9

Further to comments against Do complex sentences always need a conjunction? as recently asked on ELU (and Complex sentence without a subordinating conjunction? here on Linguistics), I'd like to know whether "sentence" is a useful and/or clearly-defined term from the perspective of "linguists" (as opposed to "pedagogic grammarians").

The specific reason I ask is this comment by John Lawler, pointing out that in speech no-one can tell the difference between a full stop and a semicolon (or dash, I assume) - [so] it's something of a moot point whether this current "sentence", for example, is actually two sentences.

I also assume highlighted [so] in the above is in principle "optional", and that from the "grammarian's" perspective it's at least credible to classify two clauses linked by a "non-full stop" punctuation mark as a complex sentence, regardless of whether a conjunction is explicitly included or not.

I found this definition in yourdictionary.com...

...in speech a sentence begins following a silence and concludes with any of various final pitches and a terminal juncture

...but I'm not sure how to relate that to, for example, my second paragraph above (if read aloud).

Community
  • 1
FumbleFingers
  • 998
  • 1
  • 8
  • 15
  • 3
    The general term used in Linguistics is utterrance, which has no syntactic or phonological restrictions, except that it has to be utterred by the same person at one time. What we call a "sentence" is simply a complete independent clause, which of course usually has other clauses appended. – jlawler Jul 20 '15 at 14:35
  • 1
    I don't understand what you mean by "same time". I understand that it's a continuous stream of time, but since people pause in utterances, what would a time be? – user6726 Jul 20 '15 at 16:13
  • I should have said 'during one conversational turn', to be more precise. – jlawler Jul 20 '15 at 16:20
  • 1
    @jlawler: Ah! I should have realized that before! For many purposes, "conversational turn" is the primary subcomponent of spoken language, for which there's usually no equivalent in written language, since it's not usually "interactive". – FumbleFingers Jul 20 '15 at 16:26
  • 2
    What do you mean by "useful" -- to whom? For what? What constitutes being "clearly-defined" -- do you mean "easy to detect where everybody agrees that the thing is an X"? Substitute any other linguistic term, and you'll get the same range of "yes" / "no" opinions, just from different people. – user6726 Jul 20 '15 at 18:11
  • @user6726: I'm supposing the "raison d'etre" for linguistics is to identify, describe, categorise the various features used in language, and the principles/rules governing how they're used. I realise that there will be people concentrating on different things within the entire field, but I suppose I'm asking how relevant "sentence" is to the average linguist. Or maybe I mean, How important is it in the context of "Linguistics 101"? – FumbleFingers Jul 21 '15 at 01:35
  • In phonetics, there is the term intonation phrase, which comes relatively close to what we generally call a sentence. – maj Jul 21 '15 at 19:51
  • @maj: Ah, right. I assume intonation phrase would come particularly close to the yourdictionary.com definition I cited in the question text. – FumbleFingers Jul 22 '15 at 11:47
  • In the slides that we used in our spontaneous speech lecture, all I could find was a reference to "Beckman & Pierrehumbert 1986". – maj Jul 22 '15 at 12:16

3 Answers3

3

Sentence is not a very useful construct for most things you'd want to study about language. It is only an artifact of a particular kind of written language and it is really hard to provide any definition more specific than 'it starts with a capital letter and ends with a period'. Even then, the beginnings and ends of 'sentences' are fairly arbitrary.

The only analytical concept I can think of where sentence is directly useful is measures of readability which look at the number of words per sentence. This can give a fairly reliable indication of the difficulty of text (providing they are long enough). Sentences are also used when dealing with concepts like subordination and anaphora accessibility but they're not really necessary for that.

The two concepts to look at as much more fundamental are:

  1. Clauses defined broadly as subject and predicate. They can be found in any language and are much more easily and reliably identified. You can usefully study relationships between clauses in speech without recourse to the notion of a sentence and arrive at very similar conclusions. Perhaps, here the term sentence could be reserved for groups of clauses that are linked through hypotaxis (subordination) but even here, these relationships occur in many non-sentence-like contexts. Like: A: It's going to be a great party. B. Yeah! If Johnny decides to show up.

  2. Utterances or texts which are units of expression that can be studied with respect to communicative intent without regard to length or internal structure. Thus, I can treat this entire response as a single utterance/text just as much as the word 'Boo' shouted into somebody's face constitutes an utterance. Looking at the types of utterances, their internal structure and relationships (such as co-reference, turn-taking, repair, etc.) can reveal a lot of useful things about language as a whole.

