I haven’t written many posts in 2025; here are the measly few I’ve managed to squeak out:

• Literacy Is Not Just for ELA: The Power of Content-Rich Teacher Tall
• More Productive Than an Hour of Instruction?: The Surprising Cognitive Science of a Walk in the Park
• AI, Mastery, and the Barbell of Cognitive Enhancement

While my bandwidth to peruse research has diminished this year (work has been busy, and I like spending time with my children) I have still encountered a fair number of compelling studies. In keeping with the tradition begun in 2023, and building on last year’s review, I am endeavoring to round up the research that has crossed my radar over the last 12 months.

This year presents a difficult juncture for research. Political aggression against academic institutions, the immigrants who power their PhD programs, and the federal contracts essential to their survival has disrupted research. Despite this, strong research continues to be published. Because research is a slow-moving endeavor, I suspect the full effects of these disruptions will manifest increasingly in future roundups; for now, the good work persists.

The research landscape of 2025 highlights a continued shift toward experience-dependent plasticity. This view treats the human mind as a dynamic ecosystem shaped by biological rhythms, cultural “software,” and technological catalysts. Learning is no longer seen as a linear accumulation of skills, but as a sophisticated orchestration of “statistical” internal models and external social and cultural and technological attunements.

Longtime readers will recognize this “ecosystem” view from my other blog on Schools as Ecosystems. It is validating to see the field increasingly adopting this ecological lens—viewing the learner not as an isolated machine, but as an organism deeply embedded in a biological and cultural context.

Our “big buckets” for this year have ended up mirroring the 2024 roundup, which means, methinks, that we have settled upon a perennial organizational structure:

• The Science of Reading and Writing
• Content Knowledge as an Anchor to Literacy
• Studies on Language Development
• Multilinguals and Multilingualism
• Rhythm, Attention, and Memory
• School, Social-Emotional, and Contextual Effects
• The Frontier of Artificial Intelligence and Neural Modeling
  

Let’s jump in!

I. The Science of Reading and Writing

The Critical Role of Morphology and Vocabulary

Readers of my 2023 roundup will recall that morphology was a major theme that year, and it remains central in 2025. Morphology refers to the smallest units of meaning in a word, and is strongly connected to the origins and evolution of words (etymology), and to vocabulary development and reading comprehension in general. It also serves as a crucial link to spelling, given the irregularities in a language such as English that cannot be resolved via phonological decoding alone.

In any given orthography, it is indeed the combination of phonology (the sounds) and morphology that enable a finite number of phonemes or symbols to be recombined into a potentially infinite number of unique words.

• “Combinatoriality enables an orthography to provide learnability and decipherability for the novice reader (via phonological transparency) as well as unitizability and automatizability for the expert (via morphemic transparency).” (Blueprint for a Universal Theory of Learning to Read: The Combinatorial Model, Reading Research Quarterly)

Early writing is a “canary in the coal mine” for future reading success. A study of 243 preschoolers found that initial levels and growth in name writing and letter writing significantly predicted later word reading and passage comprehension. This association held true for both monolingual and bilingual children identified as at-risk for reading difficulties, indicating that writing development is a universal literacy milestone.

• “Children’s initial levels of name writing, letter writing, and picture writing... predicted their later reading abilities in both word reading and passage comprehension.” (Beyond Word Recognition: The Role of Efficient Sequential Processing in Word- and Text-Reading Fluency Development, Scientific Studies of Reading)

A massive analysis of 1,116 children demonstrated that word reading and spelling are effectively a single latent trait (r = 0.96). However, spoken vocabulary knowledge acts as a bridge, allowing readers to use known word meanings to compensate for “fuzzy” or imprecise knowledge of letter-sound rules.

• “Our results suggest that word reading and spelling are one and the same, almost, but that spoken vocabulary knowledge is more closely related to reading than to spelling.” (On the relationship between word reading ability and spelling ability, Reading and Writing)

The ability to form and retrieve letter sequences (orthographic mapping) is a consistent driver across both typical and dyslexic populations:

• “Among typically developing children, orthographic mapping, phonological awareness, oral vocabulary, and working memory scores uniquely explained reading comprehension. Among children with dyslexia, only orthographic mapping and oral vocabulary scores uniquely predicted reading comprehension.” (The effects of orthography, phonology, semantics, and working memory on the reading comprehension of children with and without reading dyslexia, Annals of Dyslexia)

Longitudinal data showed that from Grade 3 to Grade 5, morphological awareness (manipulating prefixes, suffixes, roots) overtakes phonological awareness as the primary driver of reading comprehension and the mastery of complex, multi-morphemic words.

• “These results indicate continuing and pervasive roles for phonological awareness, naming speed, and morphological awareness over the later elementary school years, especially for morphological awareness in reading comprehension.” (Effects of morphological awareness, naming speed, and phonological awareness on reading skills from Grade 3 to Grade 5, Journal of Experimental Child Psychology)

Explicit and Implicit Instruction

All this leads to interesting findings this year around explicit instruction (EI) vs. statistical and implicit learning. We often pit these two against each other, but 2025 gave us some direction towards a more synergistic understanding.

Explicit instruction in alphabet instruction is critically important, regardless of modality and language status.

• “Young children benefit from explicit and systematic alphabet instruction, regardless of whether such instruction is multisensory or visual-auditory. . . EB and English monolingual children experienced a similar benefit from alphabet instruction, perhaps because they had similar socio-economic status and language backgrounds.” (An initial yet rigorous test of multisensory alphabet instruction for english monolingual and emergent bilingual children, Early Childhood Research Quarterly)

• “It is unrealistic to teach children a minimal set of letter–sound correspondences and expect them to deduce the more complex statistics of a writing system without guidance.. . . it is clear that the findings of laboratory studies of statistical learning do not generalize straightforwardly to the real world.” (Statistical learning in spelling and reading, Trends in Cognitive Sciences)

• “These findings clearly show that the learning via exposure is slow and does not guarantee successful learning of regularities in written languages, especially when there is more than one pattern in the input.” (Simultaneous learning of semantic and graphotactic regularities in spelling: An artificial orthography learning experiment, OSF Preprint)

But as word-level reading becomes increasing automatized, it moves to more “top-down, meaning-driven processes” related to language.

• “When we first learn to read, our brain heavily relies on its general problem-solving network (i.e. “multiple demand network”), but as we become skilled, it increasingly shifts to using specialized language networks.” (Contributions of the multiple demand network to emergent and skilled reading, Scientific Reports)

This shift is mirrored in the brain's “salience network,” which a large scale meta-analysis identifies as a shared foundation for both math and reading. While children rely on this network broadly for learning, adults engage it primarily for challenging, unmastered tasks, highlighting the importance of targeting attention and effort during the formative years.

• “The implication is that children and adults engage cognitive control networks for number-arithmetic tasks that are not yet automatized. . . ....the salience network might contribute to the common finding that learning difficulties in mathematics and reading are comorbid. . . . LD interventions should incorporate features that support the functions of cognitive control networks, including external factors that motivate attentional focus... and that highlight key information.” (Shared brain network acts as a foundation for both math and reading, Nature Communications)

For children with developmental language disorder (DLD), explicit instruction in meaning (“semantics”) is most important.

• “As a group, children with DLD showed significantly greater word-learning gain from explicit semantic interventions compared with explicit phonological therapy.” (Vocabulary interventions for children with developmental language disorder: a systematic review, Frontiers in Psychology)

Mechanisms of retention are equally critical; for children with DLD, vocabulary retention is specifically driven by the frequency of successful retrieval across multiple sessions, rather than just the intensity of exposure.

• “the number of sessions in which a child successfully produces a word's form or meaning relates positively to their ability to remember that word after extended delays . . “ (The Number of Sessions Children With Developmental Language Disorder Retrieve Words Relates Positively to Retrieval After Extended Post-Training Delays, Language, Speech, and Hearing Services in Schools)

• “word retention in children with DLD is influenced not only by the robustness of the initial learning phase but also meaningful retrieval and practice opportunities” (Vocabulary interventions for children with developmental language disorder: a systematic review, Frontiers in Psychology)

• While response to treatment generally improves with intensity, there can be “diminishing returns” once a certain threshold is passed, such as 48 exposures in a book-reading context” . . . Retention is further enhanced when retrieval occurs with “other words intervening,” which has been shown to help with “word learning and... retention more than if they just completed the task without intervening words.” (IJLCD Winter Lecture 2025: What makes language interventions work – exploring the active ingredients, Royal College of Speech and Language Therapists)

In an orthography such as Chinese, gaining automatization with the “sub-lexical mappings between orthography (form), phonology (sound), and semantics (meaning)” can be an even greater challenge for students with dyslexia. An RCT found that explicit instruction was necessary for abstracting rules (form-sound mappings), but implicit exposure was also key for optimizing speed and efficiency.

• “Only the explicit-SL [Statistical Learning] group showed abstraction of form-sound mappings, while only the implicit-SL group showed optimized reading processes across phonology and semantics.” (Abstraction and Optimization in Statistical Learning: A Randomized Controlled Trial of Implicit and Explicit Reading Intervention for Students with Dyslexia, Proceedings of the Annual Meeting of the Cognitive Science Society)

In other words, while explicit instruction is critical, it must be accompanied by sufficient volume for application and practice.

• “These results demonstrate that efficient processing of complex syntactic structures depends on both good statistical learning skills and exposure to a large amount of print so that these skills have the opportunity to extract the relevant statistical relationships in the language” (Statistical learning ability influences adults’ reading of complex sentences, Canadian Journal of Experimental Psychology)

(I wrote about this need for balancing explicit instruction and statistical learning in my post, LLMs, Statistical Learning, and Explicit Teaching)

When it comes to learning the more precise and challenging statistics of orthography, even skilled adults are “satisficers,” choosing the simplest or easiest pronunciations rather than the statistically optimal ones predicted by vocabulary data.

• “Despite years of exposure, English readers produce /k/ pronunciations for the letter “c” before “e” and “i” over 10% of the time, “even though /k/ pronunciations of ⟨c⟩ virtually never occur in this context in English words.” (Statistical learning in spelling and reading, Trends in Cognitive Sciences)

As literacy learning shifts more towards that language-based side of things, the importance of “usage-based” learning becomes even more important, as with students learning a new language.

• “The bulk of language acquisition is implicit learning from usage. Most knowledge is tacit knowledge; most learning is implicit; the vast majority of our cognitive processing is unconscious. . . . Explicit instruction can be ill-timed and out of synchrony with development... it can be confusing; it can be easily forgotten; it can be ignored.” (Usage-based approaches to SLA, Theories in Second Language Acquisition: An Introduction)
  

After all, learning a new language (oracy, vocabulary, comprehension) is not only about reading or writing silently, but also about communication, which is social in nature. Balancing when new vocabulary is introduced and used therefore becomes a consideration.

• “Vocabulary acquisition through interactive tasks involves a dynamic interplay between language-specific neural networks and social-cognitive processes, with their relative contributions shifting as learners progress through sequential collaborative tasks. . . Pre-task vocabulary practice led to greater learning, while post-task practice resulted in higher IBS [inter-brain synchronization] in the brain region underlying language processing.” (Timing matters for interactive task-based learning: Effects of vocabulary practice on learning multiword expressions and neural synchronization, Studies in Second Language Acquisition)

Assessing Literacy

Gaining a deeper understanding of student’s literacy profiles in order to tailor and target instruction to their needs is important. In the past, teachers relied on “miscue analysis” and “running records” to gain this understanding, but such analysis is about as useful as flipping a coin. Instead, a study suggests that error analysis using the valid and reliable CBM measure of Oral Reading Fluency (ORF) can provide key information on whether errors are phonemic, orthographic, morphemic, and high frequency in nature.

• “Analyzing student errors and the features of the words wherein these errors occur allows for a more tailored understanding of the area in which students are struggling and provides guidance on how to adjust instruction accordingly. . . The DBI [data-based instruction] process is iterative, and the ongoing analysis of student assessment data to inform the intensification and individualization of an intervention is essential to this process.” (What’s in a Word? Analyzing Students’ Oral Reading Fluency to Inform Instructional Decision-Making, Intervention in School and Clinic)
  

Automated oral reading fluency assessments often exhibit bias against English learners due to speech-to-text inaccuracies, which can be mitigated by including prosody as a core sub-construct.

• “The inclusion of prosody improves automated ORF assessment by reducing discrepancies between ELLs and English first language students.” (Investigating construct representativeness and linguistic equity of automated oral reading fluency assessment with prosody, Language Testing)

Furthermore, it is important to draw upon multiple sources of data to fully understand any student’s unique needs.

• “The results of the Delphi study highlight the complexity involved in assessing dyslexia and the need to draw upon multiple sources of information: background information, standardised test results, and qualitative observations.” (Towards a Consensus for Dyslexia Practice: Findings of a Delphi Study on Assessment and Identification, Dyslexia)

II. Content Knowledge as an Anchor to Literacy

Just as we moved from word-level phonological decoding and orthographic mapping towards the importance of semantic and language-based learning, we must pair the learning of any school language not only to social communication, but furthermore to the conceptual and topical knowledge entrenched in academic disciplines.

And that conceptual and topical knowledge – so critical for critical thinking – is founded upon facts.

• “Critical thinking cannot develop and cannot flourish without facts. You need to start with evidence, with things that are true so that you can think about causality.” (Young Minds, Smart Strategies: How Children Decide When to Use External Memory Aids, APS Podcast)

Curriculum programs are typically designed around “thematic units to build content schemas.” Yet categorization may be a better means.

• “Categories are rule-based (e.g. something is or is not in a category) and hierarchical (e.g. superordinate categories/subcategories). In this respect, they provided an organizational framework that is different from traditional theme-based instructional approach, which is reliant on associative relationships, and situations. . . . Topics which build concepts through categorization provide children with a more facilitative way to process, store and retrieve information, while promoting inferences that extend knowledge beyond past and current experiences.” (Knowledge-Building Through Categorization: Boosting Children’s Vocabulary and Content Knowledge in a Shared Book Reading Program, Early Education and Development)

OK, not part of 2025 research but a great connection on this, back in 2023 Susan Pimentel, David Liben, and Meredith Liben similarly advocated for a shift from broad thematic units toward a shift for building knowledge through specific topics, which they argued could more effectively support the development of content schemas.

• “To accelerate literacy learning... teachers need instructional tools that create sustained opportunities for reading and discussing informational texts, examining the language encountered in those texts, and building new content knowledge.” (Scaling the 'dinosaur effect': Topic vs. theme in elementary classrooms, Knowledge Matters Campaign)
  

Relatedly, while general prior knowledge facilitates basic comprehension, topic-specific knowledge is the primary driver for building the situational models required for complex knowledge transfer. This effect is mediated by the learner’s initial mastery of a base text, as the ability to apply information to new contexts depends entirely on the foundational transfer skills established during that first encounter.

• “The amount and specificity of prior knowledge influenced learning from both texts. Additionally, learning from the first text mediated the impact of topic-specific knowledge on learning from the second text. . . .Topic-general knowledge showed a stronger correlation with comprehension, while topic-specific knowledge was more closely associated with transfer.” (The effects of the topic-specific and topic-general prior knowledge on learning from multiple complementary texts, Learning and Individual Differences)

III. Studies on Language Development

Talker Variability

Building off our previous section on the importance of content knowledge, one single predictor of a multilingual child’s ability to master complex science and social studies vocabulary is driven by a core set of foundational language skills. A student’s foundational language factor (vocab/syntax) explained 58% of the variance in their ability to produce definitions for science concepts.

• “Learning content vocabulary is significantly related to student language skills in Spanish and in English. . . We find that content is learned when the language is learned. As such, all teachers are, first and foremost, language teachers of the subject matter that they present. . . This finding suggests that developing student language skills early facilitates the learning of curricular vocabulary words later.” (Predicting Science and Social Studies Vocabulary Learning in Spanish–English Bilingual Children, Language, Speech, and Hearing Services in Schools)

While we often think “more speakers = better,” it turns out variability helps children with strong language skills (1.95x more likely to learn), but it can actually “thwart the discovery” of patterns for children with weaker language skills.

• “This study suggests that children with different levels of language skills and bilingual experience may learn new words differently. More variability = good for children with more bilingual experience or strong language skills, less variability = good for children with less bilingual experience or weaker language skills.” (The graded effects of bilingualism and language ability on children’s cross-situational word learning, Bilingualism: Language and Cognition)

Yet some variability remains key, including for students with developmental language disorder (DLD).

• “Highly variable linguistic input seems to facilitate grammatical morpheme learning in children with DLD . . .Increasing the variability in how an object is represented in treatment also has the potential to improve children's ability to generalize their next lexical knowledge.” (IJLCD Winter Lecture 2025: What makes language interventions work – exploring the active ingredients, Royal College of Speech and Language Therapists)

For adults learning new words in their native language there was no evidence that either talker variability or scaffolding the talker presentation assisted learning. Instead success was almost entirely predicted by the learner's phonological working memory and general language ability.

• “In particular, exposure to multiple speakers of the same variety resulted in the largest gain. Thus, to facilitate adaptation to unfamiliar L2 pronunciation, high-intelligibility speakers and/or multiple speakers of the same language background should be used”. (The impact of talker variability and individual differences on word learning in adults, Brain Research)

All that said, in the context of second language learning, talker variability remains a vital tool—provided the variability maintains high intelligibility and stays within the same dialect or language variety. This principle resonates with my own experience learning Spanish in Peru. I spent roughly equivalent amounts of time en la costa, en los Andes, y en la selva; yet, just as I felt I was gaining fluency, moving to a new region and encountering an entirely different variety of the language made it feel as though I were learning it all over again.

• “In particular, exposure to multiple speakers of the same variety resulted in the largest gain. Thus, to facilitate adaptation to unfamiliar L2 pronunciation, high-intelligibility speakers and/or multiple speakers of the same language background should be used”. (Intelligibility and input variability influence adaptation to unfamiliar L2 pronunciation in L2, Foreign Language Annals)
  

Further considerations for a practice structure with “variability”: when learning new L2 vocabulary, interleaving different categories produced superior outcomes compared to studying them in blocks, likely due to a spacing effect that forces the brain to constantly retrieve and contrast new information.

• “Mixing different language categories during practice (interleaving) rather than studying them in separate blocks produced superior learning outcomes . . . Our findings indicate that the interleaving advantage observed in other domains extends to dual language learning.” (The effects of interleaving and rest on L2 vocabulary learning, Second Language Research)
  

Quality vs Quantity

A central question in language research is whether children primarily need a high volume of speech (quantity) or speech that is linguistically and conceptually rich (quality).

A meta-analysis found that in the home, quantity and quality of speech are highly correlated (r=0.88); “parents who talk more naturally tend to use a more diverse and complex vocabulary.”

• Furthermore, “Younger children benefit less from lexical diversity... because words that children acquire early in life are so common and so concrete that they are likely to appear in informative contexts even in the speech of parents who exhibit lower lexical diversity.” (How strong is the relationship between caregiver speech and language development? A meta-analysis, Journal of Child Language)

Conversely, in school, only quality moves the needle. This meta-analysis examined teacher language practices from preschool to third grade and found a statistically significant association between teachers’ language quality—defined as interactive scaffolding and conceptual challenge—and children’s development, but no significant association with the quantity of teacher talk.

• “Enhancing teacher language practices is not only strategically vital but also cost-effective and scalable. . . . These findings emphasize the need to focus on improving the quality of teachers’ language practices in early childhood education through enhanced teacher preparation and ongoing professional development” (Does Teacher Talk Matter Too? A Meta-Analysis of Partial Correlations Between Teachers’ Language Practices and Children’s Language Development from Preschool to Third Grade, Review of Educational Research)

The finding that quality of teacher talk trumps quantity reinforces what I have previously explored in Research Highlight 2 and Literacy Is Not Just for ELA. We know that explicit use of academic vocabulary and decontextualized language is what drives growth, not just a whole bunch of words.

Yet despite the importance of quality talk in classrooms, large-scale recordings of 97 preschool classrooms revealed a dearth of linguistically challenging interactions.

• Researchers found that 40% “Instructional time was primarily devoted to alphabetics, with a stark paucity of opportunities for children to acquire the language and content knowledge essential for later learning.”

• In contrast, time spent on vocabulary and science instruction supported the most complex and pedagogical language, yet these activities combined received less than half the time allotted to simple letter drills.

• There was a significant misalignment between beliefs and practice: while 96% of teachers felt confident in their ability to foster rich discussions, automated recordings showed they rarely used wh-questions or extended conversational turns (Preschool Teachers’ Child-Directed Talk, Early Education and Development)

From Womb to Weave: Human Language Development

In 2025, language research has deepened our understanding of the biological and evolutionary roots of communication. Language is not merely a set of learned properties and rules but a form of social, statistical, and biological attunement.

Human language, influenced by the sounds of the words of the adults around us, begins to develop while we are in the womb, and we begin to distinguish between our home languages and other languages.

• “Newborns recognize foreign languages they heard while in the womb.” (Babies’ Brains Recognize Foreign Languages They Heard before Birth, Scientific American)

Even mere exposure to the sounds of a tonal language like Mandarin creates lasting structural imprints in the brain's white matter that persist even if the language is no longer used.

• “Exposure to a tonal language like Mandarin early in life exerts lasting effects on white matter architecture in the brain... persisting even if the language is discontinued.” (Early but discontinued exposure to a language exerts lasting effects on white matter architecture in the brain, Communications Biology)

Once out in the world, infant attunement to their mother’s heartbeat during face-to-face interaction correlates with word segmentation ability.

