Notes and questions on conditionals, topicalization, and related topics


Conditionals

ASL uses the same prosody for conditionals as it does for yes-no questions, as illustrated in (1) (Hill et al. 2019:61, (4)).

(1) a. RAIN TOMORROW, GAME CANCEL
It will rain tomorrow, so the game is cancelled.
b. ________________br
RAIN TOMORROW, GAME CANCEL
If it rains tomorrow, the game will be cancelled.

Hill et al. 2019:61 assume without discussion that the examples in (1a) and (1b) are each a single sentence, but the examples actually raise a difficult and important question - namely, "What counts as a sentence?" This question doesn't generally get the attention it deserves in syntactic theory, but one measure of its importance is that two seminal early researchers on sign language, Ursula Bellugi and Ed Klima, gave it as a research project to one of their most gifted students, Scott Liddell, who went on to become a very influential sign linguist himself (he is the one who insists on the deictic aspect of sign language pronouns and indicating verbs).

Clearly, we might replace the comma in the English gloss of (1a) with a period, and then we would get two separate orthographic sentences. More seriously, we might take the position that the proper gloss for (1b) isn't (1a), but the two-sentence sequence in (2).

(2)     Will it rain tomorrow? (Then) the game will be cancelled.

(2) is admittedly awkward as a discourse fragment in English, but it might still be the right gloss for (1a). I'm not sure what the awkwardness is due to in the particular case of (2), but vernacular English does allow perfectly acceptable two-sentence sequences with conditional semantics. Can you think of some examples?

Hint: Hesh's Seafood

In fact, the first part of a question-consequence pair doesn't need to be a full sentence, as is evident from vernacular English examples like (3).

(3) a.   Need help? Just lemme know.
b.   Questions? Shoot me an email.

Question: Does ASL allow examples like (3) as well? If so, are they more or less common than in English? If more common, why might that be?

In this connection, it's worth nothing that English allows verb-initial conditional clauses that are clearly not sentences on their own, but rather truly integrated subordinate clauses. Evidence for subordinate clause status comes from word order patterns like those in (4)-(6).

Condition > consequence Consequence > condition
(4) a. Do you need help? Just let me know. b. # Just let me know. Do you need help?
(5) a. Should you need help, let me know. b. Please let me know should you need help.
(6) a. If you (should) need help, let me know.      b. Please let me know if you (should) need help.

When the condition is expressed as a yes-no question (with rising intonation that isn't indicated here), as in (4), then the condition must precede the consequence. (The '#' in (4b) indicates infelicity; in other words, the material follows the grammatical rules of English, but the way the discourse is put together is weird.) Otherwise, if the condition is expressed as a subordinate clause, the other order is also possible, and this is true regardless of whether the subordinate clause is marked by a clause-initial verb, as in (5), or by the explicit subordinating conjunction if, as in (6).

Question: Is it more sensible to treat (1a) in ASL as two sentences (parallel to (4a)) or as one (parallel to (5a))? What do we need to know before we can answer?

Three types of "topicalization"

Gap topicalization

In this section, we describe three types of "topicalization" that are well known from English and other spoken languages. Our specific intention in presenting them is to see if we can match them up with the three types of topic markers (tm1, tm2, tm3) mentioned in the section on topicalization in Hill et al. More generally, we're always interested in finding similarities and parallels between signed and spoken languages.

We'll start with an English sentence exhibiting basic SVO word order, as in (7).

(7)   Leslie likes peas.

Sentences like (7) can occur in contrastive contexts like (8). The underlining is intended to represent contrastive stress.

(8)   Leslie likes peas. (But) she doesn't like okra.

Both sentences in (8) continue to exhibit basic SVO order, but English also has a syntactic rule that moves a contrastively stressed phrase to the beginning of its clause. The effect of this rule is shown in (9), where the gap represents the original position of the moved phrase.

(9)   Leslie likes peas. But okra, she doesn't like ____.

In the syntax and pragmatics literature, this rule is usually called just plain topicalization. But for clarity in connection with the discussion in Hill et al., we'll call this rule and the resulting construction gap topicalization. Notice that in "out of the blue" contexts like (10a), gap topicalization as in (10b) is infelicitous.

(10) a. Guess what?
b. # Peas, Leslie likes ___.

