A Note on Wh-in-Situ in English

Abstract:  The purpose of this squib is to show that wh-in-situ is productive in English, even outside of echo-questions and multiple wh-questions. I also discuss the relation of the data I adduce to quiz-show questions (Jackendoff 1994).

A Note on Wh-in-Situ in English

Top Likes and Dislikes of Working in Academics (Revised)

After more than three decades working in academics (graduated from MIT 1993), this is a small list of things that I like and dislike the most. I understand that my position at a private university in the United States is very privileged. Don’t get me wrong, I am grateful for having my job. 

I believe that this list might be useful to somebody starting their career, and wondering whether academics is right for them. I would love to hear feedback from others on their experiences.

Some Scribblings on Nasal Gobbling

In my informal midwestern speech (raised in suburban Minnesota and rural Wisconsin), I can say the following:

1.

I’mʌ nʌ go

‘I’m going to go.’

(spelled dialectally as: ‘I’muh nuh go’, or maybe, ‘I’ma na go’).

In this short note, I sketch the series of phonological changes at play in deriving (1) in an attempt to understand my own personal speech patterns. 

I did an informal survey of the students in my undergraduate syntax class (18 students), and most of them said (1) is acceptable, but I do not know if there are any geographical generalizations about where it is used.

I assume that the underlying form is something like the following (written with English orthography):

2. I am going to go.

I rarely pronounce -ing with an engma [ng]. Mostly, I pronounce it with -in.

3. I am goin to go.

In my English, unstressed infinitival (and prepositional) ‘to’ is very generally pronounced as tʌ, which can also be slightly voiced as shown in (5):

4. I am goin tʌ go.

5. I am goin dʌ go.

From there, the nasal of -in gobbles up the following voiced dental:

6. I am goi nʌ go.

I find the same nasal gobbling with 'trying to'. And both seem related to wanna-contraction. But, I think that (5) itself is infrequent. Rather, (5) seems to prefer the following reduced form of ‘going’:

7. I am gʌ nʌ go.

In (7), [o] is changed to [ʌ] (o --> ʌ), and the [i] of the suffix -in is dropped. Alternatively, it may be that the progressive suffix is really -Vn (not -ing), and the V is generally filled with an epenthetic lax [i]. So in (7), the V of -Vn is simply not syllabified (not literally deleted).

Nasal gobbling in (7) involves nasal spreading auto-segmentally to the following C slot:

8. n t ʌ

| |

C V

Given the representation in (8), the nasal docs to the initial C, delinking the [t], yielding the correct form: nʌ.

The form in (7) can be stressed (indicated by underlining):

9.

Q: Are you refusing to go to your class?

A: I am gʌ nʌ go.

But stressed or not stressed, the vowel [o] is not possible:

10. *I am go nʌ go.

The facts in (7) and (10) suggest that we are dealing with a kind of verbal suppletion. The verb ‘go’ in English can be realized as [go] in most contexts, but can also be realized as [gʌ] in some contexts, and it is not explained away as a phonetic vowel reduction (because the latter can be stressed). 

The alternation between [go] and [gʌ] similar to the fact that ‘do’ [du] can be realized as [dʌ] as in the word ‘does’ (which can also be stressed).

From (7), the auxiliary ‘am’ can be contracted:

11. I’m gʌ nʌ go.

About the contraction in (11), I am assuming that the auxiliary am is bimorphemic a-m, and that in (11) only the second morpheme is being spelled out. Part of the reason for this assumption is the presence of [m] in all 1SG pronouns exceptive nominative: me, my, mine. This series suggest that 1SG in English is [m].

Then the 1SG nasal of the auxiliary gobbles up the following voiced velar:

12. I’mʌ nʌ go.

Note in (12) that the nasal gobbling is not accompanied by any change in the place of articulation of the nasal. (12) involves a bilabial nasal, not a velar nasal. So the [g] has completely disappeared from the utterance. 

I have no idea how common the form in (12) is in the United States. But for me personally, it is the most natural way of speaking.

If the auxiliary is not contracted, nasal gobbling is less acceptable:

13. a. *I am mʌ nʌ go.

b. *I a mʌ nʌ go.