• When mothers and infants had more synchronized heartbeats, the infants were better at identifying individual words within a stream of speech. . . . This biological synchrony correlated with maternal sensitivity to an infant's mental states, suggesting that an attuned emotional environment literally sets the rhythm for learning.” (Individual Differences in Infants' Speech Segmentation Performance, Infancy)

A nine-year longitudinal study furthermore found that index-finger pointing at age one is a specific developmental predictor of metaphor comprehension at age nine. This correlation reinforces the “embodied cognition” view—the idea that physical grounding in infancy serves as a required scaffold for abstract thought later in life.

• “Iconic gesture comprehension at age 3;0 was correlated with all language skills, performance on the ToM-scale and metaphor comprehension at age 9;0.” (Does early gesture usage contribute alongside oral language to later theory of mind performance and metaphor comprehension?, Language Acquisition)

You know how adults talk all silly as they goo goo and gah gah at babies? That baby talk seems to be an innate scaffolding technique that accelerates infant language development for all kids, including those with autism.

• “Infants were more likely to produce a speech-like vocalization following an adult utterance directed to them in parentese compared to an adult utterance directed to them in adult register. . . . Both neurotypical & autistic infants make more speech-like sounds when spoken to in parentese.” (Parentese Elicits Infant Speech-Like Vocalizations in Typically Developing and Autistic Infants, Infancy)

Such “parentese,” or “infant-directed speech,” is something that sets us apart from apes.

• “The rate that children heard infant-directed communication was 69 times as high as what Dr. Fryns observed among chimpanzees, and 399 times as high as what Dr. Wegdell observed among bonobos.” (Did Baby Talk Give Rise to Language?, NYTimes)

While macaque monkeys share similar visual-encoding machinery to us, they do not form “consensus color categories,” suggesting that language provides the needed cognitive and cultural framework to achieve shared conceptual agreement.

• “One animal showed evidence for a private color category, demonstrating that monkeys have the capacity to form color categories even if they do not form consensus color categories. . . Innate similarities between monkeys and humans are not sufficient to produce consensus color categories . . . This implies that human color categories are not 'hard-wired' by birth but depend on language and cultural coordination to achieve shared agreement.” (The origin of color categories, Psychological and Cognitive Sciences)

One interesting aspect of human gender differences is that girls develop more advanced language abilities than boys at an earlier age.

• “Girls learn language skills more rapidly than boys. Boys learn cognitive and fine motor skills more rapidly than girls.” (A Study of the Microdynamics of Early Childhood Learning, NBER Working Paper)

Across typologically diverse languages and cultures, children follow a universal pattern of transitioning from salient free negators (e.g. “No,” “Not”) to less salient bound negator morphemes (e.g. “-nt”).

• “In the acquisition of negation, universal mechanisms based on frequency and salience may be at work; however, individual trajectories are strongly shaped by culture and language-specific factors” (Negation in First Language Acquisition: Universal or Language-Specific?, Cognitive Science)

Furthermore, a phonemic analysis of animal onomatopoeia across 21 languages reveals that humans perceive animal sounds in ways that are similar across cultures. While cultural filters vary the spelling, the underlying sound interpretation transcends linguistic differences.

• “Phonemically the sounds made by the animals across the world are cognate in cat-speak, duck-speak, pig-speak, and so forth”. (Phonemic analysis of animal sounds as spelled in various popular languages, Language Log)

Phonemes can be viewed as “cognitive tools” that support and extend human thinking and ability. These basic sound units are predicated on physical and biological constraints but vary across cultural lineages to facilitate the efficient transmission of information.

• “Phonemes—the basic sound units of language—function as cognitive tools that shape and extend human thinking.” (The Phoneme as a Cognitive Tool, Topics in Cognitive Science)

For adults, familiar prosody is also a primary gateway to learning a new language.

• “Adults can quickly pick up on the melodic and rhythmic patterns of a completely novel language” (How to learn a language like a baby, The Conversation)

• Familiar pitch patterns (like those from a listener’s native language) significantly boost the ability to parse word boundaries and complex dependencies; without these melodic cues, complex structures remain unlearnable within short timeframes. (Prosody enhances learning of statistical dependencies from continuous speech streams in adults, Cognition)

And speaking of adults and parents: having more books in the house and parents who are knowledgeable about children’s stories independently helps a child's reading skills, even after accounting for the parents' own natural reading abilities.

• “Children’s reading in Grade 3 was predicted by mothers’ engagement in reading activities and by literacy resources at home, even after controlling for the genetic proxy of parental reading abilities. . . .The mothers of children who struggle tend to engage in more reading activities. . . Fathers' reported frequency of reading activities was not predictive.” (The intergenerational impact of mothers and fathers on children's word reading development, Journal of Child Psychology and Psychiatry)

Human and Animal Evolution

In 2025, the century-long view of Darwinian gradualism—the idea that species develop through slow, imperceptible increments—was further challenged by a new mathematical framework. This research reveals that living systems often evolve in sudden, explosive surges rather than a steady marathon of change. These “phantom bursts” of evolution suggest that spiky growth patterns are a general characteristic of any branching system of inherited modifications, whether in proteins, languages, or complex organisms. (The Sudden Surges That Forge Evolutionary Trees, Quanta Magazine)

What I find especially interesting about this idea of “spiky bursts” of growth is that in last year’s research roundup, we reviewed a study of 292 children which found that those who heard speech in intense, concentrated bursts had significantly larger vocabularies than those exposed to a more consistent, steady stream of language.

I also have a 2025 book recommendation, if you are interested in the history of language evolution: Proto: How One Ancient Language Went Global by Laure Spinney. There’s an interesting passage in it that I'm summarizing here:

In the Caucasus, dubbed 'the mountain of tongues' by a tenth-century Arab geographer, linguists describe a phenomenon called vertical bilingualism, where people in higher villages know the languages of those living lower down, but the reverse is not true. Why would people living in higher-altitude communities be more fluent in the languages of those residing at lower elevations? Perhaps because mountain dwellers had to travel down to lower villages for trade and resources, therefore they learned the languages of those below. Whereas, people in lower villages had less reason to travel to harder to reach and thus, more isolated higher-altitude communities. So they were less likely to learn those languages. This created the vertical flow of linguistic knowledge, mirroring the flow of physical movement.

IV. Multilinguals and Multilingualism

Just as our biological evolution has shaped our capacity for language, our environment continues to shape how those languages manifest. In 2025, the research landscape for multilingualism shifted toward an “experience-dependent plasticity” framework, viewing multiple languages not as competing systems, but as a dynamic, integrated repertoire.

Longitudinal data tracking Spanish-English bilinguals between ages 4 and 12 revealed that language dominance is fluid, not fixed. Researchers observed a rapid switch in dominance characterized by a steady decline in Spanish-only interactions as children aged. Crucially, this developmental shift is not merely a process of “loss” but one of complexity transfer.

• “The narrative complexity of a child’s Spanish (L1) stories significantly predicted the complexity of their English (L2) narratives one year later. . . . Bilinguals who produce nativelike L2 vowels are also able to maintain native L1 productions, suggesting that an increased L2 proficiency does not inevitably entail a decline in L1 proficiency.” (Factor structure and longitudinal changes in bilinguals’ oral narratives production, Applied Psycholinguistics)

This finding is complemented by validation of the Simple View of Reading (SVR) in Spanish Heritage learners, where linguistic comprehension (morphosyntax and vocabulary) was the primary predictor of reading success, echoing the need for strong L1 foundations.

• “Across both types of orthographies, decoding and linguistic comprehension together explain approximately 60% of the variance in RC. Variations between the so-named phonologically transparent and opaque orthographies highlight differences in the contributions of decoding and comprehension to RC and how these factors evolve during children's literacy development. The simplified nature of SVR thus provides a principled foundation for exploring these important questions.” (Can the Simple View of Reading Inform the Study of Reading Comprehension in Young Spanish Heritage Language Learners?, Reading Research Quarterly)

Furthermore, the structural relationship between languages matters. New research indicates that high structural and lexical overlap between a child's languages—a concept known as small linguistic distance—reduces the amount of exposure required to reach heritage language proficiency.

• “We found that language similarity affected the amount of exposure needed to reach a certain level of proficiency.” (The role of linguistic context and language similarity in the relationship between language exposure and language proficiency in bilingual children, Linguistic Approaches to Bilingualism)

I have explored this concept of “linguistic distance” in relation to diglossia and African American English, noting the greater challenged introduced when written forms diverge significantly from a student's spoken vernacular. This new research affirms that finding: just as greater distance requires more exposure, smaller distance facilitates quicker proficiency.

We often hear about the “bilingual advantage” in executive function, but 2025 research added necessary nuance regarding code-switching. The link between cross-speaker code-switching and cognitive control is heavily moderated by overall language ability. High frequency of switching was associated with better inhibitory control only for children with strong language skills; for those with weaker skills, switching often reflected lapses in production rather than strategic control.

• “Higher frequency of cross-speaker code-switches was associated with better inhibitory control only for children with higher levels of language ability . . . For children with weaker omnibus language skills, cross-speaker switches may reflect difficulties generating a message (in either language) and/or difficulties tracking language use. . . The same switching behavior may be rooted in different mechanisms in children with different levels of language ability.” (The influence of cross-speaker code-switching and language ability on inhibitory control in bilingual children, Bilingualism: Language and Cognition)

Perhaps the most striking finding this year comes from the other end of the lifespan. New evidence from 27 European countries has redefined multilingualism as a biological asset that actively slows the aging process. In a study of over 86,000 participants, monolingualism was associated with more than double the risk of accelerated biological aging compared to multilingual peers.

• “Monolingualism was associated with more than double the risk of accelerated biological aging (OR = 2.11). . . . Speaking two or more additional languages provided progressively stronger protection as individuals grew older.” (Multilingualism protects against accelerated aging in cross-sectional and longitudinal analyses of 27 European countries, Nature Aging)

V. Rhythm, Attention, and Memory

We are moving away from viewing music and speech as isolated auditory signals and toward a model of social and biological “attunement.” The latest studies suggest that rhythmic synchrony is a fundamental gateway for human connection and cognitive growth.

This attunement extends to the very mechanics of how the brain processes sound. Humans instantaneously distinguish talking from singing based on “amplitude modulation,” or the rate at which volume changes. While speech modulations reflect human vocal comfort at 4–5 hertz, music is slower and more regular at 1–2 hertz, potentially evolving specifically to facilitate group synchrony and bonding.

• “Audio clips with slower amplitude-modulation rates and more regular rhythms were more likely to be judged as music, and the opposite pattern applied for speech. . . . Our brain associates slower, more regular changes in amplitude with music (1–2 hertz) and faster, irregular changes with speech (4–5 hertz).” (How Your Brain Tells Speech and Music Apart, Scientific American)

The foundations of language development may actually lie in biological coregulation. When mothers and 9-month-old infants have synchronized heartbeats (measured via Respiratory Sinus Arrhythmia), the infants demonstrate advanced word segmentation skills. This suggests that an attuned emotional environment literally sets the rhythm for learning. (Note: we covered this one in a previous section, but worth repeating again here!)

• “The higher the cross recurrence rate (RR) of mother's and infant's RSA, the longer infants look... which we interpret as advanced word segmentation. . . . When mothers and infants had more synchronized heartbeats, the infants were better at identifying individual words within a stream of speech.” (Individual Differences in Infants' Speech Segmentation Performance, Infancy)

Readers may recall a similar theme from the 2024 roundup, where we discussed research indicating that “synchrony is learning”—showing that brain-to-brain synchrony predicts engagement and learning. This new research on heartbeat and blink synchrony takes that concept even deeper, into the physiological rhythms of our bodies.

One of the year's most fascinating discoveries is that our bodies synchronize with music in ways we never realized: spontaneous eye blinks align with musical beats. This “blink synchronization” occurs without instruction and improves the detection of subtle differences in pitch, indicating that motor alignment helps optimize attention and auditory perception.

• “Spontaneous eye blinks synchronize with musical beats... Blink synchronization performance was linked to white matter microstructure variation in the left superior longitudinal fasciculus.” (Eye blinks synchronize with musical beats during music listening, PLoS Biology)

However, just as synchrony can boost learning, “dys-synchrony” can derail it. It isn't just peer distraction that disrupts the rhythm of learning; it is the acoustic environment itself. New data reveals that background noise (the “cocktail party effect”) negatively impacts all levels of auditory processing—from reaction time to memory recall. Crucially, this burden is heavier for non-native speakers, whose brains must work double-time to filter signal from noise.

• Background noise negatively impacts all levels of auditory processing, from RT [Reaction Time] to speech recognition and memory recall.” (Reaction Time, Speech Recognition, and Verbal Memory Performance: Nonnative Versus Native English Speakers, Journal of Speech, Language, and Hearing Research)

(A reminder that we've covered the relationship between acoustics and learning in great depth previously.)

Research on “attention contagion” furthermore found that students implicitly pick up the inattentive states of their peers. In virtual learning environments, sitting “next to” (virtually) a distracted classmate significantly increased task-unrelated thoughts, proving that focus is a social phenomenon.

• “Students in the study did actually 'catch' inattentiveness from peers, though only when sitting next to or between inattentive classmates.” (The Effects of Attention Contagion on Task-Unrelated Thought in a Virtual Lecture, Collabra: Psychology)

Finally, as we rely more on digital tools, we face new trade-offs in how we manage memory. When external aids (like a digital list) are made slower or more “annoying” to access, children spontaneously choose to use their own memory more. It appears that cognitive effort is a calculated decision based on the efficiency of the environment.

• “Once you introduce [a lag time], they started using their memory more. It’s a trade-off... essentially the minimum that you can get away with.” (Young Minds, Smart Strategies: How Children Decide When to Use External Memory Aids, APS Podcast)

VI. School, Social-Emotional, and Contextual Effects

We are increasingly moving away from studying the brain in isolation, focusing instead on how the classroom functions as a biological ecosystem.

Researchers have proposed a new framework called “Classroom Carrying Capacity,” which conceptualizes the teacher as the leader of a sustainable biological ecosystem. A teacher’s own self-efficacy and burnout levels are primary determinants of this capacity; high-burnout environments often see a sharp decline in the quality of instructional support provided to students.

• “The quality of the classroom environment is determined, in part, by interactions between features of individual students, teachers, and the classroom, which influence one another reciprocally over time.” (Classrooms are complex host environments: An integrative theoretical measurement model of the pre-k to grade 3 classroom ecology)

While we often rush to digitize these learning environments, 2025 research suggests we should tap the brakes. A comparative study on reading mediums found that while digital reading enhances processing speed, it often compromises deep comprehension, retention, and “cognitive comfort.” The researchers suggest that the physical landscape of a book provides “spatial cues” that anchor memory—cues that vanish on a scrolling screen.

• “While digital reading enhances reading speed, it compromises comprehension, retention, engagement, and cognitive comfort.” (A comparative study on the effects of digital reading and print reading on children's reading engagement and story comprehension, International Journal of Chinese Writing Systems)

This ecosystem is further influenced by external events. In Florida, a study demonstrated that increased exposure to immigration enforcement actions led to a measurable decline in test scores for both U.S.-born and foreign-born Spanish-speaking students. The psychological burden disrupts the “cognitive bandwidth” necessary for academic performance.

• “Immigration enforcement reduced test scores for both U.S.-born and foreign-born Spanish-speaking students... these effects are more pronounced for students in middle and high schools.” (The Effects of Immigration Enforcement on Student Outcomes in a New Era of Immigration Policy in the United States, NBER Working Paper)
  

This “external weather” of politics and policy can cast a shadow that lasts a lifetime. A sobering study found that Black adults who attended segregated schools decades ago are now showing significantly higher risks of dementia. The chronic inflammation caused by the stress of discrimination appears to leave a biological scar that persists over the course of a life span.

• “When children are segregated in school, they experience discrimination... which can lead to... inflammation in the brain... even after 70 years.” (Exposure to School Racial Segregation and Late-Life Cognitive Outcomes, JAMA)

However, educational attainment itself appears to be a potent buffer. New research indicates that staying in school substantially reduces the risk of almost all studied mental disorders, suggesting that the school environment provides a critical scaffolding for resilience.

• “The finding that educational attainment is not merely a reflection of cognitive abilities suggests that educational attainment itself could be used as a unique predictor of mental disorders” (Cognitive Abilities and Educational Attainment as Antecedents of Mental Disorders: A Total Population Study of Males, Psychological Science)

Similarly, family structure plays a pivotal role. Using full population data from Denmark, researchers found that parental separation resulted in an immediate decline in reading scores (3% to 4% of a standard deviation), an effect that grew to 6.5% four years later. Notably, this decline was driven primarily by students in the middle of the skill distribution, who are often overlooked by policy.

• “Children who experience parental union dissolution are found to slow down in their biological maturation following the event... and report increased levels of stress.” (The effects of parental union dissolution on children’s test scores, OSF Preprint)

However, the social composition of the classroom can also be protective. Being exposed to a higher proportion of female peers was found to improve mental health for both boys and girls.

• “Being exposed to a higher proportion of female peers, despite only improving school satisfaction for boys, improves mental health for both boys and girls.” (More Girls, Fewer Blues: Peer Gender Ratios and Adolescent Mental Health, NBER Working Paper)

Finally, for adolescents, longitudinal neuroimaging and behavioral interviews revealed that the effort of making deeper meaning–through a cognitive process called transcendent thinking–literally sculpts the physical brain. This counteracts age-related thinning of the cerebral cortex and acted as a biological “heat shield” for those teens exposed to community violence.

• “Transcendent thinking may be to the adolescent mind and brain what exercise is to the body: most people can exercise, but only those who do will reap the benefits”. (Transcendent Thinking May Boost Teen Brains, Scientific American)

VII. The Frontier of Artificial Intelligence and Neural Modeling

The final frontier of 2025 research reveals that Artificial Intelligence is becoming a powerful mirror for human cognition. It is no longer just a tool for doing work, but a “model organism” for understanding how we think.

Groundbreaking neuroscience research is using Large Language Models (LLMs) to unlock the “black box” of the brain. Research led by Andrea de Varda demonstrated that multilingual neural networks share a “shared meaning space” with the human brain. A model trained to map brain activity in English and Tamil can accurately predict brain responses to a completely new language, like Italian, in a zero-shot transfer. This suggests that despite the vast diversity of 7,000 human languages, our brains and our most advanced models are all orbiting the same fundamental laws of meaning.

• “Encoding models can be transferred zero-shot across languages... providing evidence for a shared component of linguistic representations.” (Multilingual Computational Models Reveal Shared Brain Responses to 21 Languages, BioRxiv/Preprint)

This concept of a shared meaning space that is essentially statistical in nature provides fascinating confirmation for the hypothesis I explored in my series on AI and Language—specifically the idea that “the meaning and experiences of our world are more deeply entwined with the form and structure of our language than we previously imagined.” (See The Algebra of Language).

On a practical level, AI is proving to be a potent equalizer. An intervention in the UAE found that ChatGPT-based support significantly improved the coherence and writing scores of children with Arabic dysgraphia compared to standard instruction. Furthermore, medical students using AI-personalized pathways scored significantly higher on standardized tests, and classroom participation frequencies doubled.

• “AI tools like ChatGPT can significantly enhance writing abilities in children with dysgraphia... promoting a more inclusive and effective learning environment.” (Supplemental role of ChatGPT in enhancing writing ability for children with dysgraphia in the Arabic language, Education and Information Technologies)

However, access to AI tools is not enough. The “active ingredient” determining whether a student succeeds with AI isn't the technology, but their own belief in their ability to use it. Self-efficacy was found to be the single strongest predictor of achievement in AI-based settings, mediating the technology's effectiveness.

• “Successful achievement in AI-based settings is mediated by self-efficacy... Psychological predictors all together explained 61 percent of the variation in student achievement and persistence with self-efficacy being the most important predictor.” (Digital literacy and academic performance: the mediating roles of digital informal learning, self-efficacy, and students' digital competence, Frontiers in Education)

This self-efficacy finding provides the other half of the equation to the “barbell” theory of AI cognitive enhancement. We cannot simply hand the heavy lifting of cognition over to AI; the “weights” must still be lifted by the student to build the belief in their own capability that is required to effectively guide the technology.

Perhaps the most “sci-fi” finding of the year involves our ocean's giants. Project CETI has successfully used LLMs to decode the codas of sperm whales, discovering that whale communication contains vowels and diphthongs used in ways strikingly similar to human speech. These whales possess “culturally defined clans” with distinct dialects, suggesting that culture is a primary driver of communicative complexity across species.

• “AI analysis of sperm whale 'codas' uncovered vowel- and diphthong-like spectral patterns.” (Vowel- and Diphthong-Like Spectral Patterns in Sperm Whale Codas, Open Mind: Discoveries in Cognitive Science)

(Fans of previous roundups will appreciate the continuity here: in 2023, we highlighted Gašper Beguš's work on ANNs and whale phonology.)

Researchers have even identified a “meta-law” where the statistical patterns in the equations of physics mirror the mathematical distributions found in human language (Zipf's Law). This suggests that the same computational principles of efficiency govern both our communication and the physical laws of the universe.

• Understanding these patterns “may shed light on Nature’s modus operandi or reveal recurrent patterns in physicists’ attempts to formalise the laws of Nature . . . The patterns may arise from “communication optimisation,” where operators are defined “to describe common ideas as succinctly as possible . . These regularities could “provide crucial input for symbolic regression, potentially augmenting language models to generate symbolic models for physical phenomena.” (Statistical Patterns in the Equations of Physics and the Emergence of a Meta-Law of Nature, arXiv)

This finding of a universal statistical law of efficiency brings us back to Stephen Wolfram's concept of “computational irreducibility,” which I touched on in the AI barbell post. While language and physics may share efficient patterns (making them partially reducible), the act of learning—of internalizing these patterns into a human mind—remains an irreducible process that cannot be fully automated away.

