Pronouns as a part of speech

[Zju]

We all know that pronouns - you, us, this, some, none, that - are one of the parts of speech. But taking in consideration that personal pronoun have the additional categories of person and number that others don't*, are they really the same part of speech? A quick glance at wikipedia shows that it makes the same point:

Pronouns have traditionally been regarded as one of the parts of speech, but some modern theorists would not consider them to form a single class, in view of the variety of functions they perform cross-linguistically. An example of a pronoun is "you", which is both plural and singular.

*in some languages some of the other pronouns do have the category of number, but that's beside the point

So, if personal pronouns are their own thing, would the rest then make one cohesive part of speech, whose members all share the same categories, defective words notwithstanding? It's more or less the same thing called elsewhere correlatives, and that PoS would have the single category of relation, with grammemes such as proximal demonstrative, negative, indefinite, etc.
But even then some pronouns usually stand out, such as reciprocal and reflexive pronouns.

How to best analyse the pronoun mess? It's got to be on a language-by-language basis, but even within a single language, can a point be made that all pronouns form a single part of speech despite them having different categories, or is that an outdated view?

[KathTheDragon]

The most straightforward solution I know of is to dispense with strictly separate parts of speech as a concept, and instead work with a nested hierarchy of lexical categories. So you'll typically have one category of all nominals, of which personal pronouns and non-personal pronouns are (potentially the same, potentially different) subcategories somewhere down the hierarchy. The exact form of the hierarchy will of course depend very much on the language in question.

[Zju]

So is the difference between these lexical categories and traditional parts of speech just that lexical categories can be nested, and that a lexical subcategory can lack or have additional morphosyntactical categories?

[Pabappa]

Pronouns are a single part of speech. That part of speech has its own set of rules, unlike those of nouns. Thus a feature that would require breaking nouns into separate classes does not say anything against pronouns being a single class. That's what makes sense to me.

[zompist]

This is kind of a can of worms.

At a high level, pronouns do form a nice category: they are words that form an entire NP. They are always deictic, taking a meaning from the context, and not descriptive. This covers not only personal pronouns but interrogatives, demonstratives, and indefinite pronouns.

This does gloss over one problem: many of these words don't replace NPs at all, but modifiers. That is, they're pro-adjectives, not pro-nouns. But you generally want to learn them and discuss them at the same time, so the traditional grouping works.

In syntax, we really want to talk about anaphors, as any part of speech can have this sort of replacement word. E.g. there are verbal anaphors.

In some languages, like Japanese, pronouns overlap with nouns in a way that's hard to disentangle. This is often presented in a way that's needlessly exoticizing, IMHO: there are plenty of European equivalents to Japanese usage. But Japanese (and also Portuguese) force us to give up the idea that pronouns should be a limited class and etymologically non-transparent.

[Kuchigakatai]

The most straightforward solution I know of is to dispense with strictly separate parts of speech as a concept, and instead work with a nested hierarchy of lexical categories.

So is the difference between these lexical categories and traditional parts of speech just that lexical categories can be nested, and that a lexical subcategory can lack or have additional morphosyntactical categories?

I don't know if that's a wholly fair assessment of how traditional parts of speech have been modelled either... Donatus' Ars Minor (4th century AD) makes a primary distinction among partēs ōrātiōnis (parts of speech) separating nōmina (nouns/adjectives), verba, adverbia, conjunctiōnēs, etc. And then when dealing with nōmina, also between nōmina propria (nouns) and nōmina appellātīva (adjectives). And even this seems mostly independent from other distinctions of nōmina, like what gender they have or are inflected for (regardless if they're nouns or adjectives), although he also specifies that only the nōmina appellātīva (adjectives) can be positive/comparative/superlative.

Relevantly for this thread, Donatus also distinguishes prōnōmina fīnīta (personal pronouns) and prōnōmina īnfīnīta (other pronouns).

There's the simplistic parts of speech your high school teacher told you about, and then there's the things grammarians have actually been using... Wikipedia makes it look like it's a thing of "modern theorists" (and I don't know if the cited works portray today's linguists as going against the grain), but it is not.