Ultimately, sentence is just a theoretical leftover from earlier times of theorizing about language based on written sources alone.

However, that does not mean that in many practical contexts such as language teaching, composition, etc. sentence should be avoided as a way to communicate about language. It's a very useful, generally understood shortcut that is probably not even out of place in a textbook on syntax. But in thinking about how language works across many languages and usage contexts, it is of only marginal interest. It just happens that most of the language data available to us is in the form that makes sentences seem like an inherent organizing principle of language.

Dominik Lukes
  • 10,588
  • 27
  • 49
  • This answer makes a lot of sense to me, thanks. Because I'm somewhat "outside my pay grade" here on Linguistics, I'll wait to see how others vote before "accepting" it, but it's currently working for me. I'm particularly comfortable with the idea that broadly speaking language consists of utterances/texts, which are primarily composed of clauses (with implicit or explicit relationships between any given clause and its close neighbours). – FumbleFingers Jul 20 '15 at 17:41
  • 1
    Fine. Then "sentence", if you want to use it, just means "independent clause". And Greg's theory would still work with that definition, whereas "utterance" is probly too course-grained a unit to be useful in most syntax. – jlawler Jul 20 '15 at 17:55
  • 1
    "Sentence" has played a fundamental role in linguistic theory. So far as I can tell, it has nothing to do with written as opposed to spoken forms. What is the source of this odd idea that "sentence" is connected in some important way with writing? – Greg Lee Jul 20 '15 at 19:30
  • @Greg: I think it's the pull of the "Starts with a capital and ends with a period" definition. It's so simple in that medium. – jlawler Jul 20 '15 at 20:04
  • 1
    @GregLee For one thing if you work with previously undescribed languages with no writing tradition whatsoever then it's hard to know what 'sentence' would refer to. – Gaston Ümlaut Jul 20 '15 at 23:34
  • 1
    @GastonÜmlaut, Have you worked with such languages? Did you have that difficulty? I doubt very much that what you say is true. Interest in languages with no writing culture began over a century ago. Have we ever encountered one that didn't have easily identifiable sentences? Look at the literature of the Summer Institute of Linguistics/Wycliffe Bible Translators. Does the theory and method they use (tagmemics) have sentences? Yes, it does. I did a months field work on Cherente in Brasil, with no written literature or grammar -- I had no difficulty identifying the sentences. – Greg Lee Jul 21 '15 at 01:52
  • Yes, I've worked with numerous such languages (they are probably the majority of human languages!) in Australia and PNG. But I'm still not sure what you mean by 'sentence'. To me it is a non-theoretical term without a clear definition that can be used to refer to any discrete 'chunk' of language that constitutes an entire standalone utterance. But perhaps that's what you mean by 'sentence'? I find it most useful as a way to refer to examples presented in an analysis. – Gaston Ümlaut Jul 21 '15 at 02:01
  • @GastonÜmlaut, "sentence" is indeed useful in presenting an analysis (as you say), because it is a theoretical term (contrary to what you say). It cannot be satisfactorily defined in terms of observable facts, and the reason for that is that it is a theoretical term. That is what theoretical terms are like. Theories help you understand facts, but they are not themselves facts. It's hard for me to think of any linguistic theory of any importance of the last 50 years in which "sentence" has not played a crucial role. Syntactic Structures, The Grammar of English Nominalizations, ... – Greg Lee Jul 21 '15 at 16:37
  • You don't have to look at languages without a body of writing to see the issues. Just look at corpora of conversation transcripts. You'll find it very hard to identify sentences - in a way that clauses aren't. Even in written texts, you'd probably find relatively low interrater consistency if you stripped out all the outward sentence marking and asked them to reinsert it. More than with paragraphs but (at a guess) much less than with clauses. – Dominik Lukes Jul 23 '15 at 19:56
  • @GregLee I think you'll find that much of functionalism and certainly most of construction grammar would not find sentences as crucial. Also, many aspects of corpus linguistics, text linguistics and conversation analysis. Not that they don't use the term but it's not as key to them as if you started with S --> NP VP. – Dominik Lukes Jul 23 '15 at 19:59
2

I agree with what user6726 has written:

What do you mean by "useful" -- to whom? For what? What constitutes being "clearly-defined" -- do you mean "easy to detect where everybody agrees that the thing is an X"? Substitute any other linguistic term, and you'll get the same range of "yes" / "no" opinions, just from different people.

