1 Introduction
This article tracks down fragmentary and imperative clauses in a genre where one might least expect to find them: carefully edited journalistic writing. A closer examination of this genre reveals that such constructions are used in it sparingly but highly efficiently. The advantage of studying them in this not-so-natural biotope is that we can gain a fuller understanding of their properties, including their pragmatic and discursive ones. For this study, a small corpus has been compiled containing lead articles in The Economist, one of the oldest still-running newspapers in the world and widely recognised for its editorial excellence, in terms of both substance and style. Note that, with respect to language use, this weekly journal has published a usage guide under its name (Wroe Reference Wroe2018). The highly polished nature of the texts studied here means that we can safely exclude the possibility that the fragments and imperatives have slipped through inadvertently, something which is not necessarily the case for spontaneous oral discourse, where they are often used as ‘mere’ interjections (e.g. What?!) or discourse markers (e.g. Listen, …).
The following main question, then, is addressed in this article: to the extent that they occur in the running text of Economist editorials at all, are fragments and imperatives used in a haphazard way, or are they given privileged positions, for instance at the beginning or the end of the text or at the beginning or the end of a paragraph, where they stand out? The expectation is that, despite the long-standing, well-recorded trend of serious writing to become more colloquial and informal in style, fragments and imperatives in editorials are not randomly dispersed. If that turned out to be the case, this conspicuous placement might suggest that the writer wants to signal not only a willingness to engage with the reader by using conversational language but also an awareness that the linguistic form used is stylistically marked. At the same time, there might be certain functions of an imperative or fragment that make, say, a paragraph-initial placement for them a quite natural choice.
This is how my article is structured. In the remainder of this introduction, I first present some of the literature that shows that fragments and imperatives indeed occur most frequently in spoken language (section 1.1). I then argue why there is some justification in dealing with fragments and imperatives as somewhat similar phenomena (section 1.2). Finally, I discuss in some detail an authentic example to illustrate what I am putting forward as the main claim of this article, namely that these structures can take, quite literally, a special position in thoughtfully crafted writing (section 1.3). Fragments and imperatives occurring in these journalistic texts may look like involuntary intrusions of casual, conversational discourse, but their use is deliberate and functional. This section also allows me to spell out some of the background assumptions about fragments in particular.
Next, in section 2, I outline the research methodology, detailing the choice of corpus and search method (section 2.1) and the operationalisation of the above-mentioned central research question (section 2.2). I then address the types of fragment-like phenomena that were left out of consideration. Section 3 presents the quantitative results. In section 4, I delve into some specific uses, focusing first on the occurrence of a fragment or imperative at the start of the second paragraph (section 4.1) and then on imperatives whose function it is to move the reader through the article, as it were (section 4.2). Identifying and analyzing the functions of fragments and imperatives in these texts helps us to understand why they tend to appear in the positions we find them in. Some of the recurring structures and/or phraseologisms among the instances will also be reported on (sections 4.3 and 4.4). Section 5 synthesises the findings and formulates some avenues for further research.
1.1 Fragments and imperatives as predominantly oral-discourse phenomena
Using dialogue transcripts from the British National Corpus, Fernández & Ginzburg (Reference Fernández and Ginzburg2002) found that a staggering 11.15 percent of all utterances were ‘non-sentential’ ones, in the sense that they lack the NP+VP structure associated with canonical sentences. These include, among other types, various sorts of queries, namely ‘clarification ellipsis’ (e.g. (1a)), sluicing (e.g. (1b)) and check questions (e.g. (1c)); short answers (e.g. (1d)) and answers to polar questions (e.g. (1e)); fillers to a gap left by an interlocutor (e.g. (1f)); and adjectives or adverbs functioning as propositions (e.g. (1g)).
(1)
(a) (A: ..they used to come in here for water and bunkers you see.)
B: Water and?
(b) (A: Can I have some toast please?)
B: Which sort?
(c) A: So we get three readings. Okay?
(B: Right.)
(d) (A: Who's that?)
B: My aunt Peggy [last or full name]. My dad's sister.
(e) (A: Is that Mrs. John [last or full name]?)
B: No, Mrs. Billy.
(f) (A: And another sixteen percent is the other Ne Nestle coffee […] and twenty two percent is er <pause>)
B: Maxwell.
(g) (A: So we have proper logs? Over there? B: It's possible.)
A: Brilliant!
(Fernández & Ginzburg Reference Fernández and Ginzburg2002: 16–17; boldface added for ease of identification)
As these examples show, fragments appear to rely heavily on the dialogic nature of the discourse in which they appear. As Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan2021: 226) observe, ‘[n]on-clausal material is far more common in conversation than in the written registers’. This is due to the fact that conversation occurs in a cooperative environment, and the participants usually share the same context and know each other well. As a result, more information can be left unsaid, with the expectation that the listener will be able to infer it. The online, non-scripted circumstances of spoken conversation also lead to an increased use of fragments. A proposition that might be uttered in a single, complete statement in written discourse may come out in spoken discourse as an unfinished sentence or as one that is conveyed over multiple increments, often with the interlocutor also taking several interspersed short conversational turns at the same time, as in (2):
(2) A: Well – I got it from that travel agent's.
B: Oh.
A: er the one
B: In the precinct?
A: by, yeah, by Boots.
B: Oh yeah.
