You are viewing a javascript disabled version of the site. Please enable Javascript for this site to function properly.
Go to headerGo to navigationGo to searchGo to contentsGo to footer
In content section. Select this link to jump to navigation

Rhetorical figures, arguments, computation

Note: Rhetoric has an extensive technical vocabulary. In this introduction, we explain these terms as we go along. The authors of the papers in this special issue do the same. But it can be easy to get lost in a forest of so many novel terms. So we provide a glossary at the end of this introduction.

1.The story of a tweet

At 6:00 PM, 3 December, 2016, @OmarjSakr, a bisexual, second-generation American from Lebanese-Turkish Muslim migrant parents, a poetry editor, an author, and a self-styled “geeky progressive,” tweeted the following:

  • 1. Women in Middle East attacked for not wearing hijab. Women in the West attacked for wearing hijab. It’s almost like women aren’t the problem. [48]

Then he went about his regular, non-Twitter affairs. The next afternoon, fresh from a nap, he checked his account, stared at the outcome, and tweeted to @upulieto, an associate, “I just woke up. What a mess” [49]. His tweet had spiraled virally: tens of thousands of likes, retweets, endless streams of commentary. Why?

Well, lots of factors, of course. There is rarely a single factor to which one can point in rapid cultural propagation. But the pithiness of the tweet, its form/function iconicity, and its argumentative force are surely three of those factors – perhaps the three most important – and all of them are a direct consequence of the tweet’s rhetorical figures. Its figures notably capitalize on repetition, a stylistic move one might think is anathema to twitter. Repetitions are highly redundant, contributing very little new information. They should therefore be corrosive to the concision so necessary to effective tweets. Repetitions distend the ideational function, not condense it, making them very costly in twitter, which provides only a 140-character vehicle. And yet: the tweet was instantly and broadly recognized as a brilliant, incisive, necessarily brief and highly effective argument.

Most notable among its figuration is the compound figure, symploce, in which the same word or words are repeated at the beginnings and endings of proximal clauses or phrases (Women in and wearing hijab). Symploce is a compound figure because it is a combination of epanaphora (phrase- or clause-initial lexical repetition; example 2) and epistrophe (phrase- or clause-final lexical repetition; example 3), the two of them often sandwiching similar or identical phrasing (as in example 1 and also 5)

  • 2. On October 29, we again entered the Palace of Justice; again the crowds were large and excited; again the security was extremely tight; again the court was filled with dignitaries from many foreign embassies. ([43]: p. 354)

  • 3. Now, no matter what the mullah teaches, there is only one sin, only one. And that is theft. Every other sin is a variation of theft. ([28]: p. 17)

These figures of lexical repetition, distinguished by location, get their salience, their memorability, and their aesthetic effects because of human neurocognitive affinities. We naturally respond to repetitions because they re-activate recently active neural pathways. We naturally attend to category boundaries because our minds are tuned to categorization and because our perceptions are tuned to ‘edges.’ But figures don’t ‘merely stimulate’ our mindbrains. They aren’t ‘mere decorations’ that put a little rouge on the cheeks of otherwise ‘pure information.’ They serve communicative functions (including pragmatic, suasive, and argumentative functions), in addition to their aesthetic and mnemonic functions.

Figures of lexical repetition are types of ploce. Simple ploce, usually just called ploce, is unconstrained lexical repetition. Its primary function is to ensure stability of reference. One repeats the same word to evoke the same concept. Complex ploce is constrained by position, and includes symploce, epanaphora, and epistrophe, which all serve theme-and-variation communicative functions. Complex ploce include both stable elements (because they are varieties of ploce), to ensure a consistency of reference, and variable elements, to associate that referential stability with a circumscribed semantic range. Epanaphora is a figure (in English) that features a stable framing of variable predications; in the Mandela example (2), each of the predications is framed in terms of an earlier event (that is, an earlier trial).11 Epistrophe (in English), conversely, most often puts a stable predication into variable frames; the Hosseini example (3) makes the claim that there is only one sin, then qualifies that claim by conceding that there are many other activities which are also sins, but the epistrophe makes the point that under proper interpretation they are subcategories of theft. Argumentation scholars might recognize epanaphora and epistrophe from their relentless appearance in textbooks, whiteboards, blackboards, exam questions and student papers, as in 4.

  • 4.
    • a. All humans are mortal.

    • b. Mr. Jones is a human.

    • c. Therefore, Mr. Jones is mortal

The major premise (4a) and the conclusion (4c) constitute an epistrophe. The minor premise (4b) and the conclusion (4c) constitute an epanaphora.

Symploce includes two bands of stability, putting the medial lexis into a kind of referential vise. The following prototypical instance of symploce (5), from Malcolm X, argues for the inexorability of racism played out in American martial geopolitics: the reference of the medial lexis is irrelevant – Germany, the South Pacific, Japan, wherever – the white man uses the black man as cannon fodder.

  • 5. [As] long as the white man sent you to Korea, you bled. He sent you to Germany, you bled. He sent you to the South Pacific to fight the Japanese, you bled. [42]

@OmarjSakr’s tweet has this vise-like quality as well, but there are three other figures highly relevant to its argumentative functioning: mesodiplosis (medial lexical repetition; in the tweet, attacked for; in 6 below, what we did not, with the slight variation of what we could not; it also occurs in 8 with is a); antithesis (proximal opposing predications; in the tweet, not wearing hijab and wearing hijab; in 7, young… life and old… death); and (simple) ploce (the brute lexical repetition in the @OmarjSakr tweet of women; in 8, of Canadian).
  • 6. We used to joke that between the poor acoustics of the hall and the confused and inaccurate reports of the Special Brand detectives, we could be fined for what we did not say, imprisoned for what we could not hear, and hanged for what we did not do. ([43]: p. 232)

  • 7. The young would choose an exciting life; the old a happy death. ([1]: p. 155)

  • 8. Have faith in your fellow citizens my friends, they are kind and generous. They are open minded and optimistic and they know in their heart of hearts that a Canadian is a Canadian is a Canadian. [61]

Mesodiplosis unites the terms on either side of the repetition (fine and not say, imprison and not hear, hang and not do in 6, as well as the ploce a Canadian of 8), here emphasizing the capricious malevolence of the apartheid system (6) and the utter identity of Canadian citizenship (8). Antitheses present binaries and/or polarities to push terms apart, especially a double antithesis like 7, which drives the polarities young and old further apart by associating them respectively with the binary opposites, life and death. And the main function of ploce is to guarantee stability of reference (often full identity of reference) – in 8, the relentless insistence that there are no differences that matter with regards to citizenship if you are Canadian.

In order to see all of these figural functions at work in @OmarjSakr’s tweet, we can render its component sentences as so:

  • 9.
    • a. Women in Middle East [are] attacked for not wearing hijab.

    • b. Women in the West [are] attacked for wearing hijab.

    • c. It’s almost like women aren’t the problem.

Sentences 9a and 9b are the symploce and they also manifest the binary antithesis of wearing/not wearing the hijab, and all three sentences (9a, 9b, and 9c) manifest the ploce of women. So, we end up with an ad absurdum (distilling out Middle East and West and assuming problem throughout):
  • 10.
    • a. (x)((Wx&Hx)Ax)

    • b. (x)((Wx&Hx)Ax)

    • c. (x)(WxAx)

Rendered this way, we see a more precise distribution of figures. We have the epanaphora of (x) in 10a, b, and c, along with the more extensive epanaphora of (x)(Wx& in 10a and b. There is the epistrophe of Ax in 10a, b, and c, which, in turn gives us the compound figure of symploce in 10a, b, and c (all of them begin with (x) and end with Ax). And there is the antithesis of 10a and b (Hx and Hx respectively). The crucial components for the argumentative function of the tweet are the binary nature of the antithesis and the stability-of-reference of the ploces throughout.