[bradrn]

Bhat’s Pronouns starts with a summary of this issue. Based mostly on functional considerations, he concludes that ‘personal pronouns’ on the one hand and ‘proforms’ on the other make up separate categories, with third person pronouns belonging to different categories depending on the language. Given the obvious differences between the two groups, I am inclined to agree.

The most straightforward solution I know of is to dispense with strictly separate parts of speech as a concept, and instead work with a nested hierarchy of lexical categories. So you'll typically have one category of all nominals, of which personal pronouns and non-personal pronouns are (potentially the same, potentially different) subcategories somewhere down the hierarchy. The exact form of the hierarchy will of course depend very much on the language in question.

Personally, as I’ve mentioned before, my favourite approach to parts of speech is François’s: look at the syntactic (and morphological, I suppose) positions in which each word can occur, then define word classes such that each class can occur in different environments. This has the advantage of allowing the rigorous definition of parts of speech, relying only on emic facts. While François never mentions a hierarchy of word classes, this method still allows allows us to define a hierarchy (something which I agree is a good idea): if there happen to be several word classes which all occur in similar environments, it makes sense to group them as ‘nominals’ or ‘pronouns’ or what have you, with all the finer-grained word classes as subclasses of that larger class.

There are admittedly some problems with defining word classes purely semantically. Most prominently, there are some word classes which are typically defined semantically (‘interrogative pronouns’ are a standard example, but also ‘pronouns’ more generally) or even phonologically (the usual case with ‘ideophones’). I personally see the syntactic definitions as primary, with the other distinctions cross-cutting the syntactic word classes.

At a high level, pronouns do form a nice category: they are words that form an entire NP.

Only in English! Lots of languages (perhaps the majority) allow a single common noun to act as an entire NP, just like pronouns do.

They are always deictic, taking a meaning from the context, and not descriptive. This covers not only personal pronouns but interrogatives, demonstratives, and indefinite pronouns.

Bhat notes that this definition is problematic:

It is possible, for example, to regard a general term like human as standing for several more specific terms like man, woman, boy, girl, etc. In what sense do pronouns stand for nouns and these general terms do not? … On the other hand, the notion of 'standing for' something else is completely unsuitable for characterizing first and second person pronouns. This is evident from the fact that any other noun or pronoun that we try to use instead of these pronouns would fail to provide the crucial kind of meanings that they are meant to denote.

In some languages, like Japanese, pronouns overlap with nouns in a way that's hard to disentangle. This is often presented in a way that's needlessly exoticizing, IMHO: there are plenty of European equivalents to Japanese usage. But Japanese (and also Portuguese) force us to give up the idea that pronouns should be a limited class and etymologically non-transparent.

Also Chalcatongo Mixtec (though only for third person; first and second persons have more normal pronouns). And then there are languages which have no pronouns and rely entirely on verbal cross-referencing! (Strangely enough, I can’t find any examples, but I’m sure I read something about such systems a while ago.)

[zompist]

At a high level, pronouns do form a nice category: they are words that form an entire NP.

Only in English! Lots of languages (perhaps the majority) allow a single common noun to act as an entire NP, just like pronouns do.

That's not what I meant. Of course an N can be an NP: "Lawyers love criminals."

What I meant was that a pronoun occupies an NP slot, without modification. A noun can sometimes do this, but can be modified ("Greedy lawyers..."). It's important to see that pronouns are not anaphoric nouns, but anaphoric NPs.

They are always deictic, taking a meaning from the context, and not descriptive. This covers not only personal pronouns but interrogatives, demonstratives, and indefinite pronouns.

Bhat notes that this definition is problematic:

It is possible, for example, to regard a general term like human as standing for several more specific terms like man, woman, boy, girl, etc. In what sense do pronouns stand for nouns and these general terms do not? … On the other hand, the notion of 'standing for' something else is completely unsuitable for characterizing first and second person pronouns. This is evident from the fact that any other noun or pronoun that we try to use instead of these pronouns would fail to provide the crucial kind of meanings that they are meant to denote.

I didn't say "standing for" and agree that the concept is problematic. The idea probably goes back to classical grammarians' notion of an "antecedent", which as Bhat says never works for first and second person.

