Watch a heritage language class in the first week of the semester and you will see the same student twice.

The first student speaks without effort. She talks about her grandmother, the food at a family party, the argument she had with her cousin, what she is doing next weekend. Her pronunciation is native, her intonation is native, and a stranger on the phone would assume she grew up in the country her parents left. The second student cannot decode a two-paragraph news story, writes at the level of a beginner, and has no idea how to construct a sentence that sounds formal.

They are the same person. This split — fluent speech, missing literacy and missing register — is not a failure of effort or intelligence, and it is not a mysterious gift that faded. It is one of the most well-documented and predictable profiles in the study of bilingualism, and once you know what causes it you can do something about it.

The numbers behind it are not small. According to the US Census Bureau's American Community Survey, about 22.3% of the US population aged five and over speaks a language other than English at home, and 13.6% speaks Spanish at home. The count of people speaking a language other than English at home nearly tripled between 1980 and 2019, from 23.1 million — roughly one in ten — to 67.8 million. And roughly 62% of those speakers report speaking English "very well." That last statistic points straight at the problem: the census asks whether you speak the language. It never asks whether you can read it.

What Counts as a Heritage Speaker

The technical term is heritage speaker, and the standard definition is a bilingual who grew up hearing and often speaking a minority or immigrant language at home while being dominant in the majority language of the surrounding society. Maria Polinsky and Olga Kagan, two of the researchers who established the field, put it in the title of a 2007 paper: heritage languages are acquired "in the ‘wild’ and in the classroom." The language was learned in a kitchen, not a syllabus.

Three features define the profile, and all three matter for how you should study:

  • Strong foundations, uneven ceiling. Heritage speakers usually had thousands of hours of natural input before school took over their lives, but that input was restricted to a handful of settings. Family talk, food, feelings, errands and jokes: a narrow slice of the language, learned to a very high level.
  • There is no single level. The range runs from people who can only understand and never speak, through those fluent in the home register and nothing else, to people with near-total oral control and no literacy. Placement in a language class is a genuine research problem precisely because one test score cannot describe two tracks.
  • Grammars drift. A 2013 paper by Elabbas Benmamoun, Silvina Montrul and Maria Polinsky describes the population as bilingual speakers "whose first language often does not reach native-like attainment in adulthood," and identifies case marking and aspect as the areas where heritage grammars most often settle into a system of their own.
"Heritage speakers present a different case: they are bilingual speakers of an ethnic or immigrant minority language, whose first language often does not reach native-like attainment in adulthood." — Benmamoun, Montrul & Polinsky (2013), Theoretical Linguistics 39(3–4), 129–181

The Profile: What Is Strong and What Is Missing

The most useful single study for anyone with this profile compared heritage speakers to classroom learners directly. Silvina Montrul, Rebecca Foote and Silvia Perpiñán tested 69 heritage speakers, 72 second-language learners and 22 native speakers on Spanish grammatical gender agreement, across three tasks: oral production, written comprehension and written recognition.

The results split cleanly along the boundary you already suspected. On the written tasks, the classroom learners — who had studied the rules explicitly and practiced them on paper — outperformed the heritage speakers. On the oral task, the heritage speakers were ahead. Same language, same feature, opposite rankings depending on the channel.

Writing shows the same pattern. A 2017 study in Foreign Language Annals by Allison Gatti and Terry O'Neill collected biographical data and ACTFL Writing Proficiency Test ratings for 187 adult heritage speakers of Mandarin Chinese, Korean and Spanish. The findings led them to recommend that heritage learners be assessed and taught separately from beginners, and that instruction "leverage the learners' linguistic resources" rather than start from zero. A 2020 survey of research on Spanish heritage writers in the journal Dimension (indexed as ERIC EJ1249861) states the underlying cause plainly: "the opportunities to use heritage language (HL) skills are often oral, not written."

So the missing piece is not the language. It is a specific set of skills that a spoken home environment cannot supply: orthography, formal registers, academic and professional vocabulary, and the explicit grammar that schooled speakers got drilled into them.

Why Literacy Is the Part That Goes Missing

The mechanism is simple once it is stated. Speech is acquired; literacy is taught. Nobody in a kitchen tells a child to spell anything. For alphabetic languages, a heritage speaker may absorb some reading by accident — a children's book, subtitles, a text message — but in languages whose writing system is not phonemic, there is no accident available at all.