We haven’t fully explicated the argumentative activity of the tweet, of course, because there is also another figurative dimension (namely, the note of irony with almost like), and because the argument is enthymematic, with the covert premise that men are responsible for the attacks (so that the presence or absence of the hijab is just an excuse to exercise power), and because we have only charted the overt style. But we do have its logical force. One of the primary functions of antithesis, when the opposing predicates are binary, rather than polar, and when it couples referential-stabilizing ploces, is to signal a paradox, and that’s what it does in this tweet. ‘The situation is absurd,’ @OmarjSakr is saying. ‘Women are attacked for x and attacked for x, so we know the presence or absence of x is irrelevant. Women just get attacked.’

There are at least two morals here, in the story of @OmarjSakr’s tweet, about the role of rhetorical figures in argumentation. Firstly, figures activate neurocognitive affinities. Figures are salient, aesthetically appealing, and memorable because they excite fundamental neurocognitive principles, like repetition and boundary prominence. Secondly, structural iconicity signals function. Repetition not only gives textual elements presence, in Perelman’s sense ([52]: p. 17–18), it also guarantees a stability of reference. Endings and beginnings not only give textual elements presence, they also play crucial syntactic roles (which, of course, vary with the syntactic conventions of languages, but vary with considerable predictability). Rhetorical figures, in short, play central roles in argumentation, as the papers in this special issue amply demonstrate.

But Argument is only one of the defining terms of our journal’s title. The other is Computation.

2.Rhetorical figures, argument, and computation

Rhetorical figures, patterns that are present in all languages, reflecting both neurocognitive affinities and communicative imperatives, are linguistic devices that serve mnemonic and aesthetic purposes, alongside informational and argumentational purposes. As examples 1–10 illustrate clearly, they very frequently do so in clusters, sometimes in quite dense clusters, of the sort we find in syllogisms. All three of these characteristics: the affinities, the imperatives, and the clustering make them immensely valuable for computational research into language and argumentation.

Our opening example for instance – @OmarjSakr’s 3 December, 2016, tweet about women, oppression, and patriarchal misogyny (1, 9, 10) – incorporates the neurocognitive affinity of repetition and the communicative imperative of stabilizing reference. Repetition is such a fundamental aspect of neurocognition that we literally could not think without it. For that matter, we could not even breathe, our hearts couldn’t beat, our cells would deteriorate, without the repeated firing patterns of our neurons. It’s no wonder we resort to repetition in trying to remember a name, a phone number, an address, repeating them to ourselves vocally or subvocally. Repetition is a sine qua non of our brains. Stability of reference is so fundamental a communicative imperative that we literally could not communicate without it. Even animal communication systems with such broad referents as, in the vervet monkey call, ‘danger from above’ or, in the bumblebee dance, ‘food source lots of wing-beats hence from the hive,’ would fail utterly without a system that guaranteed the same noises or dance moves evoked the same concepts. @OmarjSakr’s tweet guarantees stability of reference through repetition: women evokes the same concept in all three sentences, attacked and wearing hijab evoke the same concepts in two of them, and so on. The links between affinities and imperatives are not always as tight as between repetition and stability, but they can always be traced, which is what makes rhetorical figures so valuable computationally. We want to be very clear about this: because the patterns are motivated by neurocognitive affinities, they are more robustly linked to functions and meanings than the patterns of familiar syntactic analyses, even Construction Grammar, which all operate on assumptions of arbitrariness; and because they can be charted computationally, in addition to their practical value, they help us to move back towards the originary goal of Artificial Intelligence, using computers to tell us fundamental truths about the mind.

The tweet also activates the neurocognitive affinities of contrast (East, West; not wearing, wearing) and demarcation (women and hijab both appear at clausal boundaries), and the communicative imperatives, among others, of negation (not, aren’t) and predication (in Middle East and in the West predicated of women, etc.).

Deploying affinities and realizing imperatives, rhetorical figures form into recurrent patterns: epanaphora, repetitions at the beginnings of clauses or phrases; epistrophe, repetitions at the ends of clauses or phrases; mesodiplosis, medial repetitions between distinct elements; antithesis, proximal opposing predications. The patterns and the functions correlate, which can provide computer algorithms with data on all aspects of discursive action, prominently including argumentation.

Rhetorical figures solidify and propagate arguments in revealing ways, as a major paper by Jeanne Fahnestock demonstrates. The paper, entitled “Preserving the Figure,” focuses on scientific argumentation and charts the ways in which certain figures are not just used by primary research scientists reporting on their work, but are also picked up and circulated through other reports of that research [19].22 One can track the ‘same argument’ through different incarnations by its figures. Equally, one might chart out entire discursive episodes in figurative terms. For instance, the 2008 US election campaign between Barack Obama, with Joe Biden, and John McCain, with Sarah Palin, was characterized by chiastic figures (figures in which constituents, such as words or phonemes, repeat in reverse order), as documented by New York Times editor and Slate columnist, Juliet Lapidos:

  • 11. We were elected to change Washington, and we let Washington change us. (John McCain [38])

  • 12. In politics, there are some candidates who use change to promote their careers. And then there are those, like John McCain, who use their careers to promote change. (Sarah Palin [38])

  • 13. We need a president who puts the Barney Smiths before the Smith Barneys. (Barney Smith [38])

  • 14. He has brought change to Washington, but Washington hasn’t changed him. (Barack Obama, of Joe Biden [38])

  • 15. In the end the true test is not the speeches a president delivers, it’s whether the president delivers on the speeches. (Hillary Clinton [38])

Chiastic figures manifest the neurocognitive affinities of repetition, contrast, sequence, and symmetry, from which they get their aesthetic and mnemonic appeal. Functionally, they manifest the communicative imperatives of reversal (as well as, in other contexts, reciprocality and irrelevance of order), so it was the natural figure for a campaign in which the policies and actions of George W. Bush, the incumbent, were so deeply unpopular. The arguments that characterized the campaign were to reverse everything Bush stood for. But the best known example of chiastic reversal is perhaps John F. Kennedy’s famous expression from his inaugural address:
  • 16. Ask not what your country can do for you. Ask what you can do for your country. [33]

This expression famously includes antimetabole (the lexical chiastic figure). Less noticed is that it includes antithesis. When two figures occur together, they often constrain each other’s rhetorical functions to achieve singular effects. In this case, antithesis contributes a function of mutual exclusivity, familiar from Parmenides principle of non-contradiction, while the antimetabole contributes inverse relations, and we end up with mutually exclusive inverse relations between citizens and their country; the grammatical imperative specifies which relation Kennedy advocates.33 All of the 2008 election examples function similarly. Example 11 depicts a topsy-turvy world in which one orientation (changing Washington) is clearly preferred, letting the positive valence of elected to carry that preference, 12 leaves that work to the formulaic epanaphora of the there are… and then there are construction to specify the preference, and 13 leverages the deontic modality of need and the put x before y construction. Examples 14 and 15 use the simple antithesis of negation to do the job. But they all endorse one of two opposed states of affairs, expressed chiastically.

The 2016 US election, on the other hand, was not so much about wholesale change: Hillary Clinton largely represented a continuation of President Barack Obama’s trajectory while Donald Trump represented a radical dismantling of the policies and programmes defining that trajectory. One suspects that antithesis was highly prevalent in the campaign, as the two candidates represented themselves as binary opposites (fit, unfit; honest, dishonest), and thinking only of Trump’s rhetoric, hyperbole (best, single greatest, largest, worst, … ever, in the country, in the world, in history) and epithets (Little Marco, Lyin’ Ted, Crooked Hillary) would seem highly prevalent as well. But, here’s the thing: the research hasn’t been done. Even the 2008 campaign, the year ‘of the reversible raincoat’ [38] as chiastic figures are sometimes nicknamed, was diagnosed with conventional humanist methods. Computationally, with the right algorithms we could characterize campaigns, speakers, genres, argument styles, virtually any discursive product, process, mode or episode, on the basis of its rhetorical figures; with the right knowledge base, of how figures work together, how they combine with grammatical constructions and lexical properties, and what their functional propensities are, we could understand those discourses at a deeper level than we have so far been able to manage computationally.