And my answer is based on your clarification:

I'm supposing the "raison d'etre" for linguistics is to identify, describe, categorise the various features used in language, and the principles/rules governing how they're used. I realise that there will be people concentrating on different things within the entire field, but I suppose I'm asking how relevant "sentence" is to the average linguist. Or maybe I mean, How important is it in the context of "Linguistics 101"?

To the last question, I'd say it's eminently important. When I searched on google for "linguistics textbook pdf", I came across Introductory Linguistics -- Bruce P. Hayes, The Study of Language -- George Yule, Introduction to English Linguistics --László Varga, and so on. In each of these, the concept of "sentence" was used without much of an explanation, relying on its audience to already have a clear conception of it. The last book I cited distinguishes between sentence and utterance as follows:

Sentences have to be distinguished from utterances. A sentence is any string of words produced by the sentence-forming rules of a language, these rules are stored in native speakers’ competence. (By competence we mean the native speaker’s intuitive knowledge of language, see Unit 2 above.) So sentences are constructs of competence, they are ideal, abstract entities. For instance, Peter smokes cheap cigars is an English sentence because it has the structure of an English sentence.

By contrast, an utterance is typically the physical realisation of a sentence in a real situation of language use, i.e. in performance. (Performance is the actual use of competence and it involves individual and situational factors, see Unit 2.) Since utterances belong to performance, in spontaneous speech they often contain imperfections, such as hesitations, false starts, lack of concord, etc., especially if the speaker is tired or excited or embarrassed.

My only point of contention with the phrasing above is that I would rather have said "an utterance is the spoken realisation of a sentence". Afterall, a sentence is physically realized whether it is spoken, written, or gesticulated (as in ASL), and none of these are any less real situations of language use.

prash
  • 3,649
  • 3
  • 25
  • 33
  • I agree from the perspective of Linguistics 101, sentences are great at that introductory level. But the closer you look at actual language use, the less they look like something that is a central organizing principle. Even if you subscribe to the competence/performance dichotomy - which is another example of useful shortcut but a not good candidate for a real principle at work. – Dominik Lukes Jul 23 '15 at 20:03
  • @DominikLukes I have also attended fairly advanced classes on syntax and on semantics, some of which were taught by people who are essentially theoretical linguists. These people did not ever let go of the concept of sentence. In fact the ones who let go of the concept of sentence were those using some very specific statistical methods, who would rather look at bag-of-words or n-grams, and cared naught for punctuation.

    I agree with you about competence/performance being a useful shortcut. Because AFAIK we still don't have a neurological model of language that explains what's going on.

     – prash Jul 23 '15 at 20:41
0

Sentences are useful because a language is a set of sentences. (And without language, where would we be?)

