Tracing the origin and trajectory of my research career requires telling two stories, one for the first twenty-five years when I focused primarily on language development in young children, and quite a different one for the next twenty-five years after I pivoted to a focus on literacy development and older learners. If there is any theme that recurs across those two epochs, it is that the larger fields in which my own work has been located were characterized by heated disagreements among the leading scholars and spokespersons. Furthermore, those disagreements (language is innate versus environmental; reading instruction should emphasize the code or meaning) somewhat puzzlingly transcended scientific inquiry, questions that could be answered with well-designed studies, to become political, almost religious, commitments. For some reason, nativist views of language acquisition were seen as left-wing and universalist, whereas those who emphasized environmental influences were identified with positivism. Similarly, proponents of structured phonics instruction in early reading (called the “phonicators” by their opponents) were identified with right-wing positions, and those arguing that phonics instruction could be overemphasized were the free-spirited progressives. It is not hard to see the potholes and pitfalls on both sides of these conflicts.

My work since about 2000 has been located firmly in what is often called Pasteur’s quadrant – balancing an interest in contributions to theory with a focus on contributions to improved practices and policies. I started out, though, in the adjacent Niels Bohr quadrant of the famously tessellated analysis of scholarly activity: interested in testing various hypotheses about the nature of language development (or, as we were required to call it in 1970, language acquisition). I shifted in the late nineties to a focus on literacy, a developmental challenge for which the practical, real-world concerns outweigh the theoretical, though both are present. More recently, though, I have come to realize the extent to which language learning is a very large part of literacy success: learning the new language forms required to understand and produce literate texts and exploiting the give and take of oral discussion to build the language and critical thinking skills needed for literacy. At every stage, the primary influences on my trajectory have been purely serendipitous; while I would in general promote planfulness as the best model for an academic career, planlessness has not served me badly.

Story Number One: How Do Children Learn to Talk?

In the study of how children become linguistically competent, the term “acquisition” is used to signal a nativist view of the process – that there exists a Language Acquisition Device (LAD), an innate, universal, inherently human language capacity that simply needs to be triggered by some minimal exposure to the environmental language. The argument for an innate Language Acquisition Device was buttressed by claims about how unhelpful the language in the child’s environment would be as a support to acquisition. The language input was characterized as chaotic, garbled, and ungrammatical, thus clearly not displaying features that could actually help a child “develop” language and leading inexorably to the nativist invocation of the Language Acquisition Device (LAD) as an explanatory principle. This nativist view, introduced by Chomsky in the 1960s, quickly gained the intellectual high ground, becoming the dominant discourse for researchers in the field.

The nativism-associated assertion about the nature of speech addressed to children was, as it happens, unsupported by direct observation of parent-child interaction. Moreover, in 1968 when I reviewed the then tiny body of work that had been done on child language for a seminar presentation to my fellow graduate students in the McGill University Psychology Department, it seemed to me flat wrong or at least wildly exaggerated. Thus, I set out to examine, for my PhD thesis, two questions: (1) How do mothers talk while interacting with their two-year-olds, using those same mothers’ interactions with their older children as a comparison? (2) Are children equally attentive to talk that is simple and repetitive (as their mothers’ speech turned out to be) versus talk that is more complex? I published a paper in Child Development in 1972 reporting the by-now-familiar finding that mothers’ speech to language-learning children is grammatically simple and lexically redundant – a language register that Jerome Bruner would subsequently refer to as the Language Acquisition Support System (LASS).

My research on how mothers spoke to children was motivated by doubts about the nativist view of language development and by curiosity about the actual nature of the linguistic information available to language-learning children – a theory-driven concern. My first academic position, furthermore, was in a linguistics department, where contributions to knowledge were more highly valued than knowledge applications. During my sojourn of seven years at the Institute for General Linguistics of the University of Amsterdam I carried out further studies of mother-child talk, but even more importantly I had the chance to learn some linguistics, a subject I had never studied even though it was central to my contemporaneous and subsequent interests.