Fahnestock’s work in “Preserving the Figure” [19] can be seen as a pilot study. As she points out, her results confirm the expectations of rhetorical theory: “[figures] are typically used in the process of conveying the core of an argument in research articles and in versions of these articles constructed for wider audiences” ([19]: p. 14). But she looked at only pairs of articles, not multiple iterations, and only at 26 pairs in total, and only in scientific argumentation, and only in two genres. The broader scope available to computational research tools could go much further in testing the role of figures in the propagation of arguments.

We could find out, for instance, if Fahnestock’s observation generalizes more broadly, if arguments frequently travel by way of figural realizations. As one test case for her thesis, we might look to see the discursive life of Gregor Mendel’s formula for the distribution of dominant and recessive traits in his pea experiments (example 17), which centrally features a chiastic inversion of the case of a variable (the majuscule and miniscule versions of A switch places in the middle two ratios).

  • 17. A/A+A/a+a/A+a/a ([26,45,46])

Certainly some of the reports on Mendel’s experiment in the early propagation of his work after its ‘rediscovery’ in the early twentieth century (e.g. [3]: p. 73–74) reproduce this formula, but how many and in which contexts? Is it still preserved, in some version, in contemporary textbooks? Corpus searches could answer questions like these for figures and their functions much more fully than traditional humanist scholarship could reveal, and tell us a good deal about argumentation in the bargain. Note that 17 is not an equation, not an expression of some relationship among theoretical entities (like, say, A=πr2 or e=mc2). It is just an elegant formal expression of Mendel’s methodology, a quasi-mathematical turn of phrase. To the extent 17, or some variation of 17, persists in the literature around Mendel’s research, it preserves a figure that in turn “preserves the research design” ([19]: p. 15), epitomizing Mendel’s thoroughness in testing all the possible combinations of what would later be called genes.

But research of this sort has not been done, and the computational research of rhetorical figures in argumentation is just now in its very earliest stages.

Despite its ancient foundations ([51]: p. 26–31) and some remarkable work in the twentieth century, with Fahnestock’s programme of Figural Logic leading the pack ([17,19,20,60], [51]: p. 169–181, [55,60]), rhetoricians and argumentation scholars have been very slow to catch on to the role of rhetorical figures in argumentation. How slow might be diagnosed by the two volumes of the Proceedings of the 1st European Conference on Argumentation [47]. The volumes include some 150 papers and commentaries, from the work of PhD students to the foremost scholars in argumentation studies, and aspire to provide “the most complete statement of the state of the art of argumentation studies today… [f]rom ancient rhetoric to Artificial Intelligence” (according its back-cover blurb). Not a single paper deals with rhetorical figures except in the most peripheral way. In Computer Science, the situation is little better. The case has been made, dating to the turn of the century, that computational argument studies have a rich potential to investigate figures ([8,9,25,54]), and there is growing success in the detection of rhetorical figures ([12,13,23,24,29,31,50,58]). But, until this volume and the groundbreaking work of Lawrence, Visser, and Reed, these two strains – the detection of the forms and the study of argumentative functions – have not been brought together.

3.What is a figure, and what is just some accidental, figure-like clump of linguistic properties?

One issue has dogged the entire history of rhetorical figures, and it is one that is particularly relevant for computational detection: when is some expression, x, a ‘real figure,’ and when is it just another arbitrary piece of language? The issue often comes up around notions of literality, often around tropes, and usually in discussions starring metaphor. For instance, such discussions might pick up on the use of dogged in the initial sentence of this section. Is it ‘really’ a metaphor, or just a verb that means something like ‘attend closely’? (Or perhaps both, going into a third category, the dead or dormant metaphor.) While metaphor and tropes more generally are the focus of such discussions, other words and phrases are equally implicated, and they help to show how artificial the whole ‘real figure’ issue is. We don’t ask whether words like rolly poly and hoi polloi ‘really’ rhyme, for instance (word-final syllable repetition), or whether vice versa and ship shape ‘really’ alliterate (word-initial consonant repetition). Rhetoricians, with a larger vocabulary of these patterns, don’t stay up nights wondering if very very or bye-bye are ‘real’ epizeuxes (immediate lexical repetition, with no intervening material), whether easy come, easy go or first come, first served are ‘really’ epanaphoras, whether clichés like we shall see what we shall see or it is what it is are ‘really’ epanalepses. There are processes of familiarity and entrenchment that can make property clumps in some expressions less visible than in others. For semantic properties, which are not actually visible at all, the figural motivations can disappear quickly into history for all but the etymologists. Semantic figural motivations for words and expressions are ground down in the constant friction of linguistic commerce. Those processes, in an attrition of salience, are what give us the perceptual phenomenon we call literality. Formal figural motivations can grind away as well, leading to a translucent blandness that gives us clichés and other routine figurations, but they are as much a part of everyday language as the vaunted semantic motivations.

Because they emerge from our cognitive dispositions, and because they satisfy functional needs, figural patterns arise naturally in all communicative contexts. Lawrence, Visser, & Reed’s contribution to this volume shows that very clearly for the communicative context we know and love as argumentation. For us, therefore, rhetorical figures are not specialty devices reserved for orators, poets, and rap artists. While craft and talent can produce distinctive figural realizations, figures do not require deliberation or design. A figure is its pattern, and no pattern is more authentic than another.

Concerns about authenticity and design, however, lead many researchers to worry about false positives in figure detection [12,13,29,58]. Dubremetz and Nivre, for instance, rank the examples they find on a scale of figural prototypicality ([12]), while Lawrence, Visser, & Reed, in this volume, talk in terms of quasi- and non-instances of figures in their results – things that aren’t ‘really’ figures, just accidentally figure-like. The concerns underlying these designations are genuine, and the algorithmic activity that finds and even ranks these differences in the texts they plumb is very important. We need to know what to call, and how to treat, the pieces of data the algorithms return. But these researchers are not finding and categorizing differences among individual figures.

In our research, we adopt the policy that figures are defined only by their linguistic properties (such as the repetition or the cross-domain mapping of words). Or, in the slogans of our lab, “Figures are what figures are,” and “a ploce is a ploce is a ploce.” If an algorithm comes back with proximal reverse occurrences of at least two words, it has come back with an antimetabole. Period. In fact, what detection researchers are finding is not better or worse versions of given figures, but sites of more or less figural action. Take the following examples, from Dubremetz and Nivre [12]:

  • 18. A comedian does funny things, a good comedian does things funny.

  • 19. We are all European citizens and we are all citizens of a European Union which is underpinned[, politically speaking, by a basic set of principles].

Dubremetz and Nivre are on the hunt for chiasmi, figures of reverse constituent repetition, and we agree with their intuitions that 18 ‘feels’ more chiastic than 19. We also agree that the notion of prototypicality seems to catch what is triggering our respective intuitions.44 Now, we are sure Dubremetz and Nivre would agree with us (since they use it as their illustrative prototype in all three of their papers) that an even more prototypical and cognitively satisfying chiastic instance of antimetabole is the following widely known expression, made famous in the Dumas novel, Les Trois Mousquetaires, and in related cultural artifacts (comic books, movies, television programmes, and so on):
  • 20.
    • [T]ous pour un, un pour tous. ([15]:129)

    • All for one, one for all. ([16]:80)

But if we map our intuitions against the data, we find they cannot be based on the presence (or absence) of a single figure, because all of these examples manifest multiple figures (multiple clusters of specific linguistic properties). Example 20, in particular, features not just the central member of the chiastic suite, antimetabole, but also isocolon (prosodic repetition), parison (syntactic-structure repetition) and mesodiplosis (as above, medial lexical repetition). Notice that the somewhat less satisfying example, 18, does not include parison: in the first clause, funny things is a Noun Phrase, featuring an adjective and noun; in the second clause things funny is a Noun Phrase (the single noun, things) and an Adverb Phrase (funny). These syntactic differences also throw the prosodic repetition out of whack, because of the different stress patterns of the phrases; so, no isocolon either. Our next example on the Dubremetz-Nivre scale, 19, falls yet further out of parallelism. The two key phrases, European citizens and citizens of a European Union have very different prosodies, the syntactic differences are greater, and the extra words, especially the head word of the last noun phrase, Union, alter the semantic reference notably. Example 19 is not, in our terms, less of an antimetabole than 18 or 20. But it is less figuratively dense (and less figuratively elegant). The antimetabole of 19 is not presented in the syntactic and prosodic settings that stage the antimetaboles of 18 and especially 20. The matter is not quite so simple as adding up the number of figures present in an expression, however. Some figures work together; some work against each other. Figures of parallelism, in particular, like isocolon and parison, seem to have a special role in augmenting other figures [18], while tropes can get in the way of other figures, sometimes degrading coherence. Example 18, for instance, includes antanaclasis, a trope in which a signans repeats but with a different signatum. The funny of the first clause is an adjective. The funny of the second clause is an adverb. They are different words (that share the same signans).

