The Moment We Try to Teach a Hard Word, We Flatten Our Sentences

When 990 minutes of speech from 33 teachers was transcribed and analyzed, what predicted children’s vocabulary was not which words they used but how they joined their sentences

Introduction: What Counts as Good Talk

An adult who wants to speak well to a child usually thinks about words. Better words, more precise words. “That isn’t just big, it’s enormous.” “That feeling is called disappointed.”

Educational practice runs the same way. Vocabulary instruction has been the center of early literacy education: select difficult words, teach them explicitly, explain their meanings, expose children to them repeatedly. There is good reason for it. Vocabulary in early childhood is a powerful predictor of later reading ability and academic achievement.

But there is more than one layer to the speech an adult directs at a child. A team at Temple University in the US put three of them side by side.

• Sophisticated vocabulary — rare words seldom used in everyday talk
• Complex syntax — long sentences with clauses nested inside them
• Decontextualized language — talk that leaves the here and now: the past, the future, conceptual explanation

Research already existed showing that each of the three predicts a child’s vocabulary development. The problem was that no study had put all three into one model.

There is a considerable gap in research exploring these features simultaneously.

Why does putting them together change anything? Because when a person speaks, the three do not operate independently. Using one drags another along — or crowds it out. And this study’s most interesting finding lies exactly in that crowding out.

The Core Question

The team posed three questions:

“When these three language features are considered all at once, which one actually predicts a child’s vocabulary growth?”

1. How much do preschool teachers actually use each of the three?
2. How do the three relate to one another? (Do they move together, or in opposition?)
3. Controlling for other background factors, which one uniquely predicts a child’s spring vocabulary score?

How It Was Measured

The weight of this study is in its measurement. Not a survey — actual classroom speech, transcribed and counted.

What Was Recorded

Literacy activity was filmed for 30 minutes per teacher per semester, in fall (September–October) and again in spring. Three activity types were captured — morning message, book reading, and small-group activity. Fifteen five-minute segments were drawn from each context. For the sophisticated-vocabulary analysis alone, 990 minutes of interaction was transcribed.

Sophisticated Vocabulary

Starting from the Dale-Chall list of 3,000 words known to a fourth grader, and adding common classroom talk and praise expressions, the team built a list of 8,148 “common words.” Anything not caught by that list counted as sophisticated vocabulary. They calculated the proportion of a teacher’s total speech made up of such words.

The result averaged 0.025 — 2.5% of all words. Put another way, roughly three rare words in every 120.

Complex Syntax

Measured with the subordination index (SI): the total number of clauses divided by the total number of utterance units. Not simply whether sentences are long, but how many layers of clauses are nested inside them.

An actual teacher utterance quoted in the paper captures the concept well:

“So maybe that’s why you saw them when you went outside at night, because that’s the time when they’re active.”

Five clauses in one sentence — causal (“so,” “because”), temporal (“when”), and relative clauses stacked on top of one another.

The teachers’ average SI was 1.31: 1.31 clauses per utterance, or roughly one complex sentence in every three.

Decontextualized Language

Watching the video, utterances were divided into idea units and sorted into four kinds.

1. Book information — questions and comments about content recall, inference, prediction
2. Vocabulary information — definitions and examples for unfamiliar words
3. Conceptual information — talk explaining relationships between concepts, going beyond a simple definition
4. Past and future experience — talk about things that have happened or are yet to happen

Classroom management talk, repetitions, and print-decoding instruction were excluded. Everything was converted to frequency per minute, and two or more raters cross-validated (kappa .711–.932).

The Children’s Vocabulary

PPVT-4 (Peabody Picture Vocabulary Test) — a receptive vocabulary measure where a word is spoken and the child picks one of four pictures. Administered twice, in fall and spring.

Participants

• 33 teachers — all with a bachelor’s degree or higher, 80% holding a master’s. 17 pre-K classrooms, 16 kindergarten classrooms
• 421 children — mean age 4.8, 56.8% pre-K, 43.2% kindergarten
• Location — two poorly funded school districts in the US mid-Atlantic
• Background — 85% African American, 15% Hispanic/Latino. All eligible for free or reduced-price meals. 20% were dual-language learners (75% of those speaking Spanish at home)
• Starting point — fall PPVT averages ran about one standard deviation below age norms

That last line matters. These are economically vulnerable children already at a disadvantage in language development, and the point is decisive for interpreting the results.

