On Genre as the Primary Unit of Language (Not Only) in Law

Abstract

Taking as points of departure modern pragmatic theory and the information-theoretic view of communication offered by Levinson, this paper re-defines the notion of “genre” as a primarily top-down functioning kind of pre-existing, conventionalized package deal in construing meaning. As a consequence, this paper argues for relativizing the role of code (langue), given information in favor of pre-existing pragmatic-functional knowledge in “making meaning”. This discussion is focused on law, which is where the issue of whether and how much meaning is “in the text”, and what it means to be “in the text” is theoretically and practically paramount.

1. Scope of the Paper: Between Law, Discourse, and Information Theory

How can we know so much, given that we know so little?

How can we know so little, given that we know so much?

This paper tries to deal with theoretical concerns raised by the near-ubiquitous notion of “genre”, with a special, but not exclusive, interest in its application to law.1 Being a functionally, language-externally and socially motivated phenomenon, discussions of the concept tend to be triggered by language—external developments, such as the rise of new media (Giltrow and Stein 2009; Anesa and Engberg 2023), generating new demands on knowledge management in new domains. Law, although intimately related to language, is such an external societal, cultural phenomenon, organized in genres. Law is an institutional and societal context characterized, if not defined, by a functionally coherent, but linguistically diverse landscape of genres. As a theoretically oriented paper, the concern here is not with a practical, functional analysis of specific genres, but it may have bearing on how and why such analyses are conducted.

The second source of theoretical interest in the concept of genre is found in Levinson’s recent paper on the “dark matter” of pragmatics (Levinson 2024). Levinson argues that communication—especially human, linguistic communication—needed to develop strategies for coping with informational “bottlenecks”, in particular, ones that enforce affordable strategies of redundancy to enable communication to take place in a narrow tunnel of capacity and time. I will argue that the concept of genre is one such strategy, evolutionarily evolved, and which is part of what Levinson would call the “known dark matter” in communication (On the other side of “known dark matter” is the “unknown dark matter” of pragmatics: the vast array of strategies we know must exist, but which are yet completely unknown). Among these strategies, the signal carried by linguistic forms has been overvalued, while the role of other, non-verbal information has gone largely unappreciated.

Law is an area that tends to be dominated by a “folk” ideology of language that sees the linguistic surface information as the main source of signal information. So, the third major theoretical angle of approach in this paper is through the ongoing debates about “meaning making” in modern, essentially neo-Gricean pragmatics, with its conception of an extraction or construction of meaning in a staged logical order that essentially involves the adduction of non-linguistically given knowledge at several stages. It is this non-linguistic knowledge that is the focus of this paper. It necessarily forms an important issue in law, due to the assumption, prevalent in law, that the content of the law resides within texts. But it must be asked: what is the “text”?

2. Language Occurs in Texts

I claim that the primary unit of the occurrence, not of “language”, but of communication involving language, is genre. Any individual, bounded piece of discourse gains its identity as an instantiation of a genre. Genre is a pivotal concept in the linguistic analysis of communication in the domain of law, which is linguistically organized as a complex, interrelated web of genres. The activities within law make continual reference to types of genres in the law, while classifications into genres are a continual topic in the discipline and in discourse at large (Kurzon 1997; Engberg 2013; Taboada 2004). It is therefore at least appropriate, if not obligatory, to look at the theoretical underpinning of the concept of genre from the vantage point of pragmatics.

While it is already well established in practical research, thanks to work by Swales and Bhatia (cf. the survey in Artemeva and Freedman 2016), I will here try to point to more theoretical issues, arguably those with the most salient practical consequences. One battle cry of the earliest “textlinguistic” turn in much of European continental linguistics was that language occurs in texts. This was a consequence of structuralism’s undue privileging of the sentence as the base unit of analysis, itself preceded by a focus on the morpheme. There were good reasons, then, for upscaling the explanatory domain of linguistics to a higher unit, because the grammaticality or otherwise of a number of phenomena could not be explained sentence-internally.

There are two competing or complementary concepts of a “text”: Text type, first, and second, genre. Each represents a different perspective on the constitutive object of research, depending on the point of departure. Traditionally, research started from what “occurs” on the “surface” of language. The classic analysis is Biber’s famous study (Biber 1995, “Dimensions of Register Variation”). Following a long tradition that explains or “accounts for” surface distributions by classification into registers, where a “functional” explanation of differential surface distribution resorts to a “notional” concept of genres. Biber’s perspective on the text “genre” is therefore “interpretive”. Genre is called upon as an explanatory dimension for the formal statistical distribution of surface forms. For our purposes, let us refer to the two basic perspectives as “form”-based and function”-based. The former can be seen as bottom-up, and the latter, function-based, as top-down. The former corresponds to “text type” while the latter refers to “genre”. In this view, genre is located at the top, and the text (text type) closer to, or even at the bottom, as the linguistic input level.

It is intuitively apparent from the above that the higher on the vertical axis the linguistic unit, the less “autonomous” the unit is: the sources of knowledge tend to come less from language, morphemes, form, or “code” (Lyons’ “system sentence”, Lyons 1981, p. 196) the “higher” you go in this concept. One issue that has seemed clear for higher units has been a much debated issue lower down, on the level of the sentence and lower: the issue of underdeterminacy. It has always been easier to accept that not only are lower to higher meanings increasingly determined by extra-linguistic information, but also that there are qualitatively different types of meanings ascribed to the global text level that are not explicable as additions of meanings that are, to a stronger extent, componentially generated lower down. So this “basic” type of decrease in autonomy and the later-to-be discussed notions of underdeterminacy are just two sides of the same problem.

What is the relationship between the two perspectives? Where does the researcher start the “explanatory” journey? Top or bottom? Notice that by “direction of explaining” we refer to the meta-decision of where to logically start.

Is there a one–one relationship between the two unit types (what I will later call the insulationist view, § 8), as often postulated for the lower end of the scale, such that a linguistic form would correspond to a higher-level unit? In particular, my question is: can genres be defined linguistically?

A famous dictum of pragmatics is that “form”, i.e., text type, “underdetermines” function. Given this assumption, it is not possible to define genres with reference to the occurrence of surface linguistic features. Or, to draw on Giltrow’s inverse formulation of the same: it is not possible to define genre by surface linguistic features alone.

If evidence were needed, we could point to Giltrow’s paradigmatic analysis of a 1760 “Treaty of Peace and Friendship” between Canadian First Nations and European settlers, where the linguistic material in the text “underdetermines” the genre. In Giltrow’s example, the treaty is just a writ, or indeed a contract, from which certain real obligations arise, with concrete political and financial consequences. As the graph postulates, “Form alone” does not determine genre, bottom-up, on the level of the whole discourse.

Viewed top-down, it is well known that a genre like a contract has so many sur—“faces”, that is to say many linguistic form possibilities, so many that there is no single way to define it micro-linguistically. Makmillen (2007) offers another example of how, like a chameleon, the same “text” can be read as instantiating several different genres, with drastic political and financial consequences.

3. Genre and Indeterminacy

The concept of genre has a history characterized by an early emphasis on didactic issues—that is, identifying genres so as to make them teachable—and on surface linguistic features, with the concept in close competition with the notion of “register”. While persisting in applied pursuits, such as the teaching of writing, the more modern theoretical approach to genre shifts focus to “…a level of genre that represents those [communicative, DS] events which have been culturally recognized” (Taboada 2004, p. 19). An indication of the abstract knowledge nature of the concept of “genre” is also found in Giltrow (2016, p. 220), which relativizes the traditional emphasis of research in the area of genre on linguistic form and which points to the same information-theoretic argument as made by Levinson (2024), which shifts the emphasis to a massive, but as yet underemphasized, share of ineluctable non-linguistic information. The underdetermination of meaning by form is sometimes an efficiency, sometimes a liability, and sometimes a resource.

