Baby-talk or motherese is the language that parents tend to use when addressing preverbal or just-starting-to-speak children. What are the fundamental features or rules that define this subset of a language? Is there a publicly available corpus/dataset for it?

I am primarily interested in English or rules and features that would apply to the motherese of most languages. However, rules or corpora from other languages are fine as answers, too.

  • 7
    It's notable that the use of motherese is not universal. You may be surprised to learn that some cultures do not talk to babies. – hippietrail Sep 16 '11 at 8:48
  • 1
    @hippietrail yeah, I might ask another question about universality later. I really hope to target this question at the ones that do, since it seems that most languages (where most is by number of speakers) do. – Artem Kaznatcheev Sep 16 '11 at 9:24
  • 3
    @downvoter there is a downvote on the question, I was curious about the rational (usually people say why they downvote) so that I could improve the question. Thanks. – Artem Kaznatcheev Sep 16 '11 at 11:55
  • In my experience, the way a baby talks is idiosyncratic. – Peter Olson Sep 16 '11 at 13:34
  • 1
    @NikhilBellarykar: Try this link which includes "Motherese is not a universal phenomenon: some cultures and communities either lack Motherese all together - parents speak to their children with no peculiar ..." google.com.au/… – hippietrail Feb 6 '12 at 12:18

The Deb Roy work prash brings up is certainly interesting, but it is just one child in one language. The standard for researchers that study language acquisition is to use the CHILDES Database (MacWhinney, 2000). CHILDES provides downloadable transcripts, video, and audio files mostly comprised of conversations between caretakers and children in many languages and at many ages.

As for Motherese itself, the seminal paper on this is Mother, I'd Rather Do it Myself (Newport, Gleitman, and Gleitman, 1977). While the topic has been revisted many times since, the general conclusions of that work still stand. It was the first paper to show that Motherese is "tuned" to the stage of acquisition of the child is in. That is, Motherese isn't a fixed style of speaking; we vary it with respect to how much the child we're talking to knows about language.

Some characteristics of Motherese pointed out in that article:

  1. While utterances are short, they are not necessarily syntactically simple. There are arguments made that the amount of movement required in a derivation of those sentences on average exceeds that of adult-to-adult utterances, largely because of the high number of questions.
  2. Utterances are heavily focused on directing the actions of children when they are younger, and change to contain more declaratives as they get older.
  3. It would appear that Motherese is optimized for easy processing rather than syntactic complexity. That is, utterances are short and contain few morphemes, but often more complex short utterances are used as opposed to simpler, longer ones.

Of course, the concept of "simple" is not straightforward; in the cited article it has to do in particular with the transformations common of syntactic theory of the time. Later papers questioned these ideas, but the consensus is that for whatever definition of "simple" you choose with respect to a grammar these general conclusions still hold.

There are of course other phonological and phonetic characteristics of Motherese. The exaggerated prosody (the intonation of a sentence) and facial expressions are an important part of it as well. So if you want to find a broad set of discussions on Motherese and rules, etc., I would recommend following the trail of papers that cite the one I list above. You can use this Google Scholar link to search among those articles.

  • 1
    Unfortunately Firefox says Server not found for the CHILDES link. – learner May 29 '17 at 17:59
  • @learner I've updated the CHILDES link as the old location is no longer in use. The new location is childes.talkbank.org – Constantine Aug 1 '17 at 20:27

Prof. Deb Roy has recorded every moment of his child growing up and acquiring a language — and that includes motherese, and everyone-else-ese. You can probably contact him for the corpus.


Prof. Aylin Kuntay in Koç University has started a project to build the corpus of Turkish babies. The project has not been completed so the corpus is not available online.


An eminent linguist Ashok Kelkar has made a rather detailed analysis of Marathi baby talk.

Take a look at point 3 onwards and you will find the vocabulary and whatnot. I agree that it is only for Marathi, but it is not restricted to one baby, and I, as a native Marathi speaker, can vouch for the ubiquity of the vocabulary given by him.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.