Result 1: Two Separate Registers Existed

Before the main results, the picture from the correlation analysis is the study’s real story.

Sophisticated vocabulary and complex syntax were unrelated (r = −.22, p = .228). If anything, the direction was negative.

Instead, each paired with something else:

The team read this as teachers moving between two different registers.

The Academic Register

Complex syntax + conceptual explanation. “The reason this happens is that when water freezes it expands, and that’s why the bottle cracks.” Explaining relationships between concepts lengthens sentences and nests clauses on its own.

The Bridging Register

Sophisticated vocabulary + definitions + past/future talk, and simplified sentences. “Enormous means very big. Remember the elephant we saw? That’s enormous.”

The team’s observation is sharp. When teachers used sophisticated vocabulary, they

frequently talked about vocabulary words during book reading, and in doing so commonly used simplified syntax.

That is: the moment a teacher tries to teach a difficult word, they make the sentence easier. They strip out the structure to help the child understand the new word. It is entirely natural, and entirely well-intentioned. And that is exactly where the problem was.

Result 2: What Predicted Vocabulary Was Sentence Structure

This is a multilevel regression controlling for fall vocabulary score, child gender, dual-language status, grade, and program participation.

The model with all three features entered at once:

Only complex syntax was significant. The standardized coefficient was .88 — a teacher one standard deviation higher in syntactic complexity predicted a child’s spring vocabulary score about 0.88 standard deviations higher.

Breaking decontextualized language into its four kinds and looking again, only conceptual explanation was significant (b = 2.50, p < .001, standardized .67). Not book talk, not vocabulary definitions, not past/future talk.

The final combined model — sophisticated vocabulary + complex syntax + conceptual explanation all entered:

Conceptual explanation’s predictive power vanished. Complex syntax stood alone.

This does not mean conceptual explanation is meaningless. It is closer to saying that the route by which conceptual explanation lifted vocabulary ran through sentence structure in the end. Explaining how concepts relate makes sentences complex, and what the child gained was those complex sentences.

One Thing to Be Clear About: This Does Not Mean Sophisticated Vocabulary Is Harmful

The coefficients came out negative, but the p-values are .922 and .696. These are numbers closer to nothing can be concluded than to there is no effect. The sophisticated-vocabulary values themselves clustered in a very narrow range of 0.009–0.053, which makes the coefficient magnitudes hard to read into.

The accurate interpretation is this: once other factors are considered together, using a lot of sophisticated vocabulary did not separately explain a child’s vocabulary growth.

Why Sentences and Not Words?

Syntactic Bootstrapping

The most persuasive explanation. When a child meets a word they don’t know, they infer its meaning using where and how the word sits inside the sentence as a cue. This is called syntactic bootstrapping.

In “the rabbit hopped nimbly away,” the child knows from its position that “nimbly” modifies how the hopping happened. The more complex the sentence, the more such cues there are. When clauses are nested, the relations of who did what and why are laid out inside the sentence, and the child can locate the unknown word’s place within them.

The team sees ages 4–6 as when this transition sharpens. As children come to recognize increasingly complex syntactic structures, syntax’s influence on vocabulary acquisition grows.

Cognitive Load Builds the Capacity

Following nested clauses requires working memory — you must hold the content of the first clause while connecting it to the next. The team argues that exposure to embedded clauses promotes the development of relational thinking, processing efficiency, and executive function.

Chop the sentence short and that load disappears. Comprehension gets easier, but the muscle that produces comprehension goes unused.

The More Vulnerable the Child, the More They Lean on Sentences

This is the interpretation that stays with you longest. The team writes that these children

benefited the most from teachers’ complex syntax despite their limited opportunities for exposure to school-related vocabulary

and adds: when vocabulary breadth is limited, children may rely more heavily on syntax to infer meaning.

Which means the fewer words a child knows, the more urgently they need the cue that sentence structure provides. And in real life the opposite is what tends to happen. When an adult thinks a child won’t follow, they chop the sentence short. The children who need help most end up hearing speech with the cues stripped out.