More broadly, the new genre theory can make a contribution to the work on the pragmatic problem of underdetermination. The gist is the following: spheres of activity constitute common ground, narrowing context to a common ground of accessible assumptions from which interlocutors can, and do, draw inferences. But the contribution comes at a cost for genre theory; for now, the main term to be reckoned with in genre theory is mutual consciousness, in all the silent trepidations detected by pragmatics: assumptions unascertainable, uneven, unstable and changeable, even when form remains unchanged (Giltrow 2016, p. 220).

A question, of course, is what the definitional features of a genre are, if not form. Giltrow makes several key points on this question. An acutely felt “cost” for genre theory is the loss, absence, or radical downgrade in the status, of purportedly “objective” surface information. An interactive view of text and genre finds little, if anything, in the morphemes on the “surface” of a text. It thus pits itself against the “insulationist” view, that genre knowledge is autonomous, extractable from code.

It is this fact—the very indirect or minimal access to genre from surface features—that causes such embarrassment and bewilderment, especially for a linguistic perspective that is used to starting from the bottom up. The tendency towards minimalism is understandable, as surface features seem to provide something that is observable, objective, and therefore more intersubjectively recognizable. As such, minimalist approaches are more congenial to legal endeavors. More countable, they are attractive for computational approaches and for lawyers in the same way. If minimalism means the provenience of only minimal knowledge that comes from contextual, language-external sources, maximalism implies the opposite.

A maximalist view of genre entails a cognitive view, as this external knowledge—the common ground—must be assumed to be present somehow in our cognition. Here, we will focus on the top-down perspective, starting with entities that are not directly observable, entities of a more cognitive nature, manifested at the linguistic surface only very partially and often only in traces. Yet, these cognitive entities are the prime explanatory dimension for forms on the linguistic surface.

There seems to be something of a complaint tradition, regretting the fact of underdetermination. While there are cases of underdeterminacies (Witczak-Plisiecka 2009) that create genuine problems for legal interpretation, the general philosophical complaint about the lack of one-to-one correspondence between form and meaning is a type of language ideology that bedevils many applications of language. Certainly, many current folk–linguistic debates and even applied linguistic pursuits, including forensic linguistics, specifically in the shape of the so-called “referential ideology“ (Ainsworth 2008), are still hamstrung by their fostering of this false dilemma. Just as it is a “design feature” of language to have few units, a multitude of realizations and expressive means, on the lower levels, so we should expect the same on the level of genres: never a costly one–one relationship between form and function.

The central background fact—to make an information-theoretic observation—is that human speech encoding is relatively very slow; the actual process of phonetic articulation is a bottleneck in a system that can otherwise run much faster. The speaker is trying to find an economical means of invoking specific ideas in the hearer, knowing that the hearer has exactly this expectation. Now, the solution to the bottleneck is just this: let not only the content, but also the metalinguistic properties of the utterance (its form) carry the message. Or, in other words, find a way to piggyback meaning on top of meaning (cf. Levinson 2000, p. 6).

And emphasizing the necessity to find a redundancy-creating way to minimize the infinite potentialities of extra-linguistic knowledge: “Because these inferences rule out a number of possible states of affairs they multiply the informational load of what has been said by a significant factor” (Levinson 2000, p. 31f).

Genre, then, is much “richer” than the text. Much more content “functions” in a text as discourse and operates on what is there in terms of the text sentence or “what is said” than what is represented at the surface:

• The highly selective filtering or selection of context or knowledge from all available knowledge: only a fraction of what could be selected is actually—and legitimately—selected to be mutually present in the cognition of communication partners.
• What there is in terms of surface expressions: the surface expressions themselves have a function within that present knowledge. They are intended by the writer or speaker to produce effects in a cognitively conceived discourse world on the receiver side. It is therefore clear that the content side is a cognitive entity, a view on which I will elaborate on in what follows. To a large extent, genre is internalized context, and context is knowledge, as it is represented in our cognition and is assumed by the speaker to be exactly there. However, “knowledge” is potentially everything that could be represented. Restricting all possible knowledge to a relevant and cognitively manageable, efficient amount is the salient function of genre with respect to the information-theoretical issue emphasized by Levinson (2024).

The issue here is that interlocutors mutually assume a tiny fraction of all possible knowledge—“mutual knowledge” by first approximation—to be present. Only the restricted and evoked relative knowledge is present and represents the relevant and legitimate “cognitive environment”, where assumed legitimacy is defined by mutual assumptions about exactly which type of genre game is being played, and consequently which knowledge is legitimately assumed to be present at any given point in the interaction. It is an essential component of genre that this element of “legitimacy” of mutual knowledge expectation is present. It is understood that miscalculations about mutual knowledge tend to have a worse effect on communication than incorrect grammar, and therefore, unlike grammatical mistakes, the preference is always for the immediate repair of mistakes about mutual knowledge. Mutual knowledge is linguistically reflected in linguistic forms that indicate, e.g., definiteness and specificity in the referential-pragmatic sense, as well as in the more general sense of the choice of more or less specific expressions with more or less intentional content. Importantly, that the receiver access the right kind of knowledge is just as much part of the speaker intention as the standard Gricean view of the intentionality of speaker meanings as that which is to be recovered. On a more general level, the co-construction of meaning is now generally recognized (Elder and Jaszczolt 2024, cf. Kuhlen and Rahman 2023 for psycholinguistic aspects).

Two questions now arise that I consider pivotal for my approach to genre. First of all: How can we know so little, given that we know so much? This question takes up the problem of “efficiency” mentioned by Giltrow: reducing the vast possibilities of contextual knowledge, in a manner that enables efficacious comprehension. The complementary version of the question would be the following: How can we know so much, given that we know so little? How is it that we perform such complex knowledge changes and reorderings in our cognitive world, given that these cognitive operations are triggered by such scant linguistic instructions? In terms of Relevance Theory, how can a faint linguistic signal cause in the hearer such massive cognitive effects? From the scantiest linguistic signals, interlocutors strengthen, revise, or abandon available assumptions, even though the linguistically given information is little more than a semi- or un-consciously remembered history of previous communicative encounters. What we have conventionally termed “linguistic meaning” is itself highly constrained by genre.

The issue of the autonomy of a text raises exactly this issue, begging the question: to what extent are these instructions for cognitive re-arrangements (“cognitive effects”) due to the “linguistic meaning” of expressions and/or the knowledge types and operation processes on which they operate?

The vulgo version, which keeps appearing in discussions of legal interpretation, of the autonomy issue is: To what extent is the knowledge contained either in the text or in the context? It is clear that the restricting and modulating information comes not only from the cognitive environment but also from the co-occurring referential expressions in the text (long known in literary studies, what literary scholar I.A. Richards colorfully called the “interanimation of words” (Richards [1937] 1964, p. 47)—or, in legal terms: noscitur a sociis):

According to the insulationist account the meaning of any one word that occurs in a particular sentence is insulated against interference from the meaning of any other word in the same sentence […] Interactionism makes the contradictory assertion: in some sentences in some languages the meaning of a word in a sentence may be determined in part by the word’s verbal context in that sentence.

(Recanati 2003, p. 132, citing Cohen)

This co-textual aspect (co-text turned into cognitively present context during the further linear progress of comprehension) is emphasized in its specific shape for legal documents by Skoczen (2016). But the concrete meaning of legal documents is, in turn, itself top-down influenced or massively co-determined by genre and the schemata and knowledge frames called up through it.

Cognitively oriented genre theory therefore holds that genres are bounded, prepackaged types of knowledge, including, but not exhaustively defining linguistic surface expressions, which we can initially characterize here as propositional, componential language knowledge or “linguistic meaning” or “code” knowledge.

It would appear, then, that the highest level of pre-packaged context and comprehension/constructing level is genre. Genre is characterized by two properties.