In order for gap topicalization to be felicitous, it must meet a certain condition - namely, that a superset of the contrastive expression must already be salient in the discourse. This condition is not satisfied in (10). By contrast, in (9), peas evokes (= implicitly introduces) the superset 'vegetables', and so the subsequent topicalization of okra is felicitous. The superset 'vegetables' can also be introduced explicitly, as in (11).

(11) a. What kind of vegetables do you like?
b. Well, peas, I like ___ fine, but okra, I find ___ repulsive.

Left dislocation

In contrast to gap topicalization, left dislocation has a pronoun corresponding to the clause-initial phrase in the position of the gap, as shown in (12).

(12) a.   Peas, I like them.
b.   Bill, I've known him since first grade.

The relation between the pronoun and the clause-initial phrase is anaphoric. By this, we mean that both expressions refer to the same discourse entity. For instance, in (12a), the pronoun them refers to the same set of objects as the clause-initial noun phrase peas, and in (12b), him refers to the same individual as Bill.1

Left dislocation is not all that common with anaphoric object pronouns, at least not in English, but it is very common with subject pronouns and possessive pronouns within subjects. (13) shows examples of each type; the anaphoric pronouns are in italics.

(13) a. My sister Linda, she can be a bit nuts.
b. My friend Danny, his neighbor can fix anything.

As an aside: Prescriptive grammar frowns on such sentences, at least in writing. Maybe you've had the experience yourself of having a teacher comment on your use of this construction. If I were wearing a copy editor's hat, I myself would make notes in the margin, urging you to omit she in (13a) and to replace Danny, his by the simple possessive Danny's in (13b). But that's not the hat I'm wearing now! I've never heard of any attempt to explain why left dislocation is prescriptively deprecated. After all, it's not "illogical" in the way that negative concord (= double negation) is said to be. Perhaps, back when parchment and paper were expensive, the construction was felt to be redundant and therefore wasteful of a scarce resource? Or perhaps redundancy is seen as a sign of lack of intellectual rigor? That would provide a single explanation for the prescription against left dislocation and negative concord in English. Those are just two guesses. However, whatever prescriptive grammar says, left dislocation is grammatical in both English and ASL!

Hanging topics

Finally, there is a construction closely related to left dislocation called the hanging topic construction. In this construction, illustrated in (14), the clause-initial phrase and its clause-internal counterpart are semantically related, but not in the anaphoric way just described.

(14)   ? Vegetables, I like peas.

The pure hanging topic construction in (14) is not terrifically natural in English, but the acceptability improves when the clause-initial phrase is prefaced with as for, as in (15). If as for sounds bookish to you, try the more vernacular speaking of in the following examples.

(15)   As for vegetables, I like peas.

Even with as for, sentences like the next one in (16) sound a bit awkward in English, because the relationship between 'circus' and 'elephants' is not as close as the superset relation between 'vegatables' and 'peas'.

(16)     As for the circus, I like the elephants best.

But in some languages, sentences like (17) are perfectly ordinary, even without an introductory as for. This is famously (famously in syntax circles, at least) the case in Japanese and Chinese.

(17)     The circus, I like the elephants best.

Question: Does ASL pattern with English or with Chinese/Japanese with respect to the acceptability of hanging topics?

Prefixing as for to the sentence-initial phrase is also possible with left dislocation, but not with gap topicalization, as shown by the contrast in acceptability in (18).

(18) a. As for peas, Leslie likes them.
b. * As for peas, Leslie likes ___.

Left dislocation and the hanging topic construction are also similar in another way - namely, in not being subject to the discourse condition that gap topicalization must meet. Recall that with gap topicalization, the superset of the clause-initial phrase needs to be already salient in the discourse, whether implicitly evoked or explicitly introduced. This is not necessary with left dislocation or the hanging topic construction. A natural way of making sense of this fact is to consider the part of the sentence after the initial phrase. With left dislocation or the hanging construction, that part can stand on its own. (Leslie likes them in (18a) is a complete sentence.) We can think of the initial phrase (Peas, As for peas, Speaking of peas) as playing the same role as the previous discourse does with gap topicalization. By contrast, with gap topicalization (Peas, Leslie likes), the part of the sentence after the initial phrase (Leslie likes) isn't a complete sentence on its own. So the sentence-initial phrase in gap topicalization is part of the subsequent part of the entire sentence in a way that it isn't in left dislocation or the hanging topic construction. In other words, if there were no discourse condition on gap topicalization, it would like asking the sentence-initial phrase to be part of the subsequent part of the sentence and the preceding discourse at the same time. Apparently, phrases can only do one or the other - they can't multitask.

