If/had I needed to express a counterfactual

[Please note that this is a scheduled post, and I am taking part in the ongoing industrial action by my union, UCU.]

In a recent tweet, which I won't quote because the subject was too serious and grim for this light and fluffy discussion of linguistics, I saw the construction had/if. Like this: 

This is who I'd have spoken to had/if I needed help. 

It's like when you use and/or, to be clear that it's inclusive. If I write You'll enjoy this if you like fantasy and/or thrillers, then you'll like it if you like just one of those things or both of them. In other words, fantasy and thrillers, or fantasy, or thrillers. It just makes explicit the 'inclusive disjunction' (one or both are ok). 

But with had/if, there's no equivalent thing to make explicit. We can indeed use either of them in this sentence: This is who I'd have spoken to if I needed help and This is who I'd have spoken to had I needed help. But although had is a bit more formal, the meaning is the same, unlike with and/or. They both give a conditional interpretation. This made me wonder about the difference between them and after some struggle with finding out what the 'had' construnction was called (honestly it's hard to google linguistic things sometimes), I found Iatridou and Embick's 1994 article 'Conditional inversion'. 

It's called conditional inversion because it's used in conditionals, and the verb had is inverted with the subject: I had needed help vs Had I needed help. We use inversion a lot, including in questions (Had I needed help? Who can say.). 

So, if and inversion both indicate a conditional: 

If Sabine had eaten the calamari, she might be better by now.

Had Sabine eaten the calamari, she might be better by now. 

(Examples are all taken from Iatridou and Embick 1994, but with the name 'John' changed to 'Sabine' because John is in way too many linguistics example sentences.) 

Here's one difference between them, though. If can be used in situations where the thing might be true, and we just don't know the facts: 

If Sabine has eaten the calamari, there'll be none left for us. 

Inversion, meanwhile, can only be used in what are called counterfactuals, where we know the thing isn't the case, and you're talking about if it had been (but it isn't). In the equivalent example to the one just above, where it may or may not be true, it doesn't sound at all natural:  

*Has Sabine eaten the calamari, there'll be none left for us. 

(Remember the convention that an asterisk means that a sentence is not grammatical for speakers of the language under discussion.) 

And while if can be used with contracted negation (n't), inversion can't: 

If he hadn't seen the car coming, he would have been killed. 

*Hadn't he seen the car coming, he would have been killed.

Had he not seen the car coming, he would have been killed.

So there are some differences. But in the sentence I saw, where we began this, either would have been fine, because it was a counterfactual (she didn't need help), so inversion is ok, and if works in all conditionals so that would have been ok. So there's no reason to include both for covering more bases. It looks like one of those times when it feels like they mean something more than just using one of them, so you include both for the feeling of completeness. 

Perhaps I'm asking a question?

I've just finished reading a book called 'Statistics without tears' by Derek Rowntree. It's a basic tutorial on statistical concepts focussing on the ideas and principles, rather than walking through actual calculations in any detail. I found it useful and would recommend. But I'm here to talk about language, not statistics! 

The book is written in what I would describe as a very 'careful' style. You know it - the way older academic writers tend to, with quite precise attention to punctuation. Even though the tone of this book was very informal, friendly and not at all stuffy, I felt that every colon and dash was considered. 

So it was interesting to me that both times Rowntree used a sentence in the form Perhaps you recall..., he ended it with a question mark: Perhaps you recall the idea of a confidence interval? (p.183). I've had a quick look around the internet and can't find much on this topic other than a few sites peeving about the use of a question mark with perhaps, saying that it is not necessary and therefore wrong. There are people asking about it in English forums, indicating that it's something that might feel natural. 

It seems likely, then, that it's a 'declarative question' - the same as if he'd written You recall the idea of a confidence interval?. These are common enough, though definitely I would say a feature of less formal writing, just as contractions like I'll or don't are, which Rowntree also uses throughout. But it is interesting that he doesn't use this form - he uses perhaps. The question mark itself is enough to allow the reader to see that it's a question, and to therefore know that they are not being told that they do recall the idea, but rather prompted to agree that yes, they do recall that idea. So perhaps adds a bit more prompting, a bit more questioning, a bit more possibility of you not in fact recalling the idea of a confidence interval but that's absolutely fine because it was a few chapters ago and it's complicated stuff so don't worry. 

How many Millennials

Our topic today is the response to this tweet: 

Millennials! I have a question for you specifically!