• Genres are “a vast association network between utterance forms, contexts and actions” (Levinson 2024, p. 18). Here, “actions”, in terms of speech act theory, refers to illocution types. “Utterance forms” are linguistic surface forms, elements of code. The elements of this interpretational process accessed in a way defined by genre are the following:
  “There are various sources of information that contribute to the main meaning conveyed by the speaker and recovered by the addressee. In the DS [default semantics, D.S.] model, we identify five main sources: word meaning and sentence structure (WS), world knowledge (WK), situation of discourse (SD), and two kinds of default information: stereotypes and presumptions about society and culture (SC) and properties of human inferential system (IS). The latter, for example, account for the fact that the strongest, most informative interpretation is the preferred one, such as the referential rather than attributive reading of definite descriptions, de re rather than de dicto reading of propositional attitude reports, or the anaphoric rather than presupposing reading of referential partial matches” (Jaszczolt 2011, p. 21).
• The meaning total (meaning NN) of a concrete text instance of a genre (a contextually fully interpreted instance of a “text-sentence”) is what Jaszczolt (2011) terms the “primary meaning” or the “default meaning”. It is arrived at instantaneously and effortlessly; Levinson indeed stresses the instantaneous speed at which “participants actually ascribe actions to utterances” (cf. Levinson 2024, p. 16f). Although Levinson’s focus is primarily on spoken conversation, his description of the situation holds for processing texts as members of genres. For spoken discourse, “speech act recognition can occur very early during the processing of an incoming turn—even within the first syllable or two” (Levinson 2024, p. 17). Even though, as in written communication, there are cases when communicants take time to reflect, in spoken situations, there is simply no time to deliberately figure out what action is intended.
  Even so, the cardinal phenomenon persists that utterances are instantaneously processed. It would seem, then, that the choice of “actions” is either extremely limited at any point of processing, or indeed pre-determined, so that no further decisions and only “very intensive cognitive reasoning” (Levinson 2024, p. 18) are required. Integral parts of genres are, in the van Dijk-ian sense, their “superstructures”: types and conventionalized functional parts, often in a conventional sequenced order. Classic examples might include the hero’s narrative and the argumentative essay. The existence of genre is therefore an essential part of the “efficiency” aspect of processing language.
  In Kahneman’s (2012, p. 20) neuro-psychological metaphor, genre is part of our “System 1”. Kahneman writes the following: “System 1 operates automatically and quickly, with little or no effort and no sense of voluntary control”. It is a fully automated, fast way of thinking that, in its default mode of operation, does not require special mental effort. This way of thinking is against “System 2”, which takes up significant mental space by requiring special and conscious processing effort. As Kahneman (2012, p. 21) again writes, “System 2 allocates attention to the effortful mental activities that demand it, including complex computations. […] The automatic operations of System 1 generate surprisingly complex patterns of ideas, but only the slower System 2 can construct thoughts in an orderly series of steps”.

4. Genre as Abstract Entities

It has been argued, in culturalist, Foucaultian discussions of language and law, that there is an even higher abstract level of genre knowledge. Such discussions treat cultural knowledge as largely packaged in terms of disciplines of knowledge, as organized in societal domains, such as sciences, religion, literature, and law—in their own non-technical sense, as genre.

Legal scholars are well aware that their field is, or can be viewed as, a genre. (FN 1) Law, though it deals with “society” and human experience, involves distinct forms (particular legal texts, specific practices, procedures, language, categories, etc.) that distinguish its manner of representing and processing such experience. The idea that law, as genre, is somehow distinct from, yet related to its surroundings has often been channeled through theoretical claims about the relations between (sometimes only vaguely defined) “law” and “society”. These analyses have preoccupied scholars for a long time now, and when they involve causal claims, they are debates about the level of independence of genres from extra-linguistic determination.

Rosenberg refers to the fundamental legal–internal controversy over whether law is externally conditioned (culturally/sociologically) or is a matter of law-internal rules and dogmatics. While we cannot pursue the question here let us note that the issue—whether genre is a self-contained system or one that is determined externally—is relevant to the question of text-type vs. genre. Can or should we consider texts products of system-internal rules or as genre, with massive external determination and interaction?

Not only is the dichotomy repeated within the domain of law, but the fact that legal genres are—like any genre—determined externally also appears to speak to a non-positivistic stance.

Abstraction away from the level of texts is taken one step further by Rosenberg, in a historical analysis of advertising:

Law, understood as a dynamic part of cultural negotiation rather than a predefined profession, discipline or institutional setting, became implicated in directing the system of advertising in the different media, defining approaches for analysing and understanding it, and determining its boundaries. Ultimately, by performing boundary work law shaped the status of advertising, which this book shows to have emerged as deeply conflicted: legal means constituted advertising as a legitimate and indeed indispensable system of modern life, but also as a disparaged, ridiculed and criticized one, considered suspect in epistemological and aesthetic terms.

(Rosenberg 2022, p. 10)

In Rosenberg’s approach, the very domain of law as such is, at its highest level of existence, elevated to an abstract context of process and procedurality, far away from any level of componentiality. But law is, as with all elements of what can define a genre, determinative of top-down processes of meaning building. It is also, in Levinson’s conceptual grid, part of the known, not even of the known unknown.

Abstraction of genre from surface forms makes any instantiation of genre much more than the sum of its surface forms. In turn, part of a discourse or discourse types receive their function.

A narrative receives its function, and the rules of the game for handling and interpreting it, from a larger genre configuration (like law) in which it is embedded. The raw, functionally uninterpreted narrative is an ineffectual concept. A literary environment and a judicial environment—or any other number of environments, for that matter—would each make different things out of the same surface narrative. A similar case are acts that only together make sense as a mobbing episode (Stein 2022), both for the analyst as well as for the awareness of the victim, who as a rule do not interpret constituent acts as part of a larger negative communicative scheme.

It is, in addition, a further instantiation of the phenomenon that genres can be organized in hierarchical form, such as is obviously the case in phenomena like harassment, where a partial process of negative actions receive their super-summative function through being part of a larger abstract unit “harassment” (Guillén-Nieto 2024).

We will perhaps cease to refer to this effect as a problem if we take into account what Bhatia has repeatedly referred to as the way the genres in a domain like law (or genre in Rosenberg’s sense) are intertextually and interdiscursively appropriated (Bhatia 2023, p. 159).

I will argue, therefore, that the reductional redundancies of knowledge as in System no. 1 are definitional for genre. The question remains: how do we reduce the globally possible knowledge types to a level that is cognitively manageable in terms of mental/psycho-ling processing capacities? Such reduction occurs on all hierarchical levels of discourse, on the level at which knowledge is called upon, as much as on the constraining of which knowledge to invoke at the level of the domain, or—in the Rosenbergian sense—at the level of genre. It would seem we must assume that these knowledge types are variously present and salient to different degrees, as manifested by their surface treatment as definite or not. But it is exactly this mechanism of moving knowledge into local saliency, and only that and no other knowledge or entities, that is the achievement of genre. So a genre is an n-tuple of very different types of knowledge involving selected types of games and actions with cultural and social knowledge (whether that is, for example, conversation or legal statute) as well as processing strategies. Linguistic expressions, including their meaning and reference potentials, as with specific linguistic meanings, receive their specific functions through top-down processes, even at the earliest stages of proposition creation. Even the first “pre-selections”, on the way from system to text sentence, are pragmatically determined and thereby determined genre-wise. Beyond the now well-researched micro-linguistic surface forms, so often studied in connection with Register studies, the following is a selection of—in the widest sense—processing conventions or pragmatic conventions associated with different genres, such as the following: rules of interpretation, presumptions about how to process, discourse rules and conversational rules. It is a specific sub-selection of these knowledge types that defines a genre and constrains our processes of meaning making.

Taking a statute as an illustrative example, part of the package deal of statute as a genre are interpretive principles like the following:

• literalist presumption
• parole evidence rule
• surplusage rule
• four corners rule
• legal canons of interpretation

Walton et al. (2021) described the types of knowledge that are presumed to be called on in resolving a concrete legal case and which provide arguments in the legal case:

• Structural presumption (a presumption related to the “structure” of a legal text or system) (Argument of economy): “Legal texts normally have no redundant expressions”.
• Linguistic, specific presumption (Systems argument): “By ‘control,’ the statute means control in the sense that the tenant has permitted access to the premises”.
• Linguistic, generic presumption (Natural meaning argument): “Implicit in the terms “household member” or “guest” is that access to the premises has been granted by the tenant”.
• Structural presumption (Historical argument): “In Memphis Housing Authority v. Thompson, the Tennessee Supreme Court noted that the statute, and the lease provisions that were derived from the statute, refer to four separate categories of people: (1) the resident; (2) household members; and (3) guests or (4) other persons under the resident’s control”.
• Pragmatic presumption (psychological argument): “In an effort to end what Congress termed the “reign of terror” imposed by drug dealers on public housing tenants, 42 U.S.C. 11901 (1994 and Supp. IV 1998), Congress enacted Section 1437d (l) (6) in 1988 and strengthened it in 1990 and 1996. A different interpretation deprives public housing authorities of an important tool to achieve safe and livable public housing, and to deprive public housing tenants of protection that Congress found to be of central importance for their security and well-being”.

