were the specific reasons for selecting these particular age ranges (4-15 months and 2-4 years) for the present study?
- Method: It is stated twice that the infants “already know the words used in our experiment” (p. 3). How can the authors be sure that every single child knows every word? Although Table 1 shows that the words are likely to be understood by many infants, it is not evident that all the words are known by all the participants.
- Item difficulties in Table 1: What exactly do the numbers show, the proportion of children understanding/using the particular word? Parental Questionnaire: do the numbers in table 1 represent the parents’ reports of their child’s understanding or use of the word? Why does Table 1 show Wordbank results for English, but not for German? The items are also part of the German data available in Wordbank. Table 1 needs more explanation.
- The Proportion Index (PI) is introduced in section 2.1, but only explained later in section 2.4. An earlier explanation would be helpful for understanding the analysis.
- Stimuli: to create items with stress violations, the speaker stressed the second syllable of trochaic words that are correctly stressed on the first syllable. Stressed second syllables were articulated longer and with more intensity. The examples in section 2.2.2 are “Auto” and “Baby”, i.e. words with a full vowel. Full vowels can easily be lengthened and intensified. However, all other items in the list contain reduced syllables with schwa (e.g. Nase), low schwa (e.g. Finger) or syllabic consonants (e.g. Vogel). Reduced vowels cannot be stressed without changing the vowel category (e.g. from /ə/ to /ɛ/ or from /ɐ/ to /a/). How did the speaker pronounce these items? All items containing final syllables with reduced vowels must have sounded really strange and unnatural in the condition with incorrect stress. How did the authors deal with this problem?
- Stimuli: The items were presented in pairs. Four of the 14 pairs fell into the same semantic category, while the other item pairs belonged to different categories. This adds a potential confounding factor to the item set that may influence the results. Indeed, figure 5 shows that some of the related pairs were more challenging than the non-related pairs. So, it turns out that the design of the material was not optimal. However, the authors seem to be aware of this limitation and address it on page 11. They argue that the semantic relatedness of the items did influence the results, but did not dramatically change the key outcomes. Nevertheless, the impact of this factor warrants further consideration in future research.
Minor comments:
- Line 204: cup-glasses, cup-tractor: second “cup” should be “mug”?
- Line 652: reference Marimon et al. 2024: Title is missing
Author Response
COMMENTS 1: This study explores the role of stress patterns in early word recognition in infants and toddlers. An eye-tracking investigation revealed that infants aged 4-5 months demonstrate some sensitivity to stress violations, although their word recognition abilities remain fragile. In contrast, toddlers exhibited robust word comprehension accompanied by a distinct and consistent stress effect. These findings indicate that German-speaking children utilize stress cues to build word form representations from an early age, contributing to our understanding of early word processing.
The manuscript is well written, and the study demonstrates careful design and thorough consideration. The eye-tracking method is implemented effectively, and the statistical analysis is appropriate and clearly presented. I have no fundamental criticisms of this interesting work, only a few follow-up questions and comments.
RESPONSE 1: We would like to thank the reviewer for their appreciative evaluation of our manuscript and the valuable comments on its further development.
COMMENTS 2: Participants: The study references prior work by Campbell et al., yet the current study employs age groups that differ from the former (younger infants and older toddlers). What were the specific reasons for selecting these particular age ranges (4-15 months and 2-4 years) for the present study?
RESPONSE 2: The focus of our study was on the function of syllable stress for first word meaning mapping in infants. That is, we intended to test whether infants, who are just learning their first words, are already processing syllable stress information when they relate nouns to respective objects. We compared the data of these infants with that of older toddlers, who (as a control) should already have stable word form representations of the tested words. We tested even older toddlers than Campbell (17-month-olds) to be sure that the toddlers knew all the tested words. We try to make this clearer on p. 5 l. 212-223 of the revised ms.
COMMENTS 3: Method: It is stated twice that the infants “already know the words used in our experiment” (p. 3). How can the authors be sure that every single child knows every word? Although Table 1 shows that the words are likely to be understood by many infants, it is not evident that all the words are known by all the participants.
RESPONSE 3: We agree that it is not evident that all the words are known by all infants, and we hope that this becomes even cleared in the revised manuscript. We rephrased several parts of the ms. including p.4 l. 180 -181, where we refer only to the toddlers (and not to the infants) “…who already should know all the words used in the experiment…”, and p.4 l.176-180, p.5 l. 219 and p.16 l.620-638, where we try to make it clear that we do not assume that all infants already know all the words.
COMMENTS 4: Item difficulties in Table 1: What exactly do the numbers show, the proportion of children understanding/using the particular word? Parental Questionnaire: do the numbers in table 1 represent the parents’ reports of their child’s understanding or use of the word? Why does Table 1 show Wordbank results for English, but not for German? The items are also part of the German data available in Wordbank. Table 1 needs more explanation.
RESPONSE 4: We added more explanation to that decision within Table1 on p. 9 and clarify the use of the samples in more detail in the text p.8 l.339-344.
COMMENTS 5: The Proportion Index (PI) is introduced in section 2.1, but only explained later in section 2.4. An earlier explanation would be helpful for understanding the analysis.
RESPONSE 5: We are very happy to implement this important information to optimize the text and include this information earlier in the text on p.4 l.203-208.
COMMENTS 6: Stimuli: to create items with stress violations, the speaker stressed the second syllable of trochaic words that are correctly stressed on the first syllable. Stressed second syllables were articulated longer and with more intensity. The examples in section 2.2.2 are “Auto” and “Baby”, i.e. words with a full vowel. Full vowels can easily be lengthened and intensified. However, all other items in the list contain reduced syllables with schwa (e.g. Nase), low schwa (e.g. Finger) or syllabic consonants (e.g. Vogel). Reduced vowels cannot be stressed without changing the vowel category (e.g. from /ə/ to /ɛ/ or from /ɐ/ to /a/). How did the speaker pronounce these items? All items containing final syllables with reduced vowels must have sounded really strange and unnatural in the condition with incorrect stress. How did the authors deal with this problem?
RESPONSE 6: Thank you for pointing out this important issue. We refer to that issue on p.6 l.289-301 and also discuss it on p.18 l.764-774. In addition, next to the example of “Auto” (car) in Figure 2 on p.7 we added an example of a word ending with schwa (“Puppe”, doll).
COMMENTS 7: Stimuli: The items were presented in pairs. Four of the 14 pairs fell into the same semantic category, while the other item pairs belonged to different categories. This adds a potential confounding factor to the item set that may influence the results. Indeed, figure 5 shows that some of the related pairs were more challenging than the non-related pairs. So, it turns out that the design of the material was not optimal. However, the authors seem to be aware of this limitation and address it on page 11. They argue that the semantic relatedness of the items did influence the results, but did not dramatically change the key outcomes. Nevertheless, the impact of this factor warrants further consideration in future research.
RESPONSE 7: Thank you for the comment. Yes, that's roughly how we put it in the (restructured) discussion on p.14 l. 550-576. However, we are not sure whether we should adapt something or elaborate more on this topic?
COMMENTS 8: Minor comments:
Line 204: cup-glasses, cup-tractor: second “cup” should be “mug”?
RESPONSE 8: Thank you for pointing this out, we corrected the typo.
COMMENTS 9: Line 652: reference Marimon et al. 2024: Title is missing
RESPONSE 9: Thank you for pointing this out, we added the missing title.
Author Response File: 
Author Response.docx
Round 2
Reviewer 1 Report
Comments and Suggestions for AuthorsThe revisions have improved the paper. So i am recommending publication. There are many typos still remaining, but they should be addressable at the time of the proofs.