Greg Lee
  • 12,466
  • 1
  • 18
  • 32
  • 2
    I was just about to say *utterance* seemed a more "useful" term to me - but even as I write, I see John has made that very same point (it was obviously just a "lucky guess" on my part! :) I'm increasingly of the opinion that "language" is essentially a spoken phenomenon, and that to a first approximation, the written format is just an attempt to approximate "real" language in another medium. – FumbleFingers Jul 20 '15 at 15:09
  • So, you're saying that though people may write in sentences, they don't speak in sentences? – Greg Lee Jul 20 '15 at 15:35
  • I don't think so. I'm saying that it looks to me as if sentences can in some contexts meaningfully be identified as "subcomponents" of utterances (usually, but maybe not always, written utterances). But whereas you could equally say letters of the alphabet are subcomponents of written utterances, I doubt anyone would ever claim it's meaningful to talk about letters as subcomponents of speech. – FumbleFingers Jul 20 '15 at 15:46
  • Certainly people can speak in sentences; many often do and everybody does sometimes. But I've come to the opinion that mostly what people speak is phrases, idioms, and deictic elements, with the odd sentence to string them together. It's so much easier to deal with large-scale objects. Chuck Fillmore's analogy was if you treat language like Tinkertoys, the familiar phrases are like constructions that have been left out in the rain and don't come apart that easily. What we call "idioms" are just the familiar ones; everything else is personal idiom. – jlawler Jul 20 '15 at 16:18
  • 1
    People do talk about letters as subcomponents of speech -- those are phonemes. Phonemes are an alphabetic theory with the details cleaned up a little. Correspondingly, when a grammarian proposes that "Candy." as a reply to "Do you want candy or broccoli?" is derived from an underlying sentence "I want candy", he is trying to clean up details. We communicate in sentences, just as we speak in phonemes. – Greg Lee Jul 20 '15 at 16:20
  • I can get with that theory; it's the way I was trained to think about syntax, too. But the question is a good one, since that theory presupposes some kind of derivation- or other-related link between the non-occurring sentences and the actual utterrances they are claimed to be related to. While it settles the question of the utility of sentences, the theory doesn't really define 'sentence' very well, especially if formality is an issue. – jlawler Jul 20 '15 at 16:23
  • 1
    Suppose the response is "Candy?" (spoken with questioning intonation)? Are we then to assume not only is it not actually a question at all, but that there's an underlying sentence "I'm very surprised to hear that candy is one of my available choices here"? I can see that the idea of "non-occurring sentences" behind "actual utterances" works well in some cases, but I get the feeling most conversational elements aren't that easily categorised. – FumbleFingers Jul 20 '15 at 16:40
  • @jlawler, my answer implicitly gives a definition of sentence. A sentence is one of the things that a language is a set of. – Greg Lee Jul 20 '15 at 16:50
  • @FumbleFingers, Well, it's just a theory. Could be wrong. But the fact that it is difficult to apply shouldn't count against it. Theories don't have to be simple or straightforward. – Greg Lee Jul 20 '15 at 16:58
  • Obviously both you and John know much more about this particular issue than me. I said earlier that I wasn't greatly interested in "terminology" as such, but I am interested in gaining a better understanding of the "distinct, identifiable phenomena" of language as addressed by modern linguists, regardless of what specific standard terms they use. The two final "sentences" in your last comment could be seen as one "complex sentence" with a deleted *since, because* between them. But is it a useful way of looking at things? That's what I'm wondering. – FumbleFingers Jul 20 '15 at 17:23
  • Well, it's easy to carry it to extremes; in fact, it's easy to carry just about any linguiistic theory to extremes, because there are usually a lot of idealized presuppositions in them that aren't based on phenomena, but on the theoretician's choices, and often these don't play well with actual evidence. – jlawler Jul 20 '15 at 17:37
  • @FumbleFingers: Don't forget the title of McCawley's grammar. He was serious about dealing with real phenomena and tossed out gods know how many theories because they couldn't. The title of one of his books is Thirty Million Theories of Syntax; he arrived at the title by dealing with 25 independent open questions in syntax (each of which would result in at least two different theories, depending on the answer to the question), and then, as he said, "by computing 2²⁵ and rounding downwards." – jlawler Jul 20 '15 at 17:52
  • I think McCawley would approve of sentences being basic and fundamental to language, because he thought of syntactic theory as a means to explicate how we reason in language. And the steps in a deduction in familiar logical theories are well-formed-formulas, which correspond to sentences. Could you have logic without sentences? I don't think so. – Greg Lee Jul 20 '15 at 18:36
  • 1
    @jlawler: Right from the get-go, and reinforced each time I pick up McCawley's weighty tome, I register the fact that he's identifying/explaining "real phenomena". That's to say I always have the feeling he's talking about usages that actually occur, regardless of whether they fit within some "prescriptive" description of how people should speak. One thing I particularly like about his style is he often summarises alternative terminology/analyses while setting out his own personal position, which often helps me get things into perspective... – FumbleFingers Jul 20 '15 at 19:18
  • 1
    That's one of the things I like about it, too. – jlawler Jul 20 '15 at 19:19
  • ...which I find much easier than many "pop science" books I read (on particle physics and other "theories of reality", for example), where I can often end up confused because I don't always know how to interrelate different approaches. To a considerable extent, McCawley comes across to me as "Here's a summary of what other people think/have thought, and here's my approach". – FumbleFingers Jul 20 '15 at 19:21
  • 1
    The reasoning in this answer seems circular to me. As far as I can tell, language is a set of sentences because that's how you choose to interpret it... is there a reason that's a useful interpretation? – Brilliand Jul 20 '15 at 20:16
  • @Brilliand, yes, it is a matter of interpretation. Theories that make a language a set of sentences could be all wrong. Whether it is a useful interpretation is not a matter that interests me. – Greg Lee Jul 20 '15 at 21:07
  • @GregLee This question explicitly asks whether it's useful, so... I'm thinking you have to explain why it's useful for this to be a good answer to this question. At the very least, you could explain how whatever theory you have in mind uses the concept of a sentence to clarify things. – Brilliand Jul 20 '15 at 21:31