The other day, I heard a man say to his young daughter: ‘Where are you putting that? Put it in a place that you know where it is.’
Although that 2nd sentence is perfectly clear, I would never write it myself. I probably wouldn’t say it either, though I quite often hear similar sentences.
Why wouldn’t I write or say that sentence?
The thing that struck me as unnatural was using that in that 2nd sentence. I would generally say ‘a place where you know where it is.’
In many contexts, I would happily use that to introduce a relative clause modifying a place. Whether that is possible depends on the relationship between the main clause (eg ‘put it in a place’) and the subordinate clause. That is fine, where for example, that relates to the implied subject of the subordinate clause (example 1, implied subject of ‘sense’) or to the subordinate clause’s implied direct object (example 2, implied object of ‘see’) or implied indirect object (example 3, implied indirect object of ‘get’).
(1) a place that makes sense.
(2) a place that you can see.
(3) a place that you can get to.
In fact, where wouldn’t even be possible in examples 1-3.
(In example 2, where is possible, but only with the meaning ‘a place where you can see [from]’.)
In the sentence I heard, that refers to something that is not the subject, object or indirect object of the subordinate clause. The meaning is something like ‘Put it in such a place that you know it is in that place.’ In the version of English that I speak and write, that is simply not enough by itself to carry the meaning ‘in such a place that …. in that place’.
Different version or performance error?
The sentence I heard made me think more generally about the distinction between different varieties of a language and performance errors within a single variety.
Different varieties
We often talk about languages as if each one were a single homogeneous entity. In reality, any language is an intersection of varieties that largely overlap (but do not fully coincide in all respects). Such varieties include:
- a standard language (if there is one), for example one codified by a nation state (or states).
- dialect: a general term for a variety (often excluding the standard language (if any), though it is perhaps better to regard the standard language itself as one dialect that happens to carry more prestige than other dialects.
- regiolect: a language variety used in a region (often a region of a country that also possesses a standard language).
- sociolect: a language variety used by one sector of society, for example people from a particular social class or educational background, people with a particular occupation, interests, pastimes or hobbies, or people of a particular age, birth cohort, gender, ethnicity or 1st language.
- idiolect: the language variety used by a particular individual, influenced (but not determined fully) by the standard language(s), dialect(s), regiolect(s) and sociolect(s) they are exposed to.
For more on language varieties, please see https://languagemiscellany.com/2026/04/language-or-dialect/
Performance errors
Generative linguists often distinguish:
- a speaker’s competence: the utterances the speaker is capable of generating under ideal conditions.
- a speaker’s performance: the utterances a speaker actually produces on particular occasions. Performance is often affected by factors such as tiredness, interruptions, distractions, background noise, lapses of concentration or memory limitations.
Generative linguists (particularly, but not only, those working in the Chomskyan tradition) often focus on competence rather than performance.
On the other hand, corpus linguists tend to focus on a corpus of actual utterances, which are instances of performance. Such linguists often have less interest in ‘competence’. They regard a corpus of actual utterances as evidence of how a language is actually used. They view a corpus as better evidence than the language as described in normative textbooks and prescriptive usage guides.
There is a danger in just using a corpus of actual utterances as evidence of how a language works. Any real corpus will undoubtedly contain an unknown proportion of obvious performance errors (eg typos or mis-statements caused by interruptions or by inattention). But trying to remove performance errors may induce bias if the person constructing the database misidentifies effects of other varieties of the language.
Was this instance a performance error?
As I have said, I would not personally have used that in the sentence I heard. To my ear, standard British English does not use that rather than where in that context.
I don’t know why the man used this form:
- perhaps he speaks a different variety of English in which that is permitted or even required there.
- perhaps he thought it would be inelegant to use where twice in quick succession (which my preferred sentence does do). I find this possible explanation the least convincing.
- perhaps he was speaking standard British English, but committed a performance error. A plausible explanation for many performance errors is starting to utter a sentence with one construction in mind, but then getting part way through and either:
– losing track of the intended construction; or
– realising that it is impossible to complete the sentence using the original structure.
On balance, I think the most likely explanation is a performance error, but it is also possible that the man speaks a variety of English in which that is acceptable in such constructions.
Could it be that the that-clause is purposive: ‘Put it in a place [so/such/in order] that you know where it is.’?
An interesting suggestion, which I hadn’t thought of.
To my ear, just ‘that’ by itself wouldn’t naturally carry that purposive meaning. The purposive meaning would be possible in ‘such a place that’, though such constructions also sound formal and very stilted—and I couldn’t imagine using them in talking to a small child.