Ideally, in any classification system, there needs to be:
- absolute clarity as to what falls within the domain of a particular class or category and what does not;
- no overlap between classes, so that nothing can readily fall within two or more classes;
- no equivalence of terminology between distinct but related classification systems;
- a super-ordinate classification of structural types when classes in a classification system only occur in one or another structural type (okay, yes, I agree, that’s pretty opaque but it will be clarified by exemplification below);
- transparency in the terminology used for individual classes, to facilitate the achievement of needs (i) and (ii) above;
- completeness of coverage, so that no item within the area of study hovers potentially unclassifiable in that it does not fit readily within any of the classification classes;
- no ‘overlumping’, where two or more distinct entities are included together within one category;
- no ‘oversplitting’, where two entities of marginal difference are unnecessarily separated from each other;
- a coherent explanatory value for each class or category within the system.
Let’s call these, above, criteria for determining the descriptive effectiveness of a system model. Of course, in the real world of people and things, the process of classifying can be so subjective that there can be little agreement in the classifying: one person’s perception of something as, say, ‘provocative’ can be matched by another person’s perception of the same as ‘reasonable’. Likewise, in the field of gardening, where ‘weeds’ can be defined as ‘plants with no use’ (in terms of their beauty, or food value or perfume), one person’s perception of a particular plant as a ‘weed’ can be matched by another’s perception of the same plant as a precious member of the garden.
As for language, however, the classification systems for the components of sentences should be able to achieve pretty close to ideal levels of, at least, clarity, discreteness, transparency and completeness. This is because, with a language, there has to be an almost total group-agreed perception of classes, even if this is significantly unconscious, for effective communication to be achieved. If the classification system is not achieving high levels of transparency, discreteness and clarity then, almost certainly, the classification system is ‘sloppy’ and not the language itself in its grammatical workings. So, how do conventional approaches to analysing and describing the synthetic grammar of English fare in terms of this kind of assessment? The short answer is: okay in parts, but not very well – overall.
One proviso, though, to perhaps bear in mind is that, when describing a synthetic system, it will inevitably be the case that each category, or functional item, in the system will be defined to some degree by its relationship to other classes or categories in the system.
Subject and predicate
The notions of the subject and the predicate, within linguistics, derive from the philosophical ideas of Plato and Aristotle in the area of categorising the things we can experience in the world. Aristotle drew a distinction between hupokeimenon (what we call a subject) and katêgouroumenon (what we call a predicate). In these terms,
- a subject is what a statement is about and
- a predicate is what is said about a subject in a statement.
So, for example, in these sentences, the subject is in red and the predicate is in green:
A) The professor’s machine is quite old.
B) It’s a thinking machine.
But, grammarians of distant yore made, I think, a big mistake in taking Aristotle’s essentially philosophical binary idea and extending it to cover all declarative sentences in a given language, such as English. What they did was decide that sentences such as these also have a subject and a predicate in much the same terms as in A and B above:
C) He created the machine some years ago.
D) He calls it the Ratiocinator 1 – or R1, for short.
E) He is planning a new machine.
The problem with this is that the predicate in each of C, D and E doesn’t actually tell us anything much about the subject. Also, C, D and E aren’t really just about the subject, they are just as much about the machine. We can show this more clearly with these examples:
F) He broke the vase. (This tells us extremely little about He. Who is he? The dog? The cat? A human? And, it tells us more about the vase [within the so-called predicate] than it does about He.)
G) The train pushed the passenger carriage up the hill. (This tells us as much about the passenger carriage as it does about the train. We know very little about the train from this: is it an old or a new train? Is it a steam train or an electric one? We don’t know.)
Overall, the problem here is that two distinctly different types of sentence have been conflated together by this inappropriate over-application of the binary contrast between subject and predicate. Criterion vii, above, has been flouted: sentences A and B are examples of one type of sentence and sentences C to G are examples of a second type; the classical binary contrast, between subject and predicate, ‘overlumps’ them together as not usefully distinguishable. But, they are usefully distinguishable; positing two distinctly different sentence types is a way to fulfill criterion iv, above, of having a super-ordinate classification system that avoids entities of different types being classed as identical. At some stage (assuming I don’t run out of energy, time, money and enthusiasm), I will seek to show that this awareness of distinct sentence types is the only way in which to describe certain semantic effects achieved in the language, very reliably albeit largely unconsciously, through the interplay between the synthetic and the analytic grammar systems.
Am I making an unnecessary fuss?
It could be said that I am making too much of a ‘song and dance’ about this. After all, it might be said, there is not much necessary difference between these two sentence types in terms of their analytic grammar, as shown in these examples:
type 1: It’s a thinking machine.
type 2: He is planning a new machine.
In the type 1 sentence, there is a pronoun (It), a verb (’s) and a noun phrase composed of a determiner (a), an adjectival progressive (thinking) and a noun (machine). Seemingly almost exactly similarly, in the type 2 sentence, we have a pronoun (He), a verb part (is planning) and a noun phrase composed of a determiner (a), an adjective (new) and a noun (machine). However, though there can be little difference or no difference in the analytic grammar, there is a significant difference in the way the sentences function semantically within their synthetic structure as I have sought to indicate above.