It is important to note, though, that the topic of parent-child interaction and of environmental language as a support to language acquisition remained distinctly déclassé during the last two decades of the 1900s. The more mainstream researchers devoted themselves to describing the rules young language learners were implicitly formulating, with no attention to their environments. The classic studies of the field (carried out by Roger Brown, Martin Braine, and Lois Bloom) proceeded from the assumption of universality, that is, that any child’s acquisition data would reveal the process as well as any other child’s. Linguistic anthropologists who were undertaking descriptions of language acquisition by children in Samoa, Papua New Guinea, Mayan villages in Mexico, and tribal Africa focused primarily on showing how little language children heard and how little linguistic interaction they had with adults in those settings, thus shoring up nativist claims. Suggestions that children from financially or educationally less well-resourced homes might show linguistic or cognitive delays because of the nature of the language interactions to which they had access were dismissed as racist within the nativism-influenced child language community, even as programs like Head Start, based on precisely such assumptions, were being launched.

As a relatively junior, generally conflict-averse scholar, one who furthermore had no collaborators or mentors to turn to for spine-strengthening advice, I was disinclined to mount the hustings in support of claims about how language input contributed to language acquisition. Furthermore, the field had not at that time advanced to include large-scale data collection, of the kind that could be used in robust efforts to demonstrate relationships between input and rate or outcome of learning. Child language research was still largely stuck in the n=3 mentality.

A corollary of the nativist position was an assumption of radical differences between first language acquisition – the product of the hypothesized innate human potential – and subsequent language learning. The nativist position was biologized as implying a critical period for language acquisition, ending in various iterations at age five or seven or twelve, after which the acquisition process was very different and the outcome much less successful. (Eric Lenneberg developed this theory in some detail in his 1967 book, Biological Foundations of Language.) I did not start out to contest this view, since I had no particular reason to reject it. I was simply interested to study in greater detail the context for second language learning available to older versus younger learners, hypothesizing that younger learners might be more successful in part because they had a more facilitative learning environment.

Since I was then living in the Netherlands and had access to many native English speakers who were learning Dutch, I undertook with Marianne Hoefnagel-Höhle a study of the progress of L2 learners of different ages, often all the children and adults in a newly arrived family. It turned out, to my amazement, that the younger learners made considerably less progress in Dutch over the course of the first several months of second language learning than their older siblings, or even their parents (Snow & Hoefnagel-Höhle, 1977). The presumption that younger, more plastic brains would enable faster and easier learning was not borne out. The fastest learners were, in fact, the adolescents, with adults not too far behind. Taking into account that the adults were in much less supportive environments (often working in contexts where the need to use Dutch was minimal), this finding led me to question the notion of a biologically constrained critical period for second language learning (Snow, Reference Snow and Bornstein1987), a critical position that was not so much hotly contested as dismissed out of hand. Subsequently, studies by David Birdsong and others have confirmed the possibility of achieving native-like control of a second language starting in adulthood (see also Marinova-Todd, Marshall, & Snow, 2000), and have clarified that even universal shifts in susceptibility to learning, such as the one that leads to difficulty discriminating sounds not phonemically distinct in one’s own language, are not as absolute as the critical period hypothesis might suggest.

Meanwhile, though, at some point between 1972 and 2000 a sea change occurred in the field of child language. Certainly, the publication of Hart and Risley’s book (1995) constituted one trigger, as did growing concerns about educational equity, evidence that children at risk of failure in school needed substantial time in excellent early childhood programs to be school-ready, and finally just the massive accumulation of evidence that children’s language outcomes were at least partially explained by the quantity and the quality of the language available to them in their immediate environments. (See Anderson, Graham, Prime, Jenkins, & Madigan, Reference Anderson, Graham, Prime, Jenkins and Madigan2021 for a meta-analysis of thirty-five studies and almost 2,000 children, showing an effect size for quality of input of .33.) In other words, gradually the default position in the field of child language was that LASS had whupped LAD’s ass; and the field had simultaneously expanded into Pasteur’s quadrant, with its greater emphasis on the design of parent-education and early-childhood programs to reduce ethnic and social disparities in language and cognition.