(Biber et al. Reference Biber, Johansson, Leech, Conrad and Finegan2021: 227)
‘Although the syntax looks chaotic’, Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan2021: 227) comment on this passage, it ‘would have been pretty much a regular independent clause, if said all in one go by one speaker: I got it from that travel agent's, the one in the precinct by Boots’. Discourse markers such as Yes, Okay, Oh, Y'know, etc., which are also non-sentential, are common in informal conversations to help speakers communicate their intended meaning efficiently (Schiffrin Reference Schiffrin1987) but are generally not accepted in formal writing contexts. Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan2021: 225–6) do discuss fragments in writing, including journalistic prose, but three of the five examples they cite from news sources are headlines (e.g. Image crisis for Clinton over haircut). Headlines are really a genre in their own right (Mårdh Reference Mårdh1980) and the use of non-sentential utterances in them is therefore to be distinguished from their much rarer use in the running text of an article. One non-headline instance Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan2021) cite is oddly telegrammatic; only one of their five examples, with a clichéd form (And now for something completely different: …), seems a legitimate one.
Imperatives, too, occur most frequently in spoken discourse. Biber et al. (Reference Biber, Johansson, Leech, Conrad and Finegan2021) report corpus results indicating that there are roughly 11,000 occurrences of imperatives per million words in conversations, a frequency that differs from their appearance in news by a full order of magnitude. Again, this is because imperatives are well-suited to the interactive and dynamic nature of spoken conversation, where speakers need to quickly convey requests, directions, and advice to their listeners. As mentioned before, imperatives such as look and listen may also function as pragmatic markers to grab the listener's attention (Van Olmen Reference Van Olmen2010).
Fragments and imperatives used in journalistic writing may be seen as examples of the overall tendency of written language to become more informal and colloquial (Fairclough Reference Fairclough, Keat, Whiteley and Abercrombie1994, Reference Fairclough, Coleman and Cameron1996, Reference Fairclough2003: ch.4; Mair Reference Mair2006; Leech et al. Reference Leech, Hundt, Mair and Smith2009). There is a large and growing body of literature that supports this. Speech-like phenomena in written journalism, specifically, have been noted by Haselow (Reference Haselow, Bös and Kornexl2015), Rühlemann & Hilpert (Reference Rühlemann and Hilpert2017) and Tottie (Reference Tottie2017), among others. Yet the fragments and imperatives found in these texts are still stylistically marked, as will be seen.
1.2 Why it makes sense to study fragments and imperatives together
Imperatives and fragments have certain formal and functional features in common, which makes it useful to address them both in a single study, even though they are undeniably different linguistic structures.
First, as noted before, fragments deviate from the standard structure of a sentence as they are, by definition, incomplete. Imperatives typically lack an explicit subject, as they are mostly directed to the interlocutor and this you can be (but doesn't have to be) left implicit. Therefore, most imperatives also depart from the standard NP+VP structure associated with canonical sentences. Being incomplete (in the case of fragments) and usually missing a subject (in the case of imperatives) means that both structures are also shorter, on average, than more ‘ordinary’ (full-clause, declarative or interrogative) sentences.
Second, imperatives have non-truth-conditional semantics: insofar as they have a directive function, they do not convey propositions that can be said to be true or false (or possibly true, possibly false, etc.). Many fragments also lack truth conditions (e.g. Good luck!, What to do?) or are used rhetorically to convey a certain tone or attitude that goes well beyond making a simple truth claim. The two constructions, fragments and imperatives, may sometimes even merge, as in the case of Welcome to Britaly (an example from the corpus): though a fully conventionalised expression and not felt to be a clipped version of anything, Welcome (to X) is treated as a fragment in this study because it can be seen as related to (if not necessarily derived from) either an optative clause (May you be welcome (to X)) or an imperative clause (Be welcome (to X)). In any case, whether or not we see Welcome as linked to Be welcome, it cannot be denied that this non-canonical expression is used in the same way that many true imperatives are (e.g. Come in! Welcome! Have a seat!).Footnote 1
Third, and related to this role in expressing emotions more than propositions, both fragments and imperatives have a firm place in what is referred to as ‘thetical grammar’ (Kaltenböck et al. Reference Kaltenböck, Heine and Kuteva2011), which is concerned with how language is used beyond just individual sentences. The focus of this area of grammar is on how language items function in the broader context of discourse, including how they help organise texts, facilitate communication between speakers and listeners, and convey the attitudes and intentions of the speaker. However, formally speaking, both types of constructions are certainly not always non-sentential discourse elements that are adjoined to structures that belong to ‘normal’ ‘sentence grammar’ in the same way that many other theticals are. Unlike vocatives, disjuncts, true parentheticals and the like, the fragments and imperatives under investigation here are, with very few exceptions, stand-alone utterances, in that they are syntactically and prosodically fully self-contained – though this is not to say, of course, that their interpretation does not require reference to the preceding discourse.
1.3 Assumptions
It is assumed here that fragments and imperatives in carefully edited prose are not interspersed among more ordinary sentences in a completely random way. Far from being mere afterthoughts, asides or interjections that we could dismiss as playing a rather secondary role with respect to canonical sentences, they regularly appear in positions where they stand out and, as such, play a central role in the overall argumentation of a text. Consider this example, which warrants some discussion:
(3) America's professional bodies acknowledge the science is low quality, but say they have a duty to alleviate patients’ mental anguish. Some patients suffer regret in all medical procedures, from knee surgery to liposuction. And they observe that the most shocking allegations about poor treatment are only anecdotes. Speaking on American radio last year, Rachel Levine, assistant secretary for health and a paediatrician, was very clear: “There is no argument among medical professionals … about the value and the importance of gender-affirming care.”
Except that there is. And when medical staff raise concerns – that teenage girls may be caught up in a social contagion, say, or that some parents see transition as a way to have a straight daughter rather than a gay son – they have been vilified as transphobic and, in some cases, suffered personal and professional opprobrium. (‘What America has got wrong about gender medicine’, The Economist, 5 April 2023; boldface added for ease of identification)
The use of the fragment starting with except appears at the start of a new paragraph. Readers tend to pay particular attention to this position, as it signals to the reader that the writer is shifting focus or is introducing a new argument. In this case, the fragment has the effect of completely refuting the truth of the quoted statement at the end of the previous paragraph. Observe, incidentally, that this particular fragment does also have a truth value itself – it could be evaluated as true or false – but, confirming what was stated in section 1.2, it also comes with strong attitudinal content, comparable with what would be conveyed by a fragment such as Nonsense! (which would have much the same function but would be less appropriate within the generally elevated register of a quality newspaper).
