Researcher in Profile: Ali Horowitz on Inferences From Words

By Chia-wa Yeh, Head Teacher and Research Coordinator

Research is an integral part of Bing Nursery School. Each year about 40 Stanford researchers work with children at Bing as participants in their studies. One of these is graduate student Ali Horowitz. Over the past five years, Horowitz was often seen in the classrooms with a big smile, listening to and talking with children, noticing and appreciating what children were doing. Her series of studies at Bing explored whether children make inferences about language that go beyond what’s explicitly stated.

Horowitz’s interest in how young children learn started early. Her parents believe that children can overcome setbacks and thrive with love, attention and support. In addition to raising three biological daughters, they took in foster children throughout her childhood, mainly infants and toddlers. While she was growing up, one to two foster children were in her home at any given time. The house was always lively and fun, with lots of singing, playing and art and science projects. Horowitz enjoyed taking on a big sister role to help care for the foster children, watching how quickly young minds develop, and helping to support their learning.   

Horowitz attended the University of Rochester where she completed a bachelor’s in brain and cognitive sciences, with minors in psychology and Spanish. She went on to be the lab manager for the Early Childhood Cognition Lab at the Massachusetts Institute of Technology before combining her interests in language and development in graduate study in the Language and Cognition Lab at Stanford with her advisor, associate professor of psychology Michael Frank, PhD. She earned her doctorate in developmental psychology this year.

In speech, adults can infer what meaning speakers want to convey by listening to not just what people say, but how and why they say it. For instance, “strawberry yogurt” conveys that yogurt can come in different flavors and “art museum” conveys that there are different types of museums.

Are young children sensitive to cues implied in speech? To answer this, Horowitz investigated children’s inferences about the use of adjectives. Do young children pick up on subtle implications when hearing adjectives that convey dimensions of contrast? For example, if one points to a new toy and say, “That’s a big ‘blicket,’” it conveys not only that the toy is called a blicket but that this blicket is big, size is relevant to this blicket, and other blickets are probably smaller. So from a single sentence, listeners can gain information about this particular blicket and also about what other blickets are like.

In Horowitz’s studies, she introduced 3- to 5-year-olds to pictures of novel shapes and described them using either size or property adjectives (e.g., “This is a special kind of zib. This is a small [or broken] zib”). She then showed the children two similar pictures, one that differed from the first only by size (e.g., big and small) and one that differed from the first only by property (broken and unbroken). Afterward, she invited the children to point to the one of the two they thought most zibs looked like. The study examined whether children would choose the picture that differed from the adjective provided. For example, whether children who heard “This is a broken zib” would select an unbroken one as what most zibs looked like.  

The researchers found that by age three and a half, children were more likely to select the picture that contrasted with the adjective named. In other words, if they heard “small,” they were more likely to select the picture that was big, and if they heard “broken,” they were more likely to select the picture that was unbroken.

The researchers’ findings suggest that preschoolers are learning to infer about what information is conveyed by speakers’ word choices. Although they could choose between a picture that matched the adjective named (e.g., hearing “small” and selecting the other small picture) or one that contrasted with the adjective named (hearing “broken” and selecting the picture that was unbroken), children selected the contrast. They seemed to be reasoning that the adjective implied a dimension of contrast with other zibs. In other words, they appeared to be reasoning about not just what was literally said, but what underlying information was implied from those descriptive modifiers.

“Adjectives are just one example of word choices that provide cues to underlying meaning. Once children can consider why speakers say things in particular ways—what knowledge, perspective and intentions they have—they can gain information about what the world is like that would lead speakers to make those descriptive choices,” Horowitz commented. Hence, explained Horowitz, the more children can pick up on information that goes beyond what’s explicitly stated, the more they can access opportunities for learning. Making inferences about language use enables children to become more efficient communicators and learners.

Horowitz’s study has been accepted for publication by the journal Child Development.