The Interplay of Innateness and Statistical Learning in Language Acquisition

In my first presidential column in October, I highlighted interdisciplinary work between cognitive psychology and other subareas of psychology, focusing on work integrating social–emotional factors with cognition. In this column, I am turning to interdisciplinary research on cognitive and language processes—together with their brain bases—that aims to uncover fundamental properties of human learning, specifically with regard to language acquisition.  

To address this, I spoke with Elissa Newport, who will be one of the plenary speakers at the 2025 APS Convention. Throughout her long and distinguished career, she has reported groundbreaking work providing evidence for both innate factors and statistical learning in children’s acquisition of language. Much of her research has involved purely behavioral studies, but beginning in 2012, she went in a new direction, examining the effects of early brain injury on language reorganization in the brain to provide insights into language learning.  

In high school, Newport was interested in both math and English but at the time couldn’t see a way to combine the two. So she decided to pursue English when she started at Wellesley College (in the same class as Hillary Clinton—quite an impressive class!). Though she appreciated the great professors and education at Wellesley, she decided to transfer to Barnard College, both because of a boyfriend at Columbia University and because of greater opportunities for political activism regarding the Vietnam War in New York. At Barnard, she started taking psychology courses, which she loved because all lecture classes required participation in a lab course where students replicated studies that they had read about, such as rat learning in a Skinner box. She graduated magna cum laude from Barnard in psychology and then accepted an offer to attend the clinical psychology program at the University of Pennsylvania. 

Martin and Newport spoke on February 27, 2025. An edited version of their conversation is below. 

MARTIN: How did you move from clinical psychology to the psycholinguistics of language development in graduate school? 

NEWPORT: I chose the University of Pennsylvania because they had a clinical program that was focused on doing experiments with groups who had different types of disorders rather than on training people to do therapy. Then, after I got there, that wasn’t what I did at all! At the time at Penn, we had fabulous proseminars in our first year. I read Eric Lenneberg’s work in a physiological psychology proseminar, and I was just stunned and taken with his view that there is really an instinct component or innate component to all kinds of behaviors that we would ordinarily think of as entirely experiential. From that, I got interested in critical periods for acquiring different behaviors, and I’ve sort of been living the life of Eric Lenneberg ever since! 

MARTIN: While at Penn, I saw that you ended up working with both Lila and Henry Gleitman. 

NEWPORT: Yes, that’s right. I worked first with Burt Rosner on autism and then spent a year working with Paul Rozin, looking at patients with aphasia, a language disorder that affects an individual’s ability to communicate. At the same time, I started talking with Lila Gleitman. I took her linguistics courses at Swarthmore College, where she was the one-woman linguistics department at the time. I ended up switching to work with Lila and Henry and did the remaining years of my dissertation with them on language acquisition.  

MARTIN: Your dissertation was on the Motherese concept, that is, the idea that the pattern of speech that mothers use with infants (for instance, high pitch variation and a lot of repetition) would influence their child’s language acquisition. I remember studying that notion when taking psycholinguistics in grad school.  Where does that issue stand now? 

NEWPORT: Yes, the title of my dissertation was “Motherese.” That was a term that Henry coined. The issue has always been controversial. A lot of the people who first got interested in mothers’ speech to infants viewed any effects as providing evidence opposite to the Chomskyan view about language. That is, they were arguing that it’s really because of the simplicity of mother’s speech to kids that they acquire language, not because of anything innate. I always found myself disagreeing with most of that literature. What I did in my dissertation was record mothers talking to their kids and compare that with their talking to me. In articles that I published with Henry and Lila, I argued that it wasn’t entirely simplification that mothers were doing. They were expanding the types of sentence structures that were used. When mothers talk to little kids, they don’t use all simple active declarative sentences, which would be the simplest from the point of view of learning the structure of the language. Instead, they use a lot of questions, a lot of imperatives, a lot of deletion—things which technically are considered quite complicated. I also argued that, although mothers repeat themselves a lot, kids don’t actually benefit from the repetition. Some of the strategies that mothers use occur because the kids don’t do what the mothers are asking them to do, not because this is a beneficial technique for language learning. After that, I started asking, what is it about children that helps them learn languages? I don’t think it’s due to the particular way that people speak to young children. And it’s been shown by many other people that mothers don’t talk to kids in this American middle-class way around the world, but kids around the world acquire languages just fine. 

MARTIN: Yes. I remember reading about some culture where the adults hardly talk to infants at all because they don’t see the point. But, of course, the children acquire language anyway. 

NEWPORT: Ever since then, I’ve been interested in exactly what the nature of the learning process is. I started being interested in learning way back at Barnard, but I have come to think about a different kind of learning—a different approach to thinking about how children learn. 

MARTIN: Something salient in my mind on the learning topic is your involvement with the statistical learning approach. I remember that Jen Saffran—the first author on the 1996 Science paper on infants’ learning to segment the speech stream into words, with you and Dick Aslin as coauthors—interviewed at Rice University after that paper came out. I also recall she got about 12 job offers!  