Closing Thoughts

If there is a single thread tying the research of 2025 together, it is connectivity. Whether it is the synchronization of a mother’s heartbeat with her infant, the shared “meaning space” between an AI model and a human brain, or the “vertical flow” of language in ancient mountain villages, the evidence confirms that we are not isolated cognitive units. We are ecologically situated, rhythmically attuned, and socially dependent learners.

Here’s to another year of learning, connecting, and—hopefully—a little more positive synchronization and interactive attunement with the world around us.

#language #literacy #research #reading #writing #multilingualism #assessment #brain #cognition #academics #curriculum #wrapup

In real life, it may seem self-evident that gaining mastery takes years of intense, deliberate, and guided work. Yet the perennial optimism of students cramming the night before an exam tells us that the pursuit of a cognitive shortcut may be an enduring human impulse.

It is unsurprising, then, that students—and many adults—increasingly use the swiftly advancing tools of AI and Large Language Models (LLMs) as a shortcut around deeper, more effortful cognitive work.

The Irreducible Nature of Effort and Mastery

In a previous post in my series on LLMs, we briefly explored Stephen Wolfram's concept of “computational irreducibility”—the idea that there are certain processes cannot be shortcut and that you have to run the entire process to get the result.

One of the provocations of LLMs has been the revelation that human language (and maybe, animal language?) is far more computationally reducible than we assumed. As AI advances, it demonstrates that other tasks and abilities previously thought to reside exclusively within the human province may also be more computationally tractable than we believed.

Actual learning by any human being—which we could operationally define as a discrete body of knowledge and skills internalized to automaticity—inevitably requires practice and effort. A student must replicate essential learning steps to genuinely own such knowledge. There is no shortcut to mastery.

That said, the great enterprise of education is to break down complex and difficult concepts and skills until they are pitched at the Goldilocks level of difficulty to accelerate a learner towards mastery. This is the work, as I've explored elsewhere of scaffolding and differentiation.

In a conversation on the Dwarkesh Podcast, Andrej Karpathy praises the “diagnostic acumen” of a human tutor who helped him learn Korean. She could “instantly... understand where I am as a student” and “probe... my world model” to serve content precisely at his “current sliver of capability.”

This is differentiation: aligning instruction to the individual's trajectory. It requires knowing exactly where a student stands and providing the necessary manner and time required for them to progress.

His tutor was then able to scaffold his learning, providing the content-aligned steps that lead to mastery, just as recruits learn the parachute landing fall in three weeks at the army jump school in Fort Benning, as described in Make It Stick.

“In my mind, education is the very difficult technical process of building ramps to knowledge. . . you have a tangle of understanding and you’re trying to lay it out in a way that creates a ramp where everything only depends on the thing before it.” — Andrej Karpathy

Crucially, neither differentiation nor scaffolding is about making learning easier in the sense of removing effort. They are both about ensuring the learner encounters the “desirable difficulty” necessary to move towards mastery.

Karpathy views a high quality human tutor as a “high bar” to set for any AI tutor, but seems to feel that though the achievement of such a tutor will take longer than expected, it is ultimately a tractable (i.e. “computationally reducible”) task. He notes that “we have machines for heavy lifting, but people still go to the gym. Education will be the same.” Just as computers can play chess better than humans, yet humans still enjoy playing chess, he imagines a future where we learn for the intrinsic joy of it, even if AI can do the thinking for us.

The Algorithmic Turn and Frictionless Design

As Carl Hendrick explored recently on “The Learning Dispatch,” there's a possibility that teaching and learning themselves are more computationally tractable than we had assumed:

“If teaching becomes demonstrably algorithmic, if learning is shown to be a process that machines can master . . . what does it mean for human expertise when the thing we most value about ourselves... turns out to be computable after all?””

The problem lies in the design of most AI tools — they are designed for user friendly efficiency and task completion. Yet such efficiency counters the friction needed for learning. The Harvard study on AI tutoring showed promise precisely because the system was engineered to resist the natural tendency of LLMs to be maximally helpful. It was constrained to scaffold rather than solve.

As Hendrick notes, the fact is that human pedagogical excellence does not scale well, while AI improvements can scale exponentially. If teaching is indeed computationally tractable, then a breakthrough in AI tutoring could be an actuality. But even with better design for learning, unless both teachers and students wield such powerful tools effectively, they could lead to a paradoxical situation in which we have the perfect tools for learning, but no learners capable of using them.

Brain Rot & the Trap of the Novice

The danger of AI, then, is that rather than leading us to the promised land of more learning, it may instead impair our ability—both individually and generationally—to learn over time. Rather than going to a gym to work out “for fun” or for perceived social status, many may elect to opt out of the rat race altogether. The power of AI thus misdirected as an avoidance strategy, deflecting as much thought and effort and care from our lives as conceivably possible.

The term “brain rot” describes a measurable cognitive decline when people only passively process information.

A study on essay writing with and without ChatGPT found that “The ChatGPT users showed the lowest brain activity” and “The vast majority of ChatGPT users (83 percent) could not recall a single sentence” of the AI-generated text submitted in their name. By automating the difficult cognitive steps, the students lost ownership of the knowledge.

Such risk is highest for novices. A novice could be defined by a need to develop automatized internal knowledge in a domain. Whereas an expert can wield AI as a cognitive enhancement, extending their own expertise, a novice tends to use it as a cognitive shortcut, bypassing the process of learning needed to stand on their own judgment.

If we could plug a Matrix-style algorithm into our brains to master Kung Fu instantly, we all surely would. As consumers, we have been conditioned to expect the highest quality we can gain with minimal effort. So is it any surprise that our students are eager to take full advantage of a tool designed for the most frictionless task completion? Why think, when a free chatbot can produce output that plausibly looks like you thought about it?

Simas Kicinskas, in University education as we know it is over, details how “take-home assignments are dead . . .[because] AI now solves university assignments perfectly in minutes,” and that students use AI as a “crutch rather than as a tutor,” getting perfect answers without understanding because “AI makes thinking optional.”

But really, why should we place all the burden of betterness on the shoulders of our students, when they are defaulting to what is clearly human nature?

The Barbell Approach

Kicinskas suggests that despite the pervasive current use of AI to shortcut thinking, “Universities are uniquely positioned to become a cognitive gym, a place to train deep thinking in the age of AI.”

He proposes “a barbell strategy: pure fundamentals (no AI) on one end, full-on AI projects on the other, with no mushy middle. . . [because] you need cognitive friction to train your mental muscles.”

The NY Times article highlighted a similar dynamic in that MIT study cited earlier: students who initially used only their brains to write drafts recorded the highest brain activity once they were allowed to use ChatGPT later. Students who started with ChatGPT never reached parity with the former group.

“The students who had originally relied only on their brains recorded the highest brain activity once they were allowed to use ChatGPT. The students who had initially used ChatGPT, on the other hand, were never on a par with the former group when they were restricted to using their brains, Dr. Kosmyna said.”

In other words, AI can enhance our abilities, but only after we have already put in the cognitive effort and work for a first draft.

So Kicinskas is onto something with the barbell strategy. We start with real learning, the learning that requires desireable difficulty, friction, and effort that is pitched at the right level for where the learner is at that moment in order to gain greater fluency with that concept or skill.

Once some level of ability and knowledge has been acquired (determined by the success criteria set for that particular task, course, subject, and domain) adding AI can accelerate and enhance the exploration of that problem space.

Using AI for Cognitive Lift, Rather than Cognitive Crutch

We must therefore design and use AI in more alignment with the “barbell” strategy.

At the beginning of a student's journey, or at the beginning of the development of our own individual products, we need to double down on the fundamentals. We must carve out that space for independent thought as well as for the analog and social interaction we require to gain new insights.. This is how we build the inner scaffold required for true expertise.

On the other side of the barbell, we can more enthusiastically embrace the capacity of AI to scale our ability for processing and communicating information. Once we have done the heavy lifting to clarify our thinking, we can use these tools to extend our reach and traverse vast landscapes of data.

The danger lies in that “mushy middle,” wherein we can all too easily follow the path of least resistance and allow others, including AI, do all our thinking for us by taking our attention away from our own goals. We must choose to think for ourselves not because we have to for survival, but because the friction of generating our own thought is what gives us our agency.

In a previous post, I explored how both language and learning is a movement from fuzziness to greater precision. It is possible that AI can greatly accelerate us in that journey, even as it is possible that it could greatly stymie our growth. The key is that we must subject our fuzzy, half formed intuitions first to greater resistance until they crystallize into more precise and communicable thought. If we bypass this struggle, we doom ourselves to perpetual fuzziness, unable to distinguish between AI automated slop and AI assisted insight.

Postscript: How I used AI for this Post

I use AI extensively in both my personal and professional life, and writing this post was no exception. I thought it might be helpful to illustrate some of the arguments I made above by detailing exactly how AI both posed a risk to my own agency and served to enhance it during the creation of this essay.

I began by collecting sources. I had come across several articles and a podcast that felt connected, sensing emerging themes that related to my previous posts on LLMs. I started sketching out some initial thoughts by hand, then uploaded my sources into Google's NotebookLM.

My first impulse was to pull on the thread of “computational irreducibility.” I knew there was an interesting tension in language between regularity and irregularity, so I used Deep Research to find more sources on the topic. This led me down a rabbit hole. By flooding my notebook with technical papers, the focus shifted to abstractions likeKolmogorov Complexity and NP-completeness—fascinating, but a distraction from the pedagogical argument I wanted to make. Realizing this, I had the AI summarize the concept of irreducibility and then deleted the technical source files to clear the noise.

I then used the notebook to explore patterns between my remaining sources. Key themes began coalescing. It was here that I made a classic mistake: I asked Google Gemini to draft a blog post based on those themes.

The result wasn't bad, but it wasn't mine. It completely missed the actual ideas that I was trying to unravel. I realized I was trying to shortcut the “irreducible” work of synthesis. To be fair to my intent at the time, however, I was really just interested in seeing whether the AI gave me any ideas I hadn't thought of, from a brainstorming stance. It wasn't very useful, however, so I discarded that approach, went back to my sources, and spent time thinking through the connections as I began drafting out something new.

I then began to draft the post in Joplin, which is what I now use for notes and blog drafts. I landed on the analogy of the Hollywood training montage as the way to begin, and I then pulled up Google Gemini in a split screen and began wordsmithing some of what I wanted to say. As I continued drafting, I used Gemini as an editorial support. It advised syntactical revisions and fixed a number of mispellings. I then used it to help me expand on a half-formed conclusion, as well as for cutting an extended naval-gazing section that was completely unnecessary.

Gemini tends to oversimplify in its recommendations, however, and I didn't take all of it's suggestions. I generated some images in NotebookLM based on all the sources, and also enhanced an image I had already made previously using Gemini. Finally, I did a few additional rounds of feedback between NotebookLM to reconsider my draft in relation to all the sources in my notebook, and then returned with that feedback in Gemini, and again went through my draft on a split screen. This additional process gave me some good suggestions for reorganization and enhancement of some of the content.

In the end, I almost misled myself by trying to automate the thinking process too early. It was only when I returned to the “gym”—drafting the core ideas myself—that the AI became useful. My experience writing this confirms the barbell strategy: draft what you want to say first to build the conceptual structure, then use AI to draw that out further, and to polish and enhance it. Be very cautious in the mushy middle.

#AI #LLMs #cognition #mastery #learning #education #tutoring #scaffolding #differentiation #barbell

The capacity for intense focus in our students is a finite resource—a cognitive fuel tank that can, and does, run low. We see the results in the classroom: irritability, impatience, and a fraying of impulse control. But what if one of the most powerful tools for refueling that tank wasn't a new pedagogical strategy, but something far more fundamental?

Five years ago, I wrote about the profound impact that greenery can have on health and learning in The Influence of Greenery on Learning. When I recently listened to Dr. Marc Berman, Director of the Environmental Neuroscience Lab at the University of Chicago, expand on this research on the Many Minds podcast, it prompted me to revisit that post. I was humbled to realize how many of his foundational studies I had completely overlooked. This new understanding reveals that nature is not just an amenity, but a necessity for cognition.

At the start of the episode, Berman unpacks one of the theories I had very briefly mentioned on why greenery might be so rejuvenative: Attention Restoration Theory. According to Berman, the theory posits that our capacity for intense focus, or directed attention, is a finite resource—a cognitive fuel tank that can, and does, run low. When it’s depleted, we can see the results at home and in the classroom: irritability, impatience, and a fraying of impulse control.

Natural environments, on the other hand, engage our involuntary attention—the effortless, bottom-up engagement of our senses captured by the gentle rustling of leaves or the movement of light through the clouds, and it allows our depleted resources for directed, intense focus to restore themselves. Berman terms this “soft fascination.” This is wholly distinct from the “harsh fascination” of a chaotic urban scene, with its blaring horns and noise, which consumes our mental resources.

The cognitive benefits are significant. One of the studies that kickstarted Berman’s research showed a 20% improvement in cognitive performance after a walk in nature. This boost occurred even when participants didn't particularly enjoy the walk, demonstrating a powerful, mood-independent effect.

This research has profound implications for educational equity. A follow-up study found that individuals with major depressive disorder (MDD) see even more significant cognitive gains from a nature walk. Conversely, a walk in an urban environment can actually worsen their cognitive performance. This suggests that the lack of green space in many under-resourced communities can be actively harmful to our most vulnerable students. Access to restorative natural environments should therefore not be seen as a luxury, but as a prerequisite for equitable learning.

But what is it about nature that is so restorative? Berman’s explication identifies specific “active ingredients.” It turns out my hunch about fractals was on the right track. His team analyzed what they call low-level visual features to quantify what makes a scene feel “natural.”

Key among these are:

• Fractalness and Compressibility: Natural scenes have high “fractalness”—the repetition of similar patterns at different scales. This visual structure means they are also more “compressible,” like a JPEG file. Our brains find this informational efficiency less demanding to process, which frees up cognitive bandwidth.
• Curved Edges: Natural environments have a high density of non-straight, curved edges, whereas our built environments are dominated more by rigid, straight lines. These curves are not only easier on the eyes, but as one study found, they are also correlated with a viewer's tendency to reflect on deeper topics like their life's journey and spirituality.

Berman furthermore points to additional sensory qualities of nature that go beyond the mere visual:

• Auditory Stimuli: Brief exposure to natural sounds like birdsong, wind, or flowing water has been shown to improve cognitive performance when compared to urban noise.
• Olfactory Stimuli: The air itself carries restorative properties. The scent of damp earth after rain or the airborne chemicals (terpenes) emitted by pine trees can impact our well-being through the olfactory pathway.

For restoration to occur, according to Attention Restoration Theory, an environment must provide a sense of “Being Away” from daily pressures, have enough richness to get lost in (“Extent”), and support a person’s intentions (“Compatibility”). When these elements combine, the mind can truly recharge.

Now pivot that to an educational setting. Imagine a school that embodies these principles. Instead of a long, featureless corridor (no “Extent”), picture a hallway that curves and uses natural materials with fractal patterns like wood grain. Imagine the school itself providing a space for “Being Away” from stressors, a place for creativity and inspiration. By incorporating more trees and natural design principles into our schools, we can improve learning.

Thankfully, we don’t need a week-long immersion in a forest; studies confirm that a restorative “dose” of nature can be as short as 20 minutes. In a world of education reform obsessed with short-term metrics, this research demands we look at a more fundamental input: the physical environment itself. It forces us to ask a provocative question: could 6 hours of instruction plus 2 hours in a park be more productive than 8 straight hours behind a brick wall? The science increasingly suggests that the answer is yes.

For a full, fascinating dive into the research, I highly recommend listening to the entire podcast episode, and then poking around into some of Berman’s studies!

#greenery #learning #attention #neuroscience #schools #ecosystems #wellbeing #AttentionRestorationTheory #environmentalneuroscience #equity

(Note: this was cross-posted on my other blog, Schools & Ecosystems)

While we meticulously map out and prepare for the activities we engage our students in, the specific linguistic structures and vocabulary we employ often remains implicit, almost accidental. This raises critical questions: which aspects of our classroom talk truly accelerate literacy – is it sheer volume, vocabulary precision, or syntactic complexity? And how can we become more deliberate and intentional architects of this vital linguistic environment for all students, including those developing multi-dialectalism and multilingualism?

My recent presentation at ResearchED in NYC ventured into this territory, examining the research on how the linguistic environment we curate can influence student literacy achievement.

The Power of Classroom Talk: More Than Just Words

Why this focus on classroom talk? Because literacy isn't built in a vacuum. While foundational skills like decoding and spelling are absolutely critical (and have for all too long been sidelined), the elementary ELA block at large all too often focuses on isolated skills.

Despite elementary schools in the U.S. dedicating significantly more time to ELA than any other subject, reading scores (like those from state ELA tests or the more nationally normed NAEP) often remain stubbornly flat, including here in NYC. This prompts a crucial question: is simply adding more ELA time the answer, or do we need to rethink how we build literacy—both within and beyond ELA?

This is where focusing on content-rich talk across the content areas becomes vital. Subjects like social studies, science, math, and the arts offer fertile ground for developing the academic language and background knowledge that underpin strong literacy. In fact, some research suggests this cross-curricular approach may be more effective for reading comprehension than simply adding more ELA time. For example, a 2020 study by the Fordham Institute found that increased instructional time in social studies—but not additional time in ELA—was associated with improved reading comprehension for elementary students (Tyner & Kabourek, 2020). Notably, the students who benefited most from additional social studies time included girls and those from lower-income and non-English-speaking homes. Tackling the challenge of building a strong foundation begins, fundamentally, with the language we choose to use and explicitly teach across all subjects.

Yet social studies—and other content areas—occupy an increasingly small portion of an elementary student’s learning (more recent RAND paper on this).

The Problem We Face

This focus is critical because many students, particularly in the K-5 grades, can encounter significant hurdles in developing robust literacy and language skills that are essential for academic success. These challenges can be particularly acute for multiidialectal or multilingual learners navigating academic language demands alongside or in addition to their home language(s). Key challenges include:

• Foundational Skills Gaps: Some students do not receive the focused instruction and practice they need in decoding and spelling to become fluent readers and writers.
• Knowledge and Language Gaps: Many students lack consistent and cohesive opportunities to build the background knowledge and language necessary to understand complex topics across different subjects, while building on and connecting to the cultures, schema, and languages they bring.
• Complex Language Exposure: The majority of students need more exposure to, and structured practice with, reading, writing, and talking using the complex language inherent in disciplinary discourse and texts.

What the Research Says: Listening In on Learning

I used the wonderful study by Jeanne Wanzek, Carla Wood, and Christopher Schatschneider, which I have highlighted in this blog before as the anchor for my presentation. Using LENA devices to record classroom instruction, they found:

• Teachers, on average, used relatively few academic or curriculum-specific vocabulary words.
• Crucially, teachers who did use more academic words had students demonstrating higher vocabulary achievement by the end of the school year. This held true even when controlling for the teachers' overall expressive vocabulary and across students with varying incoming abilities.
  • The takeaway: The specific words we choose during instruction have a measurable link to student vocabulary growth, a crucial component of academic success for all learners.

Correlation or Causation? Towards Stronger Links

Correlation, of course, isn't causation. Does using more academic language cause better outcomes, or do teachers with higher-achieving students simply use more academic language? While the Wanzek et al. study is correlational, a growing body of research points towards a causal link between targeted language exposure/instruction and improved outcomes. Here’s just a smattering:

• Conversational Turns: Interventions increasing parent-child conversational turns led to language skill improvements and predicted neurocognitive changes (Romeo et al., 2021).
• Mathematical Language: An RCT using dialogic reading to boost mathematical language positively impacted preschoolers' general math skills (Purpura et al., 2017).
• Classroom Math Talk: Teachers using more mathematical language were found to be more effective at raising student test scores in upper elementary grades (Himmelsbach et al., 2024).
• Content Literacy: A sustained literacy intervention grounded in science and social studies content led to lasting improvements in vocabulary, reading comprehension (across domains), and even math, demonstrating far transfer effects (Kim et al., 2024).
  • The takeaway: The pattern across these studies strongly suggests that actively improving the language environment through intentional instruction yields real results in student learning, with content-rich instruction showing particular promise for multilingual learners

Defining and Developing Academic Language

So, what is this “academic language” we're aiming for? It's the formal, complex, often abstract and decontextualized language common in school, texts, and professional settings (NYSED, Lesaux & Philips Galloway; Philips Galloway et al., 2019). Since this language isn't always prevalent outside school, the classroom becomes the primary place many students will learn it, making our role crucial, especially in fostering academic language development for multilingual learners.

Understanding how language typically develops—and recognizing that multilingual development adds further layers of complexity and potential cognitive benefits—helps us see where to intervene and build bridges for students:

1. Contextualized Interaction: Early conversational turns, rooted in the immediate environment.
2. Oral Storytelling: Moves towards abstraction, requiring inference and schema-building beyond the 'here and now'.
3. Shared Reading: Introduces more decontextualized language—denser vocabulary, complex sentences, formal structures typical of written text (I’ve rounded up a list of studies related to this).
4. Written Language: Characterized by rarer, more abstract words, complex syntax (like nominalizations, passive voice, relative clauses), and formal discourse structures

Our instruction aims to help students navigate this journey towards greater precision and abstraction. Leveraging students' home languages can serve as a powerful bridge along this continuum.