Despite the thousands of years that figures have been studied, organized, and theorized, there is considerable research remaining to do into how they combine, and what sorts of functions they serve in specific combinations (the discussion of antithesis-antimetabole combinations in connection with 16, is one such combination). Luckily for us, computers are the best tools for doing such research.

4.The papers in this special issue

This issue features four rather different papers, organized around the intersection of rhetorical figures and argumentation, all of them with an eye on computational research, all of them major contributions to computational rhetoric. They are nicely split among the disciplines. Two are by rhetoricians, with their emphases on context, historical development, and linguistic nuance. Two feature teams in computer science, with their attention to application, tractability, and testable results.

“Rhetorical figures as argument schemes – The proleptic suite,” by Ashley Rose Mehlenbacher, is an important bridging study that helps integrate the category of figures known as moves with the extensive research in argument schemes. Mehlenbacher works through the distended and terminologically inconsistent scholarship on figures of anticipation and rebuttal, in literature and in argumentation, to reveal a tractable set of such figures, aligning that set with the classical tradition of topoi and its contemporary descendants, argument schemes.

Ying Yuan’s “The argumentative litotes in The analects” finds not quite a suite of figures, but through its careful study of a subtle trope of negation in an ancient Chinese text, alongside Aristotle’s not-quite-as ancient semantic analysis of opposition, finds three distinct subtypes: contradictory, contrary and relative litotes. Among the many virtues of rhetorical figures for the study of argumentation is that their neurocognitive dimensions point strongly to a universalist, cross-cultural approach to the form arguments take. There is perhaps an infinite variety of argument realizations, and languages all have different grooves in which they flow, but since arguments are all the products of human minds engaging other human minds, there are also important patterns of commonality. We see those patterns in language (the presence of nouns and verbs, for instance, and subjects and predicates, and morpholexical-syntactic relations). Figures provide a way to see those patterns of commonality in argumentation. So, a study based in a language like Chinese, with such different grooves from the familiar European languages in which argumentation has been investigated, is always a welcome addition to the field.

“Ontological representations of rhetorical figures for argumentation mining,” an international collaboration by Jelena Mitrović, Cliff O’Reilly Miljana Mladenović, and Siegfried Handschuh, also has cross-linguistic virtues. It surveys several early projects ontologically modelling rhetorical figures, including the most ambitious such project, in Serbian. It also attends seriously to the application of ontologies in the detection of figures, as well as the detection of the semantic relations of Rhetorical Structure Theory, the inferential relations of argument schemes, and other relevant text relations and structures – the necessary first step in using them for argument mining. In connection with this paper in particular, we should note a discrepancy in the use of the word rhetorical in the important text classification and analysis research coming out of Mann and Thompson’s Rhetorical Structure Theory (RST) [44]. As we have noted elsewhere [56], we regard that work as vitally important, and RST more generally as a valuable set of insights into the semantic organization of texts, especially at inter- and intra-clausal levels, but the adjective rhetorical in the name of the theory is gratuitous. The theory, and the research it has sponsored to this point, has no serious connection to the rhetorical tradition at all, as it is understood to date back to the 5th century BCE. RST would be more accurately labelled if it were TST (Textual Structure Theory) or DST (Discourse Structure Theory), or the like. We certainly don’t wish Mann and Thompson now to rename their approach, but we caution readers to be aware that the use of rhetorical is wholly equivocal – between a vague and generalized signifier and a technical, disciplinary term – whenever RST and genuinely rhetorical notions are discussed together. Mitrović and her colleagues are aware of this difference, of course, and work to keep the usages distinct, but the pitfall is there all the same for readers.

All of the papers in this issue bring together rhetorical figures, argument analysis, and computation in productive ways that help set the agenda for this emerging confluence, but the final paper is groundbreaking. In “Harnessing rhetorical figures for argument mining: A pilot study in relating figures of speech to argument structure,” John Lawrence, Jacky Visser, and Chris Reed, all of them affiliated with the Centre for Argument Technology, bring for the first time the form-function alliance of rhetorical figures into serious commerce with computational argument studies. It has been a fruitful decade since Jakub Gawryjolek initiated computational figure-detection work ([23,24]). Important innovations have appeared, both in terms of the detection strategies and the diagnostic deployment. Claus Strommer [58], for instance, makes headway into document intent, finding correlations with pathos (emotional appeal), and James Java reports promising results with a range of figures for authorship attribution ([31]). Daniel Hromada added feature constraints ([29]) for a small set of repetition schemes, while Marie Dubremetz and Jakob Nivre zeroed in on the tricky but rich and fascinating chiastic figures (primarily antimetaboles, though they do not distinguish among them), charting them in terms of prototypicality ([12]), increasing precision by including syntactic structure (augmenting their shallow-feature approach [13]), and have corpus-trained a chiasmus classifier (rather than the previous hand-tuning methods [14]).

But no one before Lawrence, Visser, and Reed has directly investigated the argumentative functions of rhetorical figures computationally. It is a pilot study, but – since we’re talking about rhetorical figures, we hope you’ll forgive the conceit – we think of it more as a pilot light, one that can now ignite two tremendously important and mutually informing lines of research that are nascent in earlier work. The first line tests ancient and modern claims about the general purposes of rhetorical figures, and about the specific purposes of many individual figures; very likely, we are confident, uncovering new purposes as well. Much as computational methods and corpus studies revolutionized lexicography by validating, improving, and replacing aspects of traditional dictionary writing, computational corpus research that finds figures and investigates the surrounding text can renovate the field of rhetoric. The second line more directly implicates argumentation studies, by providing subtle and powerful new tools for argument mining. As empirical rigour grows and results accrue and we build an inventory of rhetorical figures to serve as discourse diagnostics, we can use computational methods to develop a deeper understanding of the cognitive and social dimensions of argumentation, and more robust methods to probe argumentative discourse.


The terminology of rhetorical figures, and rhetoric more generally, is arcane and, with the exception of a few terms (such as metaphor and irony) largely unfamiliar beyond the discipline of rhetoric. This glossary includes all the rhetorical terms in the volume, most prominently all of the figures, with examples (in parentheses, with relevant elements highlighted, many of them taken from the articles in this volume). A word about the examples, however: since rhetorical figures very often travel together, a passage that exemplifies one figure will also exemplify others. One needs to map the definitions against the examples to identify the elements that represent the specific figure at hand; and, equally, to disregard the other figures that might be present.

Please note that the terminology of rhetorical figures is quite complex, and many of the figures in our glossary have synonyms, near synonyms, or alternate spellings that might be encountered elsewhere; equally, the definitions you find elsewhere for some of these terms may be at variance with the definitions here. Many terms might refer to the same configuration (an embarrassment of synonyms); conversely, one term might refer to multiple configurations (an embarrassment of negligent referential inconsistency). There is, in short, a pressing need for the standardization of figurative terminology. In this volume we adhere to precise one-term-one-configuration definitions and adopt the four-way taxonomy of figures defined in the Waterloo Rhetorical Figure Ontology [5,26]: schemes, tropes, chroma, and moves. The scheme and trope categories are among the oldest in figuration, and, construed according to a simple signans/signatum division, the most basic and the easiest to see. Schemes are formal deviations, shifts away from conventional expectations in the usage of signantia. Protoypical examples include rhyme and ploce. Tropes are conceptual deviations, shifts away from conventional expectations in the usage of signata. Protoypical examples include metaphor and metonymy. Chroma are deviations from conventional expectations of speaker intention. Protoypical examples include erotema (“rhetorical question”) and sarcasm. Moves are deviations from conventional expectations of discourse patterns. Protoypical examples include anemographia (descriptive passages evoking wind) and paralipsis (pretending to ignore something while giving it salience).