How many of you have any sense or memory of the concept of a "rerun"? Do you have any meaningful recollection of an episode of TV you watched being such a thing?

— Daniel Summers, MD (@WFKARS) February 17, 2021

He clarified in the rest of his tweet thread what he meant by 'rerun' (tuning in to a weekly slot to find that the episode was one from a previous series), and that he was interested to know when exactly this ceased to be meaningful. He thought Gen Z probably don't really have much memory of this, but thought younger Millennials might, and so he asked 'how many of you [Millennials]'. 

The answer to a 'how many' question is a number or proportion (in this case, people reply 'me' or 'not me' and he gets a sense of the numbers), like '50%' or 'all of them'. There is definitely a presupposition that at least some Millennials will remember this concept. Exactly how many is mostly defined by pragmatics and shared context, I think. 

Compare asking a group of your friends the following: 

How many of you are coming to my comic book launch on Friday?

You would expect a fairly high number of them to come, because it's important and they're your friends. You could also ask your colleagues this question: 

How many of you are going to the meeting this afternoon?

That might be fairly neutral and you just want to know how many of them will be there, or not available at the alternative event at the same time, or whatever. You definitely imply that you think at least some of them will be going to the meeting. You could also ask a question like this:

How many of you learnt about grammar in English lessons at school?

Now, the answer is expected to be pretty low, and is a rhetorical device to lead into your lecture on the importance of grammar or the fact that they have tacit knowledge of grammar despite the lack of formal teaching, or whatever. If more than about one person answers 'me' now, it actually spoils your flow. 

So we can ask a 'how many' question with an expected answer of everyone, no one, or anything in between, or a neutral non-expectation. 

I now present you with a lesson in asking about Millennials on twitter: this tweet got 2,700 responses and a fair number of those were people who were pretty cross about this man supposedly thinking that Millennials are kids who don't remember this stuff when in fact they're in their 30s and maybe even 40 years old now, and of course they remember it. I've tried to think of any possible syntactic or semantic reason for the question being interpreted this way, and I can't. It's purely a result of the frequent co-occurrence of the term 'Millennial' with 'kids these days' messages and the resulting knee-jerk reaction to this that you get after the billionth time of seeing it. In this case, I think it was misplaced, but you can understand it. 

Can you buy macarons here?

Google sometimes asks me questions, via my phone, about places I've been to so that it can improve its information about businesses when people search for them. These are very revealing about what people search for, Google's insistence on not Britishising its communication, and the way that search data translates into queries.

Let's take the straightforward non-British terminology first. Google knows that these are places in the UK, and that I am a UK user. I've been annoyed before by how my phone's autocorrect only seems to know US places and spellings, and even the grammar is American (certain contractions that are more used in the UK don't come up in the suggested words, for instance). In its common questions are the following:

Is there a restroom here?
Is this place popular with travelers?
Is there a parking lot or structure?

The last one is specially for my friend and colleague Christina, who is very interested in who uses 'structure' to refer to what we (British people, I guess there might be variation but I don't know about it) would call a multi-storey (or underground) car park. A 'parking lot' is just a 'car park'.

This is lazy. It would be the easiest thing in the world for Google to translate 'restroom' to 'toilet'. It's not like it doesn't know. I included the 'travelers' one here too because I just felt that it struck slightly the wrong note. I think the word 'travellers' is too associated with the Traveller community (Gypsy or Romany people) and I would use 'tourists' in this context.

Some of the questions reveal differences in the kinds of places people might want to go. One of the questions asked me,

Does this place offer an all-you-can-drink option?

I think the last time I was somewhere with an all-you-can-drink option, with the exception of package holiday type places, was the Tuxedo Princess, a nightclub on a boat in Newcastle, where you paid a tenner entry and could drink rubbish alcopops all night. It's not really a thing in normal places. Similarly, another common question is this:

Does this place offer musical entertainment as well as dinner?

Apart from being a difficult question to answer because of the presupposition that it does offer dinner (how do you answer this about a place that offers musical entertainment but not dinner?), I don't think that dinner-plus-music is much of a thing. I certainly have never searched for it nor seen it offered as a tempting prospect anywhere except for a hideously old-fashioned restaurant that has since closed down.

Other questions, like the one above, pose problems because of the way they're asked. Take this one:

Does this place only serve vegetarian food? 

