Providence talks

« previous post | next post »

Emily Badger, "Providence Wins Mayors Challenge Prize for Early Childhood Project", The Atlantic Cities, 3/13/2013:

New York City Mayor Michael Bloomberg likes to say that cities are the new laboratories of democracy in the United States (sorry states!), particularly in an era of political paralysis in Washington. This was the premise behind the $9 million Mayor's Challenge launched last summer by Bloomberg Philanthropies, inviting any city with a population larger than 30,000 to submit a groundbreaking idea for funding. This morning, Bloomberg announced the five winners – including a $5 million grand prize to Providence, Rhode Island – for potentially replicable innovations "bubbling up" from cities in early childhood education, recycling, data analytics, civic entrepreneurship and resident wellbeing. [...]

Grand Prize ($5 million): Providence, Rhode Island: Research suggests that in just the first few years of life, low-income children hear millions fewer words than their middle- and upper-income counterparts, impacting the development of their vocabularies and setting back their long-term prospects for academic and career success. This program aims to close that "word gap."

The AP story explains further (David Klepper, "Providence, RI, wins $5M Bloomberg contest with plan to boost poor children's language skills"):

Providence's winning proposal will equip low-income children with recording devices that count the words and conversations they are exposed to. Combined with coaching lessons for parents,, the plan is designed to help poor children overcome a language skills deficit that develops before they even start kindergarten. [...]

Called "Providence Talks," Taveras' plan will make use of a pager-sized recorder put in a child's pocket that acts as a language pedometer, recording every conversation and word spoken to them through the course of their day. The city intends to offer the voluntary program to children in low-income families, as determined by newborn screening assessments. Their parents will receive monthly coaching sessions from social workers in which they learn ways to boost a child's vocabulary, and social work agencies will be given bonuses if a child's language skills improve.

The recording devices work in English, Spanish and other languages and automatically screen out conversations from television and radio. The recordings will be kept confidential and once the devices' data are analyzed, any conversations on the recordings will be deleted. To prevent a 3year-old from losing or damaging the recorders, the devices come with specially designed clothing to hold them in place.

Providence Talks would begin with a small number of children participating and gradually expand the program to 2,850 families by 2018.

I haven't been able to find out who makes the devices, or what technologies they use for word counting, speaker identification, "automatically screening out conversations from television and radio", and so on. The story in the NYT (Jennifer Steinhauer, "Providence Is Top City in Contest of Ideas") is less informative and more confusing:

Providence took the grand prize for its plans to improve early childhood literacy. The children who participate in the program would wear a small device called a digital language processor that would record their daily interactions with adults. Those would then be converted into audio files containing the day’s adult word count and the number of conversational turns. That data would be used to help parents in monthly coaching sessions improve the quality of their conversations to improve their children’s vocabulary.

Here's a presentation of the project by Angel Taveras, the mayor of Providence:

The "30 million word gap" is an idea that comes from the work of Betty Hart and Todd Risley (Betty Hart and Todd Risley, "Meaningful Differences in the Everyday Experience of Young American Children", 1995; Betty Hart, "A Natural History of Early Language Experience", Topics in Early Childhood Special Education, 20(1), 2000; Betty Hart and Todd Risley, "The Early Catastrophe: the 30 Million Word Gap", American Educator, 27(1) pp. 4-9, 2003).

When I discussed this work a few years ago ("Word counts", 11/28/2006), I worried that big inferences were being drawn from a rather small sample. From Hart and Risley 2003:

By age 3, children from privileged families have heard 30 million more words than children from underprivileged families. Longitudinal data on 42 families examined what accounted for enormous differences in rates of vocabulary growth. Children turned out to be like their parents in stature, activity level, vocabulary resources, and language and interaction styles. Follow-up data indicated that the 3-year-old measures of accomplishment predicted third grade school achievement. [...]

Our final sample consisted of 42 families who remained in the study from beginning to end. From each of these families, we have almost 2 1/2 years or more of sequential monthly hour-long observations. On the basis of occupation, 13 of the families were upper socioeconomic status (SES), 10 were middle SES, 13 were lower SES, and six were on welfare.

The "30 million word gap" compares the 13 upper SES families with the 6 welfare families, and the authors themselves caution against "extrapolating their findings to people and circumstances they did not include":

All parent-child research is based on the assumption that the data (laboratory or field) reflect what people typically do. In most studies, there are as many reasons that the averages would be higher than reported as there are that they would be lower. But all researchers caution against extrapolating their findings to people and circumstances they did not include. Our data provide us, however, a first approximation to the absolute magnitude of children’s early experience, a basis sufficient for estimating the actual size of the intervention task needed to provide equal experience and, thus, equal opportunities to children living in poverty. We depend on future studies to refine this estimate. [...]

There's some more discussion of these issues in "Nick Clegg and the Word Gap", 10/16/2010, along with references to some much larger studies which confirm striking differences in vocabulary development, but do not tie these to measures of verbal experience.

