5.1 Infinitival, gerundive and clausal complements in spoken and written language

We will start by considering the general distribution of wh-clefts across written and spoken language. This is because the occurrence of integrated vs non-integrated structures has been closely associated with spoken language and especially with spontaneous unplanned speech. Figure 1 represents the distribution of wh-clefts with do in the wh-clause.

Figure 1.

wh-clefts in COCA (per 1 million)

In the COCA, wh-clefts are over thirteen times more frequent in spoken than in written language, the spoken data being unscripted conversation of TV and radio programmes. Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan1999: 961) report, in a similar vein, that wh-clefts are most common in conversation. The authors attribute this proliferation of wh-clefts in conversation to the fact that the wh-clause often has low information content, as in phrases such as What this means. . . or What I'd like to do. . ., which can be used in the real time of spontaneous speech as a springboard to start an utterance (1999: 963). Others point out other rhetorical functions that, again, may be especially relevant in spoken interactional language: to buy time and hold the floor while the speaker formulates his point, or to give the appearance of fluency by filling in what otherwise would be awkward silences (Hopper & Thompson Reference Hopper, Thompson and Laury2008). Consider example (23), in which one possibility is that the wh-cleft, straightforwardly, foregrounds the focus constituent, but the second possibility is that the cleft construction, together with the filling and hedging expressions well, essentially and I think, allows the speaker time to think and plan what to say.

1. (23) Well, there's all sorts of speculation about what they're going to do, but essentially I think what they want to do is put power back into the hands of prosecutors (spoken, unscripted conversation)

The low information content of wh-clauses is most visible in amalgams with do, in which there is much repetition of lexical and grammatical material; typically, the wh-clause contributes very little information.

1. (24) And what they could have done is they could have given up to him. (spoken, unscripted conversation)

2. (25) Usually what these people want to do is they want to get rid of their pain. (spoken, unscripted conversation)

Spontaneous spoken English is thus the typical domain of wh-clefts. Having said that, below we will investigate the COCA's written and spoken wh-clefts in turn with respect to the degree of their integration.

In the spoken data, 43.5 per cent of all wh-clefts are of the amalgam type, the rest being accounted for by standard clefts (figure 2). Considering that infinitival cleft clauses come in two varieties – bare and to-infinitives – the amalgam variant is the most frequent complement type of the three options available. In written language, on the other hand, 90 per cent of occurrences are of the canonical type, with to-infinitives being markedly less common than bare infinitives (figure 3).

Figure 2.

wh-clefts in spoken English

Figure 3.

wh-clefts in written English

We will now examine specific non-integration effects. Firstly, we will look into infinitival clauses, considered fully integrated syntactically, to find out whether this standard complement type displays any signs of non-integration. Individual examples will be identified according to text type.

5.2 Mismatched copula tense

One feature of a minor lack of syntactic integration is when the tense of the copula does not match that of the wh-clause. This is possible when the present tense copula is used in a past tense environment:

1. (26) I think sexuality is confusing and complex, and what this process did for me is to open the door to be able to discuss my own sexuality. (written, newspaper)

2. (27) they found out that what we were doing is actually using the picture to protect that information (spoken, unscripted conversation)

Tables 1 and 2 represent the frequency of occurrence of copula tense mismatch with bare- and to-infinitives in the COCA. The pattern is more common with the bare infinitive than with the to-infinitive. With bare infinitives, when the choice between the present and past copula is available, i.e. when the wh-clause is in the past tense, the present copula is used in 15 per cent of cases. With to-infinitives, the present copula is used in 9.5 per cent of cases.

Table 1.

Consistency in tense use (bare infinitives)

Table 2.

Consistency in tense use (to-infinitives)

The figures also show that the inconsistent tense use is typically to be found in spontaneous spoken English. For example, for bare infinitives, 20 per cent of spoken wh-clefts are of the mismatching type as opposed to 7 per cent in written language. More to the point, on closer inspection, it becomes clear that, with only two exceptions, all the occurrences in the written medium are in fact quoted speech (in the press) or intended to mimic speech (in fiction). Clearly, unplanned speech is the domain of this non-integration effect. Note that this is in line with the view that speech in general, in comparison to written language, is marked by unintegrated syntax (e.g. Miller & Weinert Reference Miller and Weinert1998; Hopper & Thompson Reference Hopper, Thompson and Laury2008).