Having moved in 1979 from Linguistics in Amsterdam to a position at the Harvard Graduate School of Education (HGSE) in Cambridge, MA, I was experiencing a similar expansion of interests, often in response to and in collaboration with students focused on improving educational outcomes. Questions about the role of parent-child interaction in the development of “school readiness,” in particular early literacy skills and other cognitive and linguistic precursors of literacy success, were of greater interest to HGSE students than children’s learning of syntax.

In 1979–1980 I collaborated with HGSE doctoral students Wendy Barnes, Irene Goodman, Lowry Hemphill, and Jean Koch to carry out a smallish mixed methods study that looked at home influences on literacy outcomes and school success for elementary school students from low-income families (Unfulfilled Expectations, 1991). A subsequent much better funded study enabled Patton Tabors and me to explore home factors in the development of school-relevant skills in a large group of three- to five-year olds, while David Dickinson examined the quality of those same children’s preschool classrooms. We succumbed, at the end of the originally planned three-year longitudinal study, to the familiar temptation to keep collecting data on the same cohort of children, eventually securing funding to follow them through high school, though with much less intensive data collection as they got older (Beginning Literacy with Language in 2001 and Is Literacy Enough? Pathways to Academic Success for Adolescents in 2007). At the time I saw these various studies as natural extensions of the methods and the questions that had guided my early work on language development: What are the characteristics of environments that support and facilitate the development of socially valued linguistic and cognitive skills in children? The answers were unsurprising to a social-interactionist-inclined developmentalist: a rich language and literacy environment, loving and predictable family structures, and decent schools.

Story Number Two: How Do Children Learn to Read?

By 1995 I had, thus, published a fair amount about literacy development and literacy outcomes, most of it focused on home factors predicting those outcomes, in particular via their impact on the child’s vocabulary. That year the powers that be at the National Research Council (NRC, then the research arm of what is now called the National Academies of Science, Engineering, and Medicine) decided that a sufficient research base about reading had accumulated to write a report that would end the “reading wars.” What were these reading wars? Conflicts about how to teach reading were certainly not new, but a new skirmish in the reading wars broke out in 1955, when Rudolf Flesch published an attack on the “whole word” method entitled Why Johnny Can’t Read, which became an immediate bestseller. The “whole word” method was the foundation for the widely used Dick and Jane basal reading series, which was based on the theory that children could start learning to read by recognizing word-shapes in familiar, redundant, illustration-supported texts; thus, a child’s first encounter with reading was meaning-focused rather than code-focused. Indeed, because many of the most frequent words in English (e.g., the, one, of, to, you, said, could, would, have, some, come) are not decodable using simple phonics rules, teaching them using the whole-word method made (and still makes) perfect sense. Whether regular, decodable words like Dick, Jane, dog, cat, mom, and dad should be taught that way is more debatable.

Critiques of the Dick and Jane readers came from many sources in addition to Flesch, and focused on flaws that went beyond the whole-word focus. The stories were boring. The range of words taught in grade one was limited. The introduction of phonics, intended by the series designers to happen in grade two, was often omitted – leaving young readers trying to learn to read thousands of words with techniques meant for just the first few hundred. Flesch and like-minded individuals proposed a phonics-based approach, most versions of which required, given the depth of English orthography, explicitly teaching lots of rules and requiring lots of code-focused practice.

Conversely, critics deplored the lack of attention to meaning in this approach to early literacy instruction. The phonics-based methods were also often implemented using scripted curricula, and thus were critiqued for reducing teacher autonomy. Those who championed the value of lively and engaging authentic literature and of teacher professionalism in the primary grades thus attacked the phonics methods, proposing as an alternative a meaning-based approach with incidental phonics teaching they dubbed “whole language.” The conflict between phonics and “whole language” was a concern for educators choosing curricula, teacher education programs selecting content, and parents choosing schools. These concerns became more urgent in the 1970s and 1980s, as national assessment systems and international comparisons put a spotlight on the vulnerabilities of US schooling.

