- Rules are descriptions of what can happen in systems (including languages)
- Descriptions may be prescriptions, when used to describe some ideal state
The first point is more of a definition than a claim, but I nonetheless call it a "claim" because it may run against common usage of what a "rule" is. The second point directly addresses an issue linguists have made for themselves that you are likely to find in any introductory linguistics textbooks.
The problem with rules
Let's begin with how we typically think of "rules". The common sense notion of rules where I come from are commandments about what you can't do. The "grammar police" may issue commandments such as:
- Thou shalt not begin a sentence with the word 'but'?
- Thou shall not end a sentence in a preposition
- Thou shalt not confuse shall and shalt
Conventional linguists are quick to point out that these prescriptive rules about language use aren't so much directly about how language is actually used, but rather are stylistic imperatives. Good writers, I hear, often break such rules. So the linguists laugh, hahahaha the reign of the grammar police dies with descriptivism...
The clever linguist points at the very reason we have to have a "rule" like thou shalt not begin a sentence with the word 'but' is for the very reason that people DO this. So, rather than making up empty injunctions about how people may or may not use language, we ought to describe how people use language. So, turning the previously given commandments around, we get:
- English speakers may begin sentences with 'but'.
- English speakers may end sentences with prepositions.
- English speakers confuse shall and shalt.
Now, where the clever linguist went astray is making a normative claim against normative claims; the ought not state ought's but rather ought to state is's is an ought. Rules, according to many linguists ought to be descriptive. Why is this so? (see the first part of this series for more info on that 😎) If "rules" are just supposed to be descriptive, why even call them rules? Why not call them something else, like... ...try "descriptions" or "formulas" or something like that?
Things are getting unruly...
Many of the ideas just presented were first put forth about the confusion over "rules" in the context of linguistics by Roy Harris in various books and articles. Linguists, unfortunately, have gone down in social science history as being especially toxic to one another in many matters. This topic has been explored in many volumes, such as a book by another fellow named Harris in The Linguistic Wars (R.A. Harris 1993).
One difficulty with talking about language, I think, is that it is something everybody uses language (though, perhaps to varying degrees; a topic for another post). Therefore, it is difficult to come to talking about language from a "blank slate"; it is difficult to return to First Principles talking about language because we have so many thoughts about it already.
But, we are seeking Truth, so we are willing to take that path. In order to find a path of peace, let us clear our heads and establish what exactly we mean by rules. There is freedom in order. I will show you a way whereby rules may be both descriptive and prescriptive.
But, we are seeking Truth, so we are willing to take that path. In order to find a path of peace, let us clear our heads and establish what exactly we mean by rules. There is freedom in order. I will show you a way whereby rules may be both descriptive and prescriptive.
...then after establishing some principles and talking on the same terms, we can turn to tearing apart bad arguments, calling out authors (oh dear am I included in this?!) on unfounded claims, and relentlessly pruning bullshit 👺
Reestablishing Rule
In the previous post, I offered a definition of language, which I'll restate here: language is what is intentional and systematic in communication. Languages (plural), in turn, are those individual systems of communication. The notion of counting languages, bilingualism, and other matters of what having more than one language means will be taken up in the next post.
A grammar is a description of a language. We might also talk about grammar (c.f. a grammar) as that which is systematic in a language, as opposed to a description or theory trying to capture that system. That is, a grammar made up by linguists (or computers, or some unholy alliance) attempts to capture the grammar of some language. Many (all?) grammars may be defined in terms of lexical items and rules. Lexical items here, can roughly be understood to be the "words" or individually meaningful units of a language. Rules are formal descriptions of how lexical items and other rules interact.
Rules about what?
As grammars are theories made up by people, they can include or omit things that we talk about in a common usage of "language". For example, one might write a grammar of English that makes no reference to meaning, but is solely a syntactic description explaining what word orderings are "possible" (this would be modelling a very strong version of the principle of the "autonomy of syntax"). In such a grammar, all we would need to know about a word like cat is that it patterns with other words in a particular way. We might assign cat to a class "Noun" and show how it behaves according to identical constraints as other words like dog, which also belong to that same class.
Similarly, a grammar of English might consist of lexical items with lots of meaning information. One might say that the sentence, "trees are angry" is unacceptable because trees don't think or feel, so they can't be angry. Such information might be represented in a fancy-shmancy way like:
@Noun{rock,
animate = "False"
living = "False"
}
@Noun{tree,
animate = "False"
living = "True"
}
@Noun{cat,
animate = "True"
living = "True"
}
...
...in a text file representing the words a speaker of a language knows (their lexicon). We could then have a rule for a phrase like "____ are angry" that requires the thing in the blank to be both plural and animate.
Such a lexical item plus rules framework for describing a language is general, and need not correspond to things we think of as words and sentences. We could come up with a very simple language consisting of the lexical items {A, B} and the rule "X or XX", where X can be any of the lexical items. There are more formal ways to represent lexical items, rules, etc. but these details are not necessary to see the overall, higher level picture here. The permissible sentences in this language are therefore,
A
B
AA
BB
...but not AB, BA, AAA and so forth. As they appear in the strings {"A", "B", "AA", "BB"}, the individual A's and B's are called tokens of the lexical items A and B.
When we check if a certain string (e.g. "A", "AA", "AAB"...) belongs to a language, we must parse it. In this context, this means that we see if from the lexical items and the rules in our language we can generate that string. Our rule allows us to get "A", and "AA", but not "AAB". So, we say that "AAB" in ungrammatical because our rules disallow us from generating it.
The ideas about rules just discussed here are very general. The grammar describing the language {"A", "B", "AA", "BB"}, where language is understood to mean a set of strings (compatible with my definition, but more narrow/formal), could very well be a grammar for how your pet dog likes dog food. Suppose you observe the following facts:
Doggo accepts the following orders: one can of food A, one can of food B, two cans of food A, or two cans of food B; nothing else.
Could you come up with a (formal) model for this phenomena?