NEWPORT: That work was her first-year project with Dick and me. Not a bad first-year project! Dick and I had been interested in some work by John Hayes and Herbert Clark that turned into thinking about statistical learning. That is actually a term from computer science literature. Eugene Charniak wrote a book about statistical language learning that we were reading in our seminars at the time. In computer science, language learning doesn’t really mean “learning” but more like “parsing.” We argued that the kind of statistics that they were talking about—frequencies of words, frequencies of words appearing in certain contexts, and so forth—were a possible basis for the brain to do this kind of thing in the process of language learning. So we borrowed the term but applied it to human learning and then titled our Science paper “Statistical learning by 8-month-old infants,” which I thought would be really striking and sort of shocking. 

MARTIN: Yes, I remember that the findings were very striking because the infants learned to segment speech into words after something like 2 minutes of exposure to speech. 

NEWPORT: I still do quite a bit of work on statistical learning. In fact, I recently got a new grant with Heidi Getz (an assistant professor at Georgetown University) related to statistical learning. In the earliest work, we wanted to start with something very simple, and we were working on word segmentation. How do you find the words in the stream of speech? But that’s not really the aspect of learning that I’m most interested in. I’m really interested in how kids learn the grammar or the structure of sentences. What Heidi and I have been doing lately is looking at memory consolidation during the process of learning.  That is, what happens when you expose kids to miniature artificial languages that have short sentences and simple sentence structures? Instead of using a multiday exposure procedure to learn something complicated, we have begun to examine the learning of these simple sentence structures over a delay. We expose the kids on day one and then test them on day two or day five. And, surprisingly, we find that they’ve dramatically improved!  We’ve also been doing imaging with children, showing that hippocampal activation during and after day one exposure seems to mediate changes over time in the cortical representation of what they’re learning from these artificial languages. 

MARTIN: We’re born as little learning machines! When looking over your publications, I was reminded of your findings on critical periods in language learning (i.e., time periods in which the brain is very sensitive to learning certain types of information).  Also, you were involved in studies on the creolization of sign language in Deaf kids who hadn’t had good exposure to signing in the home but nonetheless developed sign language systems that were more complicated than the input they received. So, these lines of work indicate that there is this innateness aspect of your findings, which goes along with what you were saying about Motherese. Then there’s also the statistical learning finding, but people have often thought of the two as being in opposition. So, is your approach a combination of the two? 

NEWPORT: I’ve thought of them as fitting together. I think the right way to look at learning is to expect that a really good learner has to come to the problem with biases and preferences that shape the way the input gets stored and how it gets processed—affecting what a learner pays attention to, what parts of the input are salient,  what kinds of patterns get noticed, and what patterns don’t. But, then of course, to acquire a language, learning has to be involved. 

MARTIN: Of course, because everyone learns the language they are exposed to. 

NEWPORT: Right, so language isn’t totally innate. It’s not totally learned either. But I think of innateness as more like what a psychologist thinks of innateness, as opposed to the way innateness is thought of in theoretical linguistics. In theoretical linguistics, people have suggested that we innately know principles of languages. In my view, what is innate are the biases I mentioned because of the way the brain is organized. For example, humans have a tendency to group things and to organize things hierarchically. We do that with language, which totally shapes the way languages of the world are organized. But we do that in other domains as well.   

MARTIN: Yes, you can think of this hierarchical preference in the way music and motor execution are organized. 

Let’s turn to the work that you’re going to be talking about at the APS Convention, on the lateralization of language after perinatal stroke. How did you get interested in this topic? This seems a bit of a change for you to be doing this kind of work. 

NEWPORT: It required retraining, which was very, very significant and important. But it is actually the same question I’ve always been interested in. It’s a question that Eric Lenneberg posed in the article I read in grad school. It stems from some issues he raised in 1967 about critical periods. He suggested not only that kids are better language learners than adults, but also that kids recover from brain injury differently than adults. Lenneberg collected all the case studies he could find in his own patients and in the literature at the time—but he didn’t detail exactly what kind of injury these cases had. He classified the cases according to the age at which the kids had an injury, whether the left or the right hemisphere was affected, and what the language outcome was. What he argued is that very young children seemed to reorganize language to the opposite hemisphere and recovered well. But with increasing age, that ability went away. Of course, if adults have a serious brain injury to the language areas of the left hemisphere, they don’t recover as well. They generally don’t get back to pre-stroke abilities at all, and they don’t recover in the opposite hemisphere.  

