How often does an individual encounter a novel linguistic construction?

Question

I've heard it said that most utterances we encounter on a day-to-day basis are novel. That is, the individual has never heard that exact sentence before in their life.

My intuition is that the source of this productivity mostly comes from which words we slot into the grammatical categories (noun, verb....), not from hearing syntactic sequences we've never heard before (novel trees). Is this correct?

(To clarify, I'm talking about how often we encounter new derivations that fit with our existing syntactic rule systems, not novel grammatical constructions that have gained societal currency.)

Answer 1

This is a very interesting question, but there are two reasons it is very difficult to answer. First, there's a simple problem independent of any specific theory: there is no clear way to define the size of the trees in question. Also, if we believe the generative research of the past 20 years, there's reason to doubt there's such a thing as "slots" in the sense you speak of.

First, the size: simply following orthographic conventions regarding "sentences" doesn't necessarily tell us about the maximal relevant size of syntactic representations. If I utter two long sentences, loaded with adjuncts, and joined together by a conjunction, are they part of the same tree for our purposes? I have no idea, frankly.

We're probably interested in some very large structure, because the question becomes trivial once the tree is small, and I don't think that trivial question is what you were aiming at. In other words, if the relevant unit to look at is the XP of minimalist generative syntax, meaning any structure composed of a lexical item and whatever other lexical items its selection features require, then by hypothesis the same XP structure is (recursively) used to construct any and all syntactically-structure utterances, and then, indeed, no utterance is syntactically novel.

So presumably, to make the question interesting, we want to look at units up to some very big size, probably up to the biggest XPs uttered in normal life. But to the best of my knowledge, there's no clearly defined maximum.

The second problem is that, as I kind of already indicated, the concept of syntactic slots as such isn't really relevant in much of the current research in syntax. "Slots" are essentially created by lexical items ("heads") which carry selectional features that need to be satisfied in order for the structure to be well-formed (and interpretable at the interfaces to sound and meaning). They are filled by XPs, which are in turn also formed around the requirements of some head.

It is, however, assumed that the real structure-building selectional features are those selecting major content and functional categories (V,N,A,P,D,C,T,v,...). So if we assume these can be seen as "slots", we might be able to start to answer the question.

I've been trying to formulate some kind of attempt at an answer in terms of a thought experiment – the empirical side is not my forte – but without much result, because I keep bumping into the problems I outlined above. I guess this isn't strictly an answer, but hopefully it'll help someone understand the assumptions and implications of the question, and maybe work towards a real answer.

Answer 2

I'm not a linguist, so please double check my answer. To the best of my knowledge, you are more-or-less correct. Though grammatical rules can change over time, and new words are added to a language over time, at any given moment, speakers of a given variety of a language use a finite set of grammatical rules and a finite set of morphemes (meaningful parts of words such as roots and affixes) to generate a potentially limitless number of new utterances.

The reason that the number of new utterances is potentially limitless is because languages allow certain structures to be added to the same structures over and over again indefinitely. For example, in English, relative clauses can occur within relative clauses ad infinitum. Thus, "the cat that ate the rat" can be revised to "the cat that ate the rat that ate the cheese" to "the cat that ate the rat that ate the cheese that stank up the kitchen" to "the cat that ate the rat that ate the cheese that stank up the kitchen in which I was trying to eat the pudding" and so on indefinitely.

It is difficult for lay people like myself to understand the details of this generative view of language. However, it is worth noting that, in the course of everyday conversation, we would all be taken aback if we conversed with someone who used a small set of stock sentences to communicate with. In fact, we are all taken aback for this reason when we verbally interact with automated answering systems. This just goes to show that all of us expect novel utterances when we converse with other people, aside from certain stock utterances like "Yes!" and "Here we go again."

Answer 3

This is my question, but I've lost the login and can't respond to the questions anywhere other than here.

@Ron Maimon- Thanks! I think you stated my question better than I did.

@Alennano- Thanks for the welcome. What I meant was that most sentences an individual hears are novel in the sense that they have never encountered that exact sentence in their lives before. I don't even know if this is true for starters, but it seems to be repeated a lot.

@michaelsappir- It's been a while since I learned x-bar theory, but I think I understood everything you said. If you say that everything is just a lexical item combined together with another lexical item then no tree is novel which does seem pretty trivial (I guess, unless you talk about the particular combinations of lexically-based constraints used to combine the items). But in any event, what I meant by slots was nouns, verbs, prepositions, etc. In terms of tree size, I agree that it seems likely that we have encountered most small trees by adulthood. Regarding maximum tree size, since I'm interested in where "most" productivity comes from, I would be happy taking a frequency cutoff (like, the lower 99.9% of tree sizes an individual encounters) as capturing most day-to-day language use.

In any event, the reason I was wondering is that poverty-of-the-stimulus arguments seem to focus a lot on ("unexplainable") syntactic productivity in development, but I know there are (psychological) theories arguing that children's generalization is actually very restricted (lexically but especially syntactically). This just made me wonder if in principle, it's possible to "get by", as an adult, with no syntactic tree productivity. I don't know if I explained that very well. It's obviously a loaded question in full form which is why I asked the simplest version I could think of.