Review History
Early prediction of language and literacy problems: is 18 months too early?

Reviewer 1 · Jun 19, 2015

Additional comments

The authors have done a very good job of addressing concerns regarding their first revision of this submission, which presents interesting findings regarding the predictive utility of a vocabulary delay at 18 months for later language and reading problems. The authors have fully responded to my comments about the previous submission and clarified several issues that were previously confusing (including the puzzle of no pages numbers). I am glad that the authors added the subsidiary analysis of toddlers with <6 words, having realized that this and not <10 words was actually the 16th percentile for their sample. It is interesting that results were virtually identical to those using the 25th percentile cutpoint of <10 words. I also appreciate the addition of the footnote on the Method page, as this possible selection bias helps explain the 33% rate of SLI in the TD comparison group. I also think the revisions made near the end of the Discussion improve the manuscript by removing the debatable points about the two separate factors that influence early language ability and by adding the statement “The older a child is, the lower the probability that poor language is just due to normal maturational variation.” Finally, I think the revision of the final paragraph makes several good points and thereby improves the manuscript. In summary, the authors have been very responsive to reviewers’ suggestions and I recommend acceptance of this interesting submission.

Reviewer 1 · May 2, 2015

Additional comments

The authors have done a very complete job of addressing concerns raised by reviewers regarding the original submission, resulting in a much improved manuscript. The topic addressed is an important one, namely the predictive utility of a vocabulary delay at 18 months for later language and reading problems. Although this topic has already been the focus of a considerable amount of research, many using larger samples, this study presents some interesting findings and adds some valuable new information to our understanding. My few remaining comments, all relatively minor, are detailed below.

1) There are still no page numbers in the manuscript, which seems odd.

2) I am struck by the fact that the 30 LTs in the large sample, identified by < 10 words produced at 18 months, did not differ in words comprehended at 18 months on the CDI from their typically developing peers. This is somewhat surprising because late talkers identified at 24-30 months in the Rescorla cohort had lower comprehension scores than the typically developing children even though they scored in the normal range. So, I wonder how expressively delayed the 18-months LTs really were. The authors state that <10 words is “at least 1 SD below the mean” but I think they should state what percentile cutpoint it actually represents on the UK CDI norms.

3) In my original review, I noted that it seemed strange that 3 out of 9 children with typical language at 18 months had a language delay at age 4, the same rate as in the 18-month LTs. As I stated, a base a rate of 33% for SLI seems very high seems high. The authors do address this in their response letter but I think they should include this information in the present submission as well – namely, that there may have been a selection bias issue in this study with parents who had concerns about their children being more likely to agree to participate, Additionally, the 29% vs. 14% rates of SLI at 4 found when the same team followed up 24 LTs and 58 TD children at 4 should be reported in the Discussion. These percentages seem more reasonable.

4) In the top paragraph on the next to last page of the text, the authors postulate two separate factors that influence “early language ability,” namely the rate of acquisition and the child’s “ultimate language ability relative to others.” They add that “the first predominates in toddlers, but its influence then declines with age.” The authors make this same point in their response letter. I find this analysis to be rather unconvincing. First, I don’t see how the authors can argue that “early language ability” is “influenced” by a factor which would appear to come later (namely, “ultimate language ability.”). Second, children keep developing in language from 1 year to adulthood, which is why the norms for tests change with age. Therefore, children who make slower progress have lower scores than their peers at whatever time they are assessed, whether at age 2, 4, 7, or 16. Because language growth in young children is so dramatic, evident, and easy to measure (number of words, length of sentences, provision of morpheme, etc.), rate differences are very apparent, more so that at ages 6-7, for example. If the authors mean by “ultimate language ability” their “asymptotic level of performance in language,” then I would agree that if two people develop at different rates then the person with faster development will end up with more skills than the person with slower development by the time the developmental period ends. For example, people don’t typically get taller after age 18, so people who add more inches per year to age 18 are going to end up taller than those who add fewer inches per year. However, I don’t see how asymptotic level can have any effect on early ability. Rather I would argue that “ability” or “endowment” (or some parameter that is a mix of inherent skills plus environmental influences) determines both rate of language development and asymptotic level, at least to some degree.
What seems evident from the data in this study is that the older a child is when a language delay is identified, the more likely it is that an earlier delay has not been outgrown and thus will continue. Thus, 2-year-old LTs who make good progress are not identified with SLI at 4 or 5 and generally have good later outcomes. But LTs who are still delayed at 4 or 5 have less good later outcomes. However, it now well-established that most 4-5 year-olds with SLI were not late-talking toddlers, so for them it would appear that their skills enabled them to progress well in the basics of language acquisition but then they started making slower progress. In short, I would like to see the authors revise this section of the Discussion so as not to suggest that a later factor could affect an earlier factor and perhaps to address some of the points noted in this comment.