While this glossary no doubt feels exhaustive (if not exhausting) to non-rhetoricians, it is only comprised of the terms used in this special issue, including a relatively small (but highly representative) slice of the available figures. Also, while there is an abundance of technical terminology in this volume, and in any scientific journal, this glossary only concerns itself with rhetorical terms, on the reasonable assumption that the vocabulary of computational linguistics and argumentation theory is familiar, or easily accessible, to our readership.


A synonym for epizeuxis.


A synonym for polyptoton.


A scheme: the repetition of a consonant or consonant sequence at the beginning of proximal words (“The terrorists of the Islamic State of Iraq and Al-Sham (ISIS) have proven themselves to be a brutal, bloodthirsty band of brigands.” [27])


A scheme, and a type of (complex) ploce: the repetition of a word or word sequence on either side of a clause or phrase boundary. (“… you won’t prevent poverty. Poverty being more than income…” Example from Lawrence, Visser & Reed’s essay, in this issue.)


A move: the disruption of a narrative to relate prior episodes, sometimes very brief, sometimes quite distended; colloquially, flashback. (“In my younger and more vulnerable years my father gave me some advice that I’ve been turning over in my mind ever since. ‘Whenever you feel like criticizing any one,’ he told me, ‘just remember that all the people in this world haven’t had the advantages that you’ve had.’” [22]:1)


A synonym for epanaphora (avoided particularly for its potential confusion with linguistic anaphora)


A move, and a type of enargia: a vivid description invoking properties of wind. (“[T]he wind was raging like a shivering old devil, its roar punctuated by groans, screams, and wild laughter.” [36]:152–153)


A trope: the repetition of a signans with a distinct signatum, one of Pereleman and Olbrecht-Tyteca’s [52] quasi-logical figures. If the repetition is covert or unintentional, the antanaclasis is a logical equivocation. Overtly, it is often used for humour or presence. (”In America, there is plenty of light beer and you can always find a party. In Russia, Party always finds you.” [66])


A trope and a member of the chiastic suite, also known as a pseudo-antimetabole: the reverse repetition of at least two signantia at least one of which is paired with a different signatum, usually for humorous or enigmatic effect (”[Y]ou can prick your finger, but don’t finger your prick.” [4])


A synonym for prolepsis.


A scheme, a type of (complex) ploce, a central figure in the chiastic suite, one of Pereleman and Olbrecht-Tyteca’s [52] quasi-logical figures: the proximal occurrence of the same two words or word-strings in reverse orders. (“We were elected to change Washington, and we let Washington change us.” [38])


A scheme,5 a figure in the chiastic suite: the proximal occurrence of the same two sounds (phonemes or phoneme clusters, including syllables) in reverse orders, also known as acoustic chiasmus. (“Speaking as a Christian, I would like to say that I find the Apostle Paul appealing and the Apostle Peale appalling.” Stevenson [1952], qtd in [11]: p. 118.)


A type of metaphor, and therefore a trope: the depiction of a human as an object. (“Juliet is the sun” [57]: Romeo and Juliet, act 2, scene 2.)


A trope, both a figure and a class of figures: proximally opposed predications, through antonyms or affirmatives and negations. (“… isn’t there an element in this of Christianity fetishising poverty? Doesn’t Christianity have a problem with wealth creation?” Example from Lawrence, Visser & Reed’s essay, in this issue.)


A trope: the use of a description in place of a name, often metonymic (“Every self-mover, as the Philosopher [Aristotle] proves… is composed of mover and moved” [2]: p. 152.)


A synonym for prolepsis.


A scheme: the omission of a phoneme or syllable at the beginning of a word (“The King hath cause to plain [for complain]” [57]: King Lear, act 3, scene 1.)


A scheme: the omission of a phoneme or syllable at the end of a word (“What oft [for often] was thought but ne’er so well expressed” [53]: Alexander Pope, Essay on Criticism 2.)


A scheme: the omission of conjunctions or disjunctions in a list (“He was a bag of bones, a floppy doll, a broken stick, a maniac” [34]: p. 106.)


A type of metaphor, and therefore a trope: often called a ‘vice’ rather than a figure, catachresis is the name given to metaphors perceived as egregiously bad cross-domain mappings, sometimes called mixed metaphors (“We can’t continue to allow China to rape our country, and that’s what they’re doing.” [62])


A trope: ordering words in terms of the decrease of some semantic property (also known as decrementum). (“[H]e matured, grew out of his confining literalism, developed a spirit of informed and skeptical inquiry, advanced in wisdom and rounded judgement, got old, declined, died” [10]:143.)


Often used as a synonym antimetabole (as in [51]), sometimes used as a synonym for other members of the chiastic suite (as in [17]: p. 123 et passim). In the Waterloo Rhetorical Figure Ontology, it is recognized as a general pattern of reverse repetition of any linguistic constituent, with different figures defined by which constituents are implicated. See the chiastic suite.

chiastic suite

A group of figures, including schemes, tropes, and moves in which constituents repeat in reverse order, prominently including antimetabole.


A major category of figures, chroma are deviations of conventional expectations of speaker intention.


Also commutatio, a term Pereleman and Olbrecht-Tyteca [52] use for a compound figure of antithesis and antimetabole, one of their quasi-logical figures. (“We should not judge rules and duties by morals and customs, but we should judge customs and morals by duties and rules” [52]: p. 428.)

complex ploce

See ploce.

compound figure

A specific figure comprised of other figures, to be distinguished from a passage of text which includes, for various reasons, multiple figures. Only a few specific combinations of figures are compound figures (e.g. gradatio and symploce).

computational rhetoric

Computationally mediated research into persuasion, credibility, affect, arrangement, topical structure, style (including rhetorical figures), or other notions informed by the rhetorical tradition.


A trope: an ‘extended metaphor or simile,’ drawing out a cross-domain mapping by explicating the details (“Thicke was a breakfast buffet. Hash brown hair, egg yolk eyes, soft butter lips, and short stack of pancakes for a chin” [35]:89)


A move: the use of an opponent’s argument to support one’s own case (“I agree that training and court speed and strings all distort head-to-head comparisons. But, again, these factors cut both ways. Rod Laver didn’t have the benefit of NetJets and a phalanx of coaches and on-site masseuses. Then again, neither did his opponents.” [41])


A synonym for antithesis and other tropes of opposition.


A syonym for epistrophe.


A move, a type of procatalepsis, and so a member of the proleptic suite: the admission and rebuttal of a fault or guilt that might be used in a counter-argument. (”I was trying to get people to get into the boats in an orderly fashion. Suddenly, since the ship was at a 60–70 degree angle, I tripped and I ended up in one of the boats. That’s how I found myself in the lifeboat.” Spoken by the captain of a sinking ship who did not stay aboard until all passengers had been evacuated. [7])

dirimens copulatio

A move, the juxtaposition of two statements, one of which qualifies the other in a superficially negative environment, frequently incorporating the not only...but also construction. (“Trust not only requires moral integrity, but also competence.” Example from Lawrence, Visser & Reed’s essay, in this issue.)