Explicit Teaching Meets Implicit Learning: Achieving “Escape Velocity”

So, how do we teach this complex language effectively?

While explicit teaching of vocabulary or grammar acts as a necessary accelerator, it works best when launching students into an environment rich with coherent and cohesive implicit learning opportunities. This explicit scaffolding is vital for all learners navigating complex academic language, and particularly crucial for multidialectal and multilingual students acquiring these structures in the more formal English used in school. Mark Seidenberg calls this synergy achieving “escape velocity”—where explicit instruction scaffolds and enables students to learn powerfully from the sheer volume of language they encounter through reading, writing, and discussion. Our goal is to engineer this velocity for all learners.

As we’ve also explored on this blog, part of building this velocity is about providing our kids with more texts and more talk—”textual feasts,” as Dr. Tatum calls it.

Putting Research into Practice: Classroom Strategies

How can we intentionally weave denser and more complex academic language into our daily practice, while valuing and leveraging the linguistic diversity of our students? It involves concrete, planned actions:

1. Plan to Amplify Knowledge & Language:

   • Identify core concepts in a unit/text.
   • Pinpoint the essential academic vocabulary used to explain these concepts.
   • Explore morphology and etymology (e.g., using tools like Etymonline) to deepen understanding, including potential cross-linguistic connections.
   • Analyze how these words function in different sentences and contexts.
   • Plan structured opportunities for students to practice reading, writing, and speaking with these words.

2. Leverage Multimodal Text Sets: Immerse students in a topic through various texts (articles, books, videos, images) and modalities. This creates multiple, varied exposures to related concepts and vocabulary.

3. Structured Supplements for Read-Alouds: Don't just read; enhance read-alouds by providing concise definitions, examples, asking stimulating questions that require using target vocabulary, connecting to prior knowledge, and using concept maps (Mosher & Kim,, 2025). Consider incorporating home language previews or connections where appropriate.

1. Explicitly Teach Morphology & Leverage Cross-Linguistic Connections: Build awareness of word parts (prefixes, suffixes, roots) and connections between words across languages. This is especially powerful for multilingual learners; recognizing shared roots and patterns (like transparent/transparente) and using contrastive analysis between languages (like comparing verb forms) can unlock meaning and build metalinguistic awareness. Use a consistent multisyllabic word decoding strategy. Use tools like concept/semantic maps to help visualize connections, including across languages.

1. Structure Reading Instruction (Before, During, After): Be intentional about the purpose of each read:
   • Before: Build background, preview text and vocabulary. Activate or build relevant background knowledge, connecting to diverse student experiences.
   • During (1st Read): Focus on flow and gist, model fluency, check basic comprehension.
   • During (2nd Read): Zoom in on specific words, sentences, author's craft. Practice paraphrasing key details.
   • During (3rd Read): Analyze structure and language more deeply. Ask inferential questions.
   • After: Review, engage with target vocabulary/language, summarize, practice speaking/writing using mentor sentences and target words.
2. Zoom In and Amplify: When revisiting texts, strategically select specific words or sentences to focus on. Use routines (echo/choral reading, dictation, sentence combining, contrastive analysis) to deepen understanding and usage. (See the Zoom In and Amplify Menu resource for ideas).

These routines can often be adapted using contrastive analysis or strategic invitations to use and connect to home language for multilingual learners.

Moving Forward: The Bottom Line

The research is increasingly clear: the language we choose to use and teach matters. By consciously choosing to immerse students in rich, academic language within and across content areas, providing both explicit instruction and ample opportunities for implicit learning through meaningful interaction with texts and topics, we can significantly enhance language development and overall literacy achievement, creating more equitable opportunities for all students, including multidialectal and multilingual learners. It requires intentional planning and a shift towards seeing every teacher as a teacher of language, but the potential payoff for our students is enormous.

To effectively address the challenges and leverage the power of classroom talk, the evidence points towards these key actions:

• Recognize the crucial role academic language plays in student literacy development across all subjects, recognizing its importance most especially for students developing multilingualism
• Understand the interplay between explicit language instruction (the accelerator) and the implicit learning that occurs through rich language exposure (the fuel).
• Actively implement strategies to intentionally increase the quantity and quality of academic language used in classroom instruction and student interactions daily, leveraging students' diverse linguistic resources as assets.

#literacy #education #research #AcademicLanguage #TeacherTalk #ReadingComprehension #Vocabulary #Instruction #ResearchED #MultilingualLearners #ENL #Biliteracy

2024 was another great year filled with fascinating research.

Over the course of this year, I’ve written a few posts about some of it:

• How to Externalize Internal Language
• Research Highlight 3: The Reading Profiles of English Learners
• Research Highlight 4: Structuring Classroom Learning for Student Success and Agency
• A speculative series (7 posts so far) on AI, LLMs, and Language!
• Research Highlight 5: Learning In a New Language Takes Effort

Last year, I began a tradition that seems worth maintaining: reviewing all the sundry research that has come across my radar over the course of 2024. The method I used to create this wrap-up was to go back through my X/Twitter and Bluesky timelines starting in January, and pull all research related tweets into a doc. I then began sorting those by theme and ended up with several high-level buckets, with further sub-themes within and across those buckets.

The rough big ticket items I ended up with were:

• The Science of Reading and Writing
• Content Knowledge as an Anchor to Literacy
• Studies on Language Development
• Immigration, Multilinguals, and Multilingualism
• Rhythm, Attention, and Memory
• School, Social-Emotional, and Contextual Effects

The Science of Reading and Writing

There were some insightful, confirming, and surprising studies adding to the body of what we know about reading and writing development.

Dyslexia

There was a focus on revisiting the definition of dyslexia and considerations for both streamlining and expanding it.

• “Given the potential for the definition of dyslexia to be conflated with an eligibility category, along with other considerations, another significant theme emerged: the need to streamline the definition for more effective identification and intervention.” Annals of Dyslexia, Odegard et al.

• “A new definition of dyslexia...needs to transcend both past unitary characterizations and past assumptions based largely on the English orthography” Annals of Dyslexia, Wolf et al.

Speaking of moving past assumptions solely based on the English orthography, another study in this issue focused on how dyslexia manifested similarly and differently in children in Beijing, Hong Kong, and Taipei.

• The study indicates that while some core deficits like phonological processing are present across all locations, the manifestation of dyslexia varied due to differences in script complexity, language, and teaching methods.

• “Among the most interesting findings in the present study is that, compared to word reading, our task of character reading (fluency) was better able to distinguish children with or with-out dyslexia in Hong Kong and Taipei. This may be because characters are more difficult to recognize when presented alone than in multiple-character words.” Annals of Dyslexia, Jue Pan, et al.

Want to improve phonemic awareness in pre-readers at risk for dyslexia? Have them play Space Invaders Extreme 2!

• “More than 80% of the at-risk pre-readers in the AVG [Action Video Game] group showed an improvement in phonemic awareness that was above the mean gain observed in the combined control groups, indicating the treatment's high efficacy.” Science of Learning

• What the heck is going on here? The researchers hypothesize that action video games, which can be fast-paced and unpredictable, can support more efficient integration of sensory input, which may be less efficient or slower in children at risk for dyslexia.

Phonological and Morphological Awareness

When it comes to polysyllabic word reading (words like “dinosaur” or “construction”), this study found that kids in grade 3-5 who already knew a word were more likely to read it correctly. While this study doesn’t provide implications for students learning English, clearly ensuring that they can connect the meaning of words to the forms of words is important – more on this below in the section on multilingual learners. Journal of Experimental Child Psychology

And indeed, knowing more about the forms of words – and not only their sounds – “is an important longitudinal predictor of spelling development.” Journal of Research in Reading

When learning new words, the distinctiveness of those words helps them to be remembered.

• “Results showed that those words which exhibited distinctive characteristics – whether due to clear speech style, low frequency, or low density – were remembered better. The finding supports the Distinctiveness Hypothesis, suggesting that our capacity for remembering words relies on their distinctiveness, rather than on our capacity for recognizing them in real time.” Psychology of Language

And let’s not forget the importance of morphological awareness!

• “we have found that preschool morphological awareness, assessed prior to any formal literacy instruction, is a unique predictor of later reading comprehension but not of word reading skills.” Scientific Studies of Reading

While there are large differences within and between studies, morphology instruction appears to be effective for improving reading and spelling outcomes, and spelling effects can transfer to untrained words. Educational Psychology Review

Morphological systems are dynamic – balancing regularity and irregularity of forms.

• ”...a balance between regular structures and exceptional forms not only facilitates generalization but may also be essential for efficient linguistic performance and adaptation.” Cognitive Linguistics

Orthographic Processing

As we read, our eyes fixate briefly on the words in print. But we are not simply fixating on the center of words – we are also using what we know of the statistical structure of language to target the position in a word that minimizes uncertainty and maximizes our reading efficiency.

• “we provide causal evidence that the way in which a language distributes information affects how readers land on words.” Journal of Memory and Language

The presence of nearby words can interfere with the brain's ability to process a fixated word, suggesting that skilled reading involves a constant balancing act.

• “We conclude that skilled reading involves a constant complex interplay between the drive toward efficiency, which requires a broad attentional field, and the need to shield processing from interference, which limits attentional breadth.” PsyArXiv Preprints

Beyond Word Reading

After all, acquiring reading fluency is not only about recognizing words in isolation but also about efficiently processing them in sequence.

• “These findings suggest that, beyond individual word recognition, reading fluency development also requires efficient processing of multiple items presented in serial format (termed ‘cascaded processing’).” Scientific Studies of Reading

Language regions in the left hemisphere light up when reading uncommon sentences, while straightforward sentences elicit little response.

• ”...the sentences that elicit the highest brain response have a weird grammatical thing and/or a weird meaning.” MIT News

When it comes to reading fluency, however, we need to be cautious in interpreting oral reading fluency rates as it relates to reading comprehension. ORF measures are widely used as a proxy for reading comprehension.

• “The results of this study suggest that outcomes from oral reading fluency assessments that focus on rate and accuracy may not be valid indicators of reading comprehension when passages include complex, academic language.”

Why might this be? Many widely used tests of reading fluency may use simplified texts, which most students can comprehend more easily, thus inflating the correlation between fluency and comprehension. Journal of School Psychology

Gaining fluency in writing also leads to higher quality writing.

• “Results showed that children who had higher writing fluency...had higher quality writing, and this was explained directly by transcription skills and indirectly by executive functions such as working memory.” Journal of Educational Psychology

Improving Reading and Writing

There is a lot of improvement still needed in classroom instruction for reading comprehension, as this follow-up from a 50 year old observation study found. While research-based practices have increased, teachers continue to spend time mainly engaging in IRE styles of discourse (initiation-response-evaluation) rather than engaging students in extensive discussion of text or teaching practices and knowledge that more deeply support reading comprehension.

• “based on the findings from the observation studies reviewed, we have considerable opportunity in classroom instruction to enact the research-based practices for teaching reading comprehension that have been identified through research so far.” Scientific Studies of Reading

While we know that kindergarten reading intervention can be critical for students at risk, providing the right level of fidelity and dosage requires supporting teachers with implementation.

• “The results suggest teachers may need more systems-level support to ensure the intensity of instruction required to improve the early reading skills of students at risk for reading difficulties.” The Elementary School Journal

To further the point that teachers need systems-level support: aligning Tier 2 interventions with Tier 1 instruction leads to improved content knowledge, vocabulary, and content reading comprehension for kids who need it the most in fourth grade.

• “Findings from the present study suggest that aligned instruction may be especially beneficial for students with inattention” Journal of School Psychology

In a new report, “The Opportunity Makers,” TNTP similarly stressed the importance of instructional coherence and consistency in schools that were making a difference in students’ learning outcomes.

• “Research shows that instructional coherence in a school helps students learn, while incoherence creates confusion and saps students’ confidence. According to Newmann et al. (2001), “Students are more likely to engage in the difficult work of learning when experiences within classes, among classes, and over time are connected to one another. When faced with incoherent activities, students are more likely to feel that they are targets of apparently random events and that they have less knowledge of what should be done to succeed.” TNTP

One thing is for sure: simply adding more independent reading time to a school schedule is no guarantee of improved reading comprehension.

• “Our results from 14 primary studies comprising 5,522 participants in the treatment group and 4,966 in the control group alluded to no meaningful beneficial effects of independent reading on reading outcomes.” Reading & Writing Quarterly

Integrating reading with explicit writing instruction “can improve primary grade students’ writing, discourse knowledge, planning, oral language, and spelling skills.” Scientific Studies of Reading

Writing is a technology that has further differentiated humans from other animals.

• “writing enabled humans to think more abstractly and logically by increasing information capacity.” Nature Reviews Psychology

Writing by hand is critical to not only developing literacy – but for adults for deeper thinking and learning.

• “These visually demanding, fine motor actions bake in neural communication patterns that are really important for learning later on.” NPR

Screen Time and Literacy

Most screen time can be detrimental to language and reading development, and to deeper comprehension of what we read. And yet digital technology is increasingly ubiquitous in classrooms and in our lives.

• “Our results demonstrate a positive association between shared reading and vocabulary in both age groups, and a negative association between screen time and vocabulary in 24-month-olds.” Journal of Child Language

• “Television seems to be the medium most detrimental to children’s skills, as it is used in a passive manner and is often characterised by language and content that do not suit the child’s processing mode.” Brain Sciences

• “For every extra minute of screen time, the three-year-olds in the study were hearing seven fewer words, speaking five fewer words themselves and engaging in one less conversation.” JAMA Pediatrics

• “The results of the two meta-analyses in the present study yield a clear picture of screen inferiority, with lower reading comprehension outcomes for digital texts compared to printed texts, which corroborates and extends previous research.” Educational Research Review

All of that said, there is evidence that enhancing the interactivity of a PBS KIDS science show with conversational agents enhances their science learning. Journal of Educational Psychology

Content Knowledge as an Anchor to Literacy

Speaking of reading comprehension and science, ever since E.D. Hirsch, Jr. first proposed the concept of “core knowledge,” there has been increasing research demonstrating the importance of content knowledge to reading comprehension and literacy development – and vice versa.

Background Knowledge, Reading Comprehension, and the Novice-Expert Continuum

Hugh Catts and Alan Kamhi wrote a great piece on the importance of background knowledge to reading comprehension, stressing the understanding of reading comprehension as a constellation of skills rather than a singular component.

• “reading comprehension is one of the most complex activities that we engage in on a regular basis, and our ability to do so is dependent upon a wide range of knowledge and skills. These include relevant background knowledge and reasoning abilities. Also, like listening comprehension, it is dependent on well-developed language abilities, including not only vocabulary knowledge but also an understanding of grammar and text-level structures (e.g., pronoun referencing and story structure). In addition, it is influenced by the nature of the text being read (e.g., its topic, complexity, and cohesion) and the purpose of reading (e.g., to study for a test or evaluate an opinion piece). Finally, it is acquired not in a few short years, but over one’s lifetime. For these reasons, comprehension needs to be differentiated from skill-based components of reading and treated as the complex behavior it is.” American Educator

As with reading, it’s important for writers to remember the novice vs. expert continuum, especially in terms of their audience. This study found that journalists write mostly at the level that makes most sense to them – but their readers would far prefer reading texts that were simpler.

• “those who write the news read it differently from those who merely consume it. As observed in many other areas, expertise may undermine effective perspective-taking” Science Advances

After all, expertise and experience is a precondition for flow, as brain scans of Philly jazz musicians reveals. The Conversation

Building Interdisciplinary Knowledge

Disciplinary read-alouds can build interdisciplinary student knowledge and reading comprehension through the use of “structured supplements” that promotes transfer and connections between schema and vocabulary. In this study, students connected social studies and science content and texts.

• “The mediation results suggest that teacher language scaffolds can function as temporary dialogic supports that go above and beyond the intervention script and support students’ reading comprehension.”

• “In essence, treatment group teachers provided more opportunities for students both to hear and use academic vocabulary by engaging in discussions to make connections between known and new topics.” Scientific Studies of Reading

“This experimental study illustrates how sustaining and spiraling science schemas (background knowledge) and vocabulary from Grades 1 to 3 can improve students’ ability to comprehend passages in science, English language arts, and mathematics. Furthermore, findings suggest that systematically building background and vocabulary knowledge can sustain positive gains in elementary-grade students’ reading comprehension ability through the end of Grade 4, 14 months after the conclusion of the intervention activities.” Developmental Psychology; also see Neena Saha’s great Reading Research Recap on this study

Boosting knowledge of science vocabulary improves science knowledge.

• “Greater science vocabulary knowledge was associated with higher science test scores for children with language/literacy disorders (LLDs) and typical language development (TD). These findings indicate that increasing science vocabulary knowledge may improve science achievement outcomes for students with LLDs or TD.” ASHA Language, Speech, and Hearing Services in Schools

Another study demonstrated that a classroom-based content literacy intervention significantly improved argumentative writing skills for both English learners (ELs) and their English-proficient (EP) peers in grades 1 and 2. The intervention consisted of thematic units in social studies and science designed to build students’ content and vocabulary knowledge through informational texts and concept mapping and to transfer their schema to argumentative writing and research collaboration. Journal of Educational Psychology

If we want more literacy instruction integrated into secondary content area classrooms, then we had better consider “ease of use” for teachers to incorporate those practices successfully. Reading Psychology

Math, Language, and Literacy

Content knowledge and literacy and language development aren’t only about social studies and science, by the way. Math and reading fluency are connected!

• “variations in reading fluency predict variations in arithmetic fluency in Grades 1 to 3. Meanwhile, variations in arithmetic fluency predict variations in reading fluency in Grades 1 to 2.” PsyArXiv Preprints

In fact, language is fundamental to math.

• “We must be handed the cognitive tools of numbers before we can consistently and easily recognize higher quantities.” The Conversation

An analysis of 1,657 4th/5th grade lessons in 317 classrooms in 4 districts finds “students’ exposure to mathematical language varies substantially across lessons” and students make more progress in classrooms where teachers use more mathematical language. EdWorkingPapers

Furthermore, “Students learn more math skills when their teacher devotes more class time to individual practice and assessment. In contrast, students learn more language skills when their teacher devotes more class time to discussion and work in groups of students” Harvard GSE Ed Magazine

When it comes to supporting students at various levels of proficiency in the language of instruction (in this study’s case, German), language supports should be provided only to those at lower levels of proficiency.

• “”The findings indicate that the principle of 'more is better' does not always apply to additional language support, and that identical learning materials may not be suitable for all students.” Educational Studies in Mathematics

Another study highlighted the interconnected nature of reading and content knowledge, showing that early reading skills boost initial growth in science and math. Furthermore, as children progress through elementary school, the mutually reinforcing relationship between reading proficiency and knowledge in science and math becomes increasingly strong, with each skill continually enhancing the other.

• ”Notably, multilingual students instructed in their native languages demonstrate more robust connections between early domain knowledge and subsequent reading proficiency. These findings emphasize the benefits of native-language instruction for fostering reading and domain knowledge, providing educators with clear evidence of the importance of incorporating native-language support in early education.” Developmental Psychology

Studies on Language Development

We’ll dig far deeper into multilingualism and its relation to overall language and literacy development in our next section. Before we do, however, let’s look at some of the studies related to language development at large.

The Foundations of Language and Literacy

The acoustic environment that one is born into is important for all species.

• “exposure of birds that are in the egg to moderate levels of noise can lead to developmental problems, amounting to increased mortality and reduced life-time reproductive success. Such noisy conditions at the beginning of acoustic life may affect behavioral and cognitive development in many more species.” Science

• A reminder that we’ve explored the impact of acoustics previously inThe Influence of Acoustics on Learning.

Animals may lack language (and other human-distinctive behavioural traits) because they perform badly at remembering sequences of stimuli.

• “..the presumed absence of evolutionary continuity between animal communicative systems and human language aligns well with the view that language structure is culturally emergent rather than inborn.” Trends in Cognitive Sciences

For humans, “Language learning begins in the womb, and it begins with prosody. Exposure to speech in the womb leads to lasting changes in the brain, increasing the newborns’ sensitivity to previously heard languages.”

• Did you know that “”newborns cry in the accent of their mother tongue”? Aeon

Not only that, but how the brains of newborns respond to speech is predictive of their later literacy development.

• “Stronger neural responses measured in the brain in infancy to changes in speech sounds were associated with better pre-reading skills, such as rapid naming.” University of Helsinki News

Furthermore, the connectivity of the infant brain–specifically in the inferior frontal gyrus (IFG) strongly predicts future reading abilities. The strength of these early neural connections in infancy forecasts phonological skills at kindergarten, which in turn mediate the relationship between the infant brain's organization and school-age reading proficiency.

• “Overall, our findings illuminate the neurobiological mechanisms by which infant language capacities could scaffold long-term reading acquisition.” Developmental Cognitive Neuroscience

Sensitivity to the sounds of speech is not only important in infants. For adults, too, “”individual differences in sensitivity to phonetic categories mediates speech perception in challenging listening situations.”
The Journal of the Acoustic Society of America

The Patterns of Language

People learn patterns better when they are simple and consistent. This includes languages, but also visual, auditory, and even tactile information. This shapes not only how we learn languages but also how languages evolve over time.

• “the patterns that are more easily learned are precisely the ones that are found most frequently across languages.” Quarterly Journal of Experimental Psychology

Our brains are more aligned with AI and Large Language Models (LLMs) than we may think.

• Even without training, a simple computer model can process language much like the human brain does, if it's built with certain key features like how it breaks down words and uses context. arXiv Preprints

• “The better a model was at predicting the next word it would hear, the more likely it was to align with brain data.” PNAS

• A reminder that I did a deep dive series on AI, LLMs, and Language.