The copula has a loose syntactic connection with the cleft structure. It is a connecting point, but also a breaking point for the two clauses. At the same time, the two parts on both sides of the copula have a loose connection with each other, and are to some extent independent (especially in amalgams). Consider example (28), in which the speaker reports a tornado attack – an account of a past event, i.e. when the past tense is normally used. Note that the finite verbs on both sides of the copula are consistently in the past tense, where they need to be in the past tense, but not the copula.

1. (28) And what we did is get in the center of the house which was in the baby's room in the closet. We put some mattresses against the windows and we had a mattress against the door and that's when we heard the big roar and it just took the roof. (spoken, news report monologue)

Although the pattern of tense mismatch is relatively rare with infinitives, it shows that wh-clefts with infinitival complements may display a degree of non-integration. This is rather unsurprising if one sees the copula as an element of the construction that is to some extent independent of the rest of the construction. The purpose of the copula is to link the wh-clause and the cleft constituent, but the tense of the copula may vary between past and present. The mismatched copula tense is another non-integration feature of wh-clefts, although it has not been recognized as such in the previous literature.

Below we carry out an analogical analysis of copula tense use in amalgam wh-clefts. Table 3 represents their quantitative distribution. As can be seen in table 3, inconsistent copula tense use is far more common in amalgam wh-clefts (in 44 per cent of cases) and especially in the COCA's unscripted conversation, as can be seen in the recorded raw frequencies: only eight instances were found in written texts, all of them, once again, representing quoted speech. The greater frequency of mismatched copula tense in amalgams confirms that the connection of the copula to the cleft construction is looser than in clefts with infinitival complements. Consider examples (29) and (30) below. Both are similar accounts of sequences of past actions but the present tense copula alternates with the past tense form.

Table 3.

Consistency in tense use (amalgams)

1. (29) And what I did is I ran downstairs, and I woke up the girls, and I said, you know, that this has happened, and they called him. (spoken, unscripted conversation)

2. (30) So what I did was I went on the Internet, I made a twitter just as a sounding board and it worked. (spoken, unscripted conversation)

Because in amalgams there is a finite verb in the post-copula clause, tense mismatches are likely to be more frequent, as in (31):

1. (31) What they did right is they allow a much wider variety of payment models, (written, newspaper)

In other cases, tense mismatches may be triggered by local factors, e.g. by the proceeding discourse, as below:

1. (32) A: Madonna hired Michael Jackson's publicist.

   B: No, what Madonna did is she has a licensing company now that's going to handle all of her merchandise (spoken, unscripted conversation)

In (32), it is possible that the tense mismatch occurs because the tense of what Madonna did mirrors that of the immediately preceding statement (Madonna hired. . .). Speaker B swiftly moves on to correct speaker A's statement and the resulting combination of the past and present tenses is non-canonical, although perfectly suitable in its context.

To summarize this section, we have seen that the mismatched tense of the copula is another feature contributing to the non-integration of wh-clefts. It is especially common in spoken amalgam wh-clefts, but also observable with bare and to-infinitives, which offers grounds on which one may see even the standard wh-cleft structure as not fully integrated.

5.3 The prosodic separation of the clauses

The loose structure of the wh-cleft construction can also be seen in the frequent prosodic separation of the two clauses in speech. Collins (Reference Collins2006) notes that 94.6 per cent of his wh-clefts are intonationally divided into two or more tone units. Weinert & Miller (Reference Weinert and Miller1996) report pauses separating the two parts of the cleft although they offer no quantification of this observation. This section outlines an analysis of audio data with a view to establishing the extent to which wh-clefts are separated by unfilled (silent) pauses. We will consider the location and frequency of occurrence of such pauses as well as their duration. As the COCA corpus does not allow auditory analysis, the spoken subcorpus of the BNC has been used to retrieve relevant audio files (Pęzik Reference Pęzik2015–). Silent pauses preceding or following the copula were identified and their duration was measured with PRAAT (Boersma Reference Boersma2001). For this analysis, a silence equal to or longer than 0.2 seconds is counted as a pause; in choosing this cut-off point, we follow the common standard adopted elsewhere (e.g. Kendall Reference Kendal2013: 63). Occurrences of clefts with obvious disfluencies such as false starts or self-repairs were excluded as they would have been relevant in pausing.