Now, obviously, he didn’t have imaging data as this was 1967. He was basing these outcomes on clinical data. I was really interested in these patterns of good recovery in the kids. When I was at the University of Rochester, where I was for 24 years, I was working on all the behavioral research that you mentioned—statistical learning, creolization, critical periods. I was really interested in why kids are better at some things than adults. As chair of the Department of Brain and Cognitive Sciences at Rochester from 1998 to 2010, I interacted more and more with neuroscientists and started doing imaging with Daphné Bavelier. When I stepped down from being chair, I wanted to do a study of age effects on language acquisition, such as age at brain injury, like the studies Lenneberg talked about. I ended up being offered an appointment to start a center at Georgetown for this purpose. So I developed the Center for Brain Plasticity and Recovery, where I’m still the director. 

View past presidential columns

MARTIN: Why do this work at Georgetown and not Rochester? 

NEWPORT: I wanted to focus on stroke because you could have the same kinds of damage to children at different ages, with strokes to the same regions as adults have. Obviously, I needed training and a volume of patients, which was more available at Georgetown. Childhood stroke is very rare. That’s why I moved. I visited Georgetown for a year, met people here (especially my neurologist colleagues Alex Dromerick, Bill Gaillard, and Peter Turkeltaub) and found that I could make the right connections to make this research possible. Alex introduced me to John Lynch, a neurologist at the National Institutes of Health (NIH), who had basically defined perinatal stroke in the literature—that is, stroke at birth. In 2012 I got a K18 award, which is the only grant that I know of from NIH that’s for training a senior investigator in a new field. Through my award, I received training from stroke neurologists, and I went on rounds in the stroke service and observed rehab for adult and child stroke survivors. Lynch advised me to start with perinatal stroke because the frequency of perinatal stroke is much greater than strokes later in childhood. But strokes at birth are still very rare, one in 4,000 live births, whereas strokes after that in childhood are about one in 30,000. 

MARTIN: Do you see any evidence of language deficits for the left hemisphere perinatal stroke cases? 

NEWPORT: We certainly expected that there was going to be something early on. But they don’t, in our data, whereas in other people’s data they do. Looking at other studies it does look like the kids who have a left hemisphere perinatal stroke show some delay in language acquisition, which makes some sense. Our studies are looking at the long-term outcomes, the language abilities when the children are older. 

On the whole, our participants often do show executive function impairments. They are slower on speeded tasks. They have more restricted short-term memory. Their IQs are about a standard deviation below average but still within the normal range. There’s a range, though. They’re also honor students in school, and some have PhDs or are speech therapists. One is an architect. Many are very accomplished. One of the kids we’ve tested sent me an email last fall telling me he was starting his freshman year at Yale University. We have sent recordings of participants telling a story to Nan Bernstein Ratner and other speech therapists at the University of Maryland, and they could not predictably classify who had a stroke and who was a sibling. 

MARTIN: That’s fascinating. I know that you have looked at other fascinating questions, but I don’t want to cover everything you’ll be talking about!  

One final thing I would like to touch on is the current situation and threats to funding scientific research. A lot of students are very concerned and upset. Do you have advice for undergrads or graduate students at this point who are thinking about careers in research and psychology?  

NEWPORT: I think this will return to normal. I don’t know how long it’ll take. Certainly, the Trump administration is not going to last forever, and I think there’s already some alteration of public opinion. We work in a field that is really important to human health and to human achievement and success, and the country will not turn away from that. It’s really important to our understanding of how people develop and to our offering support for people who have impairments. It’s also important for the economy, which is not a Republican versus Democratic issue. There used to be strong bipartisan support for research. I think that that will return and continue. I think that this current administration is an aberration in our history.   

I think that we have to hang in there, but we also must become better at talking about the importance of what we do in a nonpartisan way. I think we have to make sure that people understand what we do and why it’s important. They need to understand, for example, what are indirect costs and why would that affect us?  

MARTIN: Yes. I’ve certainly become aware that many of the people I interact with in everyday life, like my hairdresser or my Pilates instructor (both of whom are college graduates), have little sense that universities carry out research that is very important for all kinds of social issues, including human health. 

NEWPORT: Jessica Cantlon and Steve Piantadosi have started a project called Science Homecoming. Their project is trying to get everyone, including scientists who might have grown up in red states, to write articles very accessible to the public about where they went to school, how they got into the field, and what they do. What’s the importance of their work to everyday people and to the economy? The goal is to get the articles published in local hometown newspapers. I grew up in St. Louis, Missouri. I’m writing one that I’m going to try to get published in the Jewish Light, which is a little tiny newspaper in the area where I grew up. I think it’s a really interesting project because, as Jessica notes, we can’t afford to have other people speak for us. We need to be able to speak to the people who pay taxes that support what we do and hope that we can communicate well enough that they understand why what we do is important to them. I think that a project like that is really important. 

I still believe that no matter what the tides look like these days, most people are good-hearted, empathetic, and care about the welfare of their neighbors and their family. So, I think this kind of administration will not last very long, and we will see the rise of those who were just worried about the price of eggs coming back to worrying about their neighbors. 

Feedback on this article? Email apsobserver@psychologicalscience.org or login to comment.