Of all the areas this magazine covers, bilingualism is the one where the gap between what is known and what is repeated is widest. The research base is large, international and reasonably consistent. The advice given to families is frequently the opposite of it, and is sometimes delivered with great authority by people who have not read any of it.
So let us separate the durable findings from the folklore.
The counting artefact that produces most of the alarm
Start with the single most important technical point, because nearly every myth downstream of it dissolves once you have it.
Vocabulary in bilingual children is distributed across their languages. Some words a child knows in both. Many they know in only one, usually the one associated with the setting where that thing happens. The words for bath time may live in one language and the words for nursery in another.
Now measure that child with a standard vocabulary checklist in a single language. You will count only part of what they know. Compare the result to a monolingual average and the child looks behind. Count both languages and the total is typically comparable, sometimes larger.
This is not a subtlety. It is the origin of a very large fraction of the concern that bilingual families are subjected to, and it is a measurement problem rather than a child problem.
Mixing languages is a skill, not a symptom
Bilingual speakers of every age move between their languages, sometimes within a single sentence. Adults do it. Fluent, highly educated, entirely unconfused adults do it constantly.
The reason linguists do not regard it as confusion is that it is rule governed. Switches happen at particular structural points and not at others, and speakers agree with one another about which switches sound acceptable. Confusion does not produce shared grammatical constraints.
Children mix for the ordinary reasons anyone mixes: the word they want is more available in one language, or the topic belongs to one language, or the person they are speaking to shares both. It is a solution, not a failure.
The one parent one language rule, and where it came from
Many families are told to enforce strict separation: each adult uses only one language, without deviation. It is presented as established practice.
It is a reasonable strategy and it is not a requirement. Children acquire languages successfully under many arrangements: one language at home and another outside, different languages for different activities, grandparents supplying one language entirely. What appears to matter is that each language is available in real, interactive, meaningful quantity from people the child cares about.
The stricter versions of the rule can also do harm, because they push families towards using a language the adult is less comfortable in, which tends to reduce exactly the responsive, playful, elaborate interaction that language development leans on. A parent operating in their strongest language is usually a better conversational partner, and conversational partnership is the mechanism.
The advice to drop the home language
This is the one worth stating flatly. Families are still told, sometimes by professionals, that a child who is finding language hard should stop hearing the home language and switch to English only.
There is no good evidence that this helps, and there are reasons to think it can hurt. It removes the language in which the adults are most fluent and expressive. It cuts the child off from extended family. And where a child has an underlying language difficulty, that difficulty will show up in both languages, because it is a property of the child rather than of the language, so removing one language does not remove it.
Professional guidance from bodies such as the Royal College of Speech and Language Therapists reflects this. If a family is given the opposite advice, it is entirely reasonable to ask what it is based on.
The cognitive advantage question, handled carefully
You will have read that bilingualism confers general cognitive advantages, particularly in executive function, the family of abilities involved in holding a goal in mind and ignoring distraction. The proposed mechanism is plausible: a bilingual speaker constantly manages two active systems.
The evidence here has had a hard decade. Early studies reported clear advantages. Later, larger and better controlled studies have often found smaller effects or none, and there is good reason to think the published record was shaped by which results got written up. This is a textbook case of what happens when a field re examines itself, and we have written about the general phenomenon in replication in plain words.
The sensible position today: the claim that bilingualism causes broad cognitive superiority is not well supported. The claim that it causes harm is not supported either, and it is the claim that families are actually pressured by. The reasons to keep a home language are family, culture, and the ability to speak to your grandmother. Those have never needed a cognitive bonus to justify them.
What changes when a child starts school in English
Children who arrive at school with a different home language typically acquire conversational fluency in the school language faster than they acquire the more formal, decontextualised language that academic work runs on. Those are different things and they move at different speeds.
The practical consequence is that a child can sound entirely fluent in the playground while still finding the language of instruction hard, and the fluency can cause the difficulty to be missed. Schools that handle this well know to look past conversational ease. Guidance and evidence summaries from the Education Endowment Foundation address the teaching side of this in some detail.
What a bilingual family can reasonably conclude
That the arrangement they have is very probably fine. That an appearance of delay should prompt a proper look at both languages together before anything else. That mixing is not a problem to be stamped out. That dropping a language on the advice of somebody who has not explained their reasoning is not required of them.
And that if there is genuine concern about a particular child, the assessment should be of the child's language as a whole, not of one of their languages in isolation. For the wider picture of how variable ordinary development is, see the range of normal.