In all, 277 spoken clefts were examined. In the majority the wh-clause constituted a separate tone unit, as reported by Collins (Reference Collins2006), and the tone unit boundary occurred at the copula. Of the 277 occurrences, 73 (26 per cent) were additionally separated by a pause of at least 0.2 seconds. These pauses typically occurred before the copula, with only nine cases (12 per cent) in which the pause occurred after the copula. Thus, examples (33–35) are prosodically typical of the 73 clefts with pauses (the numbers in brackets represent the location and duration of pauses):

1. (33) What you can do [0.38] is you could turn the graph that way up. (BNC, spoken, classroom)

2. (34) and what I want to do [0.22] is to look at it er over that particular period (BNC, spoken, a training course in a commercial context)

3. (35) now ideally what I’d like to do [0.80] is to have a chat with somebody, (BNC, spoken, private conversation)

All three types of copula complementation (bare infinitives, to-infinitives and amalgams) showed prosodic separation to approximately the same degree, i.e. 26 per cent of bare infinitives, 24 per cent of to-infinitives and 27 per cent of amalgams were found to be separated by silent pauses at the copula. Thus, in this respect, infinitival clefts can be said to be as (non-)integrated as amalgams. Table 4 shows the quantitative contribution of each complementation type to the total of the sample, i.e. 277 items.

Table 4.

Pauses in BNC

As can be seen, wh-clefts are not nearly as frequently separated by a pause as they are by a tone unit boundary. However, on average, in one in every four cases, the wh-cleft structure is perceptibly divided by a pause. In fact, pauses longer than 0.5 seconds, i.e. far longer than the minimum threshold, were most frequent. The pauses were all the more noticeable given that the pause co-occurs with the tone unit boundary. Table 5 shows the durations of the pauses.

Table 5.

Length of pauses

Functionally, these pauses may be boundary markers used for segmentation purposes, and/or they may have a non-segmenting communicative function of drawing the listener's attention to the upcoming information (cf. Murkherjee's (Reference Mukherjee2000) concept of anticipation pauses). In either case, the pausing produces prosodic segmentation that complements the separation by tone unit boundaries reported by Collins (Reference Collins2006), as well as the phonetic reduction and faster tempo of wh-clauses reported by Weinert and Miller (Reference Weinert and Miller1996) (also observed in this study). In combination, all these prosodic phenomena render the cleft construction prosodically non-integrated to a considerable degree and also facilitate the use of the wh-clause as a formulaic chunk. In clefts such as those in (33)–(35), the wh-clauses are relatively fixed, formulaic chunks allowing limited modifications and with low information content which can be used for a variety of rhetoric functions (see Hopper & Thompson Reference Hopper, Thompson and Laury2008). When such semi-fixed phrases are used for such purposes, they are likely to be prosodically separated from the focus phrase.

Another prosodic argument pointing to the formulaic status of the wh-clause is the fact that in 88 per cent of cases the copula is prosodically not part of the wh-clause, suggesting a strong tendency for the prosodic separation of the wh-clause and the copula, and indicating a formulaic template that consists of the wh-clause on its own. Interestingly, in two cases of a post-copula pause, there was an additional pre-copula pause, further indicating the tendency to separate the wh-clause and the copula. All in all, in only 7 cases (9 per cent) is the copula unseparated from the wh-clause by a pause. This is compatible with Miller (Reference Miller2011: 37–8), who observes that when the copula is prosodically part of the wh-clause (including a pause following the copula), as in (36), the prosodic characteristics of both pre- and post-copula clauses suggest an imperative clause analysis.