One might wonder why Except that there is qualifies as a fragment. I follow Fernández-Pena (Reference Fernández-Pena2021: 136–7) in treating as fragments any and all ‘stand-alone constructions which, despite their reduced, non-canonical, fragmentary structure, are still semantically, discursively and pragmatically equivalent to a complete clause construction’. Here, we could argue that Except that there is can be ‘completed’ as follows: This is true except that there is some such argument. The fact that we can supply material after there is, however, is not what makes Except that there is a fragment.Footnote 2 What does make it a fragment is that except (that) is a conjunction introducing a subordinate clause and that there is, therefore, arguably an implicit matrix clause in this sentence (but see further below for some refinement of this view).
If we make this matrix clause explicit, though, as we did above (This is true…), the resulting sentence is somehow semantically incoherent, at least if we take except literally to announce a particular circumstance, fact, person or thing to which a previous statement does not apply (as in I like every fruit except for bananas). The ‘complete refutation’ use of except in (3) may have arisen out of a well-established rhetorical trick to set up a wrong expectation: by using except, the writer suggests that what follows will be just a minor detail, which then turns out to be not so minor at all. The hearer is invited to draw the inference, probably using Gricean maxims, that irony, mockery or sarcasm is involved, as in the following web-attested cases:
(4)
(a) … we all know evolution is about improvement and ‘progress’; except, of course, it isn't at all.Footnote 3
(b) Fun Fact: Captain Man is really good at the internet! Except that he isn't. At all.
Footnote 4(c) Great argument, except for the tiny detail that it's complete bullshit.Footnote 5
It may be that over time, as except is increasingly used in this way, it gets to lose its basic meaning of announcing an item that needs to be excluded from the scope of the previous statement, or that this basic meaning no longer resonates as strongly as it initially did when utterances such as those in (4) were first used. This would be an expected evolution if we consider that the contrastive conjunction but comes from the Old English word būtan, which meant ‘outside of’, ‘without’ or ‘except’ (cf. Traugott Reference Traugott2022). Note that only, which similarly has a basic ‘excepting’ meaning (e.g. I was the only linguist in the room), can be used in the same way (e.g. He claimed to have won the championship three times. Only, it's not true. In reality, he's never won the championship even once).
In (3), it is not entirely clear whether Except that there is should be taken as a clause that is detached from an explicit matrix clause (namely the quoted statement in the preceding paragraph) or as one that has an implicit matrix clause (This is true, …). Neither analysis seems adequate. The utterance is most appropriately viewed as a case of insubordination (cf. Beijering et al. Reference Beijering, Kaltenböck and Sansiñena2019 for an overview of this phenomenon): while it has the formal signs of a subclause – which is why we can call it a fragment in the first place – it really functions as an independent clause. The fact that the clause appears as the first sentence of a new paragraph substantiates the view that it is not a subclause that really belongs to the previous sentence from which it is separated: the sentence-final punctuation, hard return and indentation before it strongly suggest that Except that there is forms a sentence all on its own. Moreover, the artificiality of making explicit an implicit matrix clause, not to mention the ensuing incoherence, speaks against an ellipsis-based view according to which this fragment can be ‘reconstructed’ as a main-clause-with-subclause structure (for proponents of the ‘sententialist’ position to fragments, see e.g. Morgan Reference Morgan, Kachru, Lees, Malkiel, Pietrangeli and Saporta1973; Hankamer Reference Hankamer1979; Stanley Reference Stanley2000; Merchant Reference Merchant2004). This article does not aspire to contribute explicitly to the theoretical debate about the status of fragments in general (on which, see e.g. Stainton Reference Stainton2006), but I lean towards a ‘direct interpretation’ or ‘non-sententialist’ approach to fragments (see e.g. Culicover & Jackendoff Reference Culicover and Jackendoff2005; Jacobson Reference Jacobson2016), which doesn't make the assumption that there is any deleted material in common (types of) fragments such as And now for NP, No worries!, Good luck (with X) or Why VP (if X)?
2 Method
2.1 Corpus and search method
I compiled a mini-corpus consisting of 88 editorial articles from The Economist, spanning a period of four months, with issues published between 3 September and 24 December 2022. These articles, which are referred to as ‘leaders’ in this weekly journal, range from 580 to 1,380 words in length. The mini-corpus has a total of 72,286 words. It is impossible to state how many different writers are behind these texts. Articles in The Economist, which are anonymous, can be written by more than one author:
Leaders are discussed and debated each week in meetings that are open to all members of the editorial staff. Journalists often co-operate on articles. And some articles are heavily edited. Accordingly, articles are often the work of The Economist's hive mind, rather than of a single author.Footnote 6
Given the wide diversity of topics addressed in the leaders, covering global politics, economics, technology, and environmental and social issues, there is little reason to assume that all draft versions are gone over by a single chief editor, or editorial assistant, who drastically changes their contents and style. In other words, though there is some uniqueness and uniformity of style of The Economist editorials, the imperatives and fragments in them are unlikely to reflect the linguistic habits of a single language user.