This paper shows our brains can effortlessly detect patterns at both fast and slow timescales (prioritizing quick changes). Remarkably, this dual-level learning process can be modeled by simple neural networks, suggesting a unified mechanism for processing complex temporal information. Journal of Cognitive Neuroscience

The sounds and rhythm of language, also known as prosody, were found to play a role in how we process syntax.

• “Our findings indicate that the neural representation of syntactic phrase boundaries is enhanced when they are aligned with strong prosodic boundaries, suggesting that prosodic cues scaffold the brain’s ability to process syntactic information.” Communications Biology

And the brain processes phonemes in parallel, meaning multiple sounds can be processed simultaneously without interference. What’s also crazy is that our brains actually retain a speech sound briefly as other sounds are coming in, so there is elapsed processing time. Also fascinatingly, the first phoneme of a word appears to be processed differently from subsequent phonemes. The neural representation of the first phoneme can be decoded earlier, and its information is maintained for a longer duration.

• Learned about this one from Stephen Wilson’s The Language Neuroscience podcast interview with Laura Gwilliams about her 2022 paper in Nature Communications.

The Role of Linguistic Input

You’ve no doubt heard of the infamous “30 million word gap.” Yet one of the key themes of more recent research – including this year’s – is that the quality of input that children receive is far more important than quantity alone.

A study introduced a novel term—“burstiness”—to describe irregular, “spiky” bursts of speech which were found to be more beneficial for vocabulary growth than a consistent stream of language. The researchers used child-centered audio recorders to track the language environments of 292 children aged 2-7 years, over 555 days.

• ““children who heard spiky, more intense bouts of input had larger vocabularies. . . Input bursts provide rich opportunities for children to learn, while ebbs give children the opportunity to consolidate the new referent information and entrench representations to facilitate later retrieval.” Developmental Science

“Together these findings highlight the fact that quality of input per se matters more than child age, grade, or language of instruction.” Psychological Bulletin

Gestures

Linguistic input is not merely confined to speech. When referents are not physically present, caregivers use multimodal cues, particularly iconic cues. Iconic cues are communicative forms, such as words, signs, or gestures, that have a resemblance to the sensory-motor or conceptual properties of their referents.

• “the affordances of multimodal, iconic cues that caregivers use in interactions can allow children to draw on prior knowledge gained through general cognitive and motor development to scaffold their vocabulary learning.” Child Development

In fact, gestures provide a critically important source of input.

• “our minds can change when we see others gesture and when we ourselves gesture. However, when pitted against each other, doing our own gesture is a more powerful learning tool than seeing someone else's gesture, at least when young children learn about mental rotation.”

• “adding gesture to a lesson can boost performance in children from less advantaged homes so that it is equal to performance in children from advantaged homes.” Topics in Cognitive Science

Speaking of gestures – the stereotype that Italians gesture more effusively than others certainly bears out when you compare them to Swedes (my heritage).

• “The results show that (1) Italians overall do gesture more than Swedes; (2) Italians produce more pragmatic gestures than Swedes who produce more referential gestures; (3) both groups show sensitivity to narrative level: referential gestures mainly occur with narrative clauses, and pragmatic gestures with meta- and paranarrative clauses.” Frontiers in Communication

Shared Reading

Of course, we also know that one of the richest sources of linguistic input, especially early in life, is via shared reading.

• “Our current analysis suggests that shared reading (or a more broadly assessed home literacy environment that includes shared reading) may play a significant role in relation to critical reading.” PsyArXiv Preprints

Shared reading is a great source of rarer or more “academic” words. Preschoolers who use more rare vocabulary words have higher vocabulary scores on norm-referenced vocabulary measures. ASHA American Journal of Speech Pathology

Brains, Bodies, and Language

But what about “everyday language”? How is that developed? Across languages, verbs are acquired in the following order: 1) vision, 2) touch, then 3) hearing. Vision verbs (see, look) are acquired earliest and produced most frequently by children of all ages. Taste and smell verbs were produced less frequently than other perception verbs across the board. Cognitive Science

Speaking of verbs and language related to physical experience: linking language with physical or imagined movement can make it easier for children to grasp what they hear. In other words, children can be taught to improve their listening comprehension skills, as this study shows. Four and five years olds were provided with a listening comprehension intervention that taught them “to align visual and motor processing with language comprehension.”

• As part of the study, they looked at the children’s brain activity using EEG and discovered that the children who improved in listening comprehension also showed changes in the parts of the brain related to movement and visio. This means the brain's motor and visual areas become involved when children are actively working to understand language. The training helped the children to use their visual cortex to imagine what the story was describing, and their motor cortex to imagine the actions suggested by the story. Behavioral Sciences

We’ve looked at some of the research on the surprising–and fascinating–separation of language and cognition in the human brain here on this blog before in Language and Cognition and Thinking Inside and Outside of Language. But clearly, there is a link to some degree between cognition and language.

In a study of people with aphasia (difficulty with language after a brain injury), they found that executive function was related to language ability, with verbal executive function and fluency more strongly linked to micro-linguistic narrative language such as grammar and word choice, while nonverbal executive function plays a more prominent role in macro-level discourse skills like coherence and organization. ASHA American Journal of Speech-Language Pathology

When children with developmental language disorder (DLD) received both cognitive and linguistic training, they improved their verbal short-term memory and verbal working memory. They also demonstrated far transfer effects of the training (far-transfer refers to the impact of an intervention on abilities that were not directly targeted by the training).

• Most interestingly, the order of interventions affected the results, suggesting that a combined linguistic and cognitive & tailored therapy may be most beneficial. Brain Sciences

“The findings of the current study indicate that the coexistence of ADHD in children with DLD does not exacerbate language and reading difficulties.” CPP Advances

Another study aimed to determine the extent to which oral language development is related to reading speed and accuracy in Spanish-speaking children with DLD. The children with DLD were indeed less accurate and slower in reading than “typically developing” (TD) children. The findings also show that the use of strategies during reading are different between the DLD and TD groups.

• “the network analyses suggest strong and stable connections between reading and oral production in the DLD group. This finding confirms the importance of language abilities for reading acquisition.” Reading Psychology

Speaking of the relationships between oral language and reading: oral language skills are both promotive and protective factors for children with lower reading fluency skills in grade 1.

• “The findings of our study further extend those of previous research, suggesting that while OL skills are important for the reading comprehension skills of all children, individuals with lower reading comprehension skills in G1 benefit the most from strong OL skills.” PsyArXiv Preprints

Poverty impacts a child’s developing brain – and this longitudinal study demonstrates this has a long-term impact on language ability. The findings indicate that the chronic stress of poverty alters the trajectory of neural pathways associated with language in adults. Even when adults from backgrounds of poverty had average language skills, their brains show differences in activation and connectivity patterns compared to adults from middle-income backgrounds. These differences suggest the use of compensatory mechanisms.

• “Interestingly income alone did not account for any significant differences in language functioning but educational attainment did. This suggests that language is an important driver in the choice to continue education after growing up in poverty” “

• “Greater activation in the poverty group may be indicative of inner speech during word recognition and phonemic decoding of pseudowords, which is a potential compensatory adaptive mechanism.”

• This inner speech may be used potentially due to less automatic processing of language. Brain and Language

There’s something interesting about inner speech as a compensatory adaptive mechanism, by the way. Not everyone has an “inner voice” or experiences inner speech in the same way – there is quite a bit of variation. In a study, those with less inner speech have poorer performance on a verbal working memory task and lower accuracy in rhyme judgment tasks. Yet when study participants reported talking out loud, the performance differences between groups disappeared! This suggests that both covert (inner) and overt speech can be used as compensatory mechanisms to support cognitive performance.

• “Understanding how inner speech develops has implications for education.” Scientific American; Psychological Science](https://doi.org/10.1177/09567976241243004)

Immigration, Multilinguals, and Multilingualism

Now let’s tackle a hot button topic: immigration.

Like so much of our national and political discourse, the topic of immigration is so heightened by emotion that facts and evidence are far removed from policy and perception.

Unfortunately, one source notes that “the contemporary opposition to immigration, and the tendency for it to be stronger among less educated people, are not a reflection of something specific to today, but continue a long-standing pattern.” Statistical Modeling, Causal Inference, and Social Science

If you really want to cut through the noise, I highly recommend reading a book released this year by Zeke Hernandez, The Truth About Immigration, to ground your understanding of immigrants and immigration in empirical evidence, rather than bias and sensationalism.

I first came across Zeke’s trenchant insights when I listened to a Freakonomics series on immigration (also recommended), “The True Story of America’s Supremely Messed-Up Immigration System.” I decided to check out his book, and am very glad I did. Whatever your priors on immigration may be, you will find something to learn that will surprise you, and educate you, in his book.

Now let’s turn to some more facts and evidence about immigration.

Immigrant children can benefit the learning of others

Newly arrived immigrant children who are English learners have “positive spillover effects” on the test scores of existing students, particularly in reading – even in a “new destination state” such as Delaware, which has seen a sevenfold increase in its EL student population over the past two decades. Educational Evaluation and Policy Analysis

• This builds off of previous research I highlighted in last year’s roundup, which found “significant benefits of having immigrant peers on the test scores of native students, especially among students from disadvantaged backgrounds.” Brookings

Immigration boosts the economy

• “Latin American immigrants are starting businesses at more than twice the rate of the U.S. population as a whole.” Marginal Revolution

• “New migrants contribute to economic growth in two ways: by working and by spending.” New Yorker

• ”...from a strictly budgetary point of view, the new arrivals are more than paying for themselves.” Bloomberg

• “Alabama wound up watering down its 2011 restrictions in part because of an outcry from businesses about the loss of workers. Crops rotted in the field. Investment in the state stalled.” NY Times

• Undocumented immigrants pay nearly $100 billion in taxes. Bloomberg

• When restrictions and deportations of undocumented immigrants are enforced this leads to a reduction in construction labor supply, decreased homebuilding, and ultimately, increased housing prices. SSRN

• “By adding millions of new workers to the labor market, the immigration surge has lifted payrolls and growth, and potentially helped keep a lid on consumer prices, according to recent research.” Semafor

While we’re at it, we should note that immigration does not increase crime levels in the communities where immigrants settle. And obtaining legal status decreases immigrants' involvement in criminal activities. Journal of Economic Perspectives

• “As a group, immigrants have had lower incarceration rates than the US-born for 150 years. Moreover, relative to the US-born, immigrants' incarceration rates have declined since 1960: immigrants today are 60% less likely to be incarcerated.” Stanford Law and Economics

Cultural and linguistic distances can impact immigrant mental health and learning

Immigrants tend to move to places where climates better match what they are accustomed to.

• “we show that climate strongly predicts the spatial distribution of immigrants in the US, both historically (1880) and more recently (2015), whereby movers select destinations with climates similar to their place of origin.” NBER

In Ontario, the greater the linguistic distance between an immigrant’s first language and English, the more elevated their risk of being diagnosed with a psychotic disorder. Journal of Psychological Medicine

Relatedly, cultural factors can influence how symptoms of psychosis are experienced and expressed.

• “findings seem to indicate that there is not a “one size fits all” approach to quantifying schizophrenia symptoms in multilinguals, but rather a complex interplay of medical and social factors that contribute to symptom expression.” Bilingualism: Language and Cognition

A common assumption made about more recent immigrants is that “acculturation”—becoming oriented towards mainstream culture–necessarily leads to a decline in heritage language skills. Yet this study found that mothers who maintain a balance of enculturation–or orientation towards their heritage culture–and acculturation in the United States also maintained greater bilingualism in their children.

• “Both mothers’ levels of enculturation and acculturation were significant predictors of the grammaticality of the Spanish utterances produced by the children between the ages of 3 and 4.” Journal of Child Language

Moving between cultural frames more frequently, in fact, may support executive functioning.

• “Bicultural switching effects on interference and inhibition-control persist even in participants at the developmental peak of their cognitive processing capabilities after controlling for a plethora of socio-linguistic variables.” International Journal of Bilingualism

“According to research that confirms past studies, the concern that immigrants and their children do not learn English is misplaced.” Forbes

Children with more diverse social networks also develop more flexible and nuanced speech categorization patterns, adapting to the variability of their linguistic environments. Importantly, whether their adaptive speech processing is perceived as a deficit or an asset depends on how it is measured and analyzed. PsyArXiv preprints

Yet “despite higher exposure to one language, children sometimes identified more with the language and culture they were exposed to less.”

• In fact, this study found that higher exposure to a language does not always align with higher-level skills in that language. High-level skills can also be observed in the language where exposure was quantitatively lower, but qualitatively rich. For example, engaging in activities like reading could provide qualitatively rich exposure and compensate for lower quantitative exposure. PsyArXix

But let’s go back to that concept of “linguistic distance.” Globally, the greater the “discordance” between the language of home and the language of school, the lower the basic literacy rates.

• “If we look at literature from the fields of literacy development and bilingual development, even for monolingual speakers, it is much easier for a child to learn to read and write if they can do that with a script that maps out to their oral language. This is because we start learning language way before we enter school, whereas if a child goes to school and they are confronted with reading a script that does not map out to their language, it is harder for them. If the teacher does not speak their language and does not explain [things] in a way that they understand, it’s harder for them.” Harvard GSE News

Providing an early oral language intervention in students’ home language when that language is more discordant with school language can improve learning.

“The findings indicate that school-based oral language interventions can enhance heritage language proficiency and facilitate skill transfer to specific domains of a second language.” EdArXiv Preprints

For low SES immigrant families in Paris, a shared book reading intervention significantly enhanced children's language skills and the effects persisted in a six month follow-up. For $5 dollars a kid, not a bad deal. Journal of Research on Educational Effectiveness

A note that we’ve examined the concept of “linguistic distance” on this blog previously, suggesting that when there is a greater distance between the forms of a language that are spoken at home and written in school, this may make it more challenging and complex for young learners to acquire literacy. This applies also to spoken dialects of a written language, such as African American or Black English, Cantonese, or Moroccan Arabic.

The Benefits of Multilingualism

A study in the UK found that although multilingual learners initially face challenges in Key Stage 2, particularly in English and Science, they achieve comparable results with–and often excel over–their monolingual peers by Key Stage 4.

• “Notably, this academic advantage was observed even among students from low socioeconomic backgrounds, suggesting that multilingualism can offset the negative effects of socioeconomic disadvantage and contribute to greater educational equity and social mobility.” International Journal of Bilingual Education and Bilingualism

A longitudinal study in Chicago Public Schools demonstrates the importance in disaggregation of English learner data, as there are ELLs who go on to outpace their monolingual peers. For students who have achieved English language proficiency, “They had higher-than-district-average outcomes: cumulative GPAs and SAT scores; high school graduation rate; two-year college enrollment rate; and two-year college persistence rate (among all college enrollees).” University of Chicago Consortium on School Research

• This corresponds to similar data on former ELLs from NYC Public Schools. The Research Alliance for New York City Schools

Learning a new language may even make you better at learning math! Adolescents who received formal instruction in a foreign language were about three times more likely to achieve higher grades in math tests than those who did not. (Note that this does not establish causation.) Bilingualism: Language and Cognition

The conversation about bilingual education programs often focuses on the benefits for students who are learning English. Yet it’s good for English proficient students, too!

• “On average, native English-speaking students in Grades 1 through 4 who win access to a DLI program score higher in reading and math by 0.12 and 0.14 SDs, respectively. The achievement gains in test scores are realized as early as first grade.” Educational Evaluation and Policy Analysis

Bilingual education isn’t only about spoken languages! In a study of an ASL bilingual program, kids at risk of language deprivation (due to having caregivers who don’t know sign language) who entered the program young achieved the same academic performance as kids who were not at risk (due to having caregivers who use sign language).

• In other words, a bilingual program can act as an early intervention to mitigate the effects of potential language deprivation on academic development! The Journal of Special Education

Yet despite the potential benefits of multilingualism and of bilingual education programs, the United States remains far beyond the rest of the world.

• According to the U.S. Census Bureau, “about 20% of the U.S. population speaks another language other than English, compared to 59% of Europeans who can speak at least a second language”. National Geographic

• The European Union states in its language policy that every European citizen should master two or more other languages in addition to their mother tongue. EU Language Policy

• Share of kids not learning a foreign language in school US 80% Germany 18% Italy 18% Finland 16% Sweden 8% Spain 4% Poland 2% France 0% Norway 0% Pew Research Center

Cognition and Multilingualism

Working Memory

When solving word problems in math, multilingual learners with a home language of Spanish draw on their working memory systems, which operate across both languages.

• “Results show increased accuracy of targets and generalisation of sounds across languages when treatment was administered only in the L1.” Journal of Educational Psychology

Importantly, the structure of working memory was found to be similar in both monolingual and bilingual children. This now allows for more valid comparisons, generalizable interventions, and can strengthen our theoretical understanding of working memory in both populations. Bilingualism: Language and Cognition

In one study, they taught bilingual children (Spanish-English) who were 4 and 5 years old new words paired to objects. In one condition, they taught the label with only English-like words, and in the other, they taught them both Spanish- and English-like words for different objects. They found that the bilingual children learned the words best in the single language condition, suggesting that competition between languages might be a factor affecting learning. Journal of Experimental Child Psychology

This fascinating study finds that better performance of older bilinguals in L2 than L1 on paired associate learning tasks “cannot be accounted for by cognitive decline, but follows straightforwardly from basic principles of learning.” Dimensions of Diffusion and Diversity

Cognitive Flexibility and Task Switching

Learning a second language in adulthood can strengthen neural connections.

• “‘The dynamic changes in brain connectivity were found to be directly correlated with the increase in performance in the language test of the Goethe-Institute,’” emphasized Alfred Anwander, the study’s last author.” Max Planck Institute

That said, there have been conflicting findings about whether learning multiple languages enhances executive function or not. This research article compares studies of the “bilingual advantage” with cognitive training studies and finds them both to be null. The authors argue that if cognitive training does not result in far transfer, then it is unlikely that bilingualism would, unless there was a special status for bilingual language control. International Journal of Bilingualism

Meanwhile, another study replicated a previous finding that bilingualism enhances cognitive flexibility in task switching, specifically by reducing the global switch cost.

• “Overall, findings contributed to the argument that bilingualism does indeed confer a bilingual advantage in task switching, as observed in young adult bilinguals with diverse language experiences.” Studies in Second Language Acquisition

Yet this study cautions that bilingual advantages in cognitive flexibility are not straightforward and can be influenced by both language-related factors and psychological stress.

• “Our findings suggest that advantages in cognitive flexibility are conditional, shedding light on the ongoing debate about the ambiguous relationship between experience and cognitive control in bilinguals.” International Journal of Bilingualism

It may be that intentional code switching may be associated with greater cognitive flexibility, while unintentional switching may be negatively associated with cognitive flexibility.

• “Altogether, our findings indicate that any training instilled by dual-language code-switching is restricted to language-specific cognitive flexibility.” Journal of Cognitive Psychology

Or, it may be that switching between languages while reading can be more or less cognitively costly depending on whether the words are more concrete (with lots of interconnections conceptually between the languages) or abstract (with fewer connections between languages).

• “We found that abstract words (e.g., 正确 [correct], wrong) did not show switching costs. . . In contrast, concrete words (e.g., 晴天 [sunny], rainy) elicited significant larger switching costs.”

• “in our experiment, the absence of nontarget language activation obviated the need for language control, resulting in no significant switching cost for abstract words, while concrete words incurred larger switching costs because of the high activation level of the nontarget languages.” Journal of Experimental Psychology: Learning, Memory, and Cognition

Neural Connections and Brain Structure

The conflicting accounts of the impact of multilingualism on the brain may be due to the fact that positive effects are more localized.

• “Our analysis … suggests that one should not expect to observe a uniform impact of bilingualism across the entire lifespan – there are time-varying effects that emerge, showing that remodeling of white matter is most clearly observed closer to the learning event.”

• This study did find that specific white matter tracts associated with language processing showed reliable differences between bilinguals and monolinguals, most particularly in adults.

• “converting an effect size for the effect of age on white matter (FA) into an equivalent for these regions from our meta-analysis, allows us to speculate that the effect of bilingualism is equivalent to having white matter that is between 2.31 and 4.65 years younger than expected, a value that neatly aligns with current estimates of bilingualism’s impact on delaying the onset of dementia.” Neuropsychologia

In another study, they found that bilingual children, unlike bilingual adults, show lower FA values in language-related white matter pathways compared to monolingual children, suggesting a slower maturation of these pathways during childhood. Human Brain Mapping

While there may not necessarily be direct cognitive advantages to multilingualism, evidence does show that learning a new language imposes a cognitive burden. I wrote about this research more in depth in my post, Research Highlight 5: Learning In a New Language Takes Effort.

Semantic Representation and Conceptual Change

Learning a new language may also change concepts in your first language. Psychological Science

Another study found that semantic brain representations are largely shared across languages but modulated by each language. These results show that between the two languages, semantic representations are not fully the same, but they’re also not separate: there is a shared semantic system that is modulated by each language! bioRxiv preprint

Multilingual Phonology and Orthography

Phonological Awareness and Speech Perception

As we noted previously, the quality, rather than mere quantity, of linguistic input is what is important. This applies equally when learning a new language. One study suggests that when teaching reading in an L2, focusing on developing clear and specific phonological representations is essential.

• “Not the sheer number of words, but their phonological representations (lexical specificity) in the mental lexicon seem to matter most in the early stages of L2 reading comprehension.” International Journal of Bilingual Education and Bilingualism

• A note that we’ve discussed the concepts of fuzziness and precision in multilingual learner previously in An Ontogenesis Model of Word Learning in a Second Language.