The Providence project will draw everyone's attention to the importance of verbal interactions in early childhood, and should give participating parents useful feedback on their own young children's experiences. But I believe that the potential scientific results could be even more important, if the data can also be used to validate the Hart/Risely results on a large enough dataset that some of the many confounds in the original small study can be untangled.

Share:



24 Comments »

  1. Dick Margulis said,

    March 13, 2013 @ 12:13 pm

    Wouldn't it be appropriate for Providence to include a control group of children from families at other income levels in order to get a better handle on the size of the gap? Perhaps the actual study design includes this and it was just omitted from the news stories.

    [(myl) From the news stories, it seems that this is designed as an intervention, and not as a "study" or experiment. Still, in a group of the size discussed, there ought to be enough variation for effects on outcomes to be evaluated.

    What concerns me is that the cited "automatic" analyses are not at all trivial to do, especially from recordings made in real-world settings with microphones somehow fastened to childrens' clothing. In fact, I wouldn't be confident that such analyses were reliably possible within the current state of the art. Among the possibilities: (1) Some really brilliant and innovative algorithms are being deployed for the first time in this project; (2) The analyses will not in fact be (entirely) automated, but will involve human annotators of some kind; (3) Someone has promised device performance that they can't in fact deliver.]

  2. Ralph Hickok said,

    March 13, 2013 @ 2:22 pm

    I'm intrigued by this: "The recording devices work in English, Spanish and other languages …"

    Are there recording devices that record only English, ignoring other languages?

  3. Keelan said,

    March 13, 2013 @ 2:39 pm

    This device sounds like the one used by the LENA Research Foundation. Their website even advertises the custom-made clothing for storing the device. This page contains links to several of their publications that specify in more detail what the system does.

  4. Ginger Yellow said,

    March 13, 2013 @ 3:23 pm

    If the aim is to count the number of words children are exposed to, why screen out TV and radio?

    Also, as a practical matter, requiring children to wear special clothing seems like a surefire way to reduce long-term participation and/or compromise data quality.

  5. Joe said,

    March 13, 2013 @ 3:29 pm

    It's not the recording that's unusual. It's the following claim:

    "Here's how it would work: we propose to equip families eligible for home visitation services through Rhode Island's Universal Newborn Screening process with a small recording device capable of measuring the number of words a child hears each day. This device, developed by the LENA Research Foundation, filters out television and background noise and develops a comprehensive picture of a child's daily auditory environment, including adult word count and the number of conversational interactions the child engages in during the course of the day."

    The LENA Research Foundation is located here:

    http://www.lenafoundation.org/

    The software looks like is LENA Pro, but I'm not sure. There's a webinaire I'm going to take a look at after registering.

  6. Joe said,

    March 13, 2013 @ 3:37 pm

    Sorry for the double-post, but LENA foundation provides a number of technical reports on LENA Pro, located here:

    http://www.lenafoundation.org/Research/TechnicalReports.aspx

  7. Mark Young said,

    March 13, 2013 @ 5:46 pm

    They use a language-dependent model to estimate adult word counts without understanding the content. Thus the works-in-English/Spanish comment. They aren't actually counting words.

  8. Kyle Gorman said,

    March 13, 2013 @ 7:50 pm

    While I do not wish to dispute the claim that there is a word gap, the most obvious explanation for why there is a word gap suggests to me that this intervention will fail. The most obvious explanation is that working class parents spend more time out of the house (perhaps working multiple jobs!) and therefore can do little to address the amount of linguistic input their child gets without imperiling family finances.

  9. Kyle Gorman said,

    March 13, 2013 @ 8:06 pm

    Though, we've been through this debate before:

    "Black children from the ghetto area are said to receive little verbal stimulation, to hear very little well-formed language, and as a result are impoverished in their means of verbal expression…Unfortunately, these notions are based upon the work of educational psychologists who know very little about language and even less about black children. The concept of verbal deprivation has no basis in social reality. In fact, black children in the urban ghettos receive a great deal of verbal stimulation…and participate fully in a highly verbal culture." – William Labov, _Language in the inner city_, 1972, p. 201

  10. Mark F. said,

    March 13, 2013 @ 8:29 pm

    Ginger Yellow – My understanding was that Hart and Risley didn't count TV and radio in their comparisons. The thinking is that words heard in the present context, spoken by actual nearby people, are much more helpful for learning.

  11. Mark said,

    March 13, 2013 @ 9:16 pm

    It's Lena; they sent out a press release about it.

  12. Sarah Creel said,

    March 13, 2013 @ 10:55 pm

    Anne Fernald's lab has been doing a lot of work with LENA and has additional data on language exposure and vocabulary development–they're trying to pull apart influences of SES and language exposure. However, I can't track down the paper at the moment. It may be on the "conference circuit" and not yet in a journal.

  13. MattF said,

    March 14, 2013 @ 8:49 am

    I agree that the sample looks small, particularly when you consider the number of possible confounding variables and the far-reaching conclusions. Also, 30 million sounds large to me– bearing in mind, e.g., that one year = 31 million seconds.