5. Legal and Linguistic Knowledge as Arguments in a Legal Case

As can be seen, what is intentionally called a legitimate “contextual” genre knowledge is an array of linguistic, pragmatic and legal knowledge far vaster than micro-linguistic surface features and linguistic markers typical of statute or legal language. As befits a cognitive interpretation of legal communication and interpretation, we must stress the genre-induced rules of interpretation as a genre-determining type of intended cognitive environment. Genre is not determined by specific selections of surface language materials.

I offer two illustrations of how global interpretive principles operate here, to demonstrate how interpretive strategies are “switched” on (long) before we approach an utterance. Consider, as our first example, prose fiction, with its narrated details. In this case, there is an underlying assumption that the factors selecting, from all the globally possible details, only a tiny fraction as worthy to be presented as stimulus for cognitive belaboring are intentionally selected by the author and will be instrumental in the progress of the story (Chafe 1992, especially “fine-grained resolution” pp. 237–39). This presumption is one of the uppermost determinants of comprehension and interpretation of prose fiction.

A second example can be drawn from Skoczen’s (2021) view of the implicatures in statutes. Skoczen (2021) argues that Grice’s cooperative principles are insufficient to make explicit what happens in legal interpretation. She postulates an additional “fixed strategic normative framework” (Skoczen 2021, p. 4), including a dominant “strategic maxim”, superimposed on the still functional Gricean maxims, as determinative of legal interpretation. So, the domesticating powers of genre do include micro-linguistic means, but they can act only in the strong company of all other kinds of more abstract genre determinants, ranging from the “strategic” maxims to the notion of “intention”. Skoczen’s (2021) own example is from a case of harassment, where she argues these principles are operative alongside the problem of uptake in this genre, as well as the register factors determining options in choice of expressions.

After all, there is already consensus that the topmost determinant of genre is “intention” or “purpose”, be it in the minds of one person or of several people—a mens rea—or a more abstract societal purpose. After all, there can be no purpose unless it is intended by individuals, whether concretely or abstractly. The individual may not even be consciously aware of having a purpose. Individuals may, for example, gang up and engage in mobbing actions, resulting in “mobbing” as an incipient genre type, defined by the purpose of jointly denigration. Or “society” (a composite of minds of individuals) may evolve such a purpose. Over time, society might package a purpose-defined configuration of language-system resources, pragmatic comprehension and interpretation conventions into a complexity-reducing, automated communication package that defines genre-relative normality expectations, such as those reflected in Horn’s I and Q baseline principles that are part of Kahneman’s System 1.

The dualism between the linguistic surface and the non-surface pragmatic strategies is explicitly recognized by Palermo and Visconti (2023, p. 250), who argue that genres “are governed by diachronically stable and generally universal principles, as they relate to general cognitive and communicative features”. For Palermo and Visconti, these features manifest in the shape of cohesion and coherence-based elements of genre definition, while the approach in the present paper gives clear logical and functional precedence to the coherence, that is to say, to the pragmatic side. My privileging of the pragmatic side seems also warranted by the inclusion of a historical perspective on the phenomenon of genre. From history, we can see that genres come and go; sometimes their purposes become opaque in retrospect. On the other side, genres may be in statu nascendi and on the way to “sedimentation” (Palermo and Visconti 2023, p. 249), a process that welds the linguistic and pragmatic elements into an unanalyzed whole to function as a composite element of System 1.

The historicity of genre formation is strikingly demonstrated by Rosenberg, through the application of the relativity of genre formation (“advertising-as-genre”) with respect to historically arising purposes on an abstract societal level. In an analysis of the history of advertising from the 19th century, Rosenberg writes the following:

As a genre, made of texts and images that dialectically forge and are forged by changing media and audiences, advertising is not an obvious object but a historical problem and ongoing effort of creation, in which law participates.

From the information-theoretic point of view, what is essential is that it is only the whole package that is called into service, not discrete elements. Only then is the package capable of solving the informational bottleneck problem, from both the production and the comprehension side in communication. The individual modularized element contributes little towards informational capacity and efficiency. Nor do isolated elements have explanatory power in a modular analytical investigation, except when seen through their relative function within the larger genre whole.

6. Making Meaning: Bottom Up

Having emphasized, in the preceding section, the top-down character and effect of contextual knowledge that is “switched on” by entering the communicative game of a specific genre, it is now time to include the bottom-up perspective in construing meaning. After all, the default mechanism of comprehending involves the sequential processing of sentences as they are transformed from physical inscriptions to sequences of morphemes and constituents. Modern pragmatics has conceptualized the way meaning is built up from, and triggered by, incoming sentences in terms of staggered processes of “meaning making”. How this process is conceptualized is of particular relevance for the societal domain of law, which has the construction of meaning on the basis of texts as a constitutive linguistic basis of its activity. For this reason, some aspects of the following discussion pay special attention to issues of legal interpretation, especially those relevant to genre-based top-down processes. The discussion will touch on a number of theory-internal issues, for which we can point to a rich body of literature. The interested reader is invited to consult the survey of pragmatic linguistic theory offered by Jaszczolt (2023).

The initial template of our discussion will assume, at the beginning of meaning extraction, the availability of a “text sentence”, or a compositionally available raw proposition, composed of language-internal morphemic, lexical information chunks, in the sense of Lyons’ (1981, p. 196) “system sentence”, but which is made more explicit by the addition of anaphorically given information. There would then be a first level of meaning extraction without explicatures. Not even Recanati’s primary enrichments as they are triggered by the componentially available “text sentence” would be needed.

Saturation is a primary pragmatic process. If the uttered sentence is ‘She is smaller than John’s sister’, then in order to work out what is said, I must (at least) determine to whom the speaker refers by the pronoun ‘she’ and what the relevant relation is between John and the mentioned sister. Were saturation a secondary pragmatic process, I would have to proceed in reverse order, i.e., to identify what is said in order to determine those things. Beside saturation, which is linguistically mandated (bottom-up), there are, I claim, other primary pragmatic processes that are optional and context-driven (top-down). The paradigm case is free enrichment, illustrated by example (1):

(1) Mary took out her key and opened the door.

In virtue of a ‘bridging inference’, we naturally understand the second conjunct as meaning that Mary opened the door with the key mentioned in the first conjunct; yet this event is not explicitly articulated in the sentence. Insofar as the bridging inference affects the intuitive truth conditions of the utterance, it does so as a result of free enrichment. In typical cases, free enrichment consists of making the interpretation of some expression in the sentence contextually more specific.

(Recanati 2003, p. 23f)

Primary processes like saturation (linguistically mandated) and free enrichment (pragmatically mandated) are essentially automatic processes, operating below conscious awareness. They are, therefore, arguably part of the speaker’s System 1, in Kahneman’s terminology. Primary processes are not defeasible. These first analytic, logical steps in meaning making result in a layer of meaning that is falsifiable, variously referred to as the proposition “what is said” and “explicature”.

Such primary stages of meaning extraction already require access to pragmatic, language-external knowledge. Only the initial construct of the “system-sentence” is, by definition, a completely a-contextual and a-pragmatic, purely logical-semantic level of analysis. The recognition of this initial level of knowledge identification would, as a logically conceived process, function as the first step in triggering the next steps. After all, if there is not a linguistic form awaiting saturation, the later processes cannot logically happen at all.

What is important, at this point, is the inclusion of selected elements of pragmatic external knowledge from the very first stages of the meaning-building process in the shape of the first inferential processes.