That said, phonological awareness as a skill seems to be more of a language-general construct, rather than only a language-specific one.

• “These findings provide evidence that phonological awareness is a language-general skill that supports reading across languages, consistent with the common underlying proficiency model of bilingual reading development.” Journal of Experimental Child Psychology

• “These findings reveal that the neural basis of PA is both shared, as evidenced by the activation of a common left perisylvian network, and language-specific, with greater modulation in the temporal regions for Spanish and in frontal regions for English.” Mind, Brain, and Education

• “The portions of the brain that control the muscles needed to make the noises we associate with language aren't especially picky about which language they're handling.” Ars Technica; Nature Biomedical Engineering

So it’s not surprising then that treating bilingual children with speech-sound disorders in their home language of Spanish facilitates progress of similar sounds in English. Clinical Linguistics & Phonetics

Though it also may be that bilingual children develop two distinct phonological systems that interact with each other, and the specific patterns of acquisition in each language are influenced by the frequency of phonological features in the input. International Journal of Bilingualism

“our findings support the idea that phonological transfer might be possible even between languages with very different phonological structures.” Reading and Writing

Sound Discrimination and Learning

Yet how we discriminate sounds between languages can be based on how we learn them.

• This study looked at how people who speak three languages (trilinguals) can tell the difference between sounds in their different languages. They found that people were better at recognizing sounds in their first language compared to their second or third languages. And unsurprisingly, the study found that the more someone knows a language, the better they are at recognizing sounds in that language.

• Those who learned languages through social immersion (like living in a country where that language is spoken) showed better sound discrimination than those in formal classrooms. Naturalistic learners processed L1 and L2 sounds similarly, unlike formal learners who showed clear differences across all three languages. Bilingualism: Language and Cognition

It’s possible that the multilingual brain processes word similarities from a new language to their first language at different speeds. Journal of Experimental Psychology: Learning, Memory, and Cognition

Speaking of learning something new: articulating a new word out loud for children facilitates learning of that word more than if you just passively receive it.

• When students are learning a new language, saying new words out loud is even more important! The researchers suspect that this is because it requires more mental effort. Memory & Cognition

Word Learning and Spelling

Similarly, word learning in a new language is further facilitated (just as it is in your first language) by pairing the sounds to the words in print.

• “In both experiments, orthographic facilitation was found in both less and more advanced readers. . . Our results can be explained by the strong interplay between orthographic and phonological processing: phonological representations are quickly and automatically activated upon the presentation of a written word. Just as with L1, L2 word learning is facilitated by pairing sounds to words in print.” Journal of Experimental Child Psychology

• We conclude that both English monolingual and bilingual children learn more novel words when the spellings of words are present, and that this benefit does not appear to be larger for bilingual children.” Reading and Writing

In terms of spelling, one study found that cross-linguistic influence of spelling errors was mostly unidirectional. Children typically made errors in one language due to influence from the other but did not make similar errors in both languages.

• “even if dual language learners did have balanced oral language skills, they may develop the spelling patterns of the two languages at different rates.”

• This is significant because it shows that spelling development is not simply a reflection of oral proficiency. It is also influenced by factors like: the characteristics of each language’s writing system, the type of instruction received, and a learner’s stage of development. Reading and Writing

Multilingual Learning and Instruction

Building on Home Languages

Translanguaging has become a ubiquitous term in the field. Yet it’s not always clear exactly what the term means in practice, nor in terms of its evidence base.

• “Translanguaging, which has taken on an air of orthodoxy in applied linguistics and language education, may now be immutably associated with deconstructivism, making a return to its earlier meaning difficult to achieve with adequate clarity.” International Journal of Bilingualism

• “the notion of translanguaging has been very successfully marketed . . . there are no diagnostic criteria against which researchers can check multilingual practices and decide whether or not these count as translanguaging.” Linguistic Approaches to Bilingualism

Yet what we do know–as research in other sections has already pointed out–is that supporting an English learner’s skills and knowledge in their home language supports their language and literacy development in English.

• “The findings further suggest that supporting heritage-language literacy may further strengthen emerging bilinguals’ literacy development across their languages.”

• In this study of Spanish-English and Chinese-English bilinguals, they found direct longitudinal transfer of phonological awareness skills from the heritage language (Spanish or Chinese) to English for both groups of bilinguals – which again suggests, as we examined previously, that phonological awareness is a language-general skill that can be readily transferred between languages.

• On the other hand, morphological awareness appeared more language-specific than phonological awareness. Morphological awareness transfer is more complex and depends on the structural similarities between the languages involved.

• “literacy instruction that includes systematic phonological, morphological and orthographic training is critical for bilingual and monolingual speakers.” Bilingualism: Language and Cognition

A study shows that for Korean-speaking adolescents, morphological awareness in Korean boosts reading comprehension in both Korean and English. Journal of Experimental Child Psychology

“Notably, oral language and reading skills in both MLs’ first language and in English were essential components of the SOR for MLs.” Educational Psychology Review

English Learner Reading Profiles

For students who are learning English in English only environments, the task of learning then becomes more challenging. The Simple View of Reading was used in one study to distinguish English learner reading profiles with a home language of Spanish from English proficient reading profiles.

• Unsurprisingly, proficient English speakers were more likely to be in the typically developing and poor decoder/good Listening Comprehension (LC) profiles, while Spanish-speaking ELs were more likely to be in the good decoder/poor LC and poor decoder/poor LC profiles. Unsurprising, because regardless of whether an EL is good at decoding or not, they are by definition learning English.

• So if they need more decoding support or intervention, they will need BOTH decoding and comprehension support at the same time. Reading and Writing

• I went far more in-depth into the reading profiles of English learners in my post, Research Highlight 3: The Reading Profiles of English Learners.

Linguistic Proficiency and Reading Intervention

Speaking of intervention, a critically important study of 6th and 7th grade multilingual learners with reading difficulties found that providing intensive intervention in English reading was only effective when students had “relatively strong English proficiency.”

• This is important because there is a tendency in the field right now to put newly arrived immigrant students into reading intervention, rather than ensuring that they are receiving comprehensive language-rich instruction through all their Tier 1 content areas.

• “These findings highlight once again the importance of linguistic proficiency to students' reading achievement and suggest that without linguistic proficiency even an intensive and extensive intervention may not meet students' reading needs. . . We interpret this suggestion as a rationale for more intensive language and literacy supports beyond the context of a tier 2 intervention and into tier 1 content area classes.” Learning and Individual Differences

Conversations and Incidental Learning

For early childhood programs, “the findings suggest the importance of improving opportunities and providing more support for emergent bilinguals to engage in conversational turn-taking with their teachers and peers.” Early Childhood Education Journal

One review of corpora, both student talk and lessons, in English classes at a university in Vietnam found that student talk is an excellent source for the incidental learning of high-frequency word families and a good source for learning core formulaic sequences, as well as provides opportunities for both spaced repetition and varied repetition, which are crucial for vocabulary learning. They found that knowledge of the most frequent 1000-word families is needed for reasonable comprehension of student talk. The Language Learning Journa

“. . . overall, interaction is a key source of L2 receptive vocabulary development.” International Review of Applied Linguistics in Language Teaching

Balancing Explicit and Implicit Learning

A study of Japanese students learning English highlights the need for pedagogy to assist second language learners in achieving both declarative (explicit, conscious understanding) and automatized phonological vocabulary knowledge.

• They found that declarative knowledge of phonological vocabulary is linked to more formal classroom-based training and working memory, while automatized knowledge is more strongly associated with extracurricular activities that expose learners to auditory materials and provide more real-world language experiences (such as study abroad).

• “For effective L2 learning, it is imperative that teachers not only emphasize explicit word comprehension but also provide abundant practice to foster knowledge automatization.” Bilingualism: Language and Cognition

• I’ve explored the importance of automatization in language learning in the post, Research Highlight 1: The Importance of Automatization in Learning a New Language.

Finding the right balance between explicit and implicit learning requires that we more precisely identify the highest leverage items that must be taught explicitly. For Spanish speakers in third grade, explicitly teaching novel suffixes was far more effective than mere exposure.

• “At both testing points (i.e., immediate and delayed post-test), explicit instruction yielded better results for the learning of the form of the suffixes compared to implicit instruction.” Journal of Experimental Child Psychology

In a study with university students learning a new language, they found a reciprocal relationship between explicit and implicit knowledge.

• “The strongest predictor of current explicit knowledge was prior explicit knowledge; the strongest predictor of current implicit knowledge was prior implicit knowledge.”

• “The results from an autoregressive cross-lag analysis suggest L2 explicit and implicit knowledge influenced each other reciprocally over time. Neither activity type predicted knowledge development. We conclude that language acquisition is a developmental process typified by a dynamic, synergistic interface between explicit and implicit knowledge.” Bilingualism: Language and Cognition

One method to support incidental vocabulary learning is through the addition of captions to videos. This benefits “intermediate-level” learners the most, suggesting that additional scaffolds would be needed for lower proficiency learners.

• “The results showed a medium effect of captioning on L2 vocabulary learning.” Language Learning

Speaking of implicit learning: you’re never too old to implicitly learn a new language!

• “Given that implicit language learning mechanisms are shown to be preserved over the lifespan, the present data provide crucial support for the assumptions underlying claims that language learning interventions in older age could be leveraged as a targeted intervention to help build or maintain resilience to age-related cognitive decline.” Bilingualism: Language and Cognition

Though it might help your learning of the new language if you deplete your cognitive resources first!

• “late-developing cognitive control abilities, and in particular attentional control, constitute an important antagonist of implicit learning behavior relevant for language acquisition.” Journal of Experimental Psychology-General

All of that said, a reminder that explicit instruction is a powerful means to direct learning and can act as a shortcut to achieving the same neural representation that would have been formed through implicit learning. Nature

And learning a new language is also aided by . . . sleep.

• “By demonstrating how specific neural processes during sleep support memory consolidation, we provide a new perspective on how sleep disruption impacts language learning...Sleep is not just restful; it’s an active, transformative state for the brain.” SciTechDaily

A note that I’ve discussed the balance between explicit and implicit learning more in-depth in relation to AI in my post, LLMs, Statistical Learning, and Explicit Teaching.

Assessing and Diagnosing Language Skills with Multilingual Learners

Gathering and analyzing the language samples of children can be a really useful way to learn more about their language use.

They can help you to better understand dialectal differences.

• “the findings from this study underscore the potential use of language sample analysis in describing linguistic patterns to support the characterisation of communication profiles for culturally and linguistically diverse children.” Clinical Linguistics & Phonetics

And they can help you to better distinguish between developmental language disorder and typical language development in multilingual learners.

• “Results of this study provide evidence of the clinical utility of LSA in differentiating between DLD and TL in bilingual children.” ASHA Journal of Speech, Language, and Hearing Research

• For Spanish-English bilingual children, mean length of utterance in words (MLUw) and percentage of grammatical utterances (PGU) seem to have the greatest diagnostic accuracy. ASHA Language, Speech, and Hearing Services in Schools

Gaining greater diagnostic accuracy with multilingual learners is important, because how they perform on a vocabulary and listening comprehension test may be due more to the specific test items, rather than differences between the children themselves!

• “These results indicate a need for careful and deep investigation into assessment and item factors that influence item response accuracies in oral language tasks.” ASHA Journal of Speech, Language, and Hearing Research

Multilingual learners in preschool who are identified with DLD may be less likely to be dominant in their home language in comparison to MLs without DLD.

• “all bilinguals with better selective attention more often had balanced vocabularies in both languages, while those with compromised selective attention coupled with poorer L1 speech tended toward L2 dominance.” Research in Developmental Disabilities

Rhythm, Attention, and Memory

In this section, we’ll continue to examine some research related to multilingualism, but there was an interesting few additional themes and other studies that came up around music, synchrony, and the role of attention and memory in learning.

We Learn Through Rhythm

The Synchrony of Learning

There are patterns of different oscillations and rhythms across the layers of the brain.

• “we suspect that different pathologies of synchrony may contribute to many brain disorders, including disorders of perception, attention, memory, and motor control.” Science Daily

Interbrain synchrony is linked with better learning.

• “The better their brain waves synchronized, the better they performed these tasks as a group.” Quanta Magazine

• “In all, the similar neural representations and interbrain synchronization between co-learners suggest that co-learning companionship offers important benefits for learning words in a new language.” Cerebral Cortex

• “in multilingual contexts, the activation of synchronization processes involving both linguistic and non-linguistic mechanisms...is necessary to enable effective linguistic communication, comprehension and translation. . . . [Furthermore] some studies have indicated heightened activation in the motor cortex during L2 processing compared to L1.” Imminent

Music

That heightened activation in the motor context suggests that gesture, movement, and music can support the learning of languages.

• “The infants who were randomly assigned to complete the music intervention showed enhanced brain responses that reflected detection of small differences in not only musical sounds, but also speech sounds.” Science

• “The available evidence suggests that musical ability is indeed positively related to second-language learning, even after factoring in publication bias revealed by the meta-analysis.” PsyArXiv Preprints

• “The musicians performed better than the non-musicians on Cantonese phonological awareness, Cantonese tone awareness, and English phonological awareness.” Journal of Experimental Child Psychology

And yet, music may not be “derivative of speech—it serves its own purpose.” Scientific American

Playing music may help keep your brain young. PLOS One

Movement and Rhythm

When it comes to rhythm, there’s a goldilocks equation: moderate syncopation makes people want to dance, while too much or too little does not. Scientific American; Science Advances

If you’ve ever thought there is a rhythm to writing, this study on how children learn to write backs you up – and shows that there is even “an internal representation of the rhythm of handwriting [that] is available before the age in which handwriting is performed automatically.” Nature

And when it comes to movement, the cerebellum–once thought to only control body movement–connects to so much more!

• “These new, groundbreaking studies show that in addition to controlling movement, the cerebellum regulates complex social and emotional behavior.” Wired

Attention and Memory

“Our work suggests that sustained attention acts like a gatekeeper, controlling what “gets in” to children’s long-term memory—and the gate to memory remains shut more often in children. These novel findings raise the possibility that differences in sustained attention may explain broad differences in cognitive performance and that to boost children’s learning we must first help them to effectively sustain attention.” Well, yeah. That's the hard part. Psychological Science

We certainly don’t help children focus with all the clutter we put on our walls in classrooms. Classroom decorations can overwhelm students’ working memory and attention. Learning and the Brain

The good news is that purely visual distractions are easy to get rid of, and researchers have found that children's working memory is not significantly more affected by multisensory distractions (visual and auditory) than by purely visual distractions.

• “children’s working memory – which is fundamental to learning – is more robust to interference than we might think.” Bold

Spacing and Interleaving Learning

One of the most robust findings in the body of science of learning is that of the “testing effect” on learning.

There were a number of studies this year further examining retrieval, spacing, and interleaving practice.

Students most typically try to cram all their studying for tests the night before. This is termed “massed practice.” While it might be fine for one-off learning, cramming won’t get you far in medical school, where you need to be able to retain and build upon that learning – and ultimately, be able to apply it in medical practice. This more distant application to novel experiences is termed “far transfer.” But is “blocking” the practice, or “interleaving” the practice more effective for far transfer?

• “giving students practice with multiple contexts seems to be particularly important for far transfer, and when that happens, interleaving the examples is better than blocking.” The Learning Scientists

Retrieval practice (i.e. flashcards) isn’t so bad with easy stuff. But when it gets more difficult, students tend to avoid it. This study shows that if you explain the benefits of retrieval practice for both easy and difficult items in the long run, students are more likely to do retrieval practice even with difficult items. Educational Psychology Review

“both spacing and variability can benefit memory, depending on what aspect of an experience you are trying to remember.” Scientific American

There is great potential for spaced retrieval to support vocabulary development for students with DLD, but there is still quite a bit to figure out to make it most effective.

• Spaced retrieval can help to prevent the erosion of phonetic details in word recall, which is particularly beneficial for children with DLD—who may otherwise experience a decline in phonetic accuracy over time.

• Spaced retrieval is most effective when it integrates immediate retrieval, provides consistent spacing, and includes feedback, helping to enhance long-term word recall and preserve phonetic details in children with DLD.

• Future research should clarify the optimal spacing between retrieval attempts and whether gradually increasing this spacing is necessary for long-term retention. Autism & Developmental Language Impairments

Individual differences play a role with testing effects. It all has to do with how much working memory is available – some of us have more WM than others.

• This paper theorizes that when we are tested on something, working memory is needed both in the attempt to retrieve the information and then to re-encode and further solidify it.

• Individuals with lower WM may find that after retrieving the information, they don’t have enough WM left for re-encoding.

• The model suggests that testing should be challenging enough to engage working memory, but not so difficult that it overwhelms it, which relates to the concept of “desirable difficulty.”

• Providing feedback after a retrieval attempt may help to reduce the working memory load, allowing those with lower WM to benefit more from testing. NPJ Science of Learning

In a study with mice, they found that rest periods after learning helps to integrate new memories with older ones. Nature

Researchers examined how mathematical procedural complexity interacts with spacing retrieval practice.

• The study found no evidence that the spacing effect is less effective for more complex material (when complexity is defined as the number of steps in a procedure).

• “The spacing effect is robust to variations in procedural complexity and supports its use in the teaching and learning of mathematics.” PsyArXiv Preprints

Testing can even be beneficial before you’ve learned something! This is called “pretesting.”

• “keep in mind that it works best when the questions are focused on information that will be covered in what you’re about to learn.”

• “take the pre-quiz shortly before engaging with the learning material. . . you can ‘turn learning objectives into questions and attempt to answer them before exploring the content.’”

• “including incorrect but closely related answer options in a multiple-choice test format can help direct your attention.”

• There was this nugget in the article that could help reframe the direct instruction vs. inquiry-based learning debate: “Another guessing-based strategy that has proven effective, often in group learning, is known as ‘productive failure’. In subjects like mathematics, it involves encouraging learners to attempt solving problems before receiving formal instruction – and again there’s evidence that this form of guessing can result in better outcomes than instruction alone.”

• In other words, inquiry-based math learning could be effective, when structured well, in the sense of this pre-testing effect – rather than being viewed as about “discovery.” Psyche

School, Social-Emotional, and Contextual Effects

School Effects

OK, I know these books by Karin Chenoweth weren’t published in 2024, but I happened to finally come around to reading them in 2024, and I highly recommend them, as well as the podcast: Schools That Succeed, Districts That Succeed.

Why do I recommend these? Because Chenoweth reminds us that schools can serve the most vulnerable students and communities and make a tremendous impact as evidenced by the hard data – and that the means to do so are not mystical: A culture of high expectations and belief in kids, transparent data-based inquiry, committed and sustained leadership, and coherent school organization and scheduling.

Illustrative quotes:

• “Nowhere are a school's values and priorities more on display than in a school's master schedule,.”

• “Schools that go...from serving mostly white middle-class students to serving mostly low-income students or new immigrants are often revealed as institutions that are not in and of themselves 'good schools'.”

But do school reforms have long-term effects?

• “We find little evidence to support improved long-run student outcomes – mostly null effects that are nearly zero in magnitude. Our results contribute to a broad call for educational researchers to examine whether school reforms meaningfully affect student outcomes beyond short-term improvements in test scores.” EdWorkingPapers

Well, getting a college degree still matters.

• Almost 70 percent of overdoses in the United States occur in people without a college degree. JAMA Health Forum

And early childhood programs have multifaceted positive effects, despite the critiques around “fade-out” effects.

• In fact, the fade-out effect is the very reason to continue to invest in early childhood programs, according to one study. That’s because the effect is linked to the share of classmates who also attended preschool, and increasing the number of children attending preschool would help reduce this fade-out effect by creating a stronger social network and support system.

• “human capital accumulation is inherently a social activity, leading early education programs to deliver their largest benefits at scale when everyone receives such programs.” NBER

• Another study suggests that the main benefit of early childhood programs is actually for parents.

• “UPK enrollment increases parent earnings by 21.7% during pre-kindergarten, and gains persist for at least six years after pre-kindergarten. Gains are largest for middle-income families.” NBER

• “Consistent with an increase in overall economic activity, places that introduced Universal Pre-K also had larger increases in new business applications and the number of establishments than places that did not” Whitehouse Issue Brief

How we measure teacher effects is important. For a long time, we have been focused on test-based effects. But according to this study, test-based measures are more aligned with high-achieving students and outcome-based measures like SAT scores and AP test performance, while non-test measures better predict outcomes related to college enrollment and high school graduation, and may be especially important for students who are at risk of not enrolling in college or not graduating from high school.

• “the results of this study suggest that it is nontest teacher quality that is especially relevant for disadvantaged students and that gaps in access to effective teachers along the nontest dimension would be even greater cause for concern.” Journal of Human Resources

If we want to decrease achievement gaps, we need to focus less on “homework help” or enrichment programs, and more on classroom management, challenging content with a high degree of support, heterogenous grouping, and tutoring. Studies in Educational Evaluation

Social-Emotional Effects

Social-emotional neglect has serious consequences for child development.

• “Over the course of 20 years, we have consistently demonstrated that even when a child’s physical needs are met, psychosocial neglect is deleterious to brain and behavioral development.” Current Directions in Psychological Science

“Being bullied as a child worsens well-being and labour market performance up to half a century later. It lowers the probability of having a job throughout adulthood and raises the probability of premature death.” Social Science & Medicine

For students with ADHD in Switzerland, targeting social-emotional skills through the Promoting Alternative Thinking Strategies (PATHS) program had persistent positive effects lasting over a decade. Treated children were more likely to complete academic high school and enroll in university. The Review of Economic Studies

Yet “ boosting social-emotional skills, like boosting cognitive skills, does not appear to be a silver-bullet solution to changing children's developmental trajectories.”