I was asked it about a pub that doesn't do food. So technically, it doesn't only serve vegetarian food, and the answer is 'no'. But that implies that it does do non-vegetarian food, as 'only' implies a sub-set of all kinds of food.

Then, take this one:

Can you drink outside on the street here? 

Well, at this pub that doesn't do food, no, you can't drink outside on the street. But it does have a really nice, big beer garden. Are people that ask this question just interested in drinking on the street, or do they more generally want to know if they can drink outdoors? That's another question, so I've been answering this literally, but I'm not sure it's the most useful thing to ask.

My favourite kind of question is the ones about weirdly specific products at random shops. I was asked of Morrison's (a low-end supermarket):

Can you get carpets here? 

And I was asked of Aldi (a cheap European supermarket):

Can you get macarons here? 

I... don't know. I suppose sometimes people ask where they can buy macarons or carpets and that translates, via algorithm, into a question for me.

The logical Cantonese

Post number 3 from The Leopard! What good value that book was.

Now, here's a nice example of linguistic FACT being used to propagate linguistic SILLINESS. Kaja asks Harry Hole a negative question, You don't take milk, do you?. There follows the kind of misunderstanding we have all the bleeding time in English, because English is ambiguous in its answers here: he answers Yes, meaning 'yes, that's true, I don't take milk', but she thinks he means 'yes, I do take milk'. The fact that English can even do this is unusual and interesting and a worthy object of study, but we'll leave it aside for now (I think I've written about it before anyway).

Languages differ with respect to how they confirm the truth of a negative question. English is confused and can answer yes (= yes, that's true, I don't take milk) or no (= no, I don't take milk). Most languages pick one or the other. Cantonese, which Hole is familiar with because he's been living in Hong Kong, confirms the truth with yes (= yes, that's true, I don't take milk). Kaja says, when Harry uses this strategy, that he's "stopped using double negatives". I assume by this she means that he doesn't echo the negation in the proposition, answering No, I don't take milk. This isn't what we normally mean by a double negative, but it's the way Norwegian (the original language of the book and the characters' language) confirms the truth of a negative proposition.

It's not more logical to do it the Cantonese way, though. I'm not sure why it would be considered to be so, and in any case, languages are not logical. They're messy and arbitrary(ish) and when there's two options, as in this case, they're just one or the other without one being more logical. If there was a good reason for doing it one way, all languages would do it that way.

I got all this information from the brilliant SSWL database, by the way, where you can read more about this and other syntactic facts to your heart's content.

That's not a dumb question, innit

In my MA syntax class this week we were talking about innit and how it's going from being a tag question to a kind of discourse particle.

Tag questions (aren't we?, don't you?, won't she?) have the following properties (among others):

1. They follow the main clause 
2. They have the form of a question
3. They contain an auxiliary verb that matches the one in the main clause or is do (are, do and will in the examples above)
4. The verb agrees in person and number with the main clause verb (e.g. 1st person plural, 2nd person singular or plural, and 3rd person singular in the examples above)
5. They have the opposite polarity to the main clause (so they're often negative, following a positive main clause, but can be positive if the main clause is negated: He's not coming, is he?) (though see NB below)
6. They have a pronoun that matches the main clause subject (we, you, she above)

Isn't it is a form of tag question, then: it has all the above properties in an example like It's a funny old world, isn't it?. But in some dialects, it has crept out of these constraints and is used in a broader range of contexts, and as well as losing a lot of its phonological properties (it's reduced to innit), some of the syntactic properties in 1-6 no longer apply to it.

1: It still follows the main clause. But that's about it.

2: It may not have the form of a question. Lots of times, it's written without a question mark (OK, that's no guarantee, but it's telling). It's attached to sentences that can't possibly be questioned, like assertions on the part of the speaker: You're fit, innit. That's the speaker's opinion. How can it be questioned?

3: The verb probably doesn't match the main clause one now - in the example I just gave, it does, but you can also say I've got no money, innit, where it doesn't.

4, 6: We no longer need the agreement in person and number. In We're late, innit, the verb in the main clause is first person plural. Innit, if it comes from isn't it, is 3rd person singular. Similarly, the pronoun is it rather than the matching one in the main clause, we.

5: Generally, the polarity is opposite to that in the main clause, but only by chance: innit is negative, and more sentences are positive than negative. But I heard an example on Gogglebox this week of a negative clause followed by supposedly negative innit:

That's not actually a dumb question, innit. 

