More on the CDC Developmental Milestones

Several weeks ago, I wrote about the CDC revisions to its developmental milestones. You can read the original post here. I got a lot of responses. One piece of this was people wanting more on topics like crawling, or receptive language development. I’ve got those on the list for future posts. But a large share of the responses focused on the expressive-language milestones, and in particular the issue of “50 words at 30 months.”

Is this the right milestone, and what does it mean? I collected feedback from Michael Frank, who runs the Wordbank that I referenced in the post; from some of the speech-language pathologists behind The Informed SLP (JoAnne Berns, Marie Bloem, Katherine Sanchez, and Karen Evans); and from the CDC authors of the journal article that informed these guidelines. Today I’m going to relay and try to organize some of that feedback. I do not think we’ll get to an answer that satisfies anyone, but I’m hoping to capture the tenor of the debate.

I’m going to start by going through a few general points of criticism, then provide the CDC response, and try to summarize.

Issue 1: Counting words is hard

In the last post, I talked about a primary issue in evaluating language development, which is that it is difficult for parents to recall all the words their child has, especially as they increase. When it’s just mama, dada, kitty, and shoes, it’s no problem. But if your child says 50 or 100 or 300 words, you will not remember them all.

This is why, when researchers evaluate this, they use a long list of words and ask parents if their child knows them. Those types of assessments are where the Wordbank data comes from. Michael Frank very kindly provided me with the graph below, which shows the expected count of assessed words by age in the general U.S. population.

Based on this data, a child at 30 months who produced only 50 words would be well under the 10th percentile.

But! Michael pointed out that the milestone isn’t about how many words kids have based on this assessment. The milestone is based on what parents say when asked how many words their kids have. People generally underestimate (Michael told me that in the 1870s people estimated that peasants only said about 200 different words. Wrong.)

Assuming some underestimation, maybe this all lines up. If you want to identify kids at the 25th percentile at 30 months, and the assessment says 300 words, 50 words works if parents underestimate by a factor of 6. But there isn’t anything in the data that would point to this particular amount of underestimation.

Issue 2: The source of “50 words” is totally unclear

The conclusion of the above is that it is hard to measure parental-recall word production, which is really what the milestones are aiming to get at. This leads to the question of where, precisely, this number came from.

It is unclear.

The supportive publication references two actual data sources (and also a set of language scales). One of the sources is an aggregation of milestones but doesn’t have anything on 50 words. The second is a paper studying 40 children, looking at language development over the second year of life. This actually does look at parent-measured acquisition of 50 words. It finds an average of 18 months for this milestone, with a range of 13 to 21 months. There may be reasons, of course, why this doesn’t directly apply, but certainly this source wouldn’t justify the 50 words at 30 months.

I posed this question directly to the team at The Informed SLP, who said the following:

We have gone through virtually all the references to try to understand this and are coming up totally empty. We’re stumped.

I am also stumped. More on the CDC response below, but I would say that the answer isn’t that we are missing something. There is no particular source that generates this number.

Issue 3: Single words is a poor milestone at this age

The most constructive criticism here is that it doesn’t make sense — perhaps for the recall reasons discussed above — to use word counts at this age. The Informed SLP team wrote me the following:

The separation of 50 words (at 30 months) from two-word phrases (at 24) is especially confusing. All of Zubler et al.’s sources that discuss both word combinations and a 50-word vocabulary have those two achievements at the same age (i.e. 24 months), which makes sense since you really don’t see kids building phrases before they have a 50-word productive vocabulary. A good reference for this (included in Zubler et al.’s reference list) is Fenson et al. (1994), which reports on the validation of the CDIs. We’ll include two very relevant passages from that publication here:

“If we accept a judgment of either ‘sometimes’ or ‘often’ as evidence for word combinations, then 20.9% of children with vocabularies under 50 words are already producing a few word combinations, and just under 50% of the sample are combining when vocabularies reach between 50 and 100 words. However, if we impose the more stringent criterion of combining ‘often,’ then only 1.3% of the children with productive vocabularies under 50 words and slightly more than half (54.1%) with vocabularies of 200-300 words meet the criterion.”