The proposition—in the above-described sense—is then what triggers, in an extremely context-dependent way, further implicatures of the kind variously called “particularized” or “conversational”. Their hallmark is that they are in this sense “postpropositional”, that they are defeasible, and that they arise on the basis of the proposition or the explicature.

To give an example, taken from Carston and Hall (2012, p. 47f):

(1)	Max: How was the party? Did it go well?
(2)	Amy: There wasn’t enough drink and everyone left early.

Amy’s answer as in (2) will be freely (secondarily) enriched to yield the following explicature (3), which we will at this point accept as the “proposition” or “what is said”:

(3)	THERE WASN’T ENOUGH ALCOHOLIC DRINK TO SATISFY THE PEOPLE AT THE PARTY AND SO EVERYONE WHO CAME TO THE PARTY LEFT IT EARLY.

It is only through the explicature as in (3) that the ultimate intended meaning will be arrived at:

(4)	THE PARTY DID NOT GO WELL.

Amy’s response triggers, as a post-propositional process, the conversational implicature in (4). (4) is the ultimate pragmatic meaning of Amy’s answer, implicated to be so construed by the speaker, and only vestigially related to the content of the proposition, or even to the system sentence.

By way of another example, taken from the rich literature on the subject, consider the following utterance in the context of the meeting of a selection committee:

• Text sentence: “She has a brain”.
  • SHE HAS A HIGHLY FUNCTIONING BRAIN (explicature and proposition, what is said)
  • SHE IS THE IDEAL CANDIDATE FOR THIS POSITION (implicature)

In both examples, the (correct) implicature is the purpose of the utterance. It must therefore be included in any realistic explanatory account of communication. The totality of these meanings goes, since Grice, by the name of “meaning nn”.

Meaning construction departs from this level of “What is said”, finally going on to a level of “Meaning nn” that subsumes all of the meaning effects (presumed to be) intended by speaker and (legitimately) construed by the hearer. From an early stage, it develops towards a final public putting-on-record that the applicant in question is the ideal candidate for the job.

What should be clear by now is that there is a long stretch of inferential distance from the code input (system sentence) to the final implicature. The length of the distance presents a challenge to any account of meaning extraction in a domain like law, which operates on its self-confidence in being able to make explicit how utterance meanings (meaning nn) are arrived at, along a stretch of linguistic communication from forensic linguistics via police work to subsumption and legal interpretation.

It should be said that the exposition given in the above section is embedded in the history of the “border war” issue between semantics and pragmatics. As it happens, ever more information in the meaning-making process was “taken over” by pragmatics, with an endpoint in the shape of “contextualism” (Recanati 2003). For contextualism, a maximum amount of information comes from text-external pragmatic knowledge, so that the emphasis is on pragmatics. To stress “emphasis” does not imply that semantics are overlooked, nor to belittle the contribution of semantics to interpretation, but to relativize it—and yet nonetheless to emphasize that semantic understanding is arrived at by pragmatic means. It is likely that none of the statements in the foregoing would be fully subscribed to by any of the theory protagonists named, which is usually the case when applying categories of theory to an applied field such as law.

Apart from the theory-internal reasons to make pragmatics the first point of departure in making explicit the construal of meanings on all levels, it would appear that there is significant added value in applying pragmatic categories of analysis to the practical conduct of law. It is very much to the credit of the ground-breaking work of Smolka and Pirker (2016, 2018) to have shown how knowledge of the issues in this field can inform the processes of legal interpretation and help shed unhelpful ideas and ideologies of language still rampant in the legal professions.

7. What Is in the Text?

A standard issue in legal theory is the question of to what degree the meaning of legal text resides or should reside “in the text”, and to what extent meaning involves accessing contextual knowledge. The branches of this discussion go under a variety of names, “textualism”, “intentionalism”, “originalism”, and others. It is natural that a theory of meaning origination that has at its center the relative shares of knowledge from language surface versus from pragmatic context should make this issue a subject of inquiry. It is important, for reasons of its own efficacy, that the field of law defines precise points in the meaning origination process at which extralinguistic knowledge can, or even has to be, legitimately accessed. For pragmatics, it is axiomatic that not all meaning comes from “the text” and certainly not from the level of the “system sentence” where linguistic meaning is self-sufficient, pure, and pragmatically undefiled.

The phenomenon of underdetermination has been amply discussed with respect to how it operates at the “bottom level”, i.e., where specific linguistic and syntactic structure undergo multiple interpretability and give rise to persistent types of vagueness (Witczak-Plisiecka 2009).

The notion of “underdeterminacy” has something of an internal history in modern Gricean, neo-Gricean, and post-Gricean theoretical discourse (Jaszczolt 2023), which does not concern us here. We will simply accept that there is no level or representation or meaning derivation that is not affected by pragmatic knowledge. Pragmatic knowledge is always present in non-linguistic knowledge mutually assumed to be present, be it through indexicalization, reference, or modulation. The exact contribution from each domain—or the location of the borderline between semantics and pragmatics—does not concern us here, beyond the statement that knowledge from both sources is available at every point, including knowledge accrued during the processing of a text.

Linguistic indeterminacy is a classic issue for language and the law. For the purposes of law, indeterminacy is always seen as pathological, a recurrent point of crisis for legal professionals. This conception—and practical experience of—indeterminacy is part and parcel of language ideology in law. Of course, lawyers’ linguistic heaven would be where the “text” is fully self-sufficient, its meaning a fully autonomous effect of surface forms. No grueling interpretive efforts would be required. A lawyer’s heaven would be imagined as the following:

a “paradise “where all words have a fixed, precisely ascertained meaning; where men may express their purposes, not only with accuracy, but with fullness; and where, if the writer has been careful, a lawyer, having a document referred to him, may sit in his chair, inspect the text, and answer all questions without raising his eyes.

(James Bradley Thayer, cited in Tiersma (2010, p. 29))

And the law would be as follows:

a code at once so flexible and so minute, as to create for every conceivable situation the just and fitting rule.

(Solan and Tiersma 2012, p. 88)

And this would be computational linguists’ heaven, too. On the lowest level, there is a “language”, with its inherently indeterminate features, such as—inherent dangling participles, scope ambiguities, and the like. Such cases are amply discussed in the linguistically oriented literature on legal interpretation, where an entire case may hinge on what is called the “ambiguity” of, say, the connectives “and” and “or”. The fact of inherent indeterminacy has long been recognized, and the issue is a key part of any competent introduction to the linguistics of statutes.

If we want to interpret a statute—or any text—the first question is the following: what is it that we “have”? The very notion that we ”have” something in the surface linguistic forms and features of a text is already problematic enough. The title of a famous article on literary interpretation reads as follows: “Words are all we have” (Schaub 2010, p. 185). To law, pragmatics must retort: What exactly do we “have”?

What we “have”, once we know we are engaging with a piece of discourse in the legal domain—let us say of a statute—is all the knowledge that resides in the genre “statute” and the super-genre “law”, such as described in the above citation by Giltrow. We also have higher order knowledge, such as referred to by Rosenberg. Essentially, what we have are the genre concepts and various kinds of interpretive principles, be they obligatory or be they moot.

It may be seen as definitional for types of legal genres that there is an asymmetry not only in types in discourse rights (like topic determination, turn-taking, etc.), but in terms of cognitive environments, as e.g., in depositions. Here, the presumption is that the deposition takers do not have the same knowledge as the deposed persons, whereas in a cross-examination, the cognitive environments likely coincide. Questions asked in cross-examination are not intended to add to the cognitive environments of known facts.

Apart from the issue of the “richness” of the external knowledge, another issue seems to be that these constitute prepackaged knowledge types. An occurrence of a concrete text or discourse is always an instantiation of a type. I would even go so far as to say that you cannot make an utterance that relies only on internal knowledge. Whenever you speak, you invariably speak within a genre.

So there is, to start with, something like a “logical” primacy of genre. What is there first is really not a sentence or utterance, but a genre. Genre is what we “have” first, even before we read the first sentence or hear the first utterance.