Cool, innit?

NB: Tag questions can have the same polarity as the main clause if they're both positive, but it has a different meaning. Compare: 
a. That's the bus we need to get, isn't it? 
b. That's the bus we need to get, is it? 
In a, the speaker thinks they have the correct information and wants their interlocutor to confirm it. In b, the speaker thinks that the interlocutor has the relevant information, and wants to confirm it. But we don't find negative assertions and negative tags:
*That's not the bus we need to get, isn't it? 

Don't you not want some points?

On this week's episode of QI (available on the iPlayer for a few days if you're in the UK), Sandi Toksvig asked Alan Davies the following question:

Don't you not want some points or not? 

The point of QI, if you don't know, is that the questions are all kind of a trick. In this case, the trick is that it's hard to work out what the right answer is, and there's a penalty for a wrong answer.

This question brought about some linguistic chat, all of which ended with Sandi saying that only 'arbitrary pedants' would worry about double negatives, and explaining why the stigma came about in the first place. Gyles Brandreth also took the opportunity to tell us at length how English has half a million words and German only a hundred thousand (I guess it depends how you count them).

But the point here is the difficulty of the piled-up negatives, of course. Let's dispense with the easy one first: the or not at the end caused both Alan Davies and Victoria Coren to claim that it wasn't a yes or no question: it's two questions. WELL. This is technically true, it is two questions, but as I discuss at length in my thesis (and as is generally known; it's not my idea), this is actually one question because yes/no questions are a choice between two alternatives. So that's OK. We can in theory give a yes or no answer to this question. So which is the right one, assuming that you do want points?

WE DON'T KNOW! That's one of the cool things about this question and it's not even because of the double negative! It's because negative questions are inherently ambiguous! Look:

Don't you want some points?
No (I don't want some points)
Yes (??)

We don't always really have a clear sense of what 'yes' means in answer to a negative question. Probably, you feel that it means yes, I do want some points in this case. I do. But it's got to have the right intonation otherwise it doesn't mean anything much, though it can't mean that I don't want points, I don't think.

Do you not want some points?
No (I don't want some points)
Yes (??)

Again, it's not clear, without a sort of contradictory emphasis, what this means. There's tons of research on this which you can read if you're interested. Kramer & Rawlins use examples like the following to show that in fact, the interpretation is likely to be negative in such questions:

Is Alfonso not coming to the party?
No (=he isn't coming)
Yes (=he isn't coming)

So we've already got an issue with this kind of negation. Adding in a double negation, as in the original question, just adds parsing difficulty to the already ambiguous question. It is a real double negative, so it cancels out, as in I don't not like him... or is it double negation as found in many varieties, where one reinforces the other, as in I didn't do nothing or Didn't you not? We just don't know.

What, you thought I was going to answer this one? Nope. Unsolved problem, my friends.

(The 'correct' answer was yes, which is actually not what I'd have gone for: I'd have said it meant something like Isn't it the case that you don't want some points, or is that wrong?, so I'd have answered No, that's not the case. But maybe they went with yes as in Yes, that's true, it isn't the case that....)

Can you not?

I've written before about negation and the difference in meaning when it takes different scope. The other day, a friend in the pub said this, directed at a football 'pundit' on telly:

You're 60, can you not wear a suit and trainers? 

By this, he meant

Because you're 60 [and should know better/are too old], please don't wear a suit and trainers. 

It can have another meaning, though, and one that I think is much more accessible, or perhaps only available, in certain dialects:

You're 60, so is it impossible for you to wear a suit and trainers?

For me, both those readings work fine for the sentence and the intonation is pretty much the same in each case. The difference is once again a matter of how 'high' the negation is in the sentence:

Is it the case that
you can
[not wear a suit and trainers]?

vs

Is it not the case that
you can
[wear a suit and trainers]?

For more on this, a very easy to read article (old now, but still good) is Bob Ladd's 1981 paper.

If you could read this blog post

Sometimes, a request is expressed like this:

I'm snowed under, so if you can just book that room for me. Thanks! 

My colleague/friend/fellow twitterlinguist Damien Hall (@EvrydayLg) says that this has an implied 'that'd be good' after the if-clause. I think he's right about this, and the full form would be this:

If you can just book that room for me, that'd be good.