Japanese is the clearest case, and the research on it is blunt. Kanji is not a code you can guess from speech: every character is a memorized unit with its own readings, and the average Japanese school student spends nine years accumulating roughly two thousand of them. A 2007 study of literacy development in young Japanese heritage learners in the United States (ERIC EJ830629) notes that oral language and literacy development depend on a joint effort between the heritage school and the parents, precisely because the hours available in a weekend school cannot carry the load alone. This is why heritage speakers of Japanese and Chinese so often sound entirely native and read like beginners, and why Korean heritage speakers who have mastered the alphabet in a week still hit a wall when academic vocabulary arrives in hanja.

The register gap works the same way. A minority language used at home serves a small number of functions — intimacy, family logistics, humor, prayer, maybe music. It is rarely used for argument, negotiation, instruction, or the news. The vocabulary of those functions was never spoken in the house, so it was never learned. Benmamoun and colleagues' observation about case and aspect is the grammar version of the same story: the parts of the system that only show up in formal, written, complex sentences are exactly the parts that were starved of input.

The Argument About the Word "Incomplete"

For years, the dominant framing was that heritage speakers had incomplete acquisition — a version of their parents' language with pieces missing. Silvina Montrul's 2008 book, Incomplete Acquisition in Bilingualism, is the definitive statement of that view, and it has real evidence on its side.

But the framing has been contested, and the objection is worth internalizing if you are the person being described. Tania Kupisch and Jason Rothman argued in International Journal of Bilingualism that "difference is not incompleteness": heritage speakers are not failed monolinguals, but speakers of a variety in its own right, shaped by the fact that they were exposed to a different distribution of input. Jason Rothman made the same case for Romance heritage languages in 2009.

This is not a semantic squabble. If your home Spanish is a broken version of "real" Spanish, your instinct will be to erase everything you know and start clean — the worst possible study plan, since it throws away your strongest asset. If it is a legitimate variety with a different distribution of strengths, the plan writes itself: keep the fluency, add the missing tracks.

What Childhood Input Actually Leaves Behind

One more body of research matters here, because it answers the question every heritage speaker asks: is there still an advantage in there somewhere, or is it gone?

Starting in 2002, Terry Au and colleagues studied a group of adults who had overheard a language as children — Spanish or Korean heard regularly in the house — without ever learning to speak it, and who then tried to learn it later. Their findings settled a long-running dispute in the field:

  • Compared with learners who had no childhood exposure, the overhearers ended up with a more native-like accent (Au, Knightly, Jun & Oh, Psychological Science, 2002).
  • The advantage was surprisingly specific. The full study — 15 native bilinguals, 15 childhood overhearers and 15 typical late learners of Spanish — found the overhearers better on every pronunciation measure (voice onset time, degree of lenition, accent ratings of phonemes and story production), and showed no benefit at all in morphosyntax (Knightly, Jun, Oh & Au, Journal of the Acoustical Society of America, 2003).
  • Later work in the same line (Au, Oh, Knightly & Jun, Journal of Memory and Language, 2008) tested whether a classroom course could recover the rest. It could — and the phonetic advantage survived the instruction.
"Results confirmed a pronunciation advantage for the childhood overhearers over the typical late L2 learners on all measures: phonetic analyses (VOT and degree of lenition), accent ratings (phoneme and story production), but no benefit in morphosyntax. … These findings illustrate the specificity of overhearers' advantage to phonological production." — Knightly, Jun, Oh & Au (2003), JASA 114(1), 465–474

The flip side is the most sobering result in this literature, and it explains why some people lose the language completely. Christophe Pallier and colleagues scanned adults who were born in Korea and adopted by French families in childhood, and who reported no recollection of Korean. On behavioral tests of Korean memory they performed no better than French speakers with no exposure to Korean, and their brains showed no response specific to Korean sentences when hearing Korean, French and unknown languages. Ventureyra, Pallier and Yoo reported the same loss at the level of phonetic perception in 2004.

Put the two lines of research together and the lesson is sharp. Childhood exposure is not a bank account that sits waiting; the advantage survives only where use continued and is thinnest exactly in the formal, written layers. A heritage speaker who still talks to family every week has kept the phonology and the fluency. The grammar, the formal register and the orthography were always going to have to be learned deliberately.

