Contextual Modulation of Adult–Child Language Interaction: Semantic Network Connectivity and Children’s Vocabulary Development
Round 1
Reviewer 1 Report
Comments and Suggestions for AuthorsThe present study focused on an underexplored topic about the extent to which different contexts of conversation may influence language development in early childhood. This study involved 62 English-speaking children at the age of 3, and data were collected in natural language learning contexts, such as the activities of book reading, toy play and mealtime. The primary findings suggest that the context of book reading contributes greatly to building deep internal connections within semantic networks; that the semantically relevant conversations in book reading play a significant role in the lexical development in early childhood.
I appreciate the view proposed by the authors that high interconnectivity of words in diverse contexts may promote word learning, and the depth and quality of vocabulary knowledge should be valued in children’s language learning. However, I indeed have some concerns about the current manuscript, mostly related to the measurement, data analysis and results that I detail as follows.
1) One concern is related to the statistical features used to analyze the structure of semantic networks. Lines 127-128 on page 3 and lines 153-154 on page 4, “Four primary statistical properties are commonly utilized to characterize the structure of semantic networks: average distance (L), diameter (D), density or sparsity (S), and clustering coefficient (C)”. However, in the Result section, “The analysis of these semantic networks focused on three properties: short path lengths (L), density (S), and clustering coefficient (C).”
Lines 154-155 on page 4, the authors seemed to explain the difference between L and S, but it is not an explicit reason for selectively not using D as the property. Would the authors please justify why there is the selective use of the three properties?
Additionally, there are no consistent names in terms of referring to the specific terms for the property of L. Average distance is used at the beginning, but later the term of short path lengths is adopted throughout the Result section. Do these two terms mean the same thing? Is it possible to consistently use a term to refer to the same property?
2) Line 163 on page 4, please add a reference to specify the origin of this computation formula.
3) Lines 182 to 190 on page 5, the content related to the computation of clustering coefficient C is not clearly described. Firstly, “…Ti represents the number of connections between the neighbors of node i, while ki(ki-1)/2 represents the number of connections between i’s neighbors…”. Readers may be confused about what actually the connections refer to in this explanation. Is “the neighbors of node i” different from “i’s neighbors”? My understanding is that the number of connections in Ti is related to the edges between non-colored nodes (neighbors of the node i); that the number of connections in ki means the number of edges attached to the node i (connections between the node i and neighbors). I would like to suggest that the explanation of the computation may be comprehensible by taking the specific connections in in one graph of Figure 3 as an example to illustrate the computation process. Moreover, the authors can consider explicitly utilizing the types of lines in Figure 3 as an aid to facilitate the explanation of the number of connections. For instance, the thin solid lines refer to the connections related to ki, while the bold solid lines mean the connections related to Ti.
4) Line 227 page 6, “… and the diagram (i.e., maximum path length: D) was only 7.79.” Please check which word, diameter or diagram, should be used here.
5) Please add the effect size to the following statistical results: F (1.74, 64.43) = 0.92, p = 0.39. on line 224 page 6; F (1.88, 69.51) = 11.11, p < 0.001. on line 235 page 7; F (1.91, 70.68) = 27.74, p < .001. on line 247 page 8.
6) Would the authors please provide a further explanation related to the statement on line 243 page 7 of “with a word being connected to only an average of 1.5% of the total number of words”?
Minor concern:
Please rephrase this statement on line 211 page 5. There are multiple resources to build semantic networks. However, in the current study, three representative sources of semantic knowledge were utilized as a way to build semantic networks.
Comments on the Quality of English LanguageMinor editing of English language required
Author Response
Please see the attachment.
Author Response File: 
Author Response.pdf
Reviewer 2 Report
Comments and Suggestions for AuthorsThis study utilized three statistical features to describe the structure of semantic networks: average distance, density, and clustering coefficient, and investigated the extent to which the use and effectiveness of semantically relevant conversations vary across activity contexts. In this sense, the topic is interesting and of much significance for the establishment of parent-child interaction environment. But so far I haven’t seen the theoretical framework and the theoretical contribution of the study and there are also some other problems related to the article writing. I would like to give them a major revision.
First, the review of the literature is insufficient. There should be a section summarizing and analyzing how much has been done in this field and what the insufficiency lies.
Second, in 2.1. dataset. the authors say “the study examined the social factors influencing language development in a group of English speaking children from low-income families.”, which seems to be contradicted with the topic, and besides semantic network factors, the authors did not do any analysis on SES factors in the study. Also, in this section, the authors should have described how the data were collected even though it is from the CHILDS database. And since in the database many settings of activities were collected, why only three in the analysis of the current study.