5) I would agree with the authors that 18 months is too early to identify LTs, because children identified at this age typically catch up quickly. This conclusion has been reached by several studies in addition to the current one. However, many LTS identified at 24-30 months take several years to catch up, suggesting an important difference between 18 months at 2 years as a screening time point. As the most interesting aspect of this manuscript is the strong prediction from age 4, even with a very small sample, I think the authors should discuss more fully the evidence for differences in prediction accuracy between 2, 3, and 4 years of age. They state in their literature review that prediction from age 3 is mediocre, but some more systematic comparison in the Discussion of predictive measures from these three time points would be useful.

If the authors can revise the manuscript to clarify the issues outlined above, I think the article will be a useful addition to the literature by highly esteemed researchers of early language delay.

Thomas Klee · Apr 24, 2015

Additional comments

This is an excellent paper and I look forward to its publication. I appreciate how carefully you documented the changes you made in response to the reviews.

Reviewer 1 · Jan 11, 2015

Additional comments

This submission is a brief report on a topic that has already been the focus of a considerable amount research, namely the predictive utility of a vocabulary delay at 18 months for later language and reading problems. Many of these other studies have used much larger samples, and the results of this submission are largely consistent with these previous studies. Nonetheless, the study presents some interesting findings and is generally well-presented. Specific comments are detailed below. Many of these comments relate to puzzling aspects of the methodology that suggest the current submission represents an effort to capitalize on existing data rather than a study that was specifically designed to address the questions posed :

1) On line 82 (there are no page numbers in the manuscript) the authors refer to a sample of 300 children in a longitudinal study. However, on line 92 they state that 153 children had 18-month OCDI data. As this number is so much smaller than 300, the authors should explain why they don’t have 18-month data for the full sample.

2) It is also not clear on line 100 why the follow-up age range was so broad (from 4 to 9 years). The overall design of the larger project needs to be outlined so that the reader can understand both the small number of children with 18-month data and why this follow-up age range is so wide.

3) On lines 114-115, the authors state that 18 out of the larger sample of children had been assessed at 4 years. Once again, this seems like a very small number given a study with 300 children and this needs to be explained.

4) It seems strange that 3 out of 9 children with typical language at 18 months had a language delay at age 4. A base a rate of 33% for SLI seems very high. It also is surprising that the rate of SLI at 4 was identical in the LT and AT groups. While is it well-known that being a late talker at 18 months is not very predictive of SLI, it does seem surprising that the rate of age 4 SLI should be no higher in the LT than in the TD group

5) Given the fact that some of the children were 4 years old at Time 2, the choice of measures on 133-138 is surprising. Few 4-year-olds would be expected to read passages.

6) I do not understand the logic of the statement on line 185. The notion of “illusory recovery” is that a group with an early delay appears to have caught up in skills by a certain age but then is revealed to have continuing delays in a related skill at an even later time. As the current submission ‘s main analyses concern a single follow-up (at ages 4 to 9) after initial delay at 18 months, I don’t see the relevance of illusory recovery (or lack thereof). That is, the study’s design does not present data that can address the hypothesis of illusory recovery, except for the 6 children who were LTs at 18 months and did not have SLI at 4, and these children are not described with any specifics in the submission.

7) The most interesting aspect of the manuscript is the strong prediction from age 4, even with a very small sample. Since the authors make the point in their literature review that prediction from age 3 is mediocre, I think the submission should briefly address why there should be such a difference between prediction at age 3 vs. age 4.

If the authors can revise the manuscript to clarify the issues outlined above regarding their methodology, I think the article would be a useful though modest addition to the literature by a very highly regarded set of researchers in the area of early language delay.

Thomas Klee · Jan 21, 2015