“The median number of words produced by children in the present study was 55 at 16 months, 225 at 23 months, and 569 at 30 months (note that this last figure is the most likely of the three to be an underestimate). Thus, the rate of vocabulary learning is 0.81 words per day between 16 and 23 months and (at least) 1.64 words per day between 23 and 30 months.”

To de-jargon a bit, the point here is that by this age, language learning is a lot about language construction and communication, and not about words per se. This would suggest that any expressive-language milestone at this age might be better based on some sense of communication ability rather than word counts. By the age of 3, the language milestone is along the lines of “can be understood most of the time.” It might be better to have a version of that at 30 months too.

Issue 4: There were no speech and language pathologists on the panel

This is a topic of some disagreement. The CDC says one of the paper authors was trained in both pediatrics and SLP. The Informed SLP team pointed out that this person isn’t working as an SLP and that it isn’t one of their main credentials.

As someone who is constantly being criticized for having the wrong credentials, I think this argument is basically unhelpful. The guidelines could be appropriate even if there were no SLPs, and they could be wrong even if there were. So I’m going to leave this one.

CDC response

I asked the CDC study authors about these issues. Because there has been so much criticism, I include my question and their verbatim response.

Is there a norming data sample that you can point me to that informed the 24-month and 30-month count of word guidelines? I see the citations in the paper, but they do not point to a particular data source. Is there something more specific that you can point me to?

The CDC checklists are meant to be health communication tools for use at well child visits (health supervision visits). They are not to be used as developmental screening or evaluation tools. The revision process was undertaken because of feedback from parents, pediatricians, and early childhood professionals who felt the lists could be improved. Some areas that were identified were adding 15- and 30-month checklists, using clear language, not repeating milestones across checklists, and using milestones most children would be expected to achieve by that age so that developmental screening would more likely be done if a child is missing milestones or if parents or professionals had concerns (rather than waiting and seeing if the child develops the skill later on).

We used normative data that was published within a limited number of studies, and supportive evidence from common developmental screening tools and parent resources. We found normative data on individual milestones difficult to find. The group revising the milestones used that information along with clinical opinion to determine if the milestone should be included and at what age. Placing certain milestones on a specific age well visit checklist, when most children may achieve it between well visit ages, is something that can be reviewed as more evidence becomes available.

The group acknowledges (also in the paper) that there are limitations to the process and best practices for developmental monitoring/surveillance. They hope more research will be done to further identify best practices for developmental surveillance/monitoring.

I read this as acknowledging that the milestones are based as much on clinical judgment as on evidence.

The Informed SLP team told me:

When the authors say “evidence and judgment” led to their recommendations, but the sources they cite don’t check the “evidence” box, it leads us to believe that “judgment” was the deciding factor.

This seems consistent with the response.

Final thoughts

I will say that, academically, I find this debate interesting. It touches on why it’s hard to produce data. It also gets into how we translate evidence into practice. How do we take a word count from a developmental screening tool and translate that to “What do parents notice?” And even more, translate it into something pediatricians can implement at a 15-minute well-child visit, when they’re doing a million other things.

Looking at all of this together, though, it does feel to me like the 50-words milestone is a miss — or, maybe better said, a missed opportunity. It’s not that the count is right or wrong; it just seems like it isn’t really the right way to summarize the language development goals at this point. Evaluating language skills at this age is difficult, when a child is somewhere between “can put two words together” and “can talk.” Word counts don’t capture that, but it isn’t easy to see what specific guidance would capture it.

I asked the Informed SLP team what they’d want here, and they shied away from defining a milestone but said the following:

Not a word count, since it would be difficult for parents to estimate the number of words most children seem to have at this age. A more useful 30-month vocabulary milestone (if having one feels necessary) might be functional: something like “using words to communicate most all of the time,” “having a word for almost everything,” but there isn’t a ton of data to base this on. We don’t know of any other data sets looking at vocabulary size that could possibly have supported the recommendation.

Perhaps a more important angle than vocabulary to consider at 30 months is emerging grammar and sentence length/complexity. SLPs evaluating children this age typically look at measures like the average length of the sentences children say, the emergence of grammar (e.g. tense and agreement), and the different types of word combinations children use. The work of Dr. Pam Hadley could be really informative here.

My end guess is that the CDC will need to revisit this particular milestone. The authors’ response hints at a willingness to do so. There is an open question, though, of whether there are any simple guidelines for this age category that would be helpful.