The if-clause is a conditional, and conditionals have to be subordinate to a main clause. So technically, this is a fragment: it's a subordinate clause set adrift, with no main clause to modify the meaning of. But we see it a lot. I do it myself, because it's a very convenient way of making a request in writing.

Written requests have become a problem, if they are from one person to another on the same level. You have to be polite and grateful, and one by one, our ways of doing it have become too abrupt. A plain request would be Please book that room for me. This is way too imperative, so we have questions. Can you book that room for me? is not technically a request: it's a question, and the askee could say I'm sorry, I'm not able to without having to say I won't. But then we feel we need to add please, and then could seems to sound more polite, and we get Could you please book that room, and then we add hedges like just as in the first example... and so on. Eventually you sit at your computer rephrasing it over and over but it all sounds like you're ordering the person around. Right now, to me, this 'unfinished conditional' construction seems to be non-imperative. It's a simple statement of fact: I'm not asking you to do anything, but if you were to happen to do it anyway, that'd be grand.

But here's why I love it most of all: it looks just like the way you ask a question in some other languages. We aren't really using it to ask a question in English: conditional-if is different from question-if, and they just happen to have the same form. But if it was question-if, look at this question in Polish (source):

Czy Basia ma kota?
'Does Basia have a cat?' (literally: 'If Basia has (a) cat')

Here, the same word czy that is used for if in an embedded question (as in English I wonder if Basia has a cat) is used for a main clause question and it looks really like the English unfinished conditionals.

We said that the English one is a conditional, not a question, and I think this is right. But imagine that it was a question: If you can book that room for me means Can you book that room for me? and we have what is apparently exactly the same syntax as Polish. (I'm not claiming it is the same - just noting the parallel.)

Yes or no what?

I was sort of half-heartedly watching ITV's 'This Morning' programme while I was working yesterday (honest) and they were talking about vigilantes. They had one of those viewer polls, where viewers have to text or tweet their opinion about an issue and either the result is an equal split or it confirms what we already knew. 

This was the question put to viewers about this 'issue': 

Perfectly reasonable question. Two opposing points of view to choose between: either vigilantes have a valid role (in the process of catching criminals) or the job of catching criminals is one that should be left to the police alone. Choose the position you agree with and tell ITV. 

But here's how the results were presented: 

Well, yes or no what? Yes, either vigilantes have a valid role or catching criminals should be left to the police? And no, neither of the above apply? 

Clearly not. Obviously, what they meant was yes or no to the first part of the question (do vigilantes have a valid role), as this clearer presentation shows:

Probably no one except me was confused by the wording. I've spent more time than most people thinking about questions over the last few years, though, so I noticed this strange construction. 

The question was presented as if it was an 'alternative question': one where the answer is one of the alternatives that are mentioned in the question. More everyday examples tend to have smaller parts of the sentence given as alternatives, as in Are you coming for lunch or tea? The answer is either 'lunch' or 'tea'. The question from ITV had two whole questions as the options: Do vigilantes have a valid role? and should catching criminals be left to the police?. 

This kind of alternative question, as far as I can tell, is hardly ever used as a real alternative question (as the lunch or tea example is). More often, the second part isn't a real alternative, and you aren't meant to pick one of them as the answer. Instead, the question is really a 'yes/no' question, or one where the answer is either yes or no. You can't answer yes or no to a real alternative question (where # means that the utterance is strange in some way): 

Did you buy that suit or hire it? #Yes.

But in the poll above, and in most cases of apparently alternative questions with two entire questions as the 'alternatives', the answer can be yes or no, showing that they must really be yes/no questions. But! You can also answer by providing one of the alternatives. This is sort of the case for yes/no questions as well, in a trivial sense which I won't go into here, but it's different with these strange halfway things.

This was actually pointed out to me pretty early on in my PhD (by Sten Vikner), who gave an example like Have you done the washing up or have you just been sitting in front of the TV all day?. The question provides two alternatives, and you can answer it by providing one of them as the answer, but it's really just asking Have you done the washing up? and it's more natural to answer with yes or no. I didn't really address this problem then, but perhaps I ought to now.

In which continent is Honolulu?