1. (36) Er, what you're going to do is go in between the camera shop and the lefthand side of the page (Miller Reference Miller2011: 37)

5.4 Omission of the copula

Another non-integration feature, and a more radical one, is the omission of the copula, which results in a fragmented structure in which the variable contained in the wh-clause lacks an overt syntactic link to its value in the focus constituent. Such omission is typically seen in amalgams rather than standard wh-clefts:

1. (37) I mean, what that ad did, it got Gil Aust a lot of press coverage in Washington and other places (spoken, unscripted conversation)

2. (38) What I would do with the glasses, I would look to see the ridge detail (spoken, unscripted conversation)

Clefts with infinitival complements are also subject to copula omission, as in (39) from Hopper & Thompson (Reference Hopper, Thompson and Laury2008: 102):

1. (39) What you oughta do though, Mar, cook all the fish.

In (39), cook all the fish might also be interpreted as the speaker's directive given to Marilyn, thus inviting an interpretation under which the phrase is an imperative clause. In this way some copula-less infinitival complements have an ambiguous status (see Weinert & Miller Reference Weinert and Miller1996: 195; Miller & Weinert Reference Miller and Weinert1998: 120–3). However, in other examples, including (40) and (41), the imperative reading is unlikely and the utterances are best seen as copula-less wh-clefts.

1. (40) Dealers aren't in the business to do social work. What they want to do, get you to come in. (spoken, news report monologue)

2. (41) First, what we have to do, curb our spending where we can. (spoken, unscripted conversation)

Admittedly, copula omission occurs only sporadically with infinitival complements, as evidenced by only eleven such instances in the COCA. Amalgams, on the other hand, allow the omission more freely (373 instances) and by virtue of this fact, once again, exhibit a less integrated structure. Table 6 represents the extent to which copula omission is observed with both types of complementation. As before, occurrences found in written language constitute quotations from unplanned speech (in the press) or mimic its style (in fiction).

Table 6.

Copula omission in COCA

5.5 Omission of subject pronouns

Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan1999: 963) note that ‘cleft constructions are syntactically integrated, making use of function words to build a close-knit structure’. However, in amalgam wh-clefts, this syntactic integration built with, for example, pronominalization can be disrupted by the ellipsis of a subject pronoun in the focus phrase, as a result of which finite clauses lack subject NPs.

1. (42) Clearly, what television does is instantly immerses an entire continental nation in a common, single experience, (spoken, unscripted conversation)

2. (43) So what the prosecutor did is tipped his hand that he's going to investigate this. (spoken, unscripted conversation)

In other cases, the pronoun is ellipted along with other function words such as auxiliaries (sentence (44)) and the negative not, as in (45):

1. (44) What we ought to have done is deployed troops from the area [. . .] (spoken, unscripted conversation)

2. (45) What he hasn't done is adopted the rhetoric of his predecessor. (spoken, unscripted conversation)

In terms of the ellipted material, the surface forms of these constructions resemble those of canonical wh-clefts in which the same function words are omitted and the infinitive is all that is left of the verb group. The difference in surface forms is that finites replace non-finites, e.g. what we ought to have done is deploytroops vs what we ought to have done is deployedtroops. In this sense then, ellipsis of pronouns improves the integration of amalgams, although the pattern is non-canonical and less integrated than classic wh-clefts. It may even be a source of confusion, as when this type of clefts are interpreted predicationally rather than specificationally. Consider (46):

1. (46) What he should have done was telegraphed at the beginning, (spoken, unscripted conversation)

Sentence (46) is ambiguous between a predicational reading (‘someone telegraphed what he should have done’) and a specificational reading (‘what he should have done is telegraph’). The finite verb telegraphed is responsible for the ambiguity as it makes possible, together with was, a passive voice reading. The use of the infinitive would resolve the ambiguity and improve the syntactic integrity of the sentence. The use of finite clauses in examples (42)–(46) brings these constructions closer to the amalgam type rather than standard wh-clefts. Consider the typology suggested below ((47) and (49) are invented examples modelled on (46)):

1. (47) What he should have done was he should have telegraphed at the beginning, (unellipted amalgam)