A complication: Silent pronouns

In many spoken languages, including English, left dislocation is clearly distinguished from gap topicalization by the presence vs. absence of an anaphoric pronoun. But in many other spoken languages, like Chinese and Japanese, this distinction is obscured by the availability of silent object pronouns. This is true of ASL as well. In such languages, what looks an instance of gap topicalization might actually be an instance of left dislocation (with an anaphoric silent pronoun). This issue arises in (19) (= Hill et al. 2019:62, example (5)), where it's impossible to tell from the string (= the sequence of signs) in (18f) whether KISS-FIST is followed by a true gap or a silent pronoun.

(19) Discussion of cheese between A and B:
Speaker A: a. IX_1 REALLY LIKE CHEESE.
'I really like cheese.'
Speaker B: b. OH-YES, IX_1 LOVE CHEESE.
'Oh yes, I love cheese.'
c. IX_1 EAT CHEESE DAILY.
'I eat cheese every day.'
____________bf
d. TYPE IX LIKE?
'What kind do you like?
Speaker A: e. IX_1 LIKE ALL, ONE IX_1 NOT FOND, GOAT CHEESE, BLEH
'I like them all, except one (kind), (and that's) goat cheese. Yuck!'
____________br
Speaker B: f. GOAT CHEESE, IX_1 KISS-FIST!
Goat cheese, I love (it)!'

The discourse context is consistent with gap topicalization, but the presence of the parenthesized pronoun it in Hill et al.'s gloss suggest either that the example is actually left dislocation, or that it is ambiguous between the two types of topicalization.

It turns out that there are contexts that allow us to distinguish the two types because the semantics of the clause-initial phrase is not compatible with left dislocation. The phrases in question contain quantifiers like 'no' or 'every'. As (20) shows, gap topicalization is possible with a fronted quantified phrase (with stress as indicated), but left dislocation is ungrammatical.

(20) Speaker A: a. What kind of cheese do you like?
Speaker B: b. i. { No, every } kind, I like ___.
ii. * { No, every } kind, I like it.

Unfortunately, the ASL counterpart of 'no', the negative quantifer NONEaltvar, is restricted to clause-final position (Hill et al. 2019:77), so a translation of (20.b.i.) into ASL that preserves the word order, as in (21), is ungrammatical for reasons having nothing to do with topicalization per se.

(21) Speaker B: * NONEaltvar TYPE IX_1 LIKE
Intended meaning: 'No kind, I like.'

We therefore need to consider examples with 'every', as in (22).

(22) Speaker B: EVERY TYPE IX_1 LIKE
'Every kind, I like.'

If (22) is grammatical, then we have conclusive evidence that ASL actually has gap topicalization, and not just left dislocation.

Question: Given the discussion so far, can you match up the three topic markers (tm1, tm2, tm3) mentioned in the Hill et al. readings with the three constructions just discussed?

The correlation between types of topicalization and intonation

Let's now consider how the three constructions from the previous section are marked prosodically, in English and in ASL. It's not that easy to determine the correlation between constructions and intonation because intonation is even more unconscious than other linguistic phenomena. Related to this and compounding the difficulty is the fact that it is represented in only the most rudimentary way in ordinary writing. Nevertheless, certain facts seems relatively clear. In particular, to my ear, left dislocation and hanging topics in English both require or very clearly prefer rising intonation. In other words, using conventional punctuation marks to represent intonation, the melodies are as in (23).

(23) a. Peas? Leslie likes them.
b. # Peas. Leslie likes them.

To maximize the parallel between spoken and signed languages, in what follows, we'll use "/" and "\" for rising and falling intonation, respectively, and mark the scope of the intonation with underscores. Using this notation, we can re-represent (23) as in (24).

___/
(24) a. Peas, Leslie likes them.
___\
b. # Peas, Leslie likes them.

The corresponding judgments for hanging topics are given in (25).