In addition, we have to assume not only a logical primacy, but also a psycholinguistically and processing primacy. Genres these pre-existing knowledge complexes, or packages, are available to the hearer, and assumed to be so by the speaker, before any other decoding takes place. Before you read or listen to a text, you know what genre you are engaging with. It is like turning on a specific light, in which you see the incoming figures. The “words” never walk alone.

The language-using organism is not a lonely organism; she is in a lot of company and cannot help being so. Our modeling of what happens in construing meaning should reflect the psycholinguistic primacy of the genre. Our perspective here, as proposed by Jaszczolt (2023), supersedes a more traditional view of meaning making, namely the so-called “pipeline” view. The pipeline metaphor aptly characterizes meaning making as a logically ordered sequentiality, where one step of meaning construction depends on the availability of the former respective “lower” level. In the pipeline view, secondary processes like implicatures are only possible after the existence of propositional explicatures, and the latter themselves only after free enrichment processes have taken place.

Seen from the point of view of the expectations, as current in law, that the meaning is “in the text”, the obvious question from the pipeline point of view is the following: exactly where in the derivational stages is “in the text” located? Each derivational stage offers a different answer, with some answers more or less attractive from the perspective of law (cf. below).

For those who accept this logical sequentiality, with information required at any later stage available earlier stages, the pipeline forms an elegant hypothesis for a real-time, orderly process. If the processing perspective itself forces a concept of genre as a psycholinguistically and informationally plausible concept, it comes at the cost of quantity–arguability. This type of “cost”, in the sense meant by Giltrow, is a cost against the background of what I have termed a language ideology that sees language predominantly as surface forms. My intention with the term “ideology” is not pejorative; I appreciate that many applied pursuits, such as computational approaches and translation, must see language this way, of necessity. For pragmatics, however, such ideology entails the unscientific default expectation that all manner of content and function needs surface signaling, which is incompatible with the arguments about informational bottlenecks advanced by Levinson (cf. above § 2). The absence of redundancy in the shape of non-linguistic knowledge, with all information signaled through code, would in fact make communication impossibly unwieldy. Arguably, 80% of what functions in communication is context, and it requires our package deals to function at all. Hence, the primacy of genre, understood as prepackaged, subconscious knowledge types.

A lot of external knowledge is adduced at several or all stages, depending on the nature of the genre. Where does it all come from? Does it come from a preselected range of genre knowledge? The major determinant of selection is the relevant genre, in the broad way we define it here. It is indeed accessed and selected in an orderly sequence, such that if one type of knowledge is called in at one stage, then it will then require another, or next, element of knowledge. Other elements are present the very moment genre is “switched on”, mostly through external, situational socio-cultural knowledge. The external triggers are contextualization cues which will affect the “modulation of lexicon, grammar, and prosody, the very contextual frame within which it should be interpreted, a bit like a snail that carries its own house around with it” (Levinson 2024, p. 34, referring to Gumperz’ notion of such cues).

One logical question to ask is the following: At each stage, as a different type of knowledge is accessed, is the prior stage discarded? Is each type of knowledge, at each sequential stage, accessed and then discarded? Is the process akin to sequentially reaching into different drawers of a cupboard, each in turn? With so many stages involved, if meaning making really operated in this way, then it would, again, be less efficient than we already know to be. It would take too much processing time and exacerbate informational bottlenecks. Processing constraints therefore force a view that different “drawers” are made salient at different points in the process, and the end result of a “discourse”, the final utterance “point”, is not computable by individual steps. Instead, it must have an element of supersummativity and may even be renegotiated interactively if the genre allows, as it does in conversations. For the time being, this supersummativity will have to remain an element of “unknown dark matter”.

Elder and Jaszczolt (2024, p. 9f) represent an advance on earlier, more or less strictly “pipeline” views, such as are still basic in Relevance Theory. In their “Default Semantics”, Elder and Jasczolt explicitly acknowledge a wide range of types of external sources of information that may be accessed selectively. Their approach ultimately results in a “flexible functional proposition”:

The merger of information coming from different sources, through the associated processes, then produces what are called in Default Semantics merger representation [….], which allows us to represent the composition of meaning that is arrived at by the interactants themselves.

Importantly, what is included in the range of “external” information are not only static types of knowledge, but also processes, including processes of interpretation and modification of knowledge.

All types of enrichments, modulations, primary and secondary, as well as all types of implicatures, require access at so many points in the comprehension process (and are calculated by the speaker to be accessed by hearer) that it suggests a top-down genre perspective. Without genre, as Levinson asks: “How does a recipient find from the forest of possibilities just the implicatures intended within just a few hundred milliseconds?” (Levinson 2024, p. 22). Levinson (2024) discusses the processing and information-theoretical perspectives primarily from the point of view of spoken conversation. But all the issues Levinson discusses apply in other kinds of language use, given some modifications for the case of written discourse and for the handling of particularized implicatures, both of which may involve some amount of conscious System 2 “thinking”. Hard legal interpretative issues of course require conscious reflection, typical both for legal statutes and literary study.

A processing perspective leads inevitably to a genre perspective: the relevant knowledge types are strongly constrained and restricted in such a way that they can plausibly and manageably be accessed at any time with minimal processing effort. So the pre-packaged knowledge is present right from the start of processing. Here is where a classical, explicit “pipeline” view of knowledge processing, while logically elegant, is implausible from a psycholinguistic point of view, for reasons of strict capacity limits and speed of processing (cf. also the argumentation in Kuhlen and Abdel Rahman).

One ramification of this psycholinguistic processing argument is the “functional pragmatic proposition”, suggested by Jaszczolt (2021, 2023):

So, we need to go a step further. We need a unit which belongs to speakers and addressees and reflects their conversational interaction in ‘meaning-making’. But; moreover, it also needs to pertain to truth-conditional content as it is understood in contextualist truth-conditional theories of meaning. For this we move to the concept of a functional proposition.

(Jaszczolt 2021)

A functional proposition is a structured proposition that reflects the composition of the main communicated meaning. It captures the primary intended, recovered, and partly co-constructed meaning as it is understood by interlocutors. As such, to reiterate, it captures the primary communicative function of the utterance. This primary meaning can be directly or indirectly communicated and may or may not correspond to the speaker’s initial intended meaning, which is negotiated interactively. In addition, the structure of the proposition relies on the varied, multimodal informational input in communication. That is, it relies on information about meaning that comes from different sources in communication, not only from the utterance itself. Rather, we communicate by immersing our utterances in a situation that exploits socio-cultural defaults, background information (i.e., common ground) and other information sources, as will be discussed in more detail in Section 8.

A functional proposition contains all interpretations, forces us to give up the “pipeline” as psycholinguistically unrealistic, implausibly cumbersome. So down the drain goes, in psycholinguistic terms, what were once thought to be the autonomous material of “what is said”. Instead, “what is said” is now identical with the functional proposition. It is the implicature “She is the most suitable candidate” that is actually “what is said”, an idea not easily entertained in the legal profession.

Even with the functional proposition understood as “what is said”, it is not the case that the sentence or utterance is prior, and that the pragmatic aspects of meaning, i.e., genre knowledge, are then “added”, incrementally called upon, operated on either in a logical model sequence or a real-time sequence. What is there first is genre and all knowledge, in real, infinitesimal time.

8. Meaning Making in the Law

For theorists, at least, the pipeline view of meaning making has a strong competitor. Yet, the classic concept of sequenced meaning building remains stimulating for linguistics and legal linguistics. “Literal meaning” is, of course, a perennial issue in law. Conceptually, the “literal” could be located after disambiguation: narrowing, modulation, and indexicalization, at the stage of the text sentence or the proposition located at the level of the explicature. If so, it begs the question: Are scope ambiguities resolved by a principle, on the spot? Are they resolved by a canon of interpretation rule, as part of the genre package? Or by larger considerations of plausibility of the resulting solution, i.e., made by a process further along the sequence?

In law, the issue of literal meaning is especially relevant for expressions like “and” and “or” (Smolka and Pirker 2018). Do we assume one literal meaning—as Smolka and Pirker do—and let pragmatic principles—in effect inferences—decide at a later stage in the derivation (as explicatures?) which meanings to apply, or do we assume a proliferation of separate langue meanings (system sentence) to be available at different points along the derivational path?