While admittedly small as corpora go, the collection of texts can serve a useful purpose to find out how often fragments and imperatives appear, and where exactly. By virtue of its limited size, it was feasible to search the entire collection of texts by hand for fragments and imperatives. Such a manual identification was meant to ensure maximum coverage, which is not guaranteed with automatic searches. Besides, it is not possible to formulate search queries if one doesn't know in advance the range of types that could be targeted by automatic searches. Whereas imperatives have fairly uniform syntax, fragments, by their very nature, can take many forms and shapes, all of which are somehow ‘deficient’ when compared with standard, ‘complete’ sentences. See Bondarenko (Reference Bondarenko2019), however, for a successful attempt at automatically extracting verbless sentences from a corpus.
2.2 Operationalisation of the research question
To answer the main research question – where do we find any fragments or imperatives in editorials such as those in The Economist? – for each fragment and imperative retrieved from the text, the following positional aspects were determined:
(a) the position of the fragment or imperative in the whole text, by number of words in that text;
(b) the position of the fragment or imperative in a paragraph, by number of words in that paragraph;
(c) the position of the fragment or imperative in the whole text, by number of paragraphs in that text;
(d) the position of the fragment or imperative in a paragraph, by number of sentences in that paragraph.
For (a) and (b), the position of the fragment/imperative was determined by taking the first word of the instance and dividing it by the total number of words in the text or paragraph, respectively. Thus, a fragment (of whatever length) at the beginning of a 100-word paragraph was assigned a position of 0.01 (i.e. one divided by hundred) in that paragraph. Values range from close to zero (fragment/imperative appearing at the beginning) to 1.00 (fragment/imperative appearing at the end). According to this metric, no fragment or imperative can have a position of exactly zero, as the numerator of the fraction is always 1. Furthermore, as the first word of the fragment or imperative determines its position, if the fragment or imperative consists of more than one word, it cannot be assigned a position of 1.00, even if it occurs at the end of the text or paragraph. For the number of words in the text, I included its headline, subheading and the words describing its general theme (e.g. ‘Renewable energy’).
For (c) and (d), what was also determined was the average number of paragraphs per text and the average number of sentences per paragraph containing one or more fragments and/or imperatives in the corpus. With these averages, it was possible to calculate the number of ‘expected’ fragments or imperatives in a given paragraph or sentence. This, in turn allowed me to find out whether there were more fragments or imperatives in, say, the first sentence of a paragraph or the last paragraph of a text than could be expected if their occurrence in a sentence or paragraph was random.
2.3 Data excluded
Given that different researchers have different conceptions of what a fragment is, this section provides the necessary details on what was included and what was not. Only fragments that appear as independent clauses were counted as true fragments for the purpose of this study. This means that in sentences like the following (which appear in the corpus), the parts in boldface (added, again, for clarity) were discarded:
(5)
(a) Inflation was much too high and tight monetary policy had taken interest rates to over 19%—problems Reagan attributed in part to rising government debt.
(‘Truss's rusty Reagonomics’, The Economist, 24 September 2022)
(b) As they have grown bigger, they have become tied to the economic cycle; a fact which the digital surge during the pandemic only temporarily masked. (‘Big tech, big trouble’, The Economist, 5 November 2022)
In such sentences, the part after the dash (or after the semicolon, in (5b)), is a fragmented clausal unit. However, in such cases, it can be analyzed as a reduced relative clause (which were problems…; which is a fact…) and the fact that they have an antecedent in the previous clause – sometimes that entire clause itself – means that they are not independent from their host clause, even if they are not syntactically connected to it by means of a conjunction. Such disintegrated increments or inserts, though excluded here, are not without interest as they can be regarded as indicative of the writer's intention to infuse the text with speech-like phenomena (Haselow Reference Haselow, Bös and Kornexl2015).
Below, I list five further types of fragment-like structures that have been excluded. Some discussion will follow.
(6) Disjuncts consisting of an adjective (phrase)
(a) Worse, the inflation that has hitherto been caused by energy prices seems to be becoming entrenched. (‘False dawn’, The Economist, 19 October 2022)
(b) Perhaps most important, Volodymyr Zelensky, Ukraine's president, continued to make nightly television broadcasts to the nation. (‘The digital front’, The Economist, 3 December 2022)
(7) Left-dislocated constituents
(a) Divorce, bereavement, problems at work—a daily pill was there to help with that, and anything else which made you sad. (‘Set patients free’, The Economist, 22 October 2022)
(b) Bloated public services, low growth, poor productivity: the problems of Italy and other southern European countries were also present in Britain. (‘Welcome to Britaly’, The Economist, 22 October 2022)
(8) Right-dislocated constituents
(a) The extraordinary challenge of the pandemic led to extraordinary actions which helped unleash today's inflation: wild government stimulus and bail-outs, temporarily skewed patterns of consumer demand and lockdown-induced supply-chain tangles. (‘What next?’, The Economist, 8 October 2022)
(b) There are few good options, but one leaps out: taxing the priciest properties more heavily. (‘How to fix a budget in 55 days’, The Economist, 12 November 2022)
(9) Structures expressing alternative or universal conditionals
(a) But Xi Jinping and the Communist Party are rushing ahead, ready or not. (‘What is the plan?’, The Economist, 10 December 2022)
(b) Whatever your politics, it is worrying that one man can choose whether to extend internet access to anywhere on Earth, can decide who can use it—and can turn it off at will. (‘How will Elon Musk use his superpowers?’, The Economist, 8 October 2022)
(10) Elliptical phenomena inside sentences
(a) It has not trained enough medical staff, nor adopted protocols on which patients to treat and where. (‘What is the plan?’, The Economist, 10 December 2022)
(b) [Conservative MPs will not wear cuts on such a scale;] neither will voters. (‘The Iceberg Lady’, The Economist, 15 October 2022)