A class of moves: the general term for vivid descriptiveness. There are many, quite specific subtypes, including anemographia. (“Never was there such a town, I thought, for the smell of fish and chips on Saturday evenings; for the Saturday afternoon cinema matinees where we shouted and hissed our threepences away; for the crowds in the streets with leeks in their hats on international nights; for the park, the inexhaustible and mysterious, bushy red-Indian hiding park where the hunchback sat alone and the groves were blue with sailors.” [59]: p. 7)


A scheme, and a type of (complex) ploce: the occurrence of the same word or word sequence at the beginning of proximal clauses or phrases. (“It can be a way of demonstrating independence. It can be risk averse. It can be conservative.” Example from Lawrence, Visser & Reed’s essay, in this issue.)


A synonym for epistrophe.


A scheme, and a type of (complex) ploce: the occurrence of the same word or word sequence at the end of proximal clauses or phrases. (“If I lend you money, I’m taking a risk. And I should carry that risk.” Example from Lawrence, Visser & Reed’s essay, in this issue.)


A move: the use of a word or phrase with a noun, often a proper noun, often metonymic, but also metaphorical, synecdochal, or ironic. Similar to antonomasia except that the name is augmented rather than replaced. (“swift-footed Achilles,” “wine-colored sea,” “laughter-loving Aphrodite.”)


A scheme, and a type of (complex) ploce: the immediate repetition of a word or word sequence with no other words intervening. (“… the system seems to be very, very heavily weighted against them.” Example from Lawrence, Visser & Reed’s essay, in this issue.)


A chroma: the use of a question to serve some other purpose besides the request for information, often to assert; colloquially, rhetorical question. (“Willis stopped in midsentence, and a smile spread across his face ‘Oh!’ he said. ‘Oh! Yeah, I get it! I ain’t supposed to know who done it. That way I’ll be tryin’ to figure it out.’ ‘Aren’t you the brilliant one, though,’ Atwood said” [32]: p. 201.)


A move: The numbering and ordering of phrases or clauses. (“The growth of what you call ‘welfarism’ is to do with, firstly the disappearance of so many skilled middle income jobs; and secondly due to the failure to build social housing.” Example from Lawrence, Visser & Reed’s essay, in this issue.)


A synonym for epizeuxis.


A scheme, and compound figure consisting of a sequence of anadiploses. (“The stories seem sadly familiar. They are becoming almost routine, which is frightening. Man behaving strangely is confronted by police. Police shoot man. Man dies. Details are sketchy. Repeat” [30]: p. A15.)


A semiotic characteristic of language and other systems of representation such that the form of the signans resembles the signatum in some way, such as in sound, in proportion, or in sequence.


A chroma, both a figure and a class of figures: deviations of conventional expectations of speaker intention, often in an exaggeratedly opposite way (“O, baš sam srećan što je Dr Igi najavio svoj povratak! (= O, how happy I am to see Dr Iggy’s comeback!) [uttered in a situation where the speaker is not happy at all to see Dr. Iggy’s comeback]” cited in by Mitrović et al. in this volume)


A trope: distending the valence of a characterization to an extreme. (“This is the greatest Witch Hunt in political history. Sad!” [63])


A scheme of syntactic interruption: the insertion of a prepositional phrase between naturally consecutive constituents. (“[I]t was suddenly possible to detect if not objections then questions about what in bloody hell Rupert was doing.” [65])


A trope which ‘reduces’ the valence of an assertion through double negations (see Yuan’s essay in this volume for considerable charting of three subtypes – contradictory, contrary and relative). (“His wife was not always out of humor, nor his home always uncomfortable.” Example from Yuan’s essay in this issue.)


A trope: the characterization of something in terms lower than its expected semantic valence; understatement; sometimes, mistakenly, meiosis is used synonymously with litotes. (“Ay, ay, a scratch, a scratch. Marry, ‘tis enough,” in reference to a mortal wound. [57] Romeo and Juliet, act 3, scene 1)


A scheme, and a type of (complex) ploce: the occurrence of the same word or word sequence in the middle of proximal clauses or phrases. (“We were elected to change Washington, and we let Washington change us.” [38].)


A trope, both a figure and a class of figures: a cross-domain property mapping with no overt comparative construction. (“if he only found the right book to suckle from, he would be saved.” [64]: p. 55)


A move, a member of the proleptic suite: the evocation and rebuttal of a known counter-argument. (“I will not make age an issue in this campaign. I will not exploit, for political purposes, my opponent’s youth and inexperience,” spoken by the elderly presidential candidate, Ronald Reagan; example cited in Mehlenbacher’s essay, this volume.)


A trope: a word correlated with a domain used to represent the domain. (“The pen is mightier than the sword.” [40]: p. 37)


A major category of figures, moves are specific discourse strategies, deviations of conventional expectations of discourse patterns.


Traditionally a synonym for prolepsis; used by Mehlenbacher in the present volume to distinguish aspects of the proleptic suite.


A syonym for eutrepismus.


A type of antithesis, and therefore a trope, one of Pereleman and Olbrecht-Tyteca’s [52] quasi-logical figures: The juxtaposition of words with opposite semantic properties, often in adjective-noun pairs. (”I like a smuggler. He is the only honest thief.” [37]: p. 154)


A move: asserting something while affecting not to assert it. (“[W]hen, by procuring the death of your former wife, you had made room in your house for another, did you not add to the enormity of that crime, by a new and unparalleled measure of guilt? But I pass over this, and choose to let it remain in silence, that the memory of so monstrous a piece of wickedness, or at least of its having been committed with impunity, may not descend to posterity. I pass over, too, the entire ruin…” [6]:1.159.)


A move: an interruption of some ordered presentation with a seemingly irrelevant story of exposition, a digression (no example provided for reasons of space).


A trope: the use of multiple terms with little ideational variation and high redundancy. (“She was young in the way an actual young person is young.” [67]: p. 97)


A synonym for antonomasia.


A scheme of syntactic repetition, often referred to as syntactic parallelism: the proximal repetition of the same syntactic pattern. (“My life is spent alone, without wealth, without status, without love, and without hope” ([39]: p. 354.)


ploce are schemes of lexical repetition (a common alternate spelling is ploche). The term ploce refers by default to simple ploce: the occurrence of the same word or word sequence two or more times proximally. complex ploce include a number of specific figures naming occurrences of the same word or word sequence two or more times constrained by relative position, syntactic position, or sequential order.


A scheme: the repetition of words derived from the same root but with different lexicomorphology (possibly null). (“Every self-mover, as the Philosopher proves… is composed of mover and moved.” [2]: p. 152)


A synonym for prolepsis.


A term from Perelman and Olbrecht-Tyteca’s [52] framework for the salience given to aspects of an argument through linguistic style, frequently correlated with rhetorical figures.


A move, a figure in the proleptic suite: the evocation and rebuttal or refutation of anticipated counter-arguments. (“Thanks to the ubiquity of text on the Internet, not to mention the popularity of text-messaging on cell phones, we may well be reading more today than we did in the 1970s or 1980s, when television was our medium of choice. But it’s a different kind of reading, and behind it lies a different kind of thinking – perhaps even a new sense of the self. ‘We are not only what we read,’ says Maryanne Wolf, …. ‘We are how we read.’” – This passage, in an argument about how the Internet reduces literacy, anticipates the objection that the quantity of reading must be improving literacy and begins the rebuttal by shifting to the quality of reading on the Internet. Example cited in Mehlenbacher’s essay in this volume.)


See proleptic suite.

proleptic suite

A set of moves pivoting on anticipation and rebuttal of arguments, charted in detail for the first time by AR Mehlenbacher in this volume. (For examples, see procatalepsis and metastasis.)


Perelman and Olbrecht-Tyteca’s [52] term for arguments that appear demonstrative but do not have the rigour of genuinely (mathematical or logical) demonstrative arguments. They designate several figures as potentially having this characteristic, including antimetabole, commutation, oxymoron, and antanaclasis.


A synonym for anadiplosis, avoided because of possible confusion with linguistic reduplication.


A synonym for epanaphora, epanalepsis, and other forms of (complex) ploce.

rhetorical question

See erotema.


Proximal repetition word-final syllables (hoi polloi and rolly poly, above).