Nasal gobbling involves the nasal spreading auto-segmentally to the following C slot:

14. m g ʌ n ʌ

| | | |

C V C V

Given the representation in (14), the 1SG nasal docks to the initial C, delinking the [g], yielding the correct form: mʌ nʌ. This is the same nasal gobbling rule as seen in (8) above. 

One thing to notice is that these reductions do not occur with the main verb use of ‘going to’:

15. a. *I am gʌ nʌ the beach.

Intended: ‘I am going to the beach.’

b. *I’mʌ nʌ the beach.

Intended: ‘I am going to the beach.’

Let’s see where the chain of reductions stops. First, the unstressed preposition ‘to’ can also be pronounced tʌ

16. I am going tʌ the beach.

But the following is completely unacceptable:

17.

a. *I am goi nʌ the beach.

b. *I am gʌ nʌ the beach.

So it seems that o-->ʌ for ‘go’ and nasal gobbling are not possible for the main for ‘go’. This suggests that the internal structure of main verb going is different from the internal structure of auxiliary verb going, in a way which blocks nasal gobbling and suppletion for main verbs, but I do not pursue that here.

Another thing to notice is that the changes only occur if ‘be’ in ‘be going to’ is finite:

18. a. He may be going to go.

b. He must be going to go (based on what I know).

c. To be going to

While the sentences in (18) are uncommon, the ones in (19) sound unacceptable to me:

19. a. *He may be gʌ nʌ go.

b. *He must be gʌ nʌ go.

c. *To be gʌ nʌ go.

If a negation (or adverbs) intervenes between the auxiliary and ‘go’, the following form is possible:

20. I’m not gʌ nʌ go

But not:

21.

a. I’m not mʌ nʌ go

b. *I’mʌ not nʌ go

So nasal gobbling seems to require phonological adjacency.

Nasal gobbling is found in many places in English phonology, which are well-known, so I will not repeat them here. Off the top of my head, I know of: ‘I don’t know’ (‘I dunno’), ‘something’ (smthn), ‘pumpkin’ (pungkin). 

In the cases that I am considering in this paper, the nasal is always part of a function word (e.g., the -ing suffix, or the auxiliary ‘am’), and the target consonant is also part of a function word (e.g., infinitival to, future going). Generalizing (8) and (14) we have, the following rule:

22. Nasal Gobbling

Structural description: N is non-syllabified nasal consonant of any place of articulation. xy are the first two segments of a non-stressed function word.

N x[stop] y[vowel]

|      |

C         V

Structural change: the nasal optionally docks to the initial C, delinking the x segment.

Top Likes and Dislikes of Working in Academics

After more than three decades working in academics (graduated from MIT 1993), this is a small list of things that like and dislike the most. I understand that my position at a private university in the United States is very privileged. Don’t get me wrong, I am grateful for having my job. The order of presentation is random.

Quantifier Raising: Entailment and Inverse Scope (Syntax II, Spring 2025)

Consider the following two sentences, each involving an existential quantifier phrase and a universal quantifier phrase.

1.

a. Some boy loves every girl.

b. Every boy loves some girl.

In both cases, we can ask what the possible logical form representations of the sentence are, and what the truth conditions of those representations are. If we adopt the syntactic operation QR, then (1a) with inverse scope yields truth conditions that clearly distinguish it from (1a) with surface scope. The question of whether inverse scope is needed for sentences like (1b) is more intricate. This note goes into these issues in detail.

Applying QR to (1a), there are two logical form representations:

2.

LF1: (surface scope)

[TP <[some boy]1> [TP <[every girl]2> [TP DP1 loves DP2]]

LF2: (inverse scope)

[TP <[every girl]2> [TP <[some boy]1> [TP DP1 loves DP2]]

In these structures, the notation <…> indicates that the occurrence is not pronounced. In LF1, for example, [some boy] has two occurrences, but only the one in subject position is pronounced. Furthermore, in these representations the possibility of VP adjunct is ignored (see Fox 2002 for discussion).

These LF representations can be paraphrased as follows:

3. LF1 (paraphrase): There is some boy who loves every girl.

LF2 (paraphrase): For every girl, there is some boy who loves her.

Now consider the following two situations:

4.

S1: John love Sue, Mary and Kathy.

(and there are no other people or relationships involved)

S2: John loves Sue, Bill loves Mary, Chris loves Kathy.