• “While it makes sense that stronger social-emotional skills should set children up for success and that boosting these skills should have enduring & cascading effects, our findings suggest that these developmental processes are likely much messier than is commonly expected. Psychological Bulletin

When physical education teachers and students took an “autonomy-supportive” workshop, the effects of autonomy-supportive teacher moved into reports of more autonomy-supportive parenting.

• “Autonomy-supportive teaching increased students’ mid-year prosocial behavior, which increased end-year autonomy-supportive parenting.” Teaching and Teacher Education

Contextual Effects

“after a boost in library capital investment, reading test scores steadily increased.” American Economic Association

An RCT in Germany gave 11-12 year olds e-book readers with free access to digital books. Their reading increased, which led to improved academic performance in reading and math, and enhanced well-being. IZA Institute of Labor Economics

On the importance of being outside

• Did you know that there is a global epidemic of myopia in children? The solution is simple: kids need to spend more time outdoors.

• “School schedules need to build in outdoor time. Schools themselves should be designed to provide outdoor space for students” Wired, Science Based Medicine

• In fact, both adults and children need to stop sitting so much!.

• “What the vast majority of adults and children need to do is move more and sit less.” Scientific American

• A reminder that I’ve done a deep dive previously into the related importance of greenery to health and learning: The Influence of Greenery on Learning.

Where You Live Matters

“Growing up in a thriving community — where the adults are employed, in good health, etc. — dramatically improves children’s outcomes, even holding fixed their own family’s situation.” NBER

“we find that neighborhood human capital at the community level has the greatest impact on mobility, followed by the street, district, county, and province levels, respectively.” Social Indicators Research

“By equalizing average neighborhood quality for Black and White families, we estimate that the Army’s quasi-random assignment reduces Black-white earnings gaps among the children of Army personnel by 23%.” NBER

“For Black students, these relationships imply that they would receive more beneficial services in a school that was more racially integrated than in one that was fully segregated, highlighting another potential negative consequence of racial segregation.” Educational Evaluation and Policy Analysis

NYC “middle school students exposed to more diverse peers apply to and enroll in high schools that are also more diverse. These effects particularly benefit Black and Hispanic students who, as a result, enroll in higher value-added high schools.” NBER

“20 years after exposure, Whites who had more Black peers of the same gender in their grade go on to live in census tracts with more Black residents...the effect on residential choice appears to come from a change in preferences among Whites.” Journal of Public Economics

Contrary to misconceptions of public housing, this paper examines the impact of growing up in public housing for NYC and finds improved economic outcomes, reduced reliance on safety nets, and a cost effective public investment.

• Furthermore, public housing developments in neighborhoods with higher household incomes or fewer renters have better outcomes for children. United States Census Bureau

Gun violence is hyperlocal.

• “Just 4% of NYC’s 120,000 blocks...account for nearly all the city's shootings” from 2020-24. Gothamist

• “Instead of people, she says, we should be looking at places. . . in study after study, South has shown that simple investments in the environment . . . lower gun violence in the surrounding blocks by as much as 29 percent.” Philly Mag

If you’ve stayed with me this far, you are a true research nerd! Wishing you a very happy new year of more learning and inquiry.

#language #literacy #research #cognition #reading #writing #multilingualism #assessment #brain #cognition #academics #curriculum #wrapup

Learning new information in L2 is more effortful than in L1. We found different functional connectivity networks of naturalistic learning through speech among adolescents, confirming this prevalent observation

–Tweet from McGill University Professor Gigi Luk

Does learning language require effort? Does it require more effort when learning a new language later in our lives? Why?

Today, we will highlight a study that shows the additional neurological networks that adolescents activate when learning in a second language – a key insight for all educators to consider.

Language Learning: Effortless for Babies, Effortful for Adults

Babies learn language with such ease that they have already begun to recognize the unique patterns of a language–even to distinguish between the unique patterns of multiple languages–while still in the womb.

We therefore tend to assume there is something wholly innate or natural to learning language.

Yet as we’ve explored previously in a series on this blog, even learning our first languages may not be as innate or natural as it can appear. Human language reflects a unique synchrony between our biological and cultural evolution, finely attuned to the social environment in which we interact. It may be argued that babies learn languages more easily because they are learning the patterns of the fabric of their entire social universe, and language is interwoven therein.

This is termed the “critical period” of learning – a period of cognitive and cultural maturation that in humans is far more protracted than that of other animals. While there is clearly something innate about human biology in our receptiveness to language, there is also the fact that this slower process of maturation allows for more complex cultural influences beyond that of any innate biology. Most animal communication systems are more innate and develop fully earlier in life.

Armed with this understanding, it is not so surprising, then, that LLMs have revealed the inseparable connection of form and meaning in human language. Our languages reflect our lives within our communities – our cultural experiences in the world over time that can be understood only in relation to each other.

As we’ve also explored in another series, learning to read and write requires more effort, because it simultaneously demands greater abstraction and greater precision. It requires greater cognitive attention and attuned fine motor ability. As a form of “decontextualized language,” reading and writing go beyond immediate social interactions, requiring structured practice, formal instruction, and sustained cognitive focus.

This greater demand of written language on cognitive resources mirrors the challenges of learning a new language later in life. This means that “cognitive load theory” may have an undervalued contribution to bring to learning a new language.

It Takes Cognitive Effort to Learn Academic Content in a Second Language

So let’s dig into this paper some more . . .

• Citation: Leon Guerrero, S., Mesite, L., & Luk, G. (2024). Distinct functional connectivity patterns during naturalistic learning by adolescent first versus second language speakers. Scientific Reports, 14(1), 18984. https://doi.org/10.1038/s41598-024-69575-1

In the study conducted by Guerrero, Mesite, and Luk, 38 middle school students—19 native Spanish speakers (of whom 14 were born outside the US) and 19 native English speakers—watched an Earth science video in English while their brains were scanned using fMRI. For an example of the type of video they watched, see here.

They found that bilingual students used more parts of their brains, especially areas related to thinking and controlling attention, compared to monolingual students. Furthermore, while learning in English (their second language), the bilingual students were found to use their Spanish skills to better understand the lesson.

Some further insights from the paper:

“Bilingual adolescents with stronger Spanish cloze comprehension displayed lower connectivity with control regions, thus suggesting that higher-order comprehension skills make L2 processing easier, regardless of whether those skills were acquired in L2 or L1, i.e., English or Spanish.”

In essence, students who had developed more advanced reading comprehension skills in Spanish expended less cognitive effort when processing the video in English. Their robust language skills in their first language served as an internal scaffold, helping them navigate the complexities of an academic lesson in a second language.

“Differences emerge particularly in regions associated with higher-order language processing and cognitive control. This finding is especially relevant to adolescents learning academic lessons in their L2 as such lessons often focus on L2 vocabulary development. The current findings provide support for the idea that, beyond L2 vocabulary, students draw upon higher-order comprehension skills developed in their L1 to integrate L2 word meanings in understanding L2 discourse. This finding highlights the potential benefits of enhancing comprehension using syntactic and integrative skills in the L1 as a linguistic resource.”

This finding illustrates that bilingual teens use similar brain regions for complex language tasks in both languages. Improving a multilingual student's understanding of complex ideas in their first language can boost their learning in another language.

Reduce Cognitive Control to Increase Implicit Learning

Another fascinating parallel comes from a study with university students, which examined how cognitive depletion impacts language learning

After engaging in a memory task that taxed their working memory, students learned novel linguistic rules more implicitly, suggesting that cognitive fatigue can paradoxically enhance language learning by suppressing rule-based, conscious learning and promoting implicit pattern recognition—similar to how children learn languages.

As we’ve explored in relation to LLMs, this kind of implicit learning is critical to understanding both human language acquisition and the development of literacy skills.

Cognitive Load Theory and Language Learning

Cognitive Load Theory (CLT), a well-established framework in educational research, explains how working memory processes information and why certain tasks demand more cognitive resources. In the context of learning in a new language, the demands on working memory are significantly increased as students must simultaneously process the content of a lesson and the language in which it’s delivered.

For a practical overview of CLT, I recommend Oliver Lovell’s “Sweller’s Cognitive Load Theory in Action.”.

As Lovell succinctly explains, “New information takes up more working memory capacity than familiar information.” For multilingual learners, this dynamic is further complicated by the need to juggle both content and language, increasing cognitive load, as our research highlight demonstrates.

Lovell also states that “We reduce the working memory load of a task by chunking and automating.” I think it’s worth further considering both of these in relation to language learning.

Chunking

“Chunking” tasks and texts is an oldie but goodie when it comes to reducing cognitive load for all kinds of students. For novice learners, breaking tasks down to their components, then sequencing them to build practice towards clearly established models and success criteria, is the essence of scaffolding and differentiation.

For older students also learning a new language at the same time, sometimes it may also mean helping them to not let their thinking get in the way. By focusing students on the content itself, we may help free up their unconscious minds for implicit learning of the language associated with the content.

Building Automaticity

As we’ve explored in our first research highlight, “automatizing” the sounds, spelling, and meaning of words in a new language is important. As I wrote in that post:

Key words need to be not merely taught, but seen, heard, and read in varying contexts – and most importantly, actively used by students in varying contexts. Within a lesson, this means drawing attention to and using key vocabulary before, during, and after reading a core text, and this is a great place to start. That key vocabulary then needs to be spaced and interwoven in practice and use throughout the remainder of the unit of study! Some of this may be explicit, especially when first introducing words, but much can also be implicit if the vocabulary is aligned to and key to understanding the topic that all the content, texts, and discussions are oriented around.

Scaffolding with Home Language

As our research highlight also shows, multilingual students have a key and often under-utilized resource they can draw upon as an internal scaffold while learning content in English: they can draw upon previous knowledge and skills in their home languages.

This doesn’t mean that everything should be translated, but rather strategic planning for what and how much can be offered in students' home language as a scaffold to the instructional language and content – and how much we can support students’ in developing metacognitive and metalinguistic awareness so they can draw more intentionally on their multilingual resources.

Practical Implications for Language Teachers

Pop quiz: Who are language teachers? A) World language teachers B) ELA teachers C) ESL teachers D) All teachers

The correct answer is D! All teachers are teachers of language – the language related to the content they are teaching – and all teachers are teachers of students learning newer, more disciplinary forms of English, including students navigating multiple dialects and languages in their homes and communities.

Based on this review, there are several things that teachers can do to help their multilingual students learn more effectively:

• Provide clear and concise instructions and consistent routines. Ambiguity increases cognitive load.
• Break tasks into smaller, manageable steps. Scaffolding reduces demands on working memory.
• Create multiple opportunities for reading, writing, and talking using key words and sentences. Repeated, meaningful use of language solidifies learning.
• Encourage students to leverage their first language. Home languages can serve as a powerful cognitive tool, bridging understanding into a new language.
• Focus learning on meaningful content, rather than on the rules of language. The older we get, the more that our conscious minds can get in the way of the implicit learning that the patterns of languages afford.

In Sum: Cognitive Insights Into Language Learning

This interesting study by Guerrero, Mesite, and Luk offers insight into how adolescent students engage cognitively when learning academic content in a second language. By identifying distinct patterns of brain connectivity, the research reveals that teens with a home language of Spanish rely on additional neural resources, especially in regions associated with attention and cognitive control, to process new information in English. Moreover, students with stronger language skills in their first language of Spanish exhibited reduced cognitive effort when learning in English, highlighting the importance of developing robust literacy skills in both languages.

This research underscores the cognitive challenges of learning academic content in a second language, but also the potential advantages that come from leveraging home language as a scaffold. By acknowledging the cognitive load that multilingual students navigate, educators can adopt strategies that not only reduce the burden on working memory but also empower students to draw upon their linguistic strengths. In doing so, we can support their ability to learn content and language simultaneously, cultivating both academic achievement and bilingual proficiency.

#language #multilingual #research #cognition

Here's where I had spitballed we might go:

• The surprisingly inseparable interconnection between form and meaning
• Blundering our way to computational precision through human communication; Or, the generative tension between regularity and randomness
• The human (and now, machine) capacity for learning and using language may simply be a matter of scale
• Is language as separable from thought (and, for that matter, from the world) as Cormac McCarthy said?
• Implicit vs. explicit learning of language and literacy

Indeed, we then went on to explore each of these areas, in that order. Cool!

Some Hypotheses from This Series

What theories have we raised through this exploration?

1) LLMs gain their uncanny powers from the statistical nature of language itself; 2) the meaning and experiences of our world are more deeply entwined with the form and structure of our language than we previously imagined; 3) LLMs may offer us an opportunity to further the convergence between human and machine language; 4) AI can potentially extend our cognitive abilities, enabling us to process and understand far more information; 5) Both human and machine learning progresses from fuzzy, imprecise representations to higher precision, and the greater the precision, the greater the effort and practice (or “compute”) that is required; and 6) LLMs challenge Chomsykan notions of innateness and suggest that implicit, statistical learning alone can lead to gaining the grammatical structure and meaning of a language.

While I’ve been mostly positive and excited about the potential of AI (aside from pointing out how it is accelerating the looming ecological catastrophe that seems to be our trajectory) I should probably pause here to acknowledge that there may be important counterpoints to many of these (perhaps somewhat starry-eyed) hypotheses.

Onto the Counterclaims

Let's take a more critical look at some of my claims:

1) I claim that language is fundamental to the generative powers of LLMs.

Yet Andrej Karpathy, who is no stranger to LLM development, tweeted:

It's a bit sad and confusing that LLMs (“Large Language Models”) have little to do with language; It's just historical. They are highly general purpose technology for statistical modeling of token streams. A better name would be Autoregressive Transformers or something.

They don't care if the tokens happen to represent little text chunks. It could just as well be little image patches, audio chunks, action choices, molecules, or whatever. If you can reduce your problem to that of modeling token streams (for any arbitrary vocabulary of some set of discrete tokens), you can “throw an LLM at it.”

I agree that LLMs are performing “statistical modeling of token streams,” and that “for any arbitrary vocabulary of some set of discrete tokens, you can ‘throw an LLM at it.’”

We now have multimodal LLMs that are modeling out of token streams of audio, visual, and text, and will no doubt have ones feeding from additional streams of sensory data as they are increasingly paired with cameras on humans, objects, and robots.

Yet I also think Karpathy undersells that when LLMs suddenly exploded into general public awareness and fascination, it was merely a “historical” fact that they were trained upon vast amounts of human generated text and were able to reproduce and generate human language. As we’ve explored in this and a previous series, there is something about human language itself uniquely adapted for our brain circuitry and the propagation of our culture within social interaction in our world. And being able to communicate with a powerful computational model through the medium of conversational human language has been a revolutionary advent. We are just in the beginning stages of grokking it.

As I tweeted in response to Karpathy, token streams may be applied to anything, but human language seems to be uniquely suited to the advancement of combined human and machine learning. Not only because we rely on it for communication – but furthermore due to the algebraic and statistical nature of our language.

Recent case in point: the viral attention currently on NotebookLM’s Audio Overview. Listening to a conversation, however artificial, resonates with us, because that's what's in our social nature. And, surprisingly, it does a fairly good job of surfacing information from across multiple multimodal sources (and soon, across languages) that we find interesting, relevant, and meaningful.

Speaking of NotebookLM Audio Overview. . . here’s one derived from all the blog posts (except this one) from this series, as well as the sources–outlined in post 1–that inspired them all: https://notebooklm.google.com/notebook/a4f35399-e288-4293-b2d2-0489e6b1f037/audio

4) I claim there is great potential for AI to extend our cognitive capabilities

Yet there is a strong case of an equal and commensurate danger that use of LLMs can reduce our cognitive capabilities.

Learning more formal content and skills, like what we learn in school or in a job, requires deliberate effort until we develop an unconscious fluency. If students learning new concepts and skills externally automate their practice of new learning (such as writing or math) to an LLM, then they will not–ironically–be able to develop the automatized internal knowledge and capacity they need to wield powerful tools like AI more effectively.

When “experts” use tools like AI, they know where the gaps are in the output and are able to use it strategically to enhance their own production and output. A few examples of this:

• Simon Willison, a programmer who is also a great communicator, uses different LLMs to support his projects, and writes and speaks about how he does so. Here’s a podcast, for example, where he explains how he uses them.

• Nicholas Carlini, a research scientist at Google DeepMind, similarly wrote about how he uses AI to support his work.

• Cal Newport, who writes extensively about how to do “deep work” in a world of distractions, recently wrote in The New Yorker how he has found ChatGPT useful to his writing.

All the people above are highly skilled at what they do – so when they explore and then figure out how to use AI to support their work, they do so in a way that does not diminish their own hard-earned ability, but rather enhances and extends their capabilities.

On the other hand, for students–who are by definition novices in the skills and knowledge they are learning–an over-reliance on AI tools may limit their ability to develop skills such as literacy, critical thinking, problem-solving, and creativity.

Recent reports on AI in education, such as from Cognitive Resonance, Center for American Progress, and Bellwether, have rightfully raised this concern.

And all educators, whether K-12 or in higher ed, are seeing an increasing use of AI by students to complete homework assignments, so this danger of truncating the development of internal capacity is real.

I think the steps we can to take to address this are two-fold:

• limit the use of digital technology for learners at the earliest stages of learning, whether learners are preK-3 or learners being introduced to a new concept

• move practice of essential skills directly into the classroom as much as possible, while considering how AI could be used to extend, rather than diminish, any practice and feedback outside of the classroom

In a post on ethical use of AI, Jacob Kaplan-Moss argues that fully automated AI is unethical in the public sector due to its inherent biases and potential for unfairness in high-stakes situations. In contrast, the assistive use of AI can enhance human decision-making.

This assistive vs. automated use of AI may be a useful frame for thinking of how AI can be used most ethically and effectively in education. We want AI to be used to assist the learning process, rather than simply automating the solving of math problems or writing essays. This view aligns with Ethan Mollick’s idea of “co-intelligence” as well.

So far, I find the most powerful and interesting assistive applications for AI are more focused on educators (“the experts”), rather than on students (“the novices”). Teachers can leverage AI to support administrative tasks, analysis of student data, and consider additional enhancements of their instruction based on student data.

That said, I don’t think the assistive use cases of AI are only limited to “experts” in a domain. AI can also help to equip those without knowledge and expertise in a specific area with the language they need to navigate learning or real-world communications more effectively. And there are some really interesting use cases of AI for feedback on student thinking and writing, when structured with specific guidelines and criteria and with the teacher in the loop.

But in the context of classroom learning, such uses must be very strategically designed and cautiously incorporated. For example, see this explanation from professor Michael Brenner on how he has begun incorporating AI into his pedagogy. But note this example is from a graduate level math class, so again, that novice vs. expert dynamic is quite different from what we would need to consider at a preK-8 level. But even at that graduate level, you can see there is quite a bit of complexity the instructor needed to consider and think through to design his course to leverage LLMs so strategically.

There’s a lot more to unpack here on all sides of the equation. I’ll leave this one here for now, accepting non-closure, and I hope to dig further into these tensions and opportunities in both this space and in my professional work.

6) I claim that LLMs have shown that language can be learned without any innate programming or structure – therefore demonstrating the power of statistical, implicit learning

I’d moved into the “Chomsky is wrong” camp for a while now, but I happened to listen to an interview of Jean-Rémi King recently, a scientist at Meta AI, by Stephen Wilson on The Language Neuroscience podcast (did I tell you I’m a nerd?). Towards the end of the conversation, King warns against writing off Chomsky too readily, and that there is something intrinsic to the human brain in its readiness for language.

I uploaded the relevant portion of the transcript from the interview, and asked Claude AI for a concise summary of King's main claims, which it willingly obliged (while I’m sure it drew upon an unconscionable amount of energy):

King argues that human brains likely don't use the same “next word prediction” principle as large language models for language acquisition, primarily because humans are exposed to far less linguistic data than these models.

He contends that while language models have shown impressive capabilities, they are extremely inefficient compared to human language learning, suggesting that we're missing some fundamental principles of how humans acquire language so efficiently.

While I think I’ve tried to temper most of my pronouncements throughout this series, I think it’s important to acknowledge that the fact that LLMs can learn language from statistical associations of word tokens alone does not mean that is exactly how we humans must also learn language.

It is rather a proof of concept that language can be learned in this way (without any innate grammar or teaching of rules). But as King points out, this is via a scale of input that is ridiculously and exponentially larger than that of any child.

That said, there are other Artificial Neural Networks (ANNs), such as in the research of Gašper Beguš, that learn from raw speech in an unsupervised manner, more closely mimicking human language acquisition. His lab has found interesting similarities between these ANNs and the human brain in processing language sounds – a parallel to King’s own research, which has found that LLM models can generate brain-like representations when predicting words from context.

And there will continue to be research into tinier models trained off sparser, and potentially richer, data.

But as King points to, there’s just so much more we need to learn. And this is exactly where I find all of this the most exciting.

Where I may be most rightfully critiqued in my last post, and perhaps in other posts, may be in extrapolating from the theoretical demonstration of LLMs to implications for classrooms.

So let me state my position a bit more clearly in case there was any confusion that I am falling onto the side of the Goodmans or something. Children need consistency, stability, clarity, and coherency in their learning experiences, and teaching what is most important to know for a given subject directly and explicitly is critical. For children at the earliest stages of learning abstract skills and content, such as learning to read, explicit and well-structured teaching is essential. At the same time, however, we need to ensure that students have abundant structured opportunities to apply and practice what they are learning – and this is where ensuring they are spending more time reading, writing, and talking–connected to the content of what we are teaching–is essential.

