“I thought X was Y” vs. “I thought Y was X”

Question

Walking with my wife the other day, I turned around and realized that the person behind me was, in fact, someone else, and my wife had stopped to look in a shop. I said to her

1a Oh, I thought that guy behind me was you.

Possibly because I had a bit of cognitive interference with "Oh, I thought you were behind me", I was tempted to say

1b Oh, I thought you were that guy behind me.

But on reflection I realize that 1 sounds right (felicitous?), and 2 sounds wrong (infelicitous).

Edit I've used these terms from now on at P Elliot's suggestion

---

As another example

2a The boy thought the woman whose hand he was holding was his mother.

Is felicitous, while

2b The boy thought his mother was the woman whose hand he was holding.

Is infelicitous

However,

3a I thought that log was a bear.

and

3b I thought that bear was a log.

Both are felicitous, although they point to different meanings.

---

My thoughts so far: definiteness might be important – playing around with how things sound, "a guy" vs. "the guy" vs. "that guy" vs. "something" seems to shift things a bit.

Alternately, maybe it has something to do with "was you" being qualitative rather than existential – that is, the meaning is not A=B but rather something closer to "A is B-like".

Or, perhaps, some basic and fundamental aspect of semantics (/pragmatics?) that I'm forgetting/never learned.

Answer 1

Copular clauses in English generally have a "Topic BE Focus" structure, rather than a "Focus BE Topic" structure. That is, the phrase that you're predicating something about precedes the copula; what you're predicating of that phrase follows the copula.

In your example, you were predicating something about "the guy behind you", namely that you were mistaken as to his identity. You weren't predicating something about your wife (as you would have been if you had said "I thought you were behind me").

This raises the non-linguistic question, which I'm not sure how to answer, of why we cognitively frame cases of mistaken identity in this way rather than the reverse: after all, if you're wrong about the identity of X because you think it's Y, aren't you also wrong about the identity of Y? But that's not how we think about it. If you were asked "Whose identity were you mistaken about, in that scenario?" you would answer "The guy behind me", not "My wife".

In a similar vein, your example 2b would be fine in a scenario (a weird one, but bear with me) where the boy has learned a wrong meaning for the word "mother", and thinks it means "any woman whose hand one happens to be holding". Describing that case you could, I think, perfectly well say 2b. This is because what the boy is mistaken about is (the meaning of) "his mother", so that goes into the Topic position, while in 2a what he is mistaken about is the identity of "the woman whose hand he was holding".

Answer 2

Although they do sound a bit odd and stilted, I personally have no trouble accepting 1b and 2b. I think your problem is that you aren't making a distinction between unacceptable and weird or awkward. Humans judge utterances based on several criteria; of course syntax and semantics but also stylistic factors and even ease of comprehension. In my [limited] experience, these are considered as extra-linguistic factors and not including in an utterances acceptability unless explicitly noted.

Note how 1 and 2 use embedded clauses (guy behind me and whose hand he was holding, respectively) while 3 doesn't. Also note how in the a versions the embedded clauses come at the end of the sentence while in b they come in the middle.

I unfortunately cannot find an academic source addressing this exact issue, embedding clauses in the middle of a sentence causes difficulty because the reader needs to figure out where the embedded clause ends. When the embedded clause comes at the end of the sentence it is trivial to see where it ends. I suspect that you, like me, don't consider 1b and 2b as wrong so much as harder to parse compared to 1a and 2a.

I was able to find multiple sources for how embedding can affect understandability but they describe different types of embedding from the one in your question.