(and there are no other people or relationships involved)

Calculating the truth value of each LF representation in each situation (actual semantic calculation omitted), we have the following results (SS = surface scope, IS = inverse scope):

5. S1 S2

LF1 (SS) true false

LF2 (IS) true true

This table shows that LF2 (inverse scope) is true in S2, whereas LF1 (surface scope) is not. These truth values provide an important argument for the existence of inverse scope. If inverse scope were not allowed, there is no way we could capture the intuition that (1a) (with structure LF2) is true in S2.

In the theory I presented in class (from May 1977, Fox 2002), inverse scope is accounted for in terms of QR. In inverse scope, the object undergoes QR to a scope position higher than the subject scope position (see LF2 above). Therefore, the data in (5) provides evidence supporting the covert movement operation QR.

This table shows that only S2 distinguishes the two LF representations, S1 does not. Therefore, it is clear that LF1 is stronger, in the sense that whenever LF1 is true, so is LF2, but not vice versa. In fact, using formal semantic tools, it is possible to prove the following:

6. a. LF1 entails LF2 (LF1 ⊨ LF2)

b. LF2 does not entail LF1 (LF2 ⊭ LF1)

When there is an entailment relation like this between two LFs, it is sometimes hard to distinguish them empirically. For example, in 5, both LF representations were true in S1. And so S1 could not be used to distinguish them.

With this background, consider now (1b) above. Applying QR to (1b), there are two logical form representations:

7.

LF1: (surface scope)

[TP <[every boy]1> [TP <[some girl]2> [TP DP1 loves DP2]]

LF2: (inverse scope)

[TP <[some girl]2> [TP <[every boy]1> [TP DP1 loves DP2]]

These LF representations can be paraphrased as follows:

8. LF1 (paraphrase): For every boy, there is some girl who he loves.

LF2 (paraphrase): There is some girl who every boy loves.

Now consider the following two situations:

9.

S1: John loves Sue, Bill loves Mary, Chris loves Kathy.

(and there are no other people or relationships involved)

S2: John love Sue, Bill loves Sue, Chris loves Sue.

(and there are no other people or relationships involved)

Calculating the truth value of each LF representation in each situation (calculation omitted), we have the following results:

10. S1 S2

LF1 (SS) true true

LF2 (IS) false true

In this case, LF2 is stronger than LF1, in the sense that LF1 is true whenever LF2 is true. In other words, we have the following entailment relations:

11. a. LF2 entails LF1 (LF2 ⊨ LF1)

b. LF1 does not entail LF2. (LF1 ⊭ LF2)

Now let’s ask what the evidence is for the inverse scope representation in LF2 of the sentence in (1b). Unfortunately, because of the entailment relations, there is no situation (such as S1 or S2) where LF2 (inverse scope) is true and LF1 (surface scope) is false. This makes it difficult to justify the inverse scope representation purely on the basis of situations where the LF representations are true. Therefore, the question is whether in this case one needs an inverse scope representation of (1b) at all.

To put matters more intuitively, concerning (1b), if it is true that there is some girl who every boy loves (inverse scope) then it is also true that for every boy, there is some girl who he loves (surface scope). So perhaps in this case we could say that only the surface scope representation in LF1 is possible, but it is vague as to whether everybody loves the same girl or not.

We can summarize as follows:

12.

a. Some boy loves every girl.

i. Inverse scope does not entail surface scope.

ii. There is a situation where inverse scope LF is true,

but surface scope LF is not true.

b. Every boy loves some girl.

i. Inverse scope entails surface scope.

ii. If inverse scope LF is true in some situation,

then so is surface scope LF.

Notice that LF1 and LF2 in (7) are not equivalent, and so they are false in different cases (see (10)). In particular LF1 is true in S1 but LF2 is false in S1. This leads to the suggestion that we could use falsity to distinguish the interpretations of LF1 and LF2.

So, given S1, consider the following dialogue:

13. (in S1)

A: Isn’t it strange that every boy loves some girl? (with structure LF2)

Intended: Isn’t it strange that there is some girl that every boy loves?

B: That is false, they all love different girls.