Although these types could in some respect be treated as fragments, none of them meets the strict requirement of being fully separated from a clause containing them (cf. also Fernández-Pena Reference Fernández-Pena2021: 140). In (6a–b), the boldfaced part can be seen as shortenings of What is {worse/perhaps most important}, but these structures haven't been counted as fragments, for the same reason that most researchers would disregard sentence-initial occurrences of first, second, last, etc. (which could also be viewed as short for the {first/second/last/…} thing is that…). Likewise, I excluded discourse-structuring constituents such as More than that, … or What's more, …. The latter is a clause but could still be seen as ‘missing’ a main verb and complementizer (What's more is that…) and so could be seen as a fragmentary structure for that reason; however, if this sequence is a shortened matrix clause at all, it is a downranked one, having the same function as Moreover, …. Non-integrated constituents like those in (7a–b) were not treated as true fragments either, as they are clearly still part of the whole sentence. Such constituents are very common in the corpus, where they generally take the form of coordinated NPs with three conjuncts. Their referents may be picked up in the rest of the sentence by a pronoun (as in (7a): that) or by a full NP (as in (7b): the problems of Italy and other southern European countries). The cases illustrated in (8a–b) are similar but have a disintegrated constituent at the right edge of the sentence (which that constituent still belongs to). Unlike left-dislocated constituents, right-dislocated ones do not typically come in groups of three (although (8a) shows this sometimes is the case); they more often name one item that provides a specific value for something. Also excluded have been the kinds of elliptical structures in (9a–b), which express two alternative possibilities (whether they are ready or not) or any circumstance under which the main-clause proposition remains valid (whatever your politics may be). While these structures have been excluded, it is worth observing that there is apparently a cluster of constructions in English with quite different lexical items – (whether) X or Y, whatever X, no matter X, regardless X – which all partake in fragmentary syntax. Finally, in (10a–b), examples are given of phenomena that can be discussed under the heading of ellipsis: sluicing in (10a) and VPE in (10b). While ellipses and fragments are closely related phenomena, not every instance where material is ‘missing’ is a true fragment. For the purposes of this study, a fragment must be used as a self-standing utterance even though it has the formal features of something that is generally part of a larger whole. In (10a), where is still part of a larger structure ([protocols on … where to treat them]). In (10b), just because there is material missing (cp. neither will voters wear cuts on such a scale) doesn't mean the remaining part is a fragment, as that part doesn't have the form of, say, a subclause or a single NP; it still has the form of a main clause, albeit one with an elided VP.
Imperatives posed fewer selection problems. I relaxed a criterion for their inclusion that was applied to fragments, however, namely that they must be used as independent sentences. In a few cases, an imperative occurred in a sentence with coordinated main clauses, as in Resist China's attempts to make the global order more autocrat-friendly, but avoid overheated martial rhetoric. In such cases, each clause with an imperative was counted separately and treated as a single ‘sentence’. This seemed the least problematic solution to determine the position of the imperatives. (The second ‘sentence’ in the example just given would then start with the word but.) I also decided to include instances of imperatives that were used as conditionals (of the type Step out of line again and you're fired), even though such a conditional imperative obviously requires the presence of the second conjunct (Russell Reference Russell2007).
Both imperatives and fragments were excluded from consideration if they appeared between brackets (as they then do not form any clearly marked choice) or were quoted (as they then do not reflect the choice of the writer of the editorial):
(11)
(a) China agreed to resume formal talks about climate change with America (see China section). (‘Green competition’, The Economist, 26 November 2022)
(b) A lone old lady, her white hair uncovered, shuffles down the street waving her headscarf in tune to the words ‘Death to Khamenei!’ (‘Is this time different?’, The Economist, 1 October 2022)
Finally, imperatives and fragments in the headlines or subheadings were ignored. Although headlines are very frequently imperatives (e.g. Know your rival and yourself), fragments (e.g. Lula will be Brazil's next president. Now for the hard part) or both (e.g. Back Bibi. Seriously), there was no principled way of distinguishing headlines with relevant types of fragments from ordinary headlines that aren't meant to be perceived as fragments, as when they are just NPs (e.g. Green competition; China's covid failure). Additionally, it's worth considering that including headlines may not provide an accurate measure of how often certain text fragments appear in news discourse. Headlines often use shortened syntax and cryptic phrasing for brevity (e.g. Little steps, many lives), which may not reflect the overall style of news content. The same reasoning applies to subheadings, even if these typically exhibit standard English syntax. So, though subheadings occasionally contained legitimate-looking fragments (e.g. Most people on antidepressants don't need them. Time to wean them off) or imperatives (e.g. Fire up Britain's economy by invigorating its second-tier cities), they often also contain uninteresting fragments, namely when they take the form of NPs (e.g. A wish-list of centrist proposals for the lame-duck Congress), gerunds (e.g. Investing in an era of higher interest rates and scarcer capital) or dependent questions (e.g. How a stable and successful country could emerge from the trauma of Russia's invasion). Therefore, since in subheadings ‘true’ fragments and irrelevant ones are also impossible to distinguish in a fully reliable way, I only looked at these constructions in the main body of a text.
3 Results
In the mini-corpus of The Economist editorials, a total of 92 fragments (n = 49) and imperatives (n = 43) could be identified (≈ 1,270 fragments and imperatives per million words). Remember that valid instances of fragments in this study are sequences that have the form of language units that are generally included into larger structures but occur as self-standing sentences. Of the 88 editorials, 54 contained one or more fragments or imperatives (see table 1).
Table 1. Number of editorials in the corpus that contain zero, one or more fragments or imperatives
The assumption that fragments and imperatives are not randomly distributed throughout an editorial article is confirmed. Figure 1 shows the distribution of both construction types in the text (X-axis) and inside a paragraph (Y-axis).
