A counter-example to the parsing rule model?

Question

The idea that we have some strict "correct" parsing rules which we use to parse sentences seems a bit wrong to me. Here's why. Consider these sentences:

• Yesterday I went to the beach.
• I, yesterday, went to the beach.
• I went, yesterday, to the beach.
• I went to, yesterday, the beach.
• I went to the beach yesterday.

Each of these sentences is perfectly understandable (if a little weird.) The only place where placing the word "yesterday" won't make sense is this:

• I went to the yesterday Beach.

Because it implies there is a type of beach called a "Yesterday Beach".

To me it seems like we have some sort of frame with a slot for the time period and it is not very important at which point in the sentence this slot gets filled with this knowledge.

On the other hand other grammar rules are important such as SVO rules and adjectives before nouns etc.

If we had a strict set or parsing rules most of these sentences wouldn't make sense. But the fact that they do suggests something else is going on.

Is there a way to explain this phenomenon in generative grammar models where sentences seemingly in the wrong order still make sense? Or is there a way to express the grammar rules where the placing of this word "yesterday" can go almost anywhere and still be understood?

e.g. The rule seems to be: "The sentence must contain Subject, Verb, Object, Time. The SVO order must apply but the Time can appear at any place". In fact I think we could also add a phrase "for a swim" at almost any place too and the sentence still be understood.

Answer 1

You could absolutely say "I went to the, uh, yesterday, beach" in casual speech. You wouldn't even hear the commas. It's not ungrammatical, you're just assuming that your listener can reconstruct the sentence you would have said if you had prepared it perfectly mentally. I don't think a speaker who says this thinks that "I went to the yesterday beach" is the best, most interpretable way to convey that information. If you didn't assume that your listener couldn't reconstruct your intent from the sentence you heard yourself saying, you would restart or find some other way to insert the information of when you went to the beach.

I originally had two paragraphs about placing the time adverbial between phrase boundaries in this response, but then I realised one of your positive examples contains "to yesterday the beach", so the above is probably the most applicable part.

Answer 2

Every single example except the first one is grammatically wrong, however that's no unsurmountable hindrance to understanding, except in your last example. The embedded phrase can be seen as a meta level: That didn't come out right, can I start over? No? Fine, then I will just leave this tiny breadcrumb here and resume where I left off: "yesterday, the beach". See, it almost sounds as if you were starting a new sentence.

That breadcrumb can become a book chapter or at least a footnote in cases where the message content is not strictly linear. Verbal language often uses tone and prosody to signal such abrupt changes, instead of meta-level explanations about what you are saying. yesterday as a word is unambiguous enough in pretty much any place that it does not need an explanation; However, beach is not in any case.

You might as well add a new clause at the end, "That was yesterday.", which is suboptimal, because it frequently signals the end of the story, when it's followed with present tense, but that's not a matter of syntax in the strict sense; perhaps style, especially if you want to confuse the recipient for dramatic effect.

Vice versa, there is no way anyone would insert yesterday between the and beach, and nobody would do the reverse the yester--beach--day. This suggestst that there's a lexicon of words, but nobody could yet agree on a precise definition of "word". It may be arguable whether longish words are puzzled together from pieces the same way phrases are, and vice versa whether phrases are recalled from fixed entities--e.g. idiomatic sayings, songs, the bible--and patterns--e.g. so called snow clones, or perhaps half memorized entities with blanks to be filled in on the go.

But you are missing one important fact: language signals social association, and deviation from the norm is rarely tolerated, nor intended by the speaker, if so assuming that the norm tended to be fit in the evolutionary sense.

Further, a degree of consistency is needed for predictability. If you had a consisten bias to deviate from the norm, a listener might be able to tolerate it. We frequently do when talking to strangers. But this takes effort.

Whether recalling patterns is energetically preferable over dynamicly creating strings is indecidable, so speech uses a bit of both. But this depends on the complexity of what you want to say, and your example is decidedly simple.