A Six-Step Protocol for Closing the Gap

Everything above converges on one practical conclusion: treat your heritage language as two tracks, and train them separately. Here is how that looks with flashcards and a study routine.

1. Assess the two tracks separately, on purpose

Do not take a single placement test and accept the verdict. Test speaking and listening with a conversation or a dictation; test reading with a real text, unassisted; test writing by producing 150 words with no translation tool. Most heritage learners discover that their listening score is two to four levels above their writing score. That gap is your curriculum.

2. Give literacy its own daily block — and start with the writing system

If your language uses a non-phonemic script, this is not optional and it is not slow: a few hundred well-chosen characters cover the overwhelming majority of everyday text. Learn them as a separate daily habit (ten to fifteen minutes), in a fixed order from most frequent to least, and accept that this block will feel like school. It is school. That is the point — it is the one input a household never gave you.

3. Build flashcards audio-first, then add production

Your strongest channel is sound, so use it as the anchor rather than the afterthought. A card with audio on the front and the written form on the back lets you attach a spelling to a word you already know how to say, which is a far faster route than learning the word from scratch. Then split your deck into two card directions and track them separately: recognition (see it, understand it) and production (mean it, say or write it). Heritage speakers usually find recognition easy and production humbling, and the production cards are where the gains are.

4. Widen the domains deliberately

Your existing vocabulary is deep in one domain and shallow everywhere else. So choose new material by domain, not by difficulty: news, school subjects, work, money, law, medicine, argument. Sentence-mine from those sources rather than from family conversation, and keep a separate deck for each domain. This is the single highest-yield change for a heritage speaker, because it converts the language from a family instrument into a general one.

5. Learn the register your family never used

Formal speech is where heritage speakers get judged — job interviews, teachers, doctors, officials. If your language marks politeness, honorifics or formality, those forms were probably the least spoken sentences in your childhood home. Learn them explicitly, as a set, the way you would learn a grammar rule: the polite question forms, the written connectors, the subjunctive or honorific verb endings, the standard greetings for strangers versus family. A handful of formulas covers most of the situations where the gap is most visible.

6. Decide which standard you are learning — and write the decision down

There is no single "correct" version of your language, and the one in the textbook may not be the one your family speaks. Pick a target deliberately — the standard variety because that is what exams and workplaces use, or your family's variety because that is the language you actually want to keep — and be consistent, while remembering that the other one is not wrong. Then record your relatives. Family speech is irreplaceable primary source material for a language that has far fewer recordings than it deserves, and the recordings double as listening material at a level you can already understand.

The Identity Tax Nobody Mentions

One last finding deserves a place in the plan, because it does as much damage as any missing grammar rule. An ethnographic study published in the Canadian Modern Language Review in 2011 (ERIC EJ931220) followed Spanish heritage learners in Canadian high school classes and documented a language ideology in which speaking ability was treated as the measure of both proficiency and heritage. A student who was reluctant to speak in class was positioned in ways that ignored her actual literacy skills, and the article notes the wider assumption it exposes: that heritage learners can and will speak the language, and that this is their main asset in a classroom.

If it has ever been suggested to you — by a teacher, a relative, or a part of your own head — that you "only really speak English," or that your Spanish or Mandarin or Tagalog is not proper, that is the ideology talking, not the linguistics. The scholarship is unambiguous that you are a legitimate speaker of a language variety formed under specific conditions. What you are missing is a set of school-taught skills. Skills are learnable, and unlike the fluency you already have, they are learnable in months.

The Point of Closing the Gap

Heritage speakers are the largest reservoir of undocumented language ability in the world, and the profile is so consistent that it can be described as a checklist: native pronunciation, native listening, narrow vocabulary, missing orthography, missing formal register, unreliable grammatical gender or case marking, and a self-assessment that is usually far too low.

The research offers a specific, encouraging answer to the obvious question. Childhood exposure leaves a durable mark on pronunciation even when the language was never spoken, which is why relearning is not the same as starting over: the Au and Knightly studies show that the foundation is still load-bearing decades later. The literacy and the register, meanwhile, were always going to be classroom work for everyone, including monolinguals. Nobody in Tokyo learns two thousand characters by overhearing them.

So the plan is not to rebuild the language. It is to add the two tracks that a kitchen could never teach, one flashcard and one paragraph at a time, until the person who can talk to her grandmother and the person who can read the news are the same person in both columns.