Figure 1. Distribution of fragments and imperatives in a text and a paragraph in a small corpus of editorials from The Economist (close to 0 = at the start; 1 = at the end)
We can observe two clusters of imperatives and fragments at the beginning of texts and at the start of a paragraph: a small cluster right at the beginning (indicated with the letter A) and a larger cluster a little further to the right of it (B). In the top right corner, there is another cluster of fragments and imperatives: these occur at the very end of the final paragraph in a text (C). Many fragments and imperatives appear to occur either at the start or at the end of a paragraph: note a concentration at the bottom (D) and a concentration at the top (E) of the figure. (The latter concentration is less conspicuous because the first word in the sentence was used to determine its position; hence, paragraph-final fragments/imperatives (i.e. those that appear in the final sentence) may still appear at somewhat different heights along the Y-axis, depending on the length of the sentence and the paragraph. Two instances that also belong to that paragraph-final zone but fall outside the indicated cluster have been connected to it by means of vertical lines; note, though, that this cluster also includes one instance that should be excluded from it.) The middle of the figure is relatively sparsely populated: not many fragments or imperatives appear in the middle of a paragraph and in the middle of a text. A straight upward slope of imperatives will not have escaped the reader's attention (F): this is a sequence of recommendations that all occur in the final paragraph of a single text.Footnote 7
Visual inspection of the data in figure 1 does not suggest any noticeable, systematic difference in placement preferences between fragments and imperatives. They will therefore be treated together when we address additional positional aspects (namely those raised in (c) and (d) in section 2.2).
If we focus on the paragraphs in which we find fragments or imperatives, we obtain further and more concrete evidence that they are not evenly distributed throughout the text space of an editorial. The articles in which we find a fragment or imperative have an average of 9.35 paragraphs.Footnote 8 Given that there are 92 fragments and imperatives, we can, in principle, expect 92/9.35 = 9.84 of them to occur in the first paragraphs, and then also 9.84 fragments or imperatives in any subsequent paragraphs. What we find is that the second and especially the final paragraph contain more fragments and imperatives than could be expected if they were spread proportionally across paragraphs (see figure 2). The observed distribution differs significantly from the expected distribution at the 0.001 significance level (χ2 = 17.656, df = 3, p-value = 0.0005178).
Figure 2. Observed and expected distribution of fragments and imperatives across paragraphs in in a small corpus of editorials from The Economist
Turning our attention to sentences within a single paragraph, we can again observe that fragments and imperatives are not evenly spread across them. Given that paragraphs with a fragment or imperative have 5.54 sentences on average in the corpus, it could be expected that there are considerably fewer of them in the first or in the last sentence than in any of the ‘middle’ sentences combined. Yet this is not what we find (see figure 3, whose orientation is made to correspond with the y-axis in figure 1, to facilitate interpretation and comparison). Paragraph-initial or paragraph-final sentences are almost twice as likely to be fragments or imperatives than could be expected under a purely chance distribution, again a statistically significant difference at the 0.001 level (χ2 = 44.58, df = 2, p-value = 2.087e-10).
Figure 3. Observed and expected distribution of fragments and imperatives across sentences within a paragraph in a small corpus of editorials from The Economist
4 Discussion
Fragments and imperatives, two constructions that thrive in spoken dialogues, are not wholly uncommon in written discourse, even in texts that have received meticulous editing. Despite their syntax being described as ‘defective’, the fragments we find in editorials from The Economist do not indicate a lack of care but are used intentionally, as appears from their frequent placement in positions where they stand out. The imperatives, too, tend to appear in conspicuous positions. This suggests that the writers of these articles have full awareness of both constructions’ special status in this genre.
4.1 A discourse-level construction involving two paragraphs
Perhaps unexpectedly, there are not conspicuously more fragments or imperatives in the very first paragraph than under a random distribution. However, we do see a slight ‘over’-representation in the second paragraph (see again figures 1 and 2). This seems to be linked to a rhetorical ploy to mislead the reader in the first paragraph with a view which is received opinion or which is ascribed to someone other than the writer. Here is an example from the corpus:
(12) Shortly before the midterm elections, Donald Trump held a rally in Ohio. “Our country is becoming third-world,” he told voters. Later he hinted—is threatened a better word?—that he would soon announce he was running for president again. What could possibly go wrong for a party with such a figurehead? Or for one whose primary voters are so keen to relitigate the 2020 election that they chose a slate of candidates in key Senate seats chiefly for being the true keepers of the Trump flame?