If this were an acceptable discourse it would provide evidence that both LF1 and LF2 are needed for (1b). Since it would show that LF2 (inverse scope) is possible, and is false in a particular situation. Unfortunately, it is not very easy to access the inverse scope interpretation in 13A, and the discourse seems forced.

Another similar way to distinguish LF1 and LF2 is through constructions like “It is false that” and “It is not the case that”. These constructions negate the embedded clause:

14. a. It is false that every boy loves some girl.

Intended: It is false that there is some girl that every boy loves.

b. It is not the case that every boy loves some girl.

Intended: It is not the case that there is some girl that every boy loves.

If LF2 (inverse scope) were a possible representation here, then the sentences would be true in S1. Once again, the facts are not that clear.

There are ways to bring out inverse scope. For example, adding the word particular makes it much easier:

15. a. It is false that every boy loves some particular girl.

Intended: It is false that there is some particular girl that every boy loves.

b. It is not the case that every boy loves some particular girl.

Intended: It is not the case that there is some particular girl that every boy loves.

This kind of sentence where one is negating the inverse scope LF representations suggests that inverse scope is also needed in examples such as (1b).

In summary, based on truth conditions, it is easier to justify an inverse scope LF representation for (1a) than it is for (1b). However, given the clear need for an inverse scope LF representation for (1a), the null hypothesis is that such an inverse scope representation should also be available for (1b). That conclusion leaves open the difficult status of the judgments in (13) and (14).

New Horizons in Morphology as Syntax: My Three Published Papers

MaS (Morphology as Syntax) is a framework which asks the question of whether putative morphological phenomena can be can be accounted for in terms of syntactic principles and operations. In MaS there is no morphological component, nor are there any post-syntactic morphological operations.

So far, I have been able to publish three papers in this framework. I list the papers, and their abstracts below, in chronological order. These three papers are enough to get a general feel for the framework, and some tools to begin working with it.

I would really love to hear from people who are doing work in a related spirt, taking syntax seriously in the analysis of putative morphological generalizations.

1. Spanish usted as an Imposter (2021, Probus, with Francisco Ordóñez)

Abstract: Across dialects, Spanish uses the third person forms usted and ustedes to refer to the addressee. In this squib, we propose an imposter analysis of these forms in the framework of Collins and Postal (2012. Imposters. MIT Press, Cambridge.). (link)

2. Towards a Theory of Morphology as Syntax (2023, Studies in Chinese Linguistics, with Richard Kayne)

Abstract: Phenomena traditionally thought of as morphological can be accounted for in terms of syntactic operations and principles, hence bringing forth questions that traditional morphology fails to ask (for instance, concerning the licensing of empty morphemes). The language faculty contains no specific morphological component, nor any post-syntactic morphological operations. (link)

3. A Syntactic Approach to Case Contiguity (2025, Continua)

Abstract: Building on the empirical results and theoretical insights of Caha (2013), I show howto derive the Case Contiguity Constraint in a syntactic theory of morphology. In particular, I show how to derive *ABA in the domain of case syncretism without appeal to late insertion. (link)

How to Syntax 3 (‘only’ Modifying Noun Phrases)

This is the third of a series of blog posts showing how I think about a syntax problem when I first notice it. For the first and second installment, see:

How to Syntax I (the now that-Construction)

How to Syntax 2 (Adverbs with Attributive Adjectives)

That first glimpse of a problem is an important period in thinking about syntax. I will occasionally choose phenomena that I notice, and talk about them in an informal fashion, breaking down the process of preliminary syntactic exploration. That is, I am just thinking off the top of my head (brainstorming), with few or no revisions. Ideally, I will give myself a time period of two to three hours maximum to prevent polishing. The focus of the discussion will be on process. I am not trying to come up with a polished analysis. Of course, if people suggest references for me to look at, I will look at them later, but that would be a second stage of thought, not the preliminary exploration.

Data Discovery

I do not recall how I first noticed the construction. I have known about it since at least Fall 2021, where I mentioned it briefly in a seminar I taught. But I kept it on the back burner, knowing at some point I would get back to it.

Combinatorial Basics

I want to investigate phrases such as the following:

(1) a. He is the only candidate.

b. The only candidate is sitting over there.

c. I met the only candidate.