But, there is something else to say here now. Type 1 sentences have, of course, an analytic-grammar option that is denied to type 2 sentences, as shown here:
type 1: It’s quite old.
type 2: He is planning quite old.
Because, in a type 1 sentence, the so-called predicate refers back to the subject, we can use an adjective or an adjective phrase (e.g. quite old) immediately after the verb. As we can see directly above, this option is not possible in a type 2 sentence.
An aside on dependency grammar and valency grammar
Two approaches to grammar description, dependency grammar and valency grammar, are very similar in the focus they give to the verb part of a sentence. In these grammar models, the verb part alone is called the predicate; what come before and after the verb part are called arguments. So, the subject is one argument of the verb. The object is another argument of the verb (or, now, predicate, within this system model). If we put the arguments in red in a type 2 sentence then, within this system, we break a sentence down thus:
type 2: He created the machine.
But, in a type 1 sentence, what comes after the verb (conventionally called a complement) is not an argument but, instead is part of the verb/predicate, thus:
type 1: The machine is old.
Valency grammar is quite fun as it compares verbs to chemical elements and shows the bonding capacities of the verb (its valency). A verb that just has a subject is monovalent:
type 2: He smiled.
A verb that has a subject and an object is bivalent:
type 2: He is turning the dials on the machine.
… and so on.
But, I would say that both these theories of grammar are unsatisfactory in their explanatory value. Neither explicitly registers the two key types of sentence structure: type 1 and type 2, as described above. Also, dependency grammar largely conflates subjects and objects as just arguments and, so, not usefully distinct which, in fact, they are. It could be said, to be fair, that these grammars do implicitly recognise the different sentence types by evicting the objects in a type 2 sentence from the predicate but not the complements in a type 1 sentence from the predicate. But, there is a further problem here, in that it is descriptively useful/essential to be able to distinguish between the complement in a type 1 sentence and the verb and not just have them lumped, rather traditionally, together in a predicate. In short, neither of these approaches solves the problems described above in relation to more traditional approaches to the predicate.
The idea of describing individual sentences as either type 1 or type 2 sentences does seem rather opaque and, thus, falls foul of criterion v, above, which extols the virtue of transparency. I’ve been toying with the idea of calling type 1 sentences ‘Being sentences’, as they do always describe states, or almost always. Also, I’ve been toying with the idea of calling type 2 sentences ‘Doing sentences’, as they do, generally, describe actions or events. But, there is a problem here, in that type 2 sentences do also describe states, as in:
type 2: He has a number of properties. OR This dish contains gluten.
and type 1 sentences occasionally describe actions or behaviours, as in:
type 1: He is being silly.
I need another more grammatically focused method of nomenclature here for basic sentence types, as it can be descriptively dangerous/ill-advised to step too quickly into semantics when seeking to design new approaches to grammar description.
‘Ditch’ the notion of the predicate, and keep that of the subject?
For the reasons given above, the predicate needs to be removed as a notion from within the grammar, because it hides important contrasts rather than makes them plain.
Interestingly, the notion of subject is now conventionally described in a rather different way from how Aristotle defined it. Its definition has shifted to accommodate its having a functional role in type 2 sentences, as well as in type 1 ones.
A subject is nowadays often described not only as what the sentence is about (as in type 1 sentences) but also as the ‘doer’ of an action (as in type 2 sentences describing actions). However, that extended notion of the subject doesn’t cover its use as the ‘enactor’ of a state, as in:
type 2: The machine is quite ancient.
Thus, the term subject runs the risk, through its generally accepted definition, of breaking criterion vi (completeness of coverage).
As a term, ‘subject’ also fails a bit in terms of criterion v (transparency); but, it kind of needs to perhaps, as it needs to cover both the idea of being essentially the ‘enactor’ of an action, event or a state in type 2 sentences and the ‘enactor’ of a state, or a role or of behaviour in type 1 sentences. These two roles of enactor of an action and enactor of a state are so significantly different, in their core, that the term to cover both needs to be fairly non-specific perhaps. It could be said that we need two terms, not one, for ‘subjects’ in these two differing contexts, but I think that we would then be being pushed too far by semantics into creating a difference that the grammar in itself does not justify and that that would, therefore, be a case of breaking criterion viii (no ‘over-splitting’). A key thing about all subjects is that they almost always determine the ‘number’ of the present-tense verb in the third person, whether it’s singular or plural (e.g. likes versus like or knows versus know), and whether it’s first person, second person or third person in the case of forms of be (am, are, is). By the by, this general grammatical role of the subject in determining the ‘number’ of the present-tense verb rather denies the rationale behind giving a super-ordinate value to the verb, as is done in dependency and valency grammars.
Anyway, as for the term ‘subject’, it’s probably as well not to try and redesign everything in one fell swoop when reviewing and re-creating system models. And, in the case of the ‘subject’, it is in a nice direct antithesis with the term ‘object’. So, in this moment, I feel in two minds about jettisoning the term subject from the syntactic lexicon.