2. (48) What he should have done washe should have telegraphed at the beginning, (ellipted amalgam)

3. (49) What he should have done was telegraph at the beginning, (standard wh-cleft)

Amalgam wh-clefts are exceptional in that they allow finite clauses to complement do as the verb of the wh-clause. These finite clauses can be complete with subject NPs and copy lexical material from the wh-clause, as in (47), or they may involve ellipsis, as in (48). In terms of structural integration, we will claim that example (48) has a more integrated structure than (47), and the reasons are the following. Firstly, the copula cannot be as easily omitted as in the case of unellipted amalgams such as (47) – no such cases have been found in the COCA. Secondly, in the 163 occurrences of finite clauses with pronoun ellipsis there are no commas separating the two clauses, which is often the case in transcribing amalgams in corpus data and which is quite universal when the copula is omitted. Although this is not a reliable diagnostic, the comma seems to reflect the perceived non-integration of amalgams. We take the absence of the comma in clefts with pronoun ellipsis to be an indication of better integration, compared to unellipted amalgams. Thirdly, the ellipsis that occurs in examples such as (48) eliminates repetition of material, thus creating a more coherent and close-knit structure. In conclusion, we propose the following scale of decreasing syntactic integration:

• standard wh-clefts > ellipted amalgams > unellipted amalgams

Of course, the scale is not complete in detail as specific representatives of one and the same type may still vary as to specific non-integration features. For example, within the ellipted amalgam type, the tense of the copula may or may not match the tense of the wh-clause, thus maintaining or undermining integrity, e.g.

1. (50) What I did is took that photographic evidence, color photographs, and shipped it to seven different experts (spoken, unscripted conversation) [past-present mismatch = integrity]

2. (51) What I did was took that photographic evidence, color photographs, and shipped it to seven different experts (modified) [tense concord = higher integrity]

Table 7 shows the extent to which subject pronoun omission is present in amalgam wh-clefts in the COCA.Footnote ⁷ As can be seen, subject pronouns are dropped in 5 per cent of eligible cases.

Table 7.

Pronoun omission in COCA

5.6 Non-coreferential subject NPs

Another non-integration effect in amalgam clefts is the non-coreferentiality of subject NPs, typically pronouns. Canonically, the two subjects should be identical, as in (52), or coreferential, as in (53) and (54):

1. (52) what he began to do was he began to separate himself from everybody (spoken, interview)

2. (53) And what South Africa did was they learned from that experience. (spoken, unscripted conversation)

3. (54) because what mathematicians do is we look at the world and we see and we identify patterns (spoken, unscripted conversation)

The coreference of the two subjects brings the two clauses together, thus improving sentence integrity. On the other hand, lack of coreference undermines integrity, as in (55) and (56):

1. (55) A: How did you guys pull this off? How did you get ants on a space shuttle?

   B: Well, actually, what we did is it started as a normal high school science project. (spoken, unscripted conversation)

2. (56) It used to be you'd say something to management and they'd say “I'm the boss. You're the worker.” Now what they do is we sit down and talk about the problem, (written, newspaper)

In (55) and (56), although the changes in subject NPs weaken the integration of the clefts themselves, they may be seen as perfectly fitting in the context of the preceding discourse. In (55) we in B's answer is used in response to you guys in A's question. Speaker B begins with the kind of wh-clause (what we did) that is a sensible response to A's question (How did you. . .?) but then switches to a non-coreferential subject to accommodate the focus phrase. Similarly, in (56) the speaker switches from they (i.e. management) to the more inclusive we, i.e. management and the worker. The implication is that it is the management's new policy now to solve problems together (‘what they do now is include the worker in the discussions’).

Similar examples of splitting groups of participants are illustrated in (57), in which they includes the father, and in (58), in which we includes I.

1. (57) So what they did was, the boy's father testified about how much the 9-year-old missed his mom. (spoken, unscripted conversation)

2. (58) But what I do do is we have wonderful discussions about princesses that really do exist. (spoken, unscripted conversation)

Still, in other cases, the use of non-coreferential subjects does not seem to be dictated by the surrounding discourse. It is in these cases that the lack of integration is most detectable. Consider the fragmentation and incoherence brought about by the lack of coreferential subjects in examples (59) and (60).