For convenience, I will occasionally refer to this as the ‘only’-NP construction. An alternative involves the expression ‘one and only’, as in ‘He is the one and only candidate.’ Another related expression is ‘the lone candidate’. I leave aside these alternatives for now.

At first glance, the construction consists of a definite determiner, followed by ‘only’, followed by a noun. 

An indefinite determiner is difficult or impossible to get:

(2) a. *He is an only candidate.

b. *An only candidate is sitting over there.

c. *I met an only candidate.

The examples in (2) all seem unacceptable to me, putting aside fixed phrases like ‘an only child’. Similarly, in the following (b) examples, the determiners with ‘only’ seem unacceptable:

(3) a. John is the only person I know.

b. John is a (*only) person I know.

(4) a. The only weird person is sitting over there.

b. That (*only) weird person is sitting over there.

(5) a. The only candidate is sitting over there.

b. Every (*only) candidate is sitting over there.

But possessors work fine:

(5) a. He is my only friend.

b. My only friend is sitting over there.

c. I met my only friend.

(6) a. his only friend

b. your only friend

c. their only friend

When there is an adjective or a numeral in the DP, it must follow ‘only’:

(7) a. The only big dog was in the corner.

b. *The big only dog was in the corner.

(8) a. The only two reporters were out for lunch.

b. *The two only reporters were out for lunch.

Although focus particles such as ‘only’ and ‘even’ overlap in their distributions (see (9) and (10)), the focus particle ‘even’ cannot be used in the same way (see (11)):

(9) a. Only John left early.

b. Even John left early.

(10) a. John only likes Mary.

b. John even likes Mary.

(11) a. *He is the even candidate.

b. *The even candidate is sitting over there.

c. *I met the even candidate.

Paraphrase and Entailment

Usually when ‘only’ modifies a constituent X, X is in focus. The general scheme is this, where (b) and (c) are entailments of (a):

(12) a. Only John came to the party.

b. John came to the party, but nobody else did.

c. It is not the case that Mary (or anybody else) came to the party.

In other words, for this example any DP substitution for John is false, as shown in (12c).

Similarly, when ‘only’ appears with a predicative constituent:

(13) a. John is only watching TV.

b. John is watching TV, but he is not doing anything else.

c. It is not the case that John is working (or doing anything else).

Now consider how such a paraphrase would work with the ‘only’-NP construction:

(14) a. The only dog was in the corner.

b. The dog was in the corner, but for no other x was the x in the corner.

c. It was not the case that the cat was in the corer (or any other animal).

These truth conditions are just wrong. Clearly, for (14a), it is possible that a cat could also be in the corner. In other words, interpreting [the only NOUN] as a focus construction where NOUN is in focus does not give the right results.

So what is the right paraphrase? It seems to be the following:

(15) a. The only dog was in the corner.

b. The x such that only x is a dog was in the corner.

(16) a. I saw the only person I know.

b. I saw the x such only x is a person I know.

In formal notation, the translation is (assuming a Fregean account of the definite article):

(17) a. the only dog

b. ιx[dog(x) ∧ ∀y[dog(y) --> x = y]]

‘The unique x such that x is a dog and for every y, 

if y is a dog, the y = x’

But such a definition poses a bit of a conundrum. Consider the definition of the definite article from Heim and Kratzer (pg. 75):

(18) a. [[the]] = λf. f ɛ D<e,t> and there is exactly one x such that f(x) = 1.

                 the unique y such that f(y) = 1.

b. [[the dog]] = Presupposition: there is exactly one x such that dog(x) = 1.

        Denotation: the unique y such that dog(y) = 1.

Comparing (17) and (18), it looks like ‘only’ is just restating the presupposition of the definite article. In other words, ‘only’ appears to be semantically vacuous when appearing with the definite article. But is this true?

(19) a. My friend is over there.

b. My only friend is over there.

If you and I are at a party, and I say (19a) pointing to the corer, you might infer that my unique friend at the party is over there. There is a uniqueness presupposition, but it is very fluid and flexible. In fact, if I say (19a), it could be the friend of mine that we were just talking about, even if I have lots of friends at the party.

On the other hand, (19b) seems less fluid and more rigid. Once again, if we are at the party, and I say (19b) pointing to the corer, your natural inference is that I don’t have any other friends in my life. If I wanted it to be clear we were talking about friends at the party, I would have to say: ‘My only friend at this party is over there.’