It turns out that the pipeline model retains some interest for the teaching of the subject, such as the importance of “stages” in this process. Different legal interpretations of language indeterminacy can be conceptualized as being resolved in a staged process. In this framework, the question becomes the following: Where is the proposition located? What part of the proposition carries a truth value? Is such a question needed for linguistic inquiry in legal procedures? A related issue is how long earlier derivational stages are still “available” in comprehension (Recanati 2003 and can be made explicit as a criterion in establishing “what happened” or “what was said” in language crimes (such as hate speech, defamation, incitement, and the like). Of course, the applicable version of “what is said” or the “proposition” would then be the “early” one as the result of primary pragmatic processes, as the explicature, or even text sentence, and not the functional proposition. Recanati (2003, p. 29) explicitly points to the possibility that “derivationally earlier” stages are in fact “skipped”, and given the fully available contextual information, a directly functional proposition in Jaszczolt’s sense is interpreted:

An important difference between the Gricean model (according to which the literal interpretation is processed first) and the parallel model just outlined is this: on the parallel model it is possible for an utterance to receive a non-literal interpretation without the literal interpretation of that utterance being ever computed. The non-literal interpretation of the global sentence does not presuppose its literal interpretation, contrary to what happens at the constituent level. If the non-literal interpretation of some constituent fits the context especially well it may be retained (and the other interpretations suppressed) before the literal interpretation of the sentence has been computed.

So the “availability” (Recanati 2003, p. 20) of earlier derivational stages is not in every case to be presupposed. In fact, it may have to be retro-construed as a task of Kahneman’s System 2.

“What is said”, in the older sense of a derivationally earlier proposition, can matter a lot in courts, especially with lawyers being so preoccupied with the “logical content” of what is said, ascribable to words as such, as something “objective and therefore, it would seem to all involved, demonstrable. The prosecution or defense in cases of language crime, such as hate speech trials, runs on shallow ground when trying to base a legal strategy on an older notion of “what was said” (Guillén-Nieto et al. 2023).

The specific language-ideological view that permeates thinking in the legal profession is aptly characterized by Smolka and Pirker:

While international law literature puts much emphasis on interpreting a “legal text […] in such a way that a reason and a meaning can be attributed to every word in the text” (Linderfalk 2007, p. 108, quoting Haraszti, emphasis—the notion of word appearing to be legalese for conceptual term (Linderfalk 2007, p. 106))—the notion of text seems not worthy of any definition altogether. This may have to do with the fact that a text is not “available” or “readable” independent of interpretation, in the process of which one may then be busy focusing on utterances. The question is as follows: given that a text typically consists of a sequence of utterances, how does this affect the interpretation process?

(Smolka and Pirker 2016, p. 30)

This, then, is the attempt to construe meanings exclusively bottom-up, as follows logically from an assumption of a full-as-possible autonomy, as is standardly assumed to apply in statutes of other legal discourse under the presumption of literalness.

The effect of genre is naturally also felt in the constraints it places on the meaning of non-propositional expressions. For instance, the semantics of “whereas” in setting of paragraphs, or the meaning of the conditional “if” in statutory regulations (Szczyrbak). Conditionals, theticals, discourse markers, and other elements, have characteristic meanings that are special to particular genres.

For instance, tense change is of particular interest in forensic linguistics. In forensic linguistics, tense change is often taken as indicative of lying, or that the statement is simply not true. But in specific types of narrative, those same markers can have the opposite meaning for truth evaluation. Instead, it is the absence of tense change, and the expression of emotion at a narrative peak—an eminent structural and highly diagnostic point in a narrative—that is indicative of lying.

The emphasis here is on what can have: genre provides the “explanation” of why we construe this meaning rather than another. It is built on the knowledge present at whatever point you are at in the particular genre in which you are involved.

This narrowing and specialization of meanings is just another example that raises the question: “How can we know so little, given that we know so much?” The plethora of meanings—the “forest of possibilities”—of words like “and”, and of grammatical forms like tense changes, is reduced by the local type of interaction.

One would expect that move-structure marking, while being the rhetorical core of suasive legal genres, will not be marked. How do hearers know? They are a priori willing to interpret next sentences or utterances as next moves, depending on the connectionist status of the sentence, as predicted by the genre. So the expectation of segmentalization, of surface marking as a default, is misguided on the lower level. To the extent that the corpus is computer-based, there is evidence that, due to the higher redundancy through the technicality of the affordances, there is even less surface marking generally in Internet language.

An interactionist view, where meanings are also top-down co-determined, is communicationally adequate. What you interact with principally is genre knowledge. Genre knowledge includes, of course, terminologies, including legal terminologies with all kinds of vagueness, like in legal standards.

We must assume that the priority of this knowledge in logical and processing terms also applies to, and is indeed a criticism of, the concept of a macrostructure, as it is manifested in the text in the Appendix A. The implication of the formation of the macrostructure is that it is linearly and sequentially constructed on the basis of the incoming propositions. While this may be an elegant concept to explicate propositional meaning, it is, as I have said, psycholinguistically and pragmatically unrealistic. On an incrementalist step-by-step assumption of m-structure building, it is clear that it would simply take far too much time. However, as a logical structure for thinking or teaching about meaning, it is very useful indeed. The text in the appendix possesses both a macro-structure and super-structure, as part of the text’s genre character: opinion text, rhetorical expressive/directive speech act.

Since functional structure is a hallmark of many genres (narrative, legal, etc.) and part of the “switched on” cognitive environment, we cannot be surprised that explicit structure boundary marking is also redundant. Genre knowledge makes us perceive the sentences at the beginnings of the four paragraphs as topic sentences, enabling an immediate expectation that what follows will be a fleshing out of the topic sentences. The reader knows many things before even the first morpheme is decoded. Contributing to this is a meta-relationship of a causal nature: the reason why I am saying/claiming this, hoping to convince is that (“topical sentence”). The reason for my assumption is then the rest of the paragraph, which is processed as giving reasons for the thesis put forward in the topic sentence. This discourse structural fact plays out on the level of relationship between sentences: sentence connection does not obtain between first sentences and what follows.

The text has no surface structure marking, as it is redundant for the reasons just set out. The writer here can well afford to not mark this structural meaning on the surface.

Given the fact that a lot of “pre-propositional” pragmatic knowledge is part of what is said, and that there is an ongoing controversy about how up or down the ”what is said” is to be located, it seems appropriate to include all manner of knowledge functioning in a new concept of what a “proposition” is. It is necessarily one that reflects all types of meaning, and which is “functional” in a bounded piece of discourse, and that includes, but not exhausts, so-called intended speaker-meaning.

Although obsolete from a scientific point of view, it may be useful to keep the notion of “literal meaning” as an “operational” concept for applied purposes, just like all other similar notions, like “basic meaning”, the notion of an autonomous “text”, and so on. But, as I have argued, it will be interesting to see how this can be profitably exploited for explicating and elucidating the handling of meaning in a legal context.

9. Logical vs. Real-Time Ordering

As has been adumbrated here, the logical ordering in the “pipeline” view is, from a real-time processing point, entirely implausible. What is actually there first, in real time, is the situation and the communicative intention, present long before any proposition or system sentence is on the table.

Even earlier ideas about meaning making, assuming that text processing and comprehension started with availability, first, of code information, also assumed the presence at the prior point of non-linguistic knowledge that would enable processing of first stages in the first place.

What is there first is the genre, not the language: this view of the primacy of genre is compatible with the idea of a “functional” proposition in the sense meant by Jaszczolt. It is also compatible with Recanati’s (2003) view, articulated in the following:

As I have been at pains to emphasize, the meaning of the whole is not constructed in a purely bottom-up manner from the meanings of the parts. The meaning of the whole is influenced by top-down, pragmatic factors, and through the meaning of the whole the meanings of the parts are also affected. So we need a more ‘interactionist’ or even ‘Gestaltist’ approach to compositionality.