  14. Elika said,

    March 14, 2013 @ 9:40 am

    So there are a few things here to unpack…one is that Hart and Risley was an excellent early attempt to get at the kind of variance, due to exposure conditions, that many language researches at the time said 'didn't matter' enough to worry about; as Anne Fernald has pointed out, this was a very different kind of poverty of the stimulus. That said, of course their fairly small sample, from almost 20 years ago didn't cover everything. But since then there have been PILES of data (from Fernald, Hurtado, Weisleder, Hoff, Rowe, inter alia), with LARGE low-ses samples, and WIDE ranges of SES, in English, Spanish, and other languages confirming the relationship between maternal input and young children's language abilities (vocab size, word comprehension speed, grammatical abilities, you name it, from toddlerhood though H.S.) Whether the 'right' thing to count is maternal types, tokens, mean-length-of-utterance, 'following in' behavior, or whatever else is hard to say since these things are highly correlated and their effects vary across samples. But the bottom line is that Providence's proposal is not a crazy idea…it's easy to say 'don't feed honey to you baby, it could cause botulism'. It's much harder to say 'talk more to your baby, and talk about more abstract things, it'll help her language skills', and some of the points Kyle raises above w.r.t. cultural differences and the amount of free time that lower ses parents are able to allocate to quality time with their children are good ones. But, insofar as such an intervention as Providence proposes is done with care, I'm sure any parent, low ses or otherwise, would be happy to hear ways they can try to improve their child's language abilities.
    On another point, the LENA system is being used by a handful of babylabs across the country (including Stanford and Rochester), and its algorithms are being compared and tested with handcoding of the data in various ways. The real shame here, for the field, is that the data is proprietary and as far as I know there are no plans for any sort of clean-up-and-release to the public.

  15. Erin said,

    March 14, 2013 @ 3:54 pm

    From a research perspective this is very interesting, and I applaud the *idea* of doing something about the problem, but practically speaking this sounds like a disaster. Let me just quote:

    "The recording devices work in English, Spanish and other languages and automatically screen out conversations from television and radio. The recordings will be kept confidential and once the devices' data are analyzed, any conversations on the recordings will be deleted. To prevent a 3-year-old from losing or damaging the recorders, the devices come with specially designed clothing to hold them in place."

    1. Have these people ever dealt with trying to keep information private in the modern day and age.
    2. HAVE THESE PEOPLE EVER DEALT WITH A CHILD.

  16. LDavidH said,

    March 14, 2013 @ 4:36 pm

    My first thought was, Is it April Fool's Day already?

  17. Ken Brown said,

    March 14, 2013 @ 4:43 pm

    30 million in 3 years? That's something like 4 hours of continuous speech a day. Every day. I'm not sure I believe it.

    And are radio, TV, and pop music not verbal?

  18. Theodore said,

    March 14, 2013 @ 4:58 pm

    I'm with Kyle Gorman above on what seems to be the obvious proximate cause. I also think they could accomplish more by eliminating the recorders and just providing the coaching.

  19. Diane said,

    March 16, 2013 @ 7:28 am

    @Theodore

    Ah, that was exactly what I was going to say. If this is an intervention, then the point is to improve the children's language skills not to find out why they their skills need improvement. In that case, why bother counting the words? Certainly not just to give the parents feedback. It would be much cheaper, easier, and more direct to just evaluate the kids' language skills beforehand, give the coaching classes to the parents, and then monitor the kids' language development to see if the coaching classes are working.

    On the other hand, I recently heard a very convincing argument that there is a funding bias towards studies/interventions that use fancy shmancy equipment. So maybe that's it.

  20. Liz said,

    March 22, 2013 @ 10:57 am

    Last week I happened to follow a young mother around a grocery store. Facing her, in the cart, was a well-dressed alert 18-month-old girl. The mother picked up, inspected, and would sometimes choose a variety of fruits and vegetables. Not one word was said to the child the whole time, not "would you like strawberries?", nothing. I could not intervene, of course.

  21. Ben Zimmer said,

    March 24, 2013 @ 11:33 am

    My latest Boston Globe column is on Providence Talks, in part inspired by the discussion here.

  22. On the Providence word gap intervention | Wellformedness said,

    March 25, 2013 @ 1:54 pm

    [...] we really count words? In Language Log post, Mark Liberman speculated that counting words might be beyond the state of the art. While I have [...]

  23. On the Providence word gap intervention | Wellformedness said,

    August 5, 2013 @ 5:32 pm

    [...] we really count words using current speech technology? In a recent Language Log post, Mark Liberman speculated that counting words might be beyond the state of the art. While I have [...]

  24. Focusing on the ‘Word Gap’ With Help from Hillary Clinton | Ed Central said,

    October 13, 2013 @ 12:13 am

    [...] and The New York Times. It has also attracted critical questions and interested onlookers from Language Log, an international blog of linguists and language [...]

RSS feed for comments on this post · TrackBack URI

Leave a Comment