So it seems at the very least there is a difference in the possibility for domain restriction between the two forms (with and without ‘only’). I leave investigating this to further work, since today’s session is just supposed to be the first glimpse of the construction.

Speculations on Structure

As part of the first glimpse into the construction, I would like to speculate on the syntactic structure. The value of the speculation is that is provides a concrete way to test further predictions (against the structure). So consider the following sentence repeated from above:

(20) a. The only dog was in the corner.

b. The x such that only x is a dog was in the corner.

My suggestion is that ‘only’ does not modify its sister semantically in the same way that the focus particle normally modifies its sister (creating a quantificational structure, quantifying over alternatives). 

First, I assume that in definite DPs like ‘the dog’, ‘the person’, there is a null category which is the argument of the noun phrase (see Collins 2024 for discussion, see also Koopman 2003, 2005). For convenience, let’s call that empty category ec:

(21) [the [ec dog]]

Second, this is enough to start to account for ‘only’-NP, if we say that ‘only’ modifies the empty category, with the interpretation in (22b).

(22) a. [the [[only ec] dog]]

b. the x such that only x is a dog.

I leave out the labels of the constituents, just giving the bare minimum for this first glace.

Combinatorial Basics Revisited

Now that I have put in place the sketch of an analysis, we can revisit some of the basic combinatorial properties listed above. For lack of time, I will not discuss all of them, but consider again the ordering of an adjective and ‘only’, repeated below:

(23) a. The only big dog was in the corner.

b. *The big only dog was in the corner.

I assume the structure is:

(24) a. [the [ec big dog]]

b. the x such that only x is a big dog

In other words, ‘big dog’ is acting as the predicate, and ec is the subject, so there is no way for ec to intervene between ‘big’ and ‘dog’. Therefore, since ‘only’ modifies ec, there is no way for ‘only’ to appear between ‘big’ and dog’.

Negative Polarity Item Licensing

A well-known property of ‘only’-phrases is that they license negative polarity items, as shown in (25):

(25) a. Only John knows any physics.

b. *John knows any physics.

But clearly, such licensing does not extend to the ‘only’-NP construction:

(26) a. *The only candidate knows any physics.

b. *The candidate knows any physics.

The difference between (25) and (26) follows from the analysis I outlined above in (22) above since on that analysis ‘only’ takes scope internal to the DP. In other words, in (25a) [only John] takes scope over the entire clause. In (26a), ‘only’ takes scope internal to the subject DP, not over the entire clause.

However, internal to the modified NP, a NPI becomes possible:

(27) a. *The candidate who has ever been to France is John.

b. The only candidate who has ever been to France is John.

This fact could also follow from the analysis is if the correct way to break it down is:

(28) a. the only candidate who has ever been to France

b. the x such that only x is a candidate and only x has ever been to France

I leave working out details to future work.

Technical Glitch

My preliminary analysis rests on the assumption that ‘only’ can modify an empty category ec. But in general, it does not look like ‘only’ can modify empty categories.

(29) a. I want (only) Mary to go.

b. I want (*only) PRO to go.

Intended: ‘I want only myself to go.’

(30) a. Only John left.

b. The man that (*only) ec left.

Example (29b) shows that ‘only’ cannot modify PRO. Example (30b) shows that ‘only’ cannot modify the trace of subject extraction in a relative clause.

It is unclear to me for the moment how to get around this problem. But acknowledging technical glitches is an important part of the discovery process. It could lead you to reject your initial analysis, or to give a much better analysis down the road.

One possibility, no considered so far, is that ‘the’ itself is the argument of the noun, so the structure would be the following:

(31) [[the only] dog]

If this were the case, it would look a bit like post focus ‘only’ in the following example:

(32) John only is allowed to enter the fort.

I put the issue aside for now.

Conclusion

I have spent a few hours analyzing the ‘only’-NP construction. The preliminary syntactic methods I have applied are the following:

(31)

a. Combinatorial basics

b. Paraphrase

c. Speculations on structure

d. Explaining facts in terms of structure

e. NPI licensing

f. Technical Glitch

The theoretical significance of the construction is that it may support the structure of DPs and the Argument Criterion as outlined in Collins 2024.