[vegfarandi]

As far as the question here, thinking in terms of part of speech and describing such labels as cohesive lands you in murky waters. Mostly nothing in language is fully cohesive, most edges are blurry and fuzzy.

I've been reading The Texture of the Lexicon and although it doesn't touch on this particular issue it presents a new framework to think about how words interrelate in a complex yet neat way. I highly recommend it.

What's called "pronouns" is the sort of thing Zompist describes above. A noun-ish thing that usually doesn't have modifiers and receives its semantics from context rather than inherently having much semantic value. But there's very generic nouns that are similar. "People" has very little semantic value and in many respects is difficult to discern from "they". In some languages, the differences between the equivalents of "people" and "they" may be greater or lesser. In Icelandic, for example, "known specific" referents cannot be referred to by an indefinite pronouns. In English, a sentence like "someone came to see you" – someone can either refer to a known specific referent (someone was Tommy, your husband) or an unknown specific referent (I don't know who it was). In Icelandic, "einhver kom að sjá þig" can only mean the latter, where the speaker doesn't know who the person is. The other type of sentence requires a generic noun and a dummy subject "það kom maður að sjá þig" (lit. it came a man to see you). So generic nouns are a little more pronounish in this instance in Icelandic than in English.

Some linguists seem to only refer to personal pronouns as pronouns, while others would include interrogtive pronouns within the category. I disagree with this practice. Even though this may have some merit, it breaks away from centuries of tradition and seeks to redefine a very commonly understood term that most elementary school graduates at least have heard of. They should really find another word. (And, in fact, some people have, and use "locuphoric pronoun" to refer to 1st and 2nd person pronouns, i.e. personal pronouns referring to speech-act participants. "Aliphoric pronouns" then refer to non-speech-act participants.)

Essentially, it's best to think of "pronoun" as an umbrella term for a variety of words which have a variety of specific uses and usage rules. From a cross-lingustic point of view, it's practically meaningless. Yet it's one of those things, despite not having a satisfying definition, we all kind of know kind of what it means, and so it remains usefuly. Ultimately, the specifics are always going to be language-specific and not generalizable to a great degree.

[KathTheDragon]

So is the difference between these lexical categories and traditional parts of speech just that lexical categories can be nested, and that a lexical subcategory can lack or have additional morphosyntactical categories?

More or less, but not thinking in terms of traditional parts of speech opens you up to splitting up the supercategories far more readily, and also helps steer you away from the UG position that parts of speech are somehow fixed, innate, and universal, none of which are really true.

Personally, as I’ve mentioned before, my favourite approach to parts of speech is François’s: look at the syntactic (and morphological, I suppose) positions in which each word can occur, then define word classes such that each class can occur in different environments. This has the advantage of allowing the rigorous definition of parts of speech, relying only on emic facts. While François never mentions a hierarchy of word classes, this method still allows allows us to define a hierarchy (something which I agree is a good idea): if there happen to be several word classes which all occur in similar environments, it makes sense to group them as ‘nominals’ or ‘pronouns’ or what have you, with all the finer-grained word classes as subclasses of that larger class.

That's pretty much the approach I meant, actually, with the addition as you observe of grouping together similar classes into a larger supercategory (and allowing for a non-tree-like hierarchy - it's entirely possible for word classes to share properties with two otherwise entirely unlike each other classes.

[bradrn]

At a high level, pronouns do form a nice category: they are words that form an entire NP.

Only in English! Lots of languages (perhaps the majority) allow a single common noun to act as an entire NP, just like pronouns do.

That's not what I meant. Of course an N can be an NP: "Lawyers love criminals."

What I meant was that a pronoun occupies an NP slot, without modification. A noun can sometimes do this, but can be modified ("Greedy lawyers..."). It's important to see that pronouns are not anaphoric nouns, but anaphoric NPs.

But pronouns can be modified! It’s not even that uncommon; e.g. “That's not the me you know” is a fairly straightforward sentence with two modifiers on the pronoun. We had a whole discussion about this earlier. (I’m sure I also saw a French example in one of your books, though I don’t feel like finding it just right now.)