Quite a lot, it turns out. [...] (‘The Trump effect’, The Economist, 12 November 2022)
The first paragraph contains a rhetorical question, What could possibly go wrong…?, whose elicited answer (Nothing, of course!) experienced language users are expected to recognise as not reflecting the writer's position but the opinion of people referred to, in this case the leaders of the US Republican Party. Indeed, this particular rhetorical question, unlike most other rhetorical wh-questions, is often used ironically or sarcastically, implying that something is likely to go wrong or that there are potential risks or problems that may arise, despite the situation seeming, to some, perfectly straightforward and under control.Footnote 9 The first paragraph thus presents a faulty view which the second paragraph, right from the start, corrects. This is comparable to how the two paragraphs are linked in example (3), the one with Except that there is. Syntactically, the fragment in (12) differs from the fragment in (3). In (3), as will be remembered, we dealt with a subordinate clause used as a main clause; in (12), the actual fragment (Quite a lot) exemplifies a short answer to a polar question, similar to (1d).Footnote 10
What this means is that we can posit a discourse-level construction (cf. Östman 2005), schematically represented in (13):
(13) Viewpoint-correcting construction
-
1: viewpoint to be contested (received opinion, the view of someone/people under discussion or a generally plausible view)
-
2: fragment/imperative (‘That view is wrong!’), followed by argumentation
-
This is a ‘construction’ in the Construction Grammar sense (Goldberg Reference Goldberg2003) in that it pairs a conventional form with a conventional function. The form, which is rather schematic, is larger than that of typical examples of constructions in the Construction Grammar literature, which are clauses, phrases or smaller. In this case, we are dealing with a full paragraph followed by another one that starts with a fragment or imperative. The function can be characterised as follows: ‘Here's a viewpoint you are supposed to recognise as familiar. Well, you should reject or adjust that view. Here is/are the reason(s) why.’ Obviously, this construction doesn't have to be used to fulfil this function, but it is one that is at the writer's disposal, and readers can see it as a somewhat familiar discursive pattern whenever they encounter it. Several concrete fragments or imperatives can be used in it: in the corpus there is also Don't count on it. Other possibilities are Not according to X, But not anymore (on not-fragments of this type, see Cappelle Reference Cappelle2020, Reference Cappelle2021), Fat chance, Don't hold your breath, Well, think again and several others.Footnote 11 Most of these fragments and imperatives, even though they occur in a written news article, are rather informal. Their use is most probably a deliberate choice, tying in with the colloquial features that Haselow (Reference Haselow, Bös and Kornexl2015) notes for contemporary journalism in general and Mullen (Reference Mullen1999) for editorials from The Economist in particular. Journalists using fragments and imperatives in their prose are likely doing so intentionally to serve rhetorical purposes. This fits with advice given by Lane Greene, the language columnist for The Economist, who says, ‘Make your prose as lean as necessary to keep your reader reading’ (Greene Reference Greene2023: 74).Footnote 12
4.2 Paragraph-initial imperatives to push the reader along
When it comes to keeping the reader reading, imperatives at the beginning of a paragraph are often used to urge the reader along in the argumentation. With such an imperative, the writer can immediately direct the reader's mental gaze to a relevant issue, as in these examples:
(14)
(a) To understand Vladimir Putin, hear what he says about his enemies. (‘Putin doubles down’, The Economist, 24 September 2023)
(b) To understand why, look beyond the hurly-burly to the long-term fundamentals.
(‘What next?’, The Economist, 8 October 2022)
(c) If this judgment sounds severe, look at gilt yields, which have this week been climbing again. (‘The Iceberg Lady’, The Economist, 15 October 2022)
(d) Start with the change in policy. (‘Truss deficit’, The Economist, 8 October 2022)
(e) Take spending cuts as an example. (‘How to fix a budget in 55 days’, The Economist, 12 November 2022)
Such paragraph-initial imperative clauses almost sound like what an oral presenter might say when showing a new slide or bullet point. They also render the use of text-structuring subheadings unnecessary. At least 12 out of the 43 of the imperatives in the data are expository directives (paragraph-initial or otherwise). This type of imperative is sometimes combined with a purpose clause, as in (14a) and (14b). Other examples of that kind are (8) and (43) in appendix B. Note also that the sluicing in the purpose clause in (14b) indicates that this directive, apart from pushing the reader forward, also strongly connects with the previous part of the discourse. These imperatives are, of course, not real orders and, as Takahashi (Reference Takahashi2004: 61) correctly observes, it is in the reader's interest to comply.
4.3 Other uses of imperatives
With the possible exception of the imperatives from (18) to (23) in appendix B – if these are imperatives indeed (see footnote 7) – which are directed at policymakers as well as ‘ordinary’ readers, all imperatives are ways of engaging more closely with the reader. Some of them exhort the reader to take a certain attitude with respect to what is being discussed (e.g. ‘Get ready’, ‘Be under no illusion(s)’ (two occurrences), ‘Expect more countries to follow’).
A special type is the imperative used as a conditional:
(15)
(a) Look beyond the boom and bust of consumer tech, though, and you see the real successes. (‘Zoom fatigue’, The Economist, 3 September 2022)
(b) Strip out volatile food and energy prices, though, and underlying ‘core’ inflation is still roaring. (‘The perils of wishful thinking’, The Economist, 17 September 2022)
(c) Add to that the misery of the millions whose lives have been robbed of their common joys by useless prescriptions, and the case for change is unanswerable. (‘Set patients free’, The Economist, 29 October 2022)
Other examples are (3), (19) and (41) in appendix B. This type again underscores the similarity between imperatives and fragments in that the imperative is shorter than a full finite condition. However, it would be wrong to consider the imperative as actually an elliptic form of conditional clause (e.g. Look beyond… in (15a) as a shortening of If you look beyond…). This is clear from the fact that with be the form is really (don't) be rather than are(n't). (The data do not contain examples, but consider, for instance, Be prepared to take inflation into account and the situation looks even grimmer.) While the token frequencies are too low to make any firm statements about their preferred appearance, four of the six occurrences appeared in either the first or the last sentence within a paragraph.
4.4 Recurrent fragment types
The corpus being rather limited in size, the purpose of this study was not to present a full taxonomy of different fragment types in written (journalistic) discourse. Nevertheless, even in this limited collection of texts it was possible to identify a number of recurrent patterns of fragments which are worth commenting on. Given their low token frequencies, I will not make any claims about where they appear in the texts and paragraphs.
One pattern, the better (not) to VP construction exemplified in (16), happens to show again the close resemblance, pragmatically speaking, between fragments and imperatives: this fragment type has the illocutionary force of giving advice on what (not) to do.
(16)
(a) Better instead to let the market set the exchange rate, and keep using interest rates to tame inflation. (‘Keep your powder dry’, The Economist, 15 October 2022)
(b) Better to phase out the ‘triple lock’, a generous formula for raising state pensions, and raise money in more sensible ways: […] (‘Welcome to Britaly’, The Economist, 22 October 2022)
(c) Better not to pretend it is an option. (‘All we want’, The Economist, 26 November 2022)
(d) Better to point teenagers to more realistic sources, such as BISH […] (‘The wounds of silence’, The Economist, 3 December 2022)
Another recurrent fragment type is the Why VP? construction:
(17)
(a) And why comply with onerous rules if you don't have to? (‘All talk, no trousers’, The Economist, 1 October 2022)
(b) If a war is raging anyway, why use exquisite code when a missile will do?