1. (59) A: Were not critical of the plans or the employers or unions that have negotiated them.

   B: No, you're – no, you are critical – you – you are critical of taxing them. You are critical of the bill for wanting to tax them.

   A: What the report does is, they look at the taxes. And there are two parts to the tax. One is an excise tax, a new health care tax that would be on pharmaceutical price – costs on devices on health insurance. (spoken, unscripted conversation)

2. (60) It is an organization that has tentacles into a lot of places. And so what they're doing in Pakistan, we are working with the Pakistani government. That's why you have military raids in the frontier region in Pakistan. (spoken, unscripted conversation)

In a number of wh-clefts in the COCA, non-coreferential subjects are used to produce an interesting effect that is not present in examples (54)–(60) above. In these clefts the NPs represent two distinct entities (X and Y) that enter a kind of cause-and-effect relationship in which the constructions may be paraphrased as ‘what X did is cause Y to do sth’. Examples (61) and (62) are two cases in point.

1. (61) So what these two science papers did was I realized how to make that happen. (spoken, unscripted conversation)

2. (62) What the Internet has done is, we have given up being the purveyor of top-secret information. (written, newspaper)

In clefts such as above, the use of non-coreferential subjects does not generate confusion, incoherence or reference problems, as is the case in (59) and (60). The use of non-coreferential subjects makes perfect sense as the two NPs refer to distinct entities which are interrelated in a way that is predictable: ‘X causes Y to do sth’; here, the two examples can be paraphrased as ‘these two science papers made me realize . . .’ and ‘the Internet has caused us to give up . . .’. In terms of logic then, the type of do-clefts represented by (61) and (62) is clearer and more transparent than examples (59) and (60); structurally, however, (61) and (62) are still non-canonical and marked by a degree of non-integration. This type is especially frequent with what it does as the wh-clause (see (63) and (64)) and also includes focus constituents introduced by that-complementizer (65).

1. (63) Well, in reality this experiments doesn't measure neutrino mass directly, what it does is we are able to distinguish between two types of neutrinos, (spoken, unscripted conversation)

2. (64) Now, what it does is the enzymes in the bacteria breaks down (spoken, unscripted conversation)

3. (65) if the police are seen as unfair in their policing, what that will do is that people will take the law into their own hands (spoken, news report monologue)

What it/that/this does is especially convenient in referring to something mentioned previously and indicating the relevance it has for something else. This suggests that what it does and similar low-content wh-clauses can be used hollistically in discourse as chunks. Additional support comes from the fact that the chunk is often not complemented by a focus phrase until later in the following sentence, as in (66), or not overtly complemented at all, as in (67). But even then one should not see it as an incomplete cleft which has been abandoned mid-sentence, but as one that has a complement derivable from context. We will return to this point in section 5.8.

1. (66) Well, high definition television, what it does is, let's say you're watching a sporting event. You can actually see the individual blades of grass. You can see the pinholes in the jersey. (spoken, unscripted conversation)

2. (67) If you listen to the television commentators, the talking heads, all of us, it sounds like what he should do is suddenly compromise, retreat, water down his agenda. I think that would be the worst thing he could possibly do. But what you can do, you know, Bill Clinton did a very good job in thwarting the Republican majority in Congress for six years, and so it's a question whether George W. Bush is going to play hardball or softball, and I don't think the jury is spoken to that one yet. (spoken, unscripted conversation)

Table 8 shows the frequency with which non-coreferential subjects occur in amalgams. As can be seen, they occur in 2.8 per cent of eligible cases.

Table 8.

Non-coreferential NPs in COCA

5.7 Tense–aspect–mood mismatches

As was pointed out above, amalgam wh-clefts often exhibit a fair amount of repetition between the wh-clause and the cleft constituent, as seen in (68) and (69) below, where some lexical material is repeated (I'm trying to, love to) and the syntactic features of tense, aspect and mood (TAM) are consistently the same (present tense, progressive aspect).