In linguistics papers, we often need to give examples of some linguistic construction or other, and so we use example sentences. It didn't strike me as at all odd until I talked to other people about linguistics, and they found it hilarious. Still. It's what we do. What I mean by this is that, say I'm writing about questions, and I need to illustrate what questions look like in various languages, I'll just give an example of a question. It doesn't matter what question, usually; it just has to be a normal question with nothing syntactically weird going on. So I might give these two examples:

Cantonese:
neih johng hohk sahn ma
‘Have you  rushed into the crane god, eh?’ (Morrison 1828, cited in Cheung 2001: 223)

Mizo:
vok i2-­‐n vulʔ doon1 em2
‘Are you going to raise pigs?’ (Chhangte 1989: 162)

I can generally be relied on to pick the silliest examples I can find, so my thesis contained lots of examples involving trousers being covered in bitumen, and expressions with rude words in them. You don't always have a lot of choice, if you don't speak the language, but if you're working on a language you speak, you can make up good examples. I despair of the number of papers I read in which John lends Mary yet another book. In my constructed examples, all sorts of things happen: gorillas get seasick, Poirot solves murders, Richard Burton is a post-mortem actor, and we have to choose whether we would rather be able to fly or be invisible. (You have to make sure you don't go too far, though. The Generative Semantics movement of the 1970s got a bit silly in this respect.)

Reading examples in other languages can be informative: I learn the words for all sorts of things in other languages, for instance. Did you know that the Cantonese for 'Christmas' is literally 'Jesus birthday'? And the Swahili for 'lion' is 'simba'?

But today I was reading a paper and was baffled. It's Hamblin's classic 1958 paper on questions. In it, he discusses how some questions include a presupposition. The (in)famous example of this is 'Have you stopped beating your wife?', which the defendant cannot answer 'yes' or 'no', because either answer presupposes that he has, in the past, beaten his wife. Hamblin notes that a question like 'In which continent is Luxembourg?' also contains a presupposition (that Luxembourg is in a continent), but that it is unimportant because the presupposition is true. He gives this question as the equivalent example where the same presupposition would be untrue: 'In which continent is Honolulu?'. You cannot, says Hamblin, answer by simply saying what continent Honolulu is in, because it is not in a continent at all.

Clearly, this is now false because Honolulu is in Hawai'i, and Hawai'i is in the United States, and the US is in America (or North America - do we class North and South as separate continents these days or not?). In 1958, when this paper was written, Hawai'i was not yet a State: it became one in 1959. But it was a Territory, and so, I thought to myself, surely still considered to be in the continent of (North) America?

Anyone any idea?

Categorisation

Sometimes, someone asks you a question that you wouldn't even have thought of.

I attended a training event on Tuesday, and the other people there were from various humanities and social science disciplines, none of them linguistics. We each had to present a talk on our research, and then there were questions afterwards. One woman, doing a fine art creative practice PhD (in ceramics), asked me this:

Is it important that everything is put into one category or another?

Typical creative type, you might think, not wanting things to be 'labelled'. She meant the labels like 'noun', 'verb', 'question particle' and so on. And it's a perfectly reasonable question to ask. Why do things have to be categorised? Without the categories, there's no problem to have to be solved.

Well, it's not just linguists doing it for the sake of it, because we like things to be rigid and ordered and follow the rules (though some of us do). These categories, although the labels are artificial, are natural classes. Consider the birds (as Brian didn't quite say).

A goldfinch is a natural thing, I think we can all agree. It occurs in nature without human intervention (all right, it's been captured, bred and sometimes escaped, and introduced to various bits of the world, but it wasn't created by humans).

This goldfinch is called Harold. That category (Harold) consists of just this one individual. Goldfinches generally, though, have a specific scientific classification: Carduelis carduelis. That's its species. Then we can identify the category of passerines, the larger category of birds, and the even larger category of animals. There are other categories in between too, each of which Harold is a member of.

The point is, although the names for these categories were chosen and imposed by humans, the categories themselves were not. Humans observed similarities between organisms, and those that share characteristics form a natural class (for example, birds, which have feathers and wings, or living things, which have the characteristics of MRS NERG - movement, reproduction, sensitivity, nutrition, excretion, respiration, growth). Everything in nature is categorisable, and not because scientists have imposed those categories, but because they really are a category.

In the study of language, we do the same. We say that all nouns are nouns because they behave like nouns, they share all the characteristics of nouns, and they are different from things that are not nouns. We called them nouns, but the category was there before we came along and noticed it and named it.

I didn't give the questioner an answer quite as full as this because I couldn't think on the spot. But I hope I got across to her the message that we're not just making all this up - language is a real, natural object. We're just describing it and trying to explain how it is the way it is.