Accordingly, in the early 1990s a Committee on Preventing Reading Difficulties in Young Children was proposed, to be funded largely by the US Office of Special Education Programs, which was concerned that many school children were identified as needing (expensive) special services simply because they could not read. The NRC faced a challenge, though, to select a chair for the distinguished group of reading researchers who had agreed to serve on the committee. Everyone who had published in the field of reading research was already aligned on one side or the other of the conflict between phonics and whole language (which is, of course, just a specific instantiation of the conflict between direct instruction and progressive pedagogy). I was identified as an ideologically untainted researcher with just enough credibility in the reading field to pass.

Thus began my baptism by fire as an “expert” in literacy. The report Preventing Reading Difficulties in Young Children was published in 1998 and was immediately adopted as a basis for guidance to schools and teachers in several states and many districts. The report focused on opportunities to learn rather than prescribing particular curricula or approaches to instruction. It emphasized that all children need opportunities to learn and to master the alphabetic principle, to develop rich language and knowledge structures, and to receive supplementary instruction as soon as possible if encountering difficulties.

The report was, of course, criticized; progressive educators argued that we had ignored the systemic challenges (poverty, racism, poor schooling) that had generated inequities in reading outcomes; whole language adherents argued that we did not attend sufficiently to the research using miscue analysis; and a few scholars aligned with Reid Lyon (whose program at NICHD had contributed a tiny amount of funding to the study), anticipated that we would not emphasize phonics enough. Strengthened in our resolve by this evidence that we had evidently been quite evenhanded, members of the committee proceeded to participate in professional development efforts in several states and at various professional gatherings. Meanwhile, though, NICHD funding was secured for a second report, one that purported to be more rigorous because it limited the evidence reviewed to experimental and quasi-experimental studies.

Ideally the NICHD report, which did an excellent job of synthesizing the research selected for inclusion, would have been presented and utilized as a complement to Preventing Reading Difficulties, providing meta-analytic support for some of the same conclusions. Its potential utility, though, was undermined by an effort to obliterate the earlier report, and a failure to acknowledge the many limitations of the research base being meta-analyzed. A few examples will suffice. Most of the experimental studies showing the effectiveness of early phonics instruction had been carried out with struggling readers; it was too easy to jump from findings showing that large doses of phonics instruction are effective with those learners to the conclusion that they are necessary for all learners, and from the length and intensity of interventions those learners needed to guidelines for all students. In addition, very few experimental studies of sustained silent reading could be located to include in the review; in the absence of attention to this limitation in the report, many educators drew the unjustified conclusion that sustained silent reading was ineffective (and thus dropped the practice), a conclusion that could have been blocked with appropriate attention to the correlational and developmental evidence showing the positive effects of sustained engagement with text. Furthermore, the experimental studies targeting comprehension outcomes were largely limited to testing the effects of strategy instruction; as a result, strategy instruction became the default approach to teaching comprehension, rather than just one of an array of useful tools in the repertoire of the secondary teacher. Despite these (and other) failings, the NICHD report was adopted by the Bush administration as its basis for Reading First legislation, which itself was a product of a brief period of bipartisan interest in education and optimism about top-down efforts to improve it.

So, several years of work on Preventing Reading Difficulties did not succeed in ending the reading wars, as is well attested by the emergence in the 2020s of controversies around the implications of the “science of reading” for reading instruction. These conflicts, like those of the previous periods, almost inevitably end up subjecting young students to instructional regimes that overemphasize either code or meaning, rather than titrating them to individual needs and integrating them with each other and with authentic engagement in learning. Furthermore, they end up focusing all the attention on reading instruction in the primary grades, leading to neglect of the literacy demands of disciplinary learning in grades four and beyond. (See the 2002 publication Reading for Understanding, written by the RAND Reading Study Group that I chaired.Footnote ¹)

I turned my attention to efforts to contribute to educational practice in ways that could not be undermined by policy clashes or political commitments. I would argue now that designing excellent tools and figuring out how to get them into the hands of teachers is a more effective way to improve learning outcomes than fighting legislative battles (though those are again heating up, precisely in the field of early reading, and I am pleased to see some of my colleagues fighting back). So I became involved in the design and development of one of the early formal research-practice partnerships focused on education.

