Is There Really a 30 Million-Word Gap?

Teacher question:

I attended one of your recent presentations. You cited the Hart & Risley canard that there is a 30 million-word gap. Aren’t you aware that study has been rejected? There is no word gap. Poverty kids have as much language support as other kids.

Shanahan response:

Research can get things wrong.

That’s why researchers—unlike practitioners and policymakers—are usually so interested in the methods of a study. Study a problem one way, you get one answer. Study it another way, perhaps a different answer emerges. Try to understand why the two studies diverged and youstart to gain a deeper understanding of the problem.

That’s why I don’t like the term “findings” in research studies. “Results” is the more accurate term. Even with qualitative studies that my claim no results, just findings, because they only watch and record and don’t intervene… yet, how one observes and records can influence outcomes, so even those kinds of studies have results.

In the 1990s, Hart and Risley published a widely disseminated study in which they collected language data on 42 families (some of these were upper socioeconomic status, some  were working class, and some were families on welfare). Monthly over 2.5 years, they audiotaped young children’s (7-9 months to 3 years) language environments.

They found that children were spoken to much more often in the upper income households than in the poverty households, and extrapolating across the children’s waking hours, they concluded that there was a 30 million-word gap. Some kids were having a lot more language experience. The researchers measured other more qualitative aspects of these children’s language environments as well, but the 30 million words became a symbol or summary of the whole study -- which has been hugely influential of policy and research.

As the questioner above notes, recently there has been some new evidence on young children’s early language environments. This summer, Child Development, published an article by Sperry, Sperry, & Miller (2018). Basically, it concluded that there were big differences in home language environments, but that these were equally distributed across the different socioeconomic strata. Some kids were definitely hearing fewer words than others, but it wasn't necessarily the kids living in poverty.

That study got a lot of play because it claimed to be a replication of the original study (it included 42 families as well). However, it was so different from Hart & Risley’s investigation that I think replication is the wrong description. For example, the biggest differences in H&R’s results were between the high SES kids and those growing up on welfare—that’s where the 30 million-word gap came from (working class and welfare differences were both much lower in this variable than the well-to-do), and yet this new study didn’t include a higher income sample. That alone could explain the lack of differences reported by Sperry and company. 

In the original study, the researchers went to the houses with tape recorders and sat there for an hour each month. There has been great concern that placing a PhD in the households of low-income families like that might suppress language use. Let’s face it. Anyone observed like that might be inhibited (I think I would be—I imagine whispering), and some speculate that this would be especially true for low income/low education parents (and there is research suggesting that this kind of inhibition does take place). If that were the case, then the welfare moms may have spoken to their children less than usual just because they were being observed.

There are things that can be done to limit or reduce such reticence (like having observers stay longer so those being observed get used to their presence or making sure the observers are of the same race and gender as those being watched, etc.), and this new study made better use of such methods. Perhaps the original observation techniques discouraged talking in some families and encouraged it in others. If so, then one might conclude that Sperry and company are correct.

The problem with that conclusion is it assumes that the only study reporting a language gap was the Hart & Risley investigation.

However, Sperry et al. weren’t the only ones re-exploring these waters and some of those other studies also reported such a gap. In my opinion, the most rigorous study on this so far is one reported by Gilkerson, et al., in 2017. They used technology instead of potentially-intrusive observers and collected a whopping 49,765 hours of recording from 329 families (more than the Hart and Sperry investigations combined). That study reported a much smaller word gap than what Hart & Risley claimed, but unlike Sperry, et al. it did identify a sizable gap (and with a much larger population, studied much less intrusively and more thoroughly).

Gilkerson claimed “only” a 4 million-word gap between those highly educated, high SES parents and those much less educated low SES ones. Four million words ain’t chopped liver! Spread over two years (these kids were observed from age 2 to 4), it would be like hearing a 55 minute a day speech from mom (spoken at 100 wpm… which is slightly slower than conversational speech, which makes sense for talking to a 2-year-old). But let’s face it…. No matter how much money or education you have, no one is going to give a 55-minute speech to a baby. What there will be is a lot more parent-child interaction…literally more hours of daily conversation between parent and toddler.

Another critical difference is that the Sperry study included the presence of ambient language. That’s the language not spoken to the children themselves, but that was in their environment. If Aunt Edna is yacking away on her iPhone while young Egbert is playing nearby his language is getting real shot in the arm, according to the Sperry methodology.

Hart & Risley (and Gilkerson) only counted words spoken to the child. One reason Sperry’s team found no difference was because of the language emanating from the TV sets and fugitive background conversations of adults who were not speaken to the kids. But as Golinkoff et al. (2018) point out, the notion that ambient language has a big effect on language development (compared to language spoken directly to the child) has been rejected by direct study. Despite the methodological problems evident in the original Hart & Risley work, analyzing the language spoken to the children instead of what might have been overheard from the TV in the next room was not one of them.

Finally, it should be pointed out that studies have shown that early language differences matter in later reading performance (Golinkoff, et al., 2018; National Early Literacy Panel, 2008). And, that the specific children exposed to low language environments in the Hart & Risley study, when followed to school, were themselves found to be at a learning disadvantage when it came to reading (Hart & Risely, 2003).

What that all means is that there is good reason to believe that many young children are not receiving sufficient language learning support during the preschool years, and that this insufficiency is implicated in later reading problems. High education, high income families appear to be more able to provide this kind of early language support, than low education, low income families. And, it is possible to provide aid and encouragement to families that allows them to narrow this significant gap (no matter what its actual size may be) (National Early Literacy Panel, 2008).

The Hart & Risley study has been under fire and it is far from a perfect study. Nevertheless, the results of this body of research continue to suggest that what parents do in the home with their children matters educationally (which is why I was using the study in my presentation). Parents need to do more to support their children’s early language learning and parents living in poverty can use some help there. Despite the methodological limitations of the Hart & Risley study, environmental differences (as opposed to genetic ones) still seems to be the best explanation of why poverty kids are underprepared when reading instruction begins.