A type of irony (and colloquially often a synonym for irony), therefore a chroma, which demeans through expressions that distend semantic properties. (“Willis stopped in midsentence, and a smile spread across his face ‘Oh!’ he said. ‘Oh! Yeah, I get it! I ain’t supposed to know who done it. That way I’ll be tryin’ to figure it out.’ ‘Aren’t you the brilliant one, though,’ Atwood said” [32]: p. 201).


A major category of figures, schemes are deviations of conventional expectations in the usage of signantia. schemes as figures are distinguished from Argumentation Schemes, from the schemas of Rhetorical Structure Theory and from such other phenomena as the image schemas of cognitive science, but they have in common with those concepts an attention to form, structure, and arrangement – abstract patterns – as opposed to content.


(plural, signantia) The material aspect of a sign (the spelling or pronunciation of a word, for instance), also known as the form, or (in Saussurean semiotics) as the signifier.


(plural, signata) The conceptual aspect of a sign (the definition of a word, for instance), also known as the meaning or the sense or (in Saussurean semiotics) as the signified.


A trope: a cross-domain property mapping that includes an overt comparative construction. (“her eyes [were] like cornered rats” [21]: p. 132).

simple ploce

See ploce.


A scheme and a compound figure, combining epanaphora and epistrophe: The occurrence of the same word or word sequence at the beginnings of two or more proximal phrases or clauses while another word or word sequence occurs at the ends of those same phrases or clauses. (“Women in Middle East attacked for not wearing hijab. Women in the West attacked for wearing hijab” [48]).


A scheme: the omission of a phoneme or syllable in the middle of a word (“What oft was thought but ne’er [for never] so well expressed” [53]: Alexander Pope, Essay on Criticism 2)

topoi (topics)

Tools of invention and analysis related to the structure of the argument, many of them related to figures. For example, argument from analogy is related to metaphor, argument from opposites is related to antithesis, argument from association is related to metonymy. Traditionally, topoi can be common (applying to arguments in any domain) and special (applying only to arguments in particular domains).


A major category of figures, tropes are conceptual deviations, shifts away from conventional expectations in the usage of signata.


A class of schemes in which one word or word sequence ‘governs’ at least two phrases, often indicated with a null pro-form. (“Trust not only requires moral integrity, but [trust] also [requires] competence.” Example from Lawrence, Visser & Reed’s essay, in this issue.)


1 Notice that for position-sensitive figures of lexical repetition language typology is important. English is a Subject-Verb-Object word-order language, so clause-initial epanaphora will tend strongly towards stabilizing subject reference, while clause-final epistrophe will tend strongly towards stabilizing object reference. Other figures, such as simile (a cross-domain comparison) and polyptoton (repetition of lexical stems with different morphology), are more uniform in their cross-linguistic functions.

2 Other work by Fahnestock that repays the attention of argumentation theorists include her “Figures of Argument” [20] and Rhetorical Figures in Science [17]. Her project of Figural Logic is an important advance in argumentation studies (see [60]: p. 69–85, [26]).

3 See [56] for some discussion.

4 Dubremetz and Nivre ([1214]) use the term chiasmus to refer to the figure antimetabole, not an uncommon usage in the rhetorical tradition, but we seek more precision in the terminology of rhetorical figures than they (or frankly, most rhetoricians, let alone computer scientists) are interested in. In our work, we have found multiple reverse-constituent-repetition patterns, which we regard as different figures, adapting and coining a variety of labels for them, grouped in our chiastic suite. For instance, their example 18 is chiastic, since it involves reverse repetition of constituents. It is not an antimetabole, however, because those constitutents are not words, but only signantia (words are signans/signatum pairings). We categorize it with the blended neologism antanametabole, since it is a combination of antanaclasis (repetition of the ‘same’ word with different senses) and antimetabole. We have also used the term pseudo-antimetabole for this pattern, since it closely mimics an antimetabole.

5 antimetathesis is one of the many figure names that are used for very different linguistic devices. The paper by Mitrović et al. in this volume, referencing Perelman and Olbrecht-Tyteca [52] use it as a synonym for antimetabole, one of their quasi-logical figures. antimetathesis is also sometimes used for a compound figure combining antimetabole and antithesis, as in the famous Kennedy/Sorenson phrase (“Ask not…”). See also commutation.


The authors would like to thank, very gratefully, Devon Moriarty for her painstaking and elegant editorial assistance; as well as Pietro Baroni, Adam James Bradley, Sean Fischer, Michael Flor, Kenneth Graham, Floriana Grasso, Nancy Green, Douglas Guilbeault, Ken Hirschkop, Daniel Hromada, Beata Beigman Klebanov, David Larochelle, Michael Macdonald, Andrew McMurry, Laura Neuhaus, Claus Strommer, Michael Ullyot, Bart Verheij, and James Wynn for their help with this issue; and Sarah Bott, Sebastian Malton, Omar Nafees, Paulo Pacheco, Katherine Tu, Yetian Wang, and the whole RhetFig family, for constant, enervating conversations about figures, computers, and argumentation. The positions, claims, and reasons offered in this introduction were most recently presented to the Artificial Intelligence and Simulated Behaviour workshop on Cognitive Ontologies and at the Computer Models of Natural Argumentation workshop. The authors are very appreciative to the organizers and participants of those workshops for their vigorously constructive feedback.

This research was partially funded by the Social Sciences and Humanities Research Council of Canada.



A. Alexis, Fifteen Dogs, Coach House Books, Toronto, (2015) .


T. Aquinas, Summa Contra Gentiles, Vol. 2: , Burns, Oates & Washbourne, London, (1923) . Translated by the English Dominican fathers.


W. Bateson, Mendel’s Principles of Heredity: A Defence, Cambridge University Press, Cambridge, (1902) .


G. Carlin, The seven words you can never say on television [recorded by George Carlin], Class Clown [record album], Little David, New York, 1972.


L. Chien and R.A. Harris, Scheme trope chroma chengyu: Figuration in Chinese four-character idioms, Cognitive Semiotics. 6: (1) ((2010) ), 155–178. doi:10.1515/cogsem.2010.6.spring2010.155.


M.T. Cicero, The Orations, Three Volumes, J & J Harper, New York, (1833) . Translated by W. Duncan.


S.F. Costa, Concordia captain tells judge how he left stricken cruise ship, The Guardian [Internet], 2012 January 18 [cited 2017 August 2], World: [about 5 screens]. Available from:


J. Crosswhite, Rhetoric and computation, in: Symposium on Argument and Computation, 2000 June 27, Bonskeid House, Perthshire, Scotland, (2000) .


J. Crosswhite, J. Fox, C.A. Reed, T. Scaltsas and S. Stumpf, Computational models of rhetorical argument, in: Argumentation Machines: New Frontiers in Argument and Computation, C.A. Reed and T.J. Norman, eds, Kluwer Academic Publishers, Dordrecht, (2004) , pp. 175–209.


D. DeLillo, White Noise, Penguin Books, London, (1999) .


B. Dole, Great Political Wit: Laughing (Almost) All the Way to the White House, Broadway Books, New York, (2000) .


M. Dubremetz and J. Nivre, Rhetorical figure detection: The case of chiasmus, in: NAACL-HLT Fourth Workshop on Computational Linguistics for Literature, 2014 June 4, Denver, CO, USA, Curran Associates, New York, (2015) , pp. 23–31. Available from:


M. Dubremetz and J. Nivre, Syntax matters for rhetorical structure: The case of chiasmus, in: Proceedings of the Fifth Workshop on Computational Linguistics for Literature, NAACL-HLT, 2016 June 1, San Diego, California, (2016) , pp. 47–53. Available from:


M. Dubremetz and J. Nivre, Machine learning for rhetorical figure detection: More chiasmus with less annotation, in: Proceedings of the 21st Nordic Conference of Computational Linguistics, 2017 May 23–24, Gothenburg, Sweden, (2017) , pp. 37–45. Available from:


A. Dumas, Les trois mousquetaires, Dufour & Mulat, Paris, (1849) .


A. Dumas, The Three Musketeers, Simon & Schuster, New York, (2010) , Translation anonymous.


J. Fahnestock, Rhetorical Figures in Science, Oxford University Press, New York, (1999) .


J. Fahnestock, Verbal and visual parallelism, Written Communication 20: (2) ((2003) ), 123–152.


J. Fahnestock, Preserving the figure: Consistency in the presentation of scientific arguments, Written Communication 21: (1) ((2004) ), 6–31. doi:10.1177/0741088303261034.