(‘The digital front’, The Economist, 3 December 2022)
(c) So why not try it? (‘The wounds of silence’, The Economist, 3 December 2022)
In the examples cited above, the Why VP? fragment is enriched with a hypothetical-volitional modal meaning: in (17a), for instance, the intended meaning is ‘Why would you want to comply…’ (see Fernández-Pena & Pérez-Guerra Reference Fernández-Pena and Pérez-Guerra2023 for details on why-fragments).
One final type with more than one token in the corpus is the afterthought-like No wonder pattern, which allows some variation in determiner choice and presence of a complementiser:
(18)
(a) No wonder that some financial institutions are getting cold feet about green alliances. (‘All talk, no trousers’, The Economist, 1 October 2022)
(b) Little wonder people have taken to the streets. (‘China's covid failure’, The Economist, 3 December 2022)
It could be argued that each of these types is a fairly conventional way of adding a more breezy and personal tone to a text that, without them, would sound more stuffy and sterile. As such, they pattern with other speech-like features increasingly encountered in journalistic prose, such as contracted verb forms or the use of dashes. This latter typographical device represents written sequences as spontaneous additions, mimicking the way they would sound in spoken language (cf. footnote 12). In this regard, see Haselow (Reference Haselow, Bös and Kornexl2015), who discusses the occurrence of and or but at the start of sentences and the use of ‘syntactic increments’ after dashes.
5 Conclusion and prospects
The positions in which we find fragments and imperatives in editorials from The Economist suggest that journalists working for at least this news outlet are aware of the special status of these constructions (which, especially in the case of fragments, have a number of subtypes). If fragments and imperatives were just ordinary sentence types that were merely less frequent and syntactically somewhat different from canonical sentences, we would expect them to occur evenly distributed in a text and among the sentences making up a paragraph. This is not what we find. Both constructions have an increased likelihood of appearing as the first or the last sentence of a paragraph. At text level, they often appear in the last paragraph. While not more frequent in the first paragraph than under a chance distribution, fragments and imperatives are more common than expected in the second paragraph. This fact can be made sense of by recognising a discourse-level construction involving two paragraphs, where the first paragraph sets the stage by presenting a view and the subsequent one forcefully corrects that view.
This study makes a case for expanding our scope to broad discourse phenomena: when we characterise constructions, not only do we need to look beyond the boundaries of a sentence but we may also have to include stylistic and rhetorical considerations of how full paragraphs relate to one another and how an entire text is structured. This is not possible when we base our analysis on corpus output that just presents a limited amount of context around the phenomenon of interest. The ‘slow linguistic’ method used here, by which constructions are identified in texts studied in their entirety, has proved fruitful for this stylistics-oriented strand of research.
Some of the imperatives, as we have seen, are in paragraph-initial position because that is where they are most practical, for instance to push the reader forward. Other imperatives, however, as well as fragments, are meant to be eye-catchers. These sentences seem to be included to bring written discourse closer to oral language, but at the same time, they still stand out and are even made to stand out. It is hard, therefore, not to find a paradox in the use of many fragments and imperatives in carefully edited written prose. It's as if professional writers producing polished texts appear to feel the need to reserve these very ‘ordinary’ sentences for special occasions, making them in effect extra-ordinary.
It would be interesting to find out whether our findings from The Economist editorials can be generalised to news writing in general and to other written genres. One curious observation is that the popular generative AI application ChatGPT, when prompted to write an article in the style of a The Economist editorial, misses opportunities to use fragments and imperatives. It might systematically come up, for instance, with It is time for… rather than with Time for…. Even when instructed very explicitly to include fragments and imperatives, this instruction is either ignored or misunderstood or the result is unnatural, as in (19):
(19) User (commenting on an AI-generated article in the style of an Economist editorial on climate change): Nice! Did you know, though, that lead articles in The Economist sometimes contain non-canonical sentences, like “What to do?,” “To see why, consider the currency markets,” or “Good luck with that”? Could you attempt to integrate such slightly ‘deviant’ syntax into your article?
ChatGPT: Indeed, The Economist is known for its unique style, which often features non-canonical sentences and unconventional grammar. These stylistic choices are intended to engage readers and make the content more accessible and entertaining. As such, I will attempt to incorporate a few of these elements into my article.
Climate change is a pressing issue that demands immediate action. But what can be done to address the challenges we face? To see why, consider the complex web of factors that contribute to global warming, from carbon emissions to deforestation and beyond.
The sentence that starts with the part highlighted in boldface cannot appear after an interrogative. If a large language model is trained on authentic writing, and fragments and imperatives have their natural place in this, further research is warranted to explain such aberrant output.
As a second avenue for future research, it would be interesting to find out to what extent fragments and imperatives are similar to or different from other non-canonical sentence patterns used in journalistic discourse. Celle (Reference Celle2024) notes how answers, which are often of a fragmentary type, can be given to questions raised by the writer in an editorial, which is a kind of ‘monologic’ discourse (as there is no addressee physically present to interact with the writer). Monologues may, furthermore, be a mode of language use where the speaker can use fragment constructions that ostentatiously signal, almost paradoxically, a sort of inner speech, as argued by Kanetani (Reference Kanetani2022) for the because X construction. While fragments are, of course, still interesting for their syntactic properties, it is clear that they have not yet given away all their pragmatic, discourse-functional secrets.
Appendices
A. Fragments in The Economist editorials (issues of 3 September – 24 December 2022)
B. Imperatives in The Economist editorials (issues of 3 September – 24 December 2022)