________/
(25) a. Vegetables, Leslie likes peas.
________\
b. # Vegetables, Leslie likes peas.

If the judgments in (24) and (25) hold up, then left dislocation and hanging topics are both marked in the same way in English and ASL, where by 'same' we mean that rising intonation in English corresponds to brow raise in ASL.

Sanity check: The "br" in Hill et al.'s examples (8) and (9) should then be obligatory, and they should be infelicitous without "br". Are they?

Before turning to gap topicalization, we'd like to note that rising intonation and brow raise on left dislocation and the hanging topic construction fit well with what we said earlier about how these constructions contrast with gap topicalization. With gap topicalization, a sender (whether speaker or signer) can be sure that the recipient (hearer or viewer) is familiar with the topic under discussion because of the condition requiring the topic to have already been evoked or explicitly introduced. But the discourse condition on gap topicalization doesn't apply in the other two cases. So it makes sense for the sender to check that the recipient is familiar with the topic. This can be done explicitly, as in (26).

(26) a. (Do) you know Bill?
(possible followup if recipient indicates 'no': My neighbor Bill? The one in the house on the corner?)
b. Well, he recently got a puppy.

But just as we don't need a full question in (2) to convey conditional semantics, so we also don't need a full question in (26) to allow a recipient to confirm or disconfirm our assumptions about our shared knowledge. In both cases, the full question can be reduced to a phrase. In the case of left dislocation, the integration can go even further, and the sentence fragment can clearly be integrated with the following discourse, with both parts forming a single sentence.

Let's turn now to gap topicalization and how it's marked intonationally. The facts here are more complex than for left dislocation or the hanging topic construction because the intonation contours depend on whether the clause-initial phrase is quantified or not. Let's start with non-quantified phrases in English. Here, it seems to me that both melodies are possible in English. Perhaps there's a slight preference for falling intonation, as indicated by the question mark for the rising intonation in (27a).

___/
(27) a. ? Leslie likes peas, but okra, she doesn't like ___.
___\
b. Leslie likes peas, but okra, she doesn't like ___.

Given (27), and assuming analogous intonation across modality, we might expect ASL gap topicalization to allow both "br", corresponding to the rising intonation of English in (27a), and a more "declarative" intonation (perhaps "bf", or even more likely "hn"), corresponding to the falling intonation in (27b).

However, we need to keep in mind that the word-for-word counterpart to (27) in ASL would be ambiguous between gap topicalization and left dislocation. In order to make sure we are dealing with gap topicalization in ASL, the sentence-initial phrase needs to be a quantified phrase. More specifically, it needs to be an 'every' phrase (because of the irrelevant complication arising from the word order restrction on NONEaltvar). Before addressing the intonation in the ASL examples, let's take a look at the relevant English example, which is given in (28).

(28) Speaker A: a. What kind of cheese do you like?
________/
Speaker B: b. i. * Every kind, I like ___.
________\
ii. Every kind, I like ___.

Here, rising intonation is clearly ruled out, and only falling intonation is possible.

So now the $64,000 question is: What is the intonation on (22), repeated for convenience in (29)?

(29) Speaker B: EVERY TYPE IX_1 LIKE
'Every kind, I like.'

Depending on the answer, our conclusion about cross-modal similarity in intonation will differ. If (29) doesn't allow brow raise, then intonation works the same way in ASL and in English. But if (29) does allow brow raise, then we are forced to a different conclusion. The generalization would then be that in ASL, brow raise marks all types of topicalization across the board, whereas in English, the intonation contour depends on the type of topicalization. To be continued on the discussion board!

A final puzzle

I'll end these notes by drawing attention to a puzzle that arises in connection with example (2a) from Hill et al. 2019:59, repeated here as (30).

______________________br
(30) a. CHEW-UP POSS_1 SHOES IX DOG
______________________br
b. POSS_1 SHOES CHEW-UP IX DOG
'The dog chewed up my shoes.'

Question: There is something unexpected about one of the variants in (30). What is it?

Some hints: If there's really a single "br" in both variants, what does say about the basic word order of ASL? Conversely, if there's only a single basic word order, as Hill et al. assume, what are the consequences for the intonation in (30)? How many brow raises would you expect?


Notes

1. The same anaphoric function as pronouns can also be served by fairly meaningless epithets, as illustrated in (i).

(i) a.   Peas, I like those little suckers.
b.   Bill, I've known that dude since first grade.