J. Fahnestock, Figures of argument (OSSA 2005 Keynote Address), Informal Logic. 24: (2) ((2004) ), 115–135. doi:10.22329/il.v24i2.2139.


W. Faulkner, The Sound and the Fury, Indigo Books, Toronto, (2014) .


F.S. Fitzgerald, The Great Gatsby, Charles Scribener’s Sons, New York, (1925) .


J.J. Gawryjolek, Automated annotation and visualization of rhetorical figures, Master of Mathematics thesis, 2009.


J.J. Gawryjolek, R.A. Harris and C. Di Marco, An annotation tool for automatically detecting rhetorical figures, in: CMNA IX (Computational Models of Natural Argument), Held with IJCAI-09, 2009 July 13, Pasadena, CA, USA, (2009) . Available from:


F. Grasso, Towards computational rhetoric, Informal Logic. 22: (3) ((2002) ), 195–229. doi:10.22329/il.v22i3.2589.


R.A. Harris, Figural logic in Mendel’s experiments on plant hybrids, Philosophy and Rhetoric. 46: (4) ((2013) ), 570–602. doi:10.5325/philrhet.46.4.0570.


M. Higgins, Bloodthirsty, depraved and tactically brilliant: Why ISIS is the most formidable terrorist group yet, The National Post, 2014 August 16, Sect. A:16.


K. Hosseini, The Kite Runner, Riverhead Books, New York, (2003) .


D.D. Hromada, Initial experiments with multilingual extraction of rhetoric figures by means of PERL-compatible regular expressions, in: Proceedings of the Second Student Research Workshop Associated with RANLP 2011, 2011 September 13, Hissar, Bulgaria, Incoma Ltd., Shoumen, (2011) , pp. 85–90. Available from:


B. Hutchinson, BC man survives violent encounter with RCMP, The National Post, 2015 July 26, Sect. A:15, A17.


J. Java, Characterization of prose by rhetorical structure for machine learning classification, PhD thesis, College of Engineering and Computing, Nova Southeastern University, 2015.


W.M. Johnstone and J.A. Johnstone, Brutal Night of the Mountain Man, Pinnacle Books, New York, (2016) .


J.F. Kennedy and T. Sorenson, The American Presidency Project [Internet], Inagural address, 1961 January 20 [cited 2017 July 31], [about 4 screens]. Available from:


J. Kerouac, On the Road, Penguin Books, London, (1976) .


T. King, The Back of the Turtle, Harper Collins, Toronto, (2014) .


A. Kuprin, The Garnet Bracelet and Other Stories, Foreign Languages Publishing House, Moscow, (1956) . Translated by S. Apresyan.


C. Lamb, The Complete Works in Prose and Verse of Charles Lamb, R.H. Shepherd, ed., Chatto and Windus, London, (1875) .


J. Lapidos, The hottest rhetorical device of ’08, Slate [Internet], 2008, September 12 [cited 31 July 2017], Life: [about 4 screens]. Available from:


C. Liu, The Three-Body Problem, Head of Zeus, London, (2006) .


E.B. Lytton, Richelieu, or, the Conspiracy: A Play in Five Acts, Routledge, London, (1874) .


W.J. Mailbag, Sports Illustrated [Internet], 2017 August 2 [cited 2017 August 4]. Available from:


Malcom X Reference Archive [Internet], Marxist Internet Archive, The Black revolution requires bloodshed, 1963 [cited 31 July 2017], [about 1 screen]. Available from:


N. Mandela, The Long Walk to Freedom, Little Brown & Co., London, (1994) .


W.C. Mann and S.A. Thompson, Rhetorical structure theory: A theory of text organization, Technical report, USC Information Sciences Institute, 1987. Available from:


G. Mendel, Versuche über Pflanzen-Hybriden, Verhandlungen des naturforschenden Vereines in Brünn 4: ((1865) ), 3–47. doi: 10.1093/oxfordjournals.jhered.a106151.


G. Mendel, Experiments on plant hybrids, in: The Origin of Genetics: A Mendel Sourcebook, C. Stern and E.R. Sherwood, eds, WH Freeman and Company, London, (1966) [1866], pp. 1–48. Translated by E.R. Sherwood. Also available (among other sources) from


D. Mohammed and M. Lewiński (eds), Argumentation and reasoned action, in: Proceedings of the 1st European Conference on Argumentation, 2015 June 9–12, Lisbon, Portugal, College Publications, London, (2016) . Available from:


S. Omar, Women in Middle East,, 2016 December 3 6:00 PM [cited August 9 2017] [Tweet], Available from @OmarjSakr.


S. Omar, I just woke up,, 2016 December 4 3:04 PM [cited August 9 2017] [Tweet]. Available from @OmarjSakr.


C. O’Reilly and S. Paurobally, Lassoing rhetoric with OWL and SWRL, Unpublished, 2010. Available from:


C. Perelman, The New Rhetoric and the Humanities: Essays on Rhetoric and Its Applications, D. Reidel, Boston, (1979) .


C. Perelman and L. Olbrechts-Tyteca, The New Rhetoric. A Treatise on Argumentation, University of Notre Dame Press, Notre Dame, IN, (1969) . Translated by John Wilkinson and Purcell Weaver.


Poetry Foundation [Internet], [cited 2017 August 02]. Available from:


R.A. Harris, C. Di Marco, Constructing a rhetorical figuration ontology, in: Persuasive Technology and Digital Behaviour Intervention Symposium. AISB (Artificial Intelligence and Simulation of Behaviour) Convention, 2009 April 08, SSAIB, Edinburgh, Scotland, (2009) , pp. 47–52. Available from:


O. Reboul, The figure and the argument, in: From Metaphysics to Rhetoric, M. Meyer, ed., Kluwer, Dordrecht, Netherlands, (1989) , pp. 169–181. doi:10.1007/978-94-009-2593-9_12.


S. Ruan, C. Di Marco and R.A. Harris, Rhetorical figure annotation with XML, in: CMNA XVI (Computational Models of Natural Argument), 2016 July 9, New York, NY, USA. Available from:


W. Shakespeare, The complete works of William Shakespeare [Internet], MIT, 2003, [updated 2000 October 16, 2017 August 2]. Available from:


C. Strommer, Using rhetorical figures and shallow attributes as a metric of intent in text, Doctoral dissertation, Cheriton, School of Computing, University of Waterloo, 2011.


D. Thomas, Quite Early One Morning, New Directions Publishing, New York, (1968) .


C.W. Tindale, Acts of Arguing: A Rhetorical Model of Argument, State University of New York Press, Albany, NY, (2000) .


J. Trudeau, Justin Trudeau, for the record: ‘We beat fear with hope’ [transcript], Macleans [Internet], 2015 October 20 [cited 2017 July 31], Politics: [about 10 screens]. Available from:


D.J. Trump, We can’t continue to allow China to rape our country, CNN [Internet], 2016 May 2 [cited 2017 August 2], Politics: [about 4 screens]. Available from:


D.J. Trump, My son Donald did a good job, 2017,, 2017 July 12 [cited 2017 August 2] [Tweet]. Available from: @realDonaldTrump.


W.T. Vollmann, The Royal Family, Viking Penguin, New York, (2000) .


M. Wolff, Murdoch’s private game, Vanity Fair [Internet], 2007 July 23 [cited 2017 August 2], [about 15 screens]. Available from:


Yakov Smirnoff Miller Lite Commercial, 1985, [Internet], [place unknown], mycommercials, 2007 November 11 [cited 25 July 2017], Video: 30 sec. Available from:


N. Zink, The Wallkeeper, Fourth Estate, London, (2014) .