If you have more critiques that I am missing in any of the above, please do share!

Egads, I think I may actually have ANOTHER post left in me after all of this. Who knew LLMs would be such an interesting topic?!

#language #literacy #AI #LLMs #cognition #research #computation #models

The Surprising Success of Large Language Models

“The success of large language models is the biggest surprise in my intellectual life. We learned that a lot of what we used to believe may be false and what I used to believe may be false. I used to really accept, to a large degree, the Chomskyan argument that the structures of language are too complex and not manifest in input so that you need to have innate machinery to learn them. You need to have a language module or language instinct, and it’s impossible to learn them simply by observing statistics in the environment.

If it’s true — and I think it is true — that the LLMs learn language through statistical analysis, this shows the Chomskyan view is wrong. This shows that, at least in theory, it’s possible to learn languages just by observing a billion tokens of language.”

–Paul Bloom, in an interview with Tyler Cowen

Challenging the Hypothesis of Innateness

For decades, the Chomskyan view has dominated our understanding of language development. This view argues that language structures are too complex to be learned solely from environmental input and therefore must require some kind of innate linguistic machinery in the brain (a “universal grammar”).

Yet as the quote above from Paul Bloom makes explicit, what LLMs have demonstrated–as a proof of concept–is that grammatical structures for language does not need to be innate. That machines can learn language via statistical associations alone, rather than explicitly programmed grammatical rules.

We have explored in a previous series on this blog the idea that language may not be a completely innate property of our brains, but rather more of a cultural phenomenon. This parallels the insight–much more widely accepted now–that learning to read is not innate.

The success of LLMs in acquiring language-like abilities through mere statistical analysis of texts demonstrates that it's possible to learn languages via statistical associations alone.

The Power of Statistical Learning

This revelation–that LLMs can learn language via statistical associations alone, rather than through any explicitly programmed rules–challenges our traditional understanding of language development and points to the power of implicit statistical learning.

However, unlike human children, who can rapidly learn language from relatively sparse input, current frontier LLMs require astronomical amounts of data to be trained. Yet the fact that machines can learn in this way suggests that the structure of language itself lends itself to such implicit learning.

This insight extends beyond language development and into literacy. We have previously examined seminal papers by Philip Gough and co arguing that learning to read words is more akin to learning a cipher than breaking a code. Rather than learning explicit rules, as from a codebook, we internalize patterns of sounds, letters, and meanings in an algorithmic fashion.

There is a fascinating line of research focused on “statistical learning,” and while there remains much to be learned about this domain, there seems to be an interesting convergence between this research as it relates to reading and as it relates to LLMs.

Reading nerds are already well acquainted with Mark Seidenberg, as he is a steady presence in the public sphere of communication and debates about reading instruction. What may be somewhat less known about him is that his oeuvre of research has been into computational, connectionist models of reading that have demonstrated how learning to read is a process of statistical learning between sounds, spelling, and meaning. It’s not that he hides this, by the way, but rather that the community of educators that are deep into the “science of reading” stuff don’t seem to be as enticed by abstract stuff like computational models and statistical learning.

But the convergence between connectionist accounts of learning language and learning to read and the advent of LLMs are important to understand. Not just from a nerdy stance, which has been mine throughout all these posts, but rather because LLMs have–again, as a proof of concept–demonstrated that implicit learning of statistical associations are fundamental not only to language and to reading, but to our knowledge and experience of the world.

Connectionist Models: Bridging AI and Human Learning

In fact, Seidenberg himself has repeatedly attempted to communicate the understanding that implicit statistical learning is just as fundamental to learning to read as it is to learning language.

He stirred up some recent controversy on this topic when he suggested that the “SOR” movement has over-corrected in response to previous squishy balanced literacy approaches by focusing too hard on explicit instruction as the cure-all for everything. See his provocative presentation and writing on this topic here: https://seidenbergreading.net/2024/06/24/where-does-the-science-of-reading-go-from-here-2/

To summarize his argument, which dovetails with where we started with LLMs, learning to read can not all be taught explicitly, and there is an opportunity cost to an over-reliance on the explicit teaching of “rules” over providing more opportunity for actual reading and writing to build up the statistical associations needed to become fluent:

“The purpose of explicit learning is to scaffold implicit learning about print, sound, meaning. Explicit instruction is the tip of the iceberg. The larger part under the surface is learned implicitly instead of teaching the whole iceberg.”

—slides on “Where does the Science of Reading go from here?”

In other words – only provide enough explicit instruction as needed to successfully spend more time engaged in an increasing volume of reading, writing, and talking.

Balancing Explicit and Implicit Learning in Language and Reading Instruction

In a paper, “The Impact of Language Experience on Language and Reading,” Seidenberg and Maryellen MacDonald also point to the fact that learning to read is easier for children with more advanced spoken language skills, while those with less exposure (due to greater variability of linguistic input) face greater challenges. This is because children exposed to multiple dialects or languages are learning to navigate multiple language systems, each with its own set of statistical linguistic patterns.

For multilingual and multidialectal learners, it is therefore especially critical to find the right combination of statistical learning and explicit teaching. According to the paper, consistent and increased exposure to the language of instruction is important. This exposure should be complemented by explicit teaching of both oral and written language patterns. And by explicitly comparing and contrasting home languages and dialects with the language used at school–both orally and in writing–students can develop metalinguistic awareness and a deeper understanding of varying language structures. This approach, implemented strategically within a welcoming and supportive classroom, allows students to leverage their existing linguistic knowledge while acquiring new language skills.

Another way of thinking about this, as we’ve explored in another post, is the movement from fuzziness to precision. By seeing, hearing, speaking, and writing an increasing volume of language, students can rapidly begin to make statistical associations. However, especially in the initial stages of learning a new language or learning to read, more effort will be required to gain greater precision, and thus, more mistakes will be a part of the learning process, and thus more feedback is needed to course correct at the very beginning.

I’ve written elsewhere about the importance of striking a balance between close reading of shared grade-level texts that are worth reading, while ensuring that each and every student reads a steady volume of texts that are more accessible. I’ve also written here about the need for “daily textual feasts” to increase the volume of rich language, knowledge, and critical thinking, as per Dr. Alfred Tatum.

Rethinking Language and Literacy Instruction

In sum, the surprising and awesome ability of LLMs, derived from mere statistical associations, has challenged traditional assumptions about the innate nature of language and, potentially, the role of explicit and implicit instruction in language and literacy learning.

This underscores the need for a comprehensive approach to teaching of reading and language, in which explicit teaching is strategically counterbalanced alongside implicit learning opportunities.

#AI #learning #language #LLMs #reading #explicit #implicit

1) LLMs gain their uncanny powers from the statistical nature of language itself; 2) the meaning and experiences of our world are more deeply entwined with the form and structure of our language than we previously imagined; 3) LLMs offer an opportunity for further convergence between human and machine language; and 4) LLMs can potentially extend our cognitive abilities, enabling us to process far more information.

In a previous series, “Innate vs. Developed,” we’ve also challenged the idea that language is entirely hardwired in our brains, highlighting the tension between our more recent linguistic innovations and our more ancient brain structures. Cormac McCarthy, the famed author of some of the most powerful literature ever written, did some fascinating pontificating on this very issue.

In this post, we’ll continue picking away at these tensions, considering implications for AI and LLMs.

Fuzziness and Precision in Language Development and Use

To start us off, I want to ground our exploration in two concepts we’ve covered previously in “An Ontogenesis Model of Word Learning in a Second Language”:

• Fuzziness: “inexact or ambiguous encoding of different components or dimensions of the lexical representation that can be caused by several linguistic, cognitive, and learning-induced factors. These factors include, among others, changes in neural plasticity, the complexity of mapping L2 semantic representations on the existing L1 semantic representations and of mapping L2 forms on the semantic representations, and problems with L2 phonological encoding”

• Optimum: “the ultimate attainment of a representation (or its individual components), i.e., the highest level of its acquisition, when the representation is properly encoded and no longer fuzzy”

I think these concepts are useful not only for thinking of learning new words in a language, but also for how we interact with LLMs and the language they are trained upon.

From Fuzziness → Optimum

When we first learn a language, whether while in the womb, in school, or after moving to a new community, what we hear and understand is fuzzy. The first thing we attune to is the prosody of the language: its tones, volume, and duration. We can’t yet fully distinguish words and sentences within a stream of speech, nor syllables from phonemes, nor vowels from consonants. Let alone connect those sounds (or signs) to meaning and communicate with them to others.

Yet as we gain greater discernment across hearing, vision, movement, and speaking, our representations of a language becomes more flexible and more precise. As I’ve written about elsewhere, connecting speech directly to its form in writing can enhance language and reading and writing development simultaneously. Oral and written language – and reading and writing – can develop reciprocally. Developing one supports refining the other.

Why would that be, given we didn’t invent the technology of writing until far down the timescale of human evolution?

Precision in Language and Cognition

Maybe it’s because the written form of a language requires greater precision in the representation in our minds. When greater precision is required, it takes more time and effort, at least initially, to produce.

As an example, you may have heard of the term “receptive bilinguals.” These are individuals who can understand the gist of an everyday conversation in another language, but may struggle to speak or produce it fluently. This is because they may have had fairly significant exposure to the language, especially in childhood, but their mental representations remain “fuzzy” because they rarely produce the language either orally or in written form.

The more that we hear and read AND produce a word – and particularly when we produce it both orally and in writing – the more likely and quickly we are to reach optimum.

We see this process play out in real time with babies. They listen to our sounds and watch our faces, then begin to babble, mimicking us. They begin connecting those sounds to things and ideas. And then they begin to gain a more precise understanding and use of a word, from there stringing multiple words together into sentences, again starting haphazardly and working towards greater flexibility and precision.

Fuzziness, Precision, and Specialization in Language, Cognition, Computation, and Literacy

LLMs have demonstrated that there is far more knowledge, meaning, and comprehension of the world embedded within the statistical relationships of the words and phrases we use than we previously suspected.

As we’ve also explored, there are fuzzier and more precise terms and concepts in a language. The more abstract and “decontextualized” an event or idea (meaning that the event or idea is not readily available in the context of that environment or moment) the more precise, vivid, or specialized our language becomes in the effort to describe it. This can lead us all the way to the extreme of computational language, which is highly precise, much harder for humans to learn, and quite alien in comparison to the general fuzziness of our everyday language used to communicate about everyday things.

The reason read-alouds are so very powerful in the beginning of childhood (and arguably, through adolescence, perhaps even beyond) is because they provide children with exposure to and immersion in this more decontextualized type of language and more abstract and broad understandings of the world. This helps prepare them for when they later engage with written forms of language and increasingly discipline-specific forms of discourse.

As language learning develops towards greater precision, networks in the brain are forged and strengthened. One of the reasons why early childhood is so incredibly important to language and literacy and motor development is because the brain supercharges the neural connections it is forming in all directions. Dendrites spring up like fungus after a rain. But learning new things requires a bit more effort as we age because we work far more on pruning our existing connections for efficiency.

Yet no matter our age, developing these increasingly robust cross-brain connections, and then increasingly specializing and refining them for specific domains and uses, can increase our mental resilience.

We can see this process of specialization play out in real time with young children as they learn to read and write. As they gain greater precision with representations of language through spelling, writing, and volume of reading, their brains increasingly forge further connections between the architecture used with executive function, speech, vision, and motor control, while then specializing and refining them.

Developing language and literacy in multiple languages – to the point of optimum – even further connects, specializes, and refines those networks. And when one is bi- or multi-literate on disciplinary topics – with the specialized and precise language required for communicating flexibly about those topics – then those networks are yet further refined.

This is similar, arguably, as with the development of cognition. Cognition—a fancy way of saying “awareness, knowledge, and understanding”—includes the facets of executive function and memory that are also tapped into when developing language, yet are surprisingly separable from language in the brain, in terms of the processes identified through brain scans, at the same time.

I think a useful way to think of this distinction may be the difference between the unconsciousness or the lack of awareness we may have about something PRIOR to learning it, and the unconsciousness and lack of awareness we have AFTER learning it to optimum. When we have attained fluency with a skill or pushed our knowledge into long-term memory, we no longer need to apply much effort – nor thought – to drawing upon it. It is the degree of effort that is required in order to learn or use something that determines the level of cognition we need to initially draw upon. And while we can certainly expand our cognitive ability and other aspects of our learning potential, there are also hard upper limits – such as the bottlenecks of our working memory and our attention.

We overcome those bottlenecks by committing important information to long-term memory through regular use and communication, automatizing regularly used skills through practice, and leveraging the institutionalization of knowledge-based communities and the technologies of writing (texts) and digitization to process and communicate and further refine larger volumes of information.

The Limitations and Potential of LLMs

While human children rapidly develop language and literacy from comparably minimal amounts of input and interaction in their world, LLMs are trained on vast bodies of text, the majority in written form (thus far). Their training is developed to refine and make more precise their abilities to predict the concatenations of continued tokens and words from what we have fed them.

Similar to human brains, LLMs move from a fuzzy-to-precise spectrum as they refine the “weights” they assign to linguistic tokens across their many layers. Early or small models of LLMs, akin to our “receptive bilingual” example earlier, demonstrate some receptive capabilities, but their generated outputs are highly fuzzy, as they did not have sufficient neural layers, training, and feedback (i.e. sufficient input and production) to achieve something close to optimum in their generation of human-like language.

But to state the obvious, LLMs do not experience the world as we do. They have no bodies, no sensory input, no social interactions (unless you count the part of their training that requires humans to provide them with corrective feedback). As a reminder, the fact that they have the capabilities they do–derived merely from the accumulated statistical relationships of parts of words–is remarkable. They do not “think,” at least, not in the manner in which our own cognition functions, and they do not continuously build and further refine their knowledge–yet–from ongoing interactions and input from other AI and with us.

LLMs are like if we took away all the other parts of our brain—those more ancient parts that continue solving problems and help us steer our way home and keeps our hearts beating—and only left the parts dedicated to language. That they are able to do all they can from mere statistical relationships forged from language alone is–again–remarkable, but it also shows us their limitations.

To be frank, that the dialogue has been so singularly focused on the “intelligence” of LLMs, with the goal of forming “artificial general intelligence” (AGI) seems remarkably off base to me. What I am far more interested in is the potential of these models to teach us something about our own development of language and literacy–and thus, how we can better teach those abilities–and to extend our own cognitive abilities.

Enhancing Cognition with AI

Towards this end, I want to suggest some implications for education that takes us away from fears about AI making kids dumber or taking away jobs from teachers.

AI and LLMs can enhance our cognitive abilities by helping us to:

• Process Large Amounts of Information to Gain Knowledge: AI and LLMs are getting better and better (seemingly every week) in sifting through vast amounts of information, such as databases, research, transcripts, and other documents, to help us summarize, answer questions, paraphrase, and understand the relevant knowledge contained in them. Furthermore, they are getting better and better at translating across multiple languages and in reading multiple modalities. You can feed an LLM an image with text in another language and it can read it.

• Augment Our Own Thinking and Writing: LLMs work really well in helping us spitball ideas or redraft our own writing. The fear that they will stop kids from being taught to write is misplaced – the writing produced by LLMs is only as good as what they are given. Yes, they are great at boilerplate forms of writing! But that’s the exact kind of writing that we do want to automate and reduce our own time and thinking on. When it comes to deeper writing and thinking like this series and post, it ain’t writing it for me. But I do find it really helpful when I get stuck or when I want to get suggestions for revision.

  In Sum

The effectiveness of our use of AI and LLMs hinges on the quality of our input.

As with previous tools like Google Search, the more precise and informed our prompts, the more powerful and accurate their responses.

Another way of framing this idea: LLMs can help us further widen or refine our own ideas and language. They are far less useful in just handing them to us. They mirror and leverage what we provide to them.

There is a lot of talk about the “hallucinations” of LLMs, but perhaps a better way to frame it is as “pixelation,” or grain size. There are larger and smaller grain sizes of pixels. The coarser the grain, the less clear it is. The finer the grainer, the sharper it becomes. The more vague and broad the grain size we feed them, the more BS they will spit. The more precise and narrow grain sizes we provide, the more accurate and useful their responses will be. They can then help us move into different grain sizes from there (either widen our lens, or narrow our lens).

This means that we need to keep teaching our kids stuff. The more knowledge they have, the more precise and flexible their ability to wield language, the better they can use powerful tools like AI.

We can help kids to use AI in this way, and we can create tech-free spaces in our schools where they need to put in the cognitive effort and time they need to build their fluency with language and literacy and read texts that build their knowledge. And then when we engage them with the tech, we teach them how to use it to extend, rather than diminish, their own potential.

There’s implications here for teachers too – in fact, I think the most exciting potential for AI is actually freeing teachers up to spend more time teaching, and less time marking up papers and analyzing data. But that’s for another post.

#AI #LLMs #cognition #language #literacy #learning #education Discuss...

“Over cultural evolution, the human species was so pressured for increased information capacity that they invented writing, a revolutionary leap forward in the development of our species that enables information capacity to be externalized, frees up internal processing and affords the development of more complex concepts. In other words, writing enabled humans to think more abstractly and logically by increasing information capacity. Today, humans have gone to even greater lengths: the Internet, computers and smartphones are testaments to the substantial pressure humans currently face — and probably faced in the past — to increase information capacity.”

—Uniquely human intelligence arose from expanded information capacity, Jessica Cantlon & Steven Piantadosi

According to the perspectives of the authors in the paper quoted above, the capacity to process and manage vast quantities of information is a defining characteristic of human intelligence. This ability has been extended over time through the development of tools and techniques for externalizing information, such as via language, writing, and digital technology. These advancements have, in turn, allowed for increasingly abstract and complex thought and technologies.

The paper by Jessica Cantlon & Steven Piantadosi further proposes that the power of scaling is what lies behind human intelligence, and that this power of scaling is what further lies behind the remarkable results achieved by artificial neural networks in areas such as speech recognition, LLMs, and computer vision, and that these accomplishments have not been achieved through specialized representations and domain-specific development, but rather through the use of simpler techniques combined with increased computational power and data capacity.

I think the authors may be overselling scaling as the main factor behind intelligence, but scale most definitely plays a leading role alongside brain and neural network architecture and specialized data, and it most definitely plays a role in how human language is used and developed.

The Potential of Scale

“LLMs give us a very effective way of accessing information from other humans.”

–Alison Gopnik in an interview with Julien Crockett in the Los Angeles Review of Books

In our previous explorations of language, cognition, and Large Language Models (LLMs), the recurring theme of the power of scale has certainly emerged.

We've delved into the statistical nature of language, where the vast interconnectedness of word combinations and their contextual relationships drive LLMs' generative abilities. We've pondered the inherent imprecision of human language and the journey towards computational precision in LLMs. And throughout, the concept of scale has remained central – the scale of data, the scale of computation, and the scale of language itself.

It's intriguing to consider the possibility, as this paper suggests, that the capacity to process increasing amounts of information may have been a key factor in the development of human intelligence. This idea extends to how, as a species, we have continually sought ways to expand our ability to store and access information, from the invention of writing to the development of computers, the internet, and smartphones.

This suggests that the most exciting potential of artificial neural networks such as LLMs may lie not only in their ability to respond to and generate human language, but furthermore in their ability to help us to process and manage vast quantities of information, and thus further extend our cognitive capabilities. When framed in this manner, it shifts the debate from whether LLMs already demonstrate human intelligence and whether they will soon achieve superhuman intelligence, to whether LLMs will indeed equip us with superhuman abilities. And – as always with advancements in powerful technologies – the question is who among us will gain the most from those abilities and whether the new tools will further increase or diminish disparities between groups (i.e. “the future is already here — it's just not very evenly distributed”).

So we’ve explored a few implications of LLMs relating to language and literacy development so far, then: 1) LLMs gain the base for their uncanny powers from the statistical nature of language itself; 2) LLMs present us with an opportunity for further convergence between human and machine language; and 3) LLMs present us with an opportunity to further extend our cognitive abilities by allowing us to process far more information.

The Dark Side of Scale

All of this said, there is a dark side to scale, as Geoffrey West elucidates in his book, Scale (more on this on my other blog, Schools & Ecosystems), which is that as we consume far more energy and create far more waste beyond our biological needs and functions than any other creature on earth as we continue to scale our technologies. As West describes it, we humans are energy-guzzling behemoths, using thirty times more energy than nature intended for creatures our size. Our outsized energy footprint makes our 7.3 billion population act as if it were in excess of 200 billion people. And we are hitting the upper limits on ecological constraints of the earth as we do so.

Similarly, as LLMs extend our capabilities, they consume ever more power as they consume and produce ever more data. So at the very same time that our earth is rapidly accelerating towards critical thresholds of environmental change and wreaking havoc on insect, animal, soil, and plant life, we are rapidly accelerating our consumption of energy and production of waste.

It’s hard to see a clear end in sight to this. It’s possible that the greedy demands of continuing to scale AI model training and use ends up leading to rapid development of greener technologies and accelerated efficiency in digital computation and compression. It’s just as possible that in our short-sighted endeavors we put a half-life on human civilization via no longer containable war, famine, disaster, and disease.

Not to end this post on such a sour note, but it is important to bear a healthy skepticism about a new technology and its attendant powers, even as we seek to gain from it. And from what I see in the discourse, it seems to me that there has been a pretty healthy mix of boosterism and critique and excitement and paranoia about it all, so I’m enjoying the ride, nonetheless.

#cognition #language #AI #LLMs #technology #brains #scale