top of page

Research corner: Making sense of our study

Words by: Monica Ellwood-Lowe

In this post, we explain some of our thoughts behind the design of the study, and feature data visualizations from our preliminary data! So just a heads up: it is graph heavy. We'll try to walk you through it step by step. As always, feel free to reach out with questions!


Scientific research often talks about parenting as if it happens in a bubble. It sometimes sounds like parenting is this clear, consistent behavior—that it all comes down to things like a parent’s skill, their effort, or their beliefs.

But anyone charged with raising a small child knows no two days are the same. Couple that with the instability of 2020—a year of sudden job losses and health scares and intense social unrest, as the structure of society softened and shifted shape more frequently and the pressures associated with parenting amplified—and it has become more clear than ever that this story of isolated parenting is incomplete.

Part of the problem is that a lot of our scientific research has compared parents to one another. This parent talks more to their child than this parent, the research says. And then the research asks, why are these parents different from one another?

What we want to understand is something different: why this parent[1] talked more to their child yesterday than they did the day before. In other words: what affects the same parents each day?

(We are primarily language researchers, so we start out by looking at parents’ words. But this is true in so many different areas!)

In the early days of the pandemic, we began to collect repeated audio recordings from families of a regular routine, young children’s bath times, along with information about how the parent was feeling and whether there had been any larger, structural changes in their lives. The goal was to collect 30 to 60 recordings per family, allowing us to compare each family to themselves, and perhaps capture how these larger structural changes happening outside of the household affected individual families and their speech to their young children.

[1] We use the term parent loosely to refer to any adult who assumes a large caretaking role in an infant’s life.

Our study

Since May 2020, we have collected over 2,000 recordings and surveys from families of 1–2-year-old children in the U.S. In addition to fluctuating government restrictions with regard to the pandemic, we have captured high-profile police murders and the nationwide protests to follow, the wildfires that devastated parts of the West coast, the election period and subsequent insurrection at the capital. And finally, we captured increasing vaccination rates and a stumble toward a new normal. As we work to transcribe and analyze the data, there are some early insights that have become clear.

Each day is different

We asked families to audio record their child’s bath times because bath times are a routine that happens most days for most families of young children, and that require some caregiver involvement. Even so, each bath time was surprisingly different. For example, the graph below shows one child’s language environment during bath time over a 60-day period. While this child bathed, a number of people were often present: their mother, their father, and a sibling. The language they are exposed to reflects that, as they are spoken to often but also frequently overhear speech to others.

On the y-axis, you can see the number of distinct words[2] spoken during the child's bath time each recording, colored by who those words were spoken to. (This doesn't include the words the child themself spoke.) Each bar on the x-axis represents a different bath time. The bottom layer of the bars, colored purple, shows how many words were directed to the target child—the one of interest in our study. Just above that, the teal color indicates words spoken to another child—the child's sibling. Green shows speech to adults, and in one bar red shows up on top to indicate the speaker was talking to herself. You can hover over the bars to get more information about who the speaker was ("mot" indicates the mother, "fat" indicates the father) and how many words they spoke to whom during each bath time.

As this graph makes clear, the child's bath time language environment varied widely. For example, during bath time recording #34, both mother and father only spoke to the child's sibling and to one another. But during recording #6, the roles were reversed—the mother spoke almost exclusively to the target child, and just a little bit to the sibling.

[2] To better understand, “scrub scrub scrub” would be counted as one word on this graph. "Scrub scrub ducky" would be two. Language researchers think of the number of distinct words as an indicator of word exposure, which relates to children’s vocabularies.

Speech varies... a lot

Let's narrow in for a second on three things. First, it's possible that the variation in words we're seeing each day has to do with the length of bath time alone. To rule out this possibility, we're going to switch to looking at the rate of words: the number of distinct words spoken per minute of bath time. Second, we're going to home in on the parent who filled out the survey for our study, since they are the one we have more information about. In this case, it was the child's mother. Finally, we're going to look specifically at that parent's words to the target child—what we call child-directed language—a kind of language that researchers have looked at a lot because it relates to children’s vocabulary development.

In the graph below, the mother's words to the child per minute are shown on the y-axis, and bath time recording is shown on the x-axis.

Animated graph featuring variability in child-directed words per minute for one family, across 60 bath times

Again, it is clear how much variability there is each day!

Is this family unique in terms of variability? Our early data say no. Below, you'll see the same graph, but now it includes data from 28 families, each of whom has a line of a different color. (Note that the number assigned to each family is random and does not actually correspond to their Study ID.)

Animated graph featuring variability in child-directed words per minute for each family, across 60 bath times

If this graph is a lot to take in, it's because there is so much variability within each family, and overlap across families! An interactive version is below. You can hover over lines to see more precise information, or double click on the legend to see just families you select.

If there is anything these graphs make clear, it is that comparing families against one another might not be the most scientifically accurate or practically helpful. Despite these families coming from all different backgrounds and having all different bath time routines, there is not a single family whose rate of words to their child doesn’t overlap with other families. Variability is the norm, not the exception.

Each family is unique

Finally, while families seem to be united in their variability, the early data gives us a hint that the underlying cause of that variability may be different across families. In other words: what leads one family to talk more one day and less the next is likely to be completely different than another family.

For example, one thing that has centered heavily in conversations during the pandemic is the burden on caregivers to balance work with parenting.

One hypothesis might be that parents talk less with their children after a long day of work, because they are tired out. Another might be that parents are more engaged, because they have missed spending time with their child.

In our data, we see that the effect of work varies for different families. The graphs below shows the link between the percent of time parents spent working on the x-axis, and the rate of their speech to their children that day at bath time on the y-axis. You can toggle between graphs for 8 different families.

In some families, the parent appears to talk less when they have worked more that day (like 17). But that certainly doesn't seem true for all families (like 27). And in other families, the parent doesn't work at all during the study period (like 18)!

While preliminary, eyeballing these graphs highlights how important it is to understand each individual family dynamic.

Wrapping up

What we've learned so far is:

  1. Most research focuses on parenting differences by comparing one parent to another

  2. Our research instead focuses on day-to-day differences in the same parent's parenting over time

  3. There is a ton of day-to-day variability in how much the same parents talk to their kids during bath time, and it's not explained by bath time length

  4. The underlying cause of this variability may be different across families

The funny thing about the insights from our data is that they should come as a surprise to no one—parents or researchers. And yet, our understanding of what really does affect individual families and their interactions with their children is sparse.

Stay tuned as we continue to delve into the ins and outs of bath time, language, and parenting in the pandemic!


bottom of page