It was again a National Academies study that led to this career pivot. Bruce Alberts, then the president of the Academies, was concerned that researchers and practitioners had not formed the sort of mutually productive working relationships in education that were common in the field of medicine and agriculture. He appointed Suzanne Donovan to lead a committee which explored various models for educational research partnerships, and which in 2003 published a report entitled Strategic Education Research Partnership (SERP). Suzanne became the executive director of the fledgling SERP Institute, which took on as its first active partnership work in the Boston Public Schools, where over the next few years the SERP team worked with teachers in designing curricular supports for academic language and literacyFootnote ² and a literacy intervention for students reading far behind their grade level (STARI), as well as working with ETS to develop assessments for literacy.

My contribution to some of the curricular tools developed by SERP was derived, I now realize, from insights I had gained while watching children learn to talk through interactions with their mothers about topics of great interest to them. The key ingredients were interest, social interaction, and engagement in authentic tasks. These ingredients are key elements in the curricular resources we have developed for fourth to eight graders; they promote literacy through authentic engagement with consequential, debatable questions. These questions impel students to read, write, and think critically about important topics; in the process they expand their knowledge and hone their literacy skills. I argue that the programs we have developed work well to teach reading and writing because they were not designed to focus on reading or writing, but to use reading and writing as tools to help achieve other aims. One of those aims is winning arguments, so every unit in the curricula includes a debate. Classroom discussion, a vanishingly rare activity in many US classrooms, constitutes a rich context for developing language,Footnote ³ integrating others’ perspectives into one’s own thinking, and building arguments. Supporting classroom discussion is not easy, given the many structural factors that militate against it, but it is a satisfyingly coherent focus for the second half of a career that started with careful study of the social contexts for children’s early language development.

Next Steps

Though I have moved definitively out of Bohr’s quadrant, the basic questions about how children learn that motivated my dissertation research at McGill University in 1967–69 are still at the center of my interest. I have discovered, though, that one can ask more nuanced and interesting versions of those questions if the children involved are in the authentic and multifaceted settings of classrooms and peer groups and are engaged in the challenging tasks of school-related learning.

I am more and more intrigued to explore the gaps between what we teach and what children learn, how much less they learn than is taught under some circumstances, and how much more under others. Part of the interest in this topic derives from my sense that certain abilities can be guided or coached but ultimately have to be developed through use rather than through structured instruction. Just as at some point acrobats have to release their hold on one trapeze and hope they will catch the next one, sixth grader readers simply have to jump in, teaching themselves to comprehend by struggling to comprehend. However, well-designed prior experiences greatly increase the chances of success for trapeze artists and for comprehenders. Authentic reasons to comprehend, access to relevant background knowledge, and opportunities to discuss the topic of the reading with others are the young reader’s equivalent of the trapeze artists’ exercises to promote balance and upper body strength.

I have emphasized in this brief recap of my career the dominant theoretical positions which my work undermined: nativism in early language acquisition, critical periods in second language learning, and the need to choose between code and meaning in teaching reading. The first two of those positions have declined in power, as counterevidence has accumulated from multiple sources. The third position remains unresolved in the public discourse, with detrimental consequences for children and their teachers. In every one of these cases, though, progress has come from paying serious attention to the developmental forces revealed in real-world settings – children in homes, preschools, playgrounds, and elementary classrooms. (See Grøver, Uccelli, Rowe, & Lieven, 2019, for extended examplesFootnote ⁴.)

If there is any advice to young researchers to be distilled from these experiences, it is this: follow lines of work that fascinate you and cultivate the courage of your convictions when your common sense or early data suggest that received wisdom is wrong.