What Has to Be Said Honestly

Before translating this study into “talk in complex sentences and your child’s vocabulary grows,” several conditions must travel with it.

This Is Observation, Not Experiment

Teachers were not randomly assigned and told “you speak complexly, you speak simply.” Teachers who already spoke that way were observed.

So causality is not established. The reverse direction is possible — teachers may have naturally spoken more complexly in classrooms of higher-language children. Controlling for fall scores substantially reduces that possibility, but does not eliminate it.

Thirty-Three Teachers Is Few

The key predictor (syntactic complexity) is a teacher-level variable. Even with 421 children, the data points supporting that variable number 33. The final model’s teacher-level explained variance of R² = .97 has to be read in that context — a figure that arises easily when data points are few, and hard to expect to replicate in another sample.

And the syntax analysis used 600 words per teacher: the first 100 words from each of six contexts. That is a thin sample for representing one person’s speech habits.

The Population Is Very Specific

Poorly funded US school districts, all children on free meals, starting a full standard deviation below norms. The team itself explains that children with limited vocabulary breadth rely more on syntax — so whether this result applies as-is to children in vocabulary-rich environments is unknown.

Limitations the Team Named

• They did not measure the children’s syntactic ability. There is no data confirming whether teachers’ complex sentences actually grew children’s syntactic comprehension or production. The proposed mechanism is untested inference
• The books read were not controlled. Some teachers read storybooks, others informational books. Book type may have shaped teacher register
• Only structured activities were observed. Free play, sharing time, snack time — other contexts where decontextualized talk would occur — were not analyzed
• The definition of “complex syntax” varies across studies. This study measured clause density; others use mean length of utterance. That may be one reason findings conflict with prior work

Practical Takeaways

Notice What the Effort to Simplify Strips Out

The most practical finding here is not the result but the correlation analysis. Teachers simplified their sentences precisely when trying to teach good words.

The same thing happens at home. “That’s, um… it’s a really big one. You get it?” In front of a hard concept we break the sentence apart. It comes from good intentions. But if what the child takes away is sentence structure, this good intention is clearing away the very thing that matters.

The alternative is to make the words easy while keeping the sentence structure alive. “That’s enormous” is worse than “look how that ship is bigger than our house — a person could sleep inside it.” Not one difficult word, and three clauses.

“Why” and “So” Make Sentences Complex

There is no need to force long sentences. What clung most tightly to complex syntax in this study was conceptual explanation (r = .62). Try to explain a relationship and sentences nest on their own.

• “Take an umbrella.” → one clause
• “When it rains your clothes get wet, and if you’re wet you’ll catch a cold, so let’s take an umbrella.” → four clauses

There is no hard word in the second sentence. It was only trying to explain cause.

Themes Over Word Lists

The team’s own recommendation runs this direction. Instead of pulling out individual words to teach,

using content-focused approaches such as thematic units that introduce vocabulary under a single conceptual theme, provide extended explanations, and build connections between concepts may help teachers naturally integrate more advanced language structures — particularly complex syntax — into interactions.

Not “five words of the day” but “this week we talk about water.” Hold onto a theme and explanations lengthen; as explanations lengthen the structure grows complex; and the words find their place inside it.

Following Along and Learning Are Different

If we take a child’s nod as the signal of success, we will keep making our speech easier. But there is nothing to learn from speech that is fully understood. The load of straining to follow a slightly-too-hard sentence is the thing this study connected to growth.

Closing Thoughts

When we ask what we are giving a child, we usually picture a list. So many words, so many books, so many concepts.

This study looks at what lies beneath the list: the shape of the vessel holding the words. The same word placed in a different structure changes how many cues a child has to grab onto.

What a child needs may not be better words, but sentences that show how words are tangled up with one another.

This is an observational study, the story of 33 teachers, and a result drawn from children already at a linguistic disadvantage. It is not a settled law.

But one thing is worth thinking about. In the moment we cut a sentence short for fear the child won’t follow — what is it, exactly, that our good intentions are clearing away?

Source: Farrow, J., Wasik, B. A., & Hindman, A. H. (2025). Exploring the relations between teachers’ high-quality language features and preschoolers and kindergarteners’ vocabulary learning. Journal of Child Language, 52(6), 1338–1366. https://doi.org/10.1017/S0305000924000485