Third, the discussions and implications can be put together, but conclusion should be split into different sections.
Comments on the Quality of English Languagethe English language is quite fluently, but still moderate revision should be done in order to express more clearly.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Reviewer 3 Report
Comments and Suggestions for AuthorsThe paper aims at investigating how contextual factors affect adults’ conversations with children (in terms of semantic networks) and their role on child vocabulary development. Specifically, the author(s) takes into consideration adult-child conversations in three different contexts (toy play, mealtime, and book reading), analysing the characteristics of the semantic networks developed in such contexts and their effectiveness on children’s lexical development. It is found that the density of semantic networks in toy play is higher than that of the other two contexts, which means that fewer concepts are involved in this setting. Book reading, on the other hand, exhibits more densely connected word communities, as results from a higher clustering coefficient. Finally, adults’ use of semantically relevant conversations during book reading is positively correlated to children’s vocabulary development.
The paper is generally well written and clearly presented. The topic is interesting, and the author(s) proposes an original way to look at vocabulary development, by focusing on vocabulary depth (operationalized in terms of interconnected networks of semantic knowledge), which has traditionally received less attention. The study also discusses the educational implications that the findings could have, which fits well the scope of the journal.
I do, however, have some observations about the way the author(s) describe the methods and a concern about the validity of one of the measures being used, which, in my view, makes the argument less compelling and the conclusions drawn by the author(s) not entirely supported by the data presented.
As for the description of the methods, I understand that the data was not collected by the author(s) but taken from the CHILDES corpus, however, I find it a bit surprising that, given the focus of the paper (contextual factors), very little information about the contexts is given. How much data per child was used (data from how many home visits)? Are the data collected from each child quantitatively comparable? How many books were read during book reading? What was the book(s) about (I believe knowing the topic(s) of the book(s) might be especially relevant for a study which proposes to focus on conversational contexts)? What kind of toys were used during toy play (some may involve more repetitive routines than others)? Could the adults choose from a range of toys or were they all instructed to use the same ones? How was mealtime data collected (from what I saw on the CHILDES website, this might be the less controlled context)? This information is of course available on the CHILDES website (by the way, the author(s) should check the link they provided, as it does not seem to be working), but I think minimal information about such aspects should be provided in the article (and possibly taken into consideration in the discussion, if relevant), given their potential role in determining the results.
Also, the author(s) does not explain how semantic associations between nodes were determined in the building of the semantic networks. Some other studies frequently mention relying on previously collected semantic relatedness judgments, vocabulary, and thesaurus information, and so on. I think adding this information might be useful here too.
The other observation I have concerns the measure for determining vocabulary development. First, I wouldn’t use the term ‘language development’ (lines 129 and 341), which is very broad, and might be misleading. In fact, this is a measure of vocabulary development (despite its importance, language is not only made up of vocabulary). More importantly, I am not quite sure about the validity of measuring vocabulary development based on data coming from one recording. As I understand from the details given on the CHILDES webpage, there was only one elicited reported per home visit, and, based on what it is said in the article, I believe the author(s) only used the data from Home Visit 1 (but again, providing more information about the methods would help clear this up). Are we sure we can make this kind of inferences about a child’s knowledge based on only one session? Given that we are talking about three-year old children, who at times might simply not be willing to do the task, I am not sure one elicited report can give reliable information about their lexical variety. Also, from what I understand, the reports could potentially deal with very different events (and therefore, very different contexts), so it doesn’t seem to be a very controlled measure. Not to mention the fact that even adults might exhibit little lexical variation in their everyday conversations if not explicitly prompted to in some sort of way (but this also, once again, also depends on the topic of conversation). I don’t know what kind of instructions (if any) the mothers in the CHILDES project were given in order to be sure to elicit as much lexical variety as possible.
I understand, of course, that this does not depend on the author(s) of the article, who did not collect the data, but I think this might overall weaken their argument and should at least be addressed in the discussion.
Minor remarks:
If I understood correctly, the original CHILDES corpus provides complete data for 64 children (for Home visit 1, which I believe is the one considered here), while here 62 children were considered. Was there a particular reason for excluding two participants? I think this could be explained in a footnote.
Line 304: “only a limited number of studies have explored …” I believe there should be some references here.
Author Response
Please see the attachment.
Author Response File: Author Response.pdf
Round 2
Reviewer 3 Report
Comments and Suggestions for AuthorsThe authors addressed all the comments provided and made the necessary adjustments.