(Recanati 2003, p. 132)

Genre, with its superstructural internal redundancies, will so strictly constrain any possible next move in a sequentially ordered type of discourse that it is unnecessary to mark it on the surface. Indeed, offering additional signaling materiality would go against expectations in such a constrained local environment. It is part of the interpretive schema in this genre that such semantically conditional structures will be interpreted as conditional in this genre. It would be a waste of signaling materiality to specially mark them on the surface and potentially would be a violation of a communicational maxim of quantity.

This semiotic-economic interpretation accords with a more general view of the relationship between surface and cognition/interpretation in genres. Following the discussion of underdetermination above, it is a well-known fact that certain genres are only minimally defined by linguistic surface information, a phenomenon we designated as “genre-maximalism” above and with Giltrow’s dictum that “form alone” (meaning surface, code marking) does not define genre.

Levinson (2024, p. 30) refers to a more general principle formulated by Sacks and Schegloff: “oversuppose and undertell”. While originally with reference to spoken conversation, the principle can be taken as operant in managing any knowledge with the help of genres. Sacks and Schegloff’s principle implies an economical procedure, achieved by “amplifying coded content by virtue of prearranged rules of thumb or pre-packaging of default assumptions” (Levinson 2024, p. 21)—rules and assumptions such as constitute genres. What this amounts to is a major relativization—and, comparatively, an acute downgrading—of the role of information supplied by code, i.e., by language. In other work, the process has been referred to as the “de-surfacing” (Nicklaus and Stein 2022, pp. 157–77) of analytic concepts, especially as far as the discourse level is concerned. It is a misguided expectation that functional structure parts, as in genres with a pronounced superstructure or conversational moves, would imply that a central information management technique as a design feature of language could not work in the efficient way it does. Still, such beliefs are widespread elements of ideologies of language, such those ideologies still dominant in law.

In some genres “undertelling” seems to apply with a vengeance, especially in artistic, or literary genres, where a minimal definition appears to have taken maximal recourse to non-surface definitional parameters. Just as contracts, for example, are defined by speech-act pragmatic categories, the same is true for literature. Literature is not defined by surface features, but by a specific type of “cooperation” with what is “in the text”, however difficult this has been to conceptualize (cf. above). The location of this meaning level (the proposition) may itself be specifically definitional for literature (one is tempted to say “genre” here). Literature may in fact be a special case for extreme minimalism, as it cannot be surface-linguistically defined even in the most limited way.

Further support for this assumption might be derived from the fact that literature can be generated by machines like ChatGPT. It is not possible to define surface constraints for verbal material that would be definitional for what is machine-generated or what is human-made. It is—as is normal for modern artistic production—the interpretive act that defines what is art. Even the nature of the type of inferential processes (like their non-finiteness in principles, the non-availability of an end state, as in principle underlying relevance theory) is open and non-determinate.

A similar perspective applies to the discussion whether ChatGPT in fact generates new genres. The question is an extrapolation of the issue whether mediality is definitional for genre boundaries. To some extent, mediality surely is definitional. The blog is a child of the Internet (Puschmann 2010), with parentage in the form of the written diary. Arguably, the law, as an abstract canonical body, does not change its nature and its canonical content only by ”packaged” in different new media (Greineder and Stein 2023). Interesting questions are nonetheless being raised for a number of legal genres relating to new genres or genre identity in a new medial garb (Anesa and Engberg 2023).

To the extent that the notion of “intention” is part of a genre characteristic, the issue is slightly different. The question is, instead, the following: can the recovery of the producer intention be at the center of meaning making in ChatGPT-produced text? Given that the producer intention (in legal theory, “intentionalism”) is an important dimension in analyzing texts in law, what does this imply for machine-generated text, given that our notion of intention is that it is something only a human can have? What does it imply for different types of genres in the legal domain? What about statutes? And what about “criminal” texts, such as those under suspicion as “hate speech” (Guillén-Nieto et al. 2023), where a mens rea is crucial for establishing a “fact” and actionability? Hate speech itself is arguably not a genre, although it is strongly tied to intention, and the “linguistic” (surface, code) share of meaning determination can be less than minimal in some cases. By way of a speculation, does “declaring it” (linguistically analyzing it as) a separate genre, tied to contextually documentable intentions, make it easier to hold humans who have used a machine accountable? If so, forensic linguistics would then face a very different type of task.

After the foregoing discussion, the very notion of “genre” naturally raises some more fundamental issues. As I have said, genre, as we understand it, is a concept from cognitively oriented pragmatics. If “text type” considered pragmatic knowledge as a patching-up repository of “context”, genre theory in turn considers surface information as only residually essential, stronger idiomatically, and idiosyncratically determined. Surface information is computed not as types, but only as tokens.

To summarize, there are several related notions of “genre”, with (in the order given) descending degrees of “surface-relatedness”:

• A macro-sense (Rosenbergian) in the sense of law or science being genres.
• A slightly lower level that ties genre more specifically to activity types in specific social or institutional situations (in the sense of Engberg, or activity types as Levinson 1992).
• A mid-level sense (as implicitly advocated here), which would agglomeratively be representative for the macro-level.
• A low-level notion, closer to individual surface form-oriented notions, that often deals with didactic considerations (how to teach genre and or about genre) and which is in a way pre-pragmatic, like the early work on Genre by Bhatia (2023) and at which more surface-dependent notions, such as computational work, must operate. Dorgeloh and Wanner (2022), for example, contribute a valid, corpus-based analysis of surface forms together with their genre privileges of occurrence.

Generally, surface forms are epiphenomena of pragmatic parameters, types of interaction and embeddings in situational, actional contexts. Illocution types would seem to be very important, especially for genres that make up the world of the law in the Rosenbergian sense.

A basic determining constellation for a genre type 2 might well be a type of situation (as part of a no 1 dimension) with its own types of illocution and specific interactive and asymmetric discourse parameters, like a police interview. This function in turn determines the type of “possible content” or ”possible move”. What can come after what, and with which function, would be quite predictable and would require little if any surface marking. Levinson (2024, p. 18) points out that sequentiality is a key indicator: “A central observation in CA is that sequential context in a series of actions is often a powerful heuristic”.

Another direction of radical questioning would ask if it is possible to have genre without any verbal share at all. In contrast to the so-called Sorites paradox (how many grains do you need to minimally have to form a heap? Answer: just one) it would have to be argued that, given the generally subordinate function of surface materials, the limiting case would indeed have to have genre without any verbal share whatsoever, which of course does exist and not only in the field of arts.

If this kind of argumentative radicalization of the theoretical issues is useful for heuristic argumentation, like in any other conceptual context, it yet does not damage the practical ineluctability and the theoretical soundness of the concept of genre. After all, you can also ask the following: “Can you have a phone without representing a phoneme?”

The pervasive share of gradience raises the issue of the “operationalizability” of the concept. The point can and has been raised that what is a genre cannot be “strictly” so, with categorical yes–no boundaries. Neither can a genre be algorithmically defined, as appears to be a precondition for the “new discreteness” postulate of much modern science. Genre cannot be defined under the dictatorship of quantificational approaches, those giving priority to discrete boundary scientific categories, no matter what sense it might make in any individual case.

Many, if not all, of our linguistic categories are fuzzy and gradient, from the phoneme all the way to concept of an NP (“nouniness”). So this cannot be used as a “principled” objection against a gradient concept of genre.

It is perfectly true that the operationalizability issue may be an insurmountable difficulty, except for genres that do have a high degree of surface representation. It would appear that genres in the domain of law, because of their high degree of ritualization indeed do not offer themselves readily for analysis in terms of quantifiable surface markers. Ritualization implies contextualization, and this in turn that less surface signaling is necessary. The “lower”, and well-embedded the units are, such as conversational moves, the higher the level or redundancy of what is possible and expected. Another feasible and common procedure is to work with something like a first derivative or watered-down version of the genre concept and to posit a type of phenomenon closer to a text-type as genre-akin (close to no 4 above), making it amenable to automatic analysis; however, there are some caveats as to a functional interpretability of statistical results and preferences in terms of genre.