They are always deictic, taking a meaning from the context, and not descriptive. This covers not only personal pronouns but interrogatives, demonstratives, and indefinite pronouns.

Bhat notes that this definition is problematic:

It is possible, for example, to regard a general term like human as standing for several more specific terms like man, woman, boy, girl, etc. In what sense do pronouns stand for nouns and these general terms do not? … On the other hand, the notion of 'standing for' something else is completely unsuitable for characterizing first and second person pronouns. This is evident from the fact that any other noun or pronoun that we try to use instead of these pronouns would fail to provide the crucial kind of meanings that they are meant to denote.

I didn't say "standing for" and agree that the concept is problematic. The idea probably goes back to classical grammarians' notion of an "antecedent", which as Bhat says never works for first and second person.

So what exactly is the difference between ‘standing for’ and ‘deictic’? (I’m genuinely curious about the difference, I’m not just being argumentative here.)

As far as the question here, thinking in terms of part of speech and describing such labels as cohesive lands you in murky waters. Mostly nothing in language is fully cohesive, most edges are blurry and fuzzy.

In general, I completely agree with this sentiment. But I’d say that parts of speech are one of the few areas of language which can be demarcated more precisely! It’s not too difficult to look at a bunch of words, see which environments they can be used grammatically in, and on that basis divide them up into parts of speech. The problems come when you start defining your classes using inconsistent criteria: whereas other classes are defined on the basis of morphosyntax, pronouns are traditionally defined on the basis of their semantics and pragmatics, so it’s not surprising that there’s some overlap between the groups.

[vegfarandi]

As far as the question here, thinking in terms of part of speech and describing such labels as cohesive lands you in murky waters. Mostly nothing in language is fully cohesive, most edges are blurry and fuzzy.

In general, I completely agree with this sentiment. But I’d say that parts of speech are one of the few areas of language which can be demarcated more precisely! It’s not too difficult to look at a bunch of words, see which environments they can be used grammatically in, and on that basis divide them up into parts of speech. The problems come when you start defining your classes using inconsistent criteria: whereas other classes are defined on the basis of morphosyntax, pronouns are traditionally defined on the basis of their semantics and pragmatics, so it’s not surprising that there’s some overlap between the groups.

Yes, you can within the grammar of each language, demarcate each type of pronoun and probably the super-group of pronouns in general. But there will always be differences between the rule surrounding personal pronouns vs. indefinite pronouns vs. interrogative pronouns – each of those, in a sense, is a "part of speech". And really, in many languages there are significant enough differences between just the first person pronoun(s) and the second person pronoun(s) that it becomes arguable whether each constitutes a part of speech on its own.

[vegfarandi]

Personally, as I’ve mentioned before, my favourite approach to parts of speech is François’s: look at the syntactic (and morphological, I suppose) positions in which each word can occur, then define word classes such that each class can occur in different environments. This has the advantage of allowing the rigorous definition of parts of speech, relying only on emic facts. While François never mentions a hierarchy of word classes, this method still allows allows us to define a hierarchy (something which I agree is a good idea): if there happen to be several word classes which all occur in similar environments, it makes sense to group them as ‘nominals’ or ‘pronouns’ or what have you, with all the finer-grained word classes as subclasses of that larger class.

That's pretty much the approach I meant, actually, with the addition as you observe of grouping together similar classes into a larger supercategory (and allowing for a non-tree-like hierarchy - it's entirely possible for word classes to share properties with two otherwise entirely unlike each other classes.

Yes – there can be horizontal overlaps. A common thing is for adjectives to have things in common with verbs and nouns, for example.

[zompist]

So what exactly is the difference between ‘standing for’ and ‘deictic’? (I’m genuinely curious about the difference, I’m not just being argumentative here.)

Deictic means "pointing to"; it's used for words that have reference only in relation to who is speaking, and when and where they are located. They're defined pragmatically, not semantically.

(Naturally, there are NPs whose referents are pragmatically determined, like "my father". But the meaning of "father" can be specified independent of context.)

(And because language categories are fuzzy, there are pronouns that convey some semantic information— e.g. "she" indicates "female". IIRC one of the SE Asian languages has a pronoun meaning "I" but which can only be used by the emperor.)