1. (68) What I’m trying to do here is I'm trying to get around the corner (spoken, unscripted conversation)

2. (69) Typically, what people love to do is they love to have these big dishes. (spoken, unscripted conversation)

Likewise, negation is repeated in both clauses:

1. (70) But what we're not doing is we're not going to have a big government takeover (spoken, unscripted conversation)

2. (71) What this bureau doesn't do well is it doesn't account for its own blind (spoken, unscripted conversation)

However, as pointed out in section 2, in amalgam wh-clefts, the cleft constituent does not necessarily mirror the syntax of the wh-clause resulting in mismatches of TAM values. This is illustrated in (72):

1. (72) Well, what I wanna do is, I don't wanna have a debate where it simply degenerates. (spoken, unscripted conversation)

What is striking in (72) is the conjunction of two clauses – one positive (what I wanna do), the other negative (I don't wanna have) – joined by the copula. In such examples, it is especially visible that wh-clauses can be distinctly formulaic and detached from their be-complements.

In other such cases, the negative clause is merely an introduction to the genuine be-complement which appears later in the text. In example (73), the complement for the wh-clause appears in the next sentence, thus marking the separation of the two clauses especially heavily. As has been noted elsewhere (Carter & McCarthy Reference Carter and McCarthy2006; Hopper & Thompson Reference Hopper, Thompson and Laury2008; Koops & Ross-Hagebaum Reference Koops and Ross-Hagebaum2008) this spread of the projectibility of the wh-clause over several clauses is rather common in amalgam wh-clefts. We will return to this point in the next section.

1. (73) Yes, it sounds a little oversensitive to me. My – what I was doing, I didn't speak about anybody by name. I talked about members of Congress. (spoken, unscripted conversation)

Sentence (73) is additionally marked for lack of syntactic integrity in that the use of TAM values is inconsistent across the relevant clauses. The Past Continuous tense in the wh-clause is mismatched with the Past Simple tense in both clauses following it. Other types of TAM inconsistencies are exemplified below:

1. (74) (Present Simple+Past Simple) What it also does is, it helped move what Keynes called the animal spirits of the market. (spoken, unscripted conversation)

2. (75) (Present Continuous+like to) Michael, what these people are doing is they like to take the word ‘Democrat’ and make it synonymous with the word ‘liberal’. (spoken, unscripted conversation)

3. (76) (Present Perfect+Present Continuous) I think what Jim has done is that he is looking at it through the lens of today (written, newspaper)

4. (77) (Past Simple+Present Simple) So what we did in this one is we go back in time to 1969 in the second act (spoken, unscripted conversation)

5. (78) (Present Simple+will) what this does is he'll have to have a roll call vote (spoken, unscripted conversation)

6. (79) (Present Simple+can) What you do, you can marinate it. (spoken, unscripted conversation)

7. (80) (have to+need) Now what we have to do, we need to tidy up a little bit (spoken, unscripted conversation)

8. (81) (could+would) what it could do, it would put one image over another, it would remove enough image so you could be right there. (spoken, unscripted conversation)

9. (82) (like to+could) what I like to do, you could add a little bit of parmesan cheese (spoken, unscripted conversation)

10. (83) (would like to+will) What I’d like to do, we will just rejoin the Barry Gray show (spoken, unscripted conversation)

11. (84) (going to+like to) But what we’re going to do, I like to use my hands when I cook (spoken, unscripted conversation)

12. (85) (would+might) So, what we’d do is we might show that information to your friends a little bit (spoken, unscripted conversation)

Of course, in some of the sentences above, several non-integration effects occur simultaneously, e.g. in example (82) the copula is dropped, there is a mismatch in lexical material, and the subject NPs are non-coreferential. The sentence is also of the amalgam type, which is less integrated than the infinitival type. As a result of this combination of features, the cleft is heavily disjointed and the wh-clause itself (what I like to do) seems to be more of an extra comment on the focus clause rather than an integral part of the assertion. Table 9 shows the extent to which TAM mismatches occur in the COCA.

Table 9.

TAM inconsistencies in COCA

5.8 Multi-clause focus constituents, delayed focus constituents, and lack of overt focus constituents

As noted above, in amalgam wh-clefts the focusing effect over the post-copula constituent may be placed over longer stretches of text: several clauses or even sentences, as in example (86) below:

1. (86) I didn't use any scientific method to determine it. What I did was I took the circle and I divided it in half and then I divided it in half again to a quarter. And I just cut out a piece. And that's how I did it. (spoken, unscripted conversation)

This results in a looser link between wh-clauses and their focus constituents as in many cases, as in (87) below, it is difficult to ascertain the exact length of text over which the focus is placed.Footnote ⁸ Instead, in (87), what I did acts as a formulaic chunk introducing the speaker's account of what happened. Thus the phrase what I did is functionally similar to what I mean or what I'm saying.

1. (87) And so, basically what I did is I couldn't live with the pain any longer, I mean, I had to start giving up sports. I mean, I love to play basketball, play tennis, I couldn't do that any longer because every time I played, my knee – it would take me two weeks to recover from it. (spoken, unscripted conversation)

Extra material between the wh-clause and the focus constituent may also occur, which further exacerbates the fragmentation effect of such constructions. Consider examples (88) and (89), in which wh-clauses and focus constituents (in bold) are separated by parenthetical information. Note that in these examples, precisely because of the intervening material, infinitival complementation in focus constituents is highly unlikely.

1. (88) And what they did was, within three months of our going into Iraq while our troops were in harm's way, they attacked the commander in chief as a liar, a fraud. (spoken, unscripted conversation)

2. (89) What they do is, we have a form in the back of the catalog and they can fill out an absentee bid form. (spoken, unscripted conversation)

In example (90) below, the separation of the two elements is quite extreme. In the excerpt, the speakers discuss research into using bees to detect explosives and the pronoun they in the wh-clause refers to bees; the focus constituent complementing the wh-clause appears much later in the next sentence (in bold).

1. (90) A: What we're doing is we're training bees much like you would train a dog, to search out a certain odor.

   B: Right.

   A: And what they do is you take a feeder that is doped with sugar water, which the bees like. . .

   B: Mm-hmm.

A: and then you, basically, sprinkle, or you put some of the odor that you want them to search for near that feeder. Then as the bees come and feed, they smell the odor and then they begin to associate the odor with a food source. (spoken, unscripted conversation)

In other cases, focus constituents are not overtly present and need to be deduced from context. In most cases the deduction is fairly straightforward, as in (91) and (92), although the lack of an overt complementation for wh-clauses (along with copula omission) makes for a strong sense of structural non-integration (translations into canonical wh-clefts are given in brackets).

1. (91) And what you also do, I am for stimulating, I am for the fact that you have to give people incentives to make jobs in the economy. (‘What you also do is stimulate people to make jobs’) (spoken, unscripted conversation)

2. (92) What we’re going to do tonight, I've got Lucianne Goldberg just itching to come out backstage. Of course, she was one of the original players in this drama. And also, Michael Moore, who comes at it from a different side. (‘What we're going to do tonight is talk to Lucianne Goldberg and Michael Moore’) (spoken, monologue)

In other cases, deducing the focus element can be slightly problematic, as in (93), where the speakers are talking about how long the economical recession will continue.

1. (93) A: Well, here's the thing, let's just – let's just assume that this is going to continue, OK?

   B: Mm-hmm.

   A: Because all indicators show that this will. What we have to do is we've already been in it for a while, we've made our adjustments, but the key point here is we've adjusted to a time that is very, very difficult, this is going to take about a year to go. None of us can call the bottom, none of us are going to be able to call the top, but this is going to be at least a year, we can all assume. (spoken, unscripted conversation)

In (93), it seems that the wh-clause (What we have to do) can optionally be complemented by three ideas which appear in the text, i.e. (a) ‘assume that it is going to continue’, (b) ‘assume that it is going to be at least a year’ or (c) ‘make our adjustments’, or by a conclusion that can be inferred from the speaker's words, i.e. ‘